LTE optimization engineering handbook (Zhang, Xincheng) (z-lib.org)

LTE Optimization Engineering Handbook LTE Optimization Engineering Handbook Xincheng Zhang China Mobile Group Design Institute Co., Ltd. Beijing, China This edition first published 2018 © 2018 John Wiley & Sons Singapore Pte. Ltd All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/ permissions. The right of Xincheng Zhang to be identified as the author of this work has been asserted in accordance with law. Registered Offices John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA John Wiley & Sons Singapore Pte. Ltd, 1 Fusionopolis Walk, #07-01 Solaris South Tower, Singapore 138628 Editorial Office 1 Fusionopolis Walk, #07-01 Solaris South Tower, Singapore 138628 For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com. Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in standard print versions of this book may not be available in other formats. Limit of Liability/Disclaimer of Warranty While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Library of Congress Cataloging‐in‐Publication Data Names: Zhang, Xincheng, 1970– author. Title: LTE optimization engineering handbook / Xincheng Zhang, China Mobile Group Design Institute Co., Beijing, China. Description: Hoboken, NJ, USA : Wiley, [2017] | Includes bibliographical references and index. | Identifiers: LCCN 2017019394 (print) | LCCN 2017022857 (ebook) | ISBN 9781119159001 (pdf ) | ISBN 9781119158998 (epub) | ISBN 9781119158974 (cloth) Subjects: LCSH: Long-Term Evolution (Telecommunications)–Handbooks, manuals, etc. | Wireless communication systems–Handbooks, manuals, etc. | Computer network protocols–Handbooks, manuals, etc. Classification: LCC TK5103.48325 (ebook) | LCC TK5103.48325 .Z4325 2017 (print) | DDC 621.3845/6–dc23 LC record available at https://lccn.loc.gov/2017019394 Cover Design: Wiley Cover Images: (Yin Yang) © alengo/Gettyimages; (Feng shui compass) © Liuhsihsiang/Gettyimages Set in 10/12pt Warnock by SPi Global, Pondicherry, India 10 9 8 7 6 5 4 3 2 1 v Contents About the Author xvi Preface xvii Part 1 LTE Basics and Optimization Overview 1 1 LTE Basement 1.1 1.1.1 1.1.2 1.2 1.2.1 1.2.2 1.2.3 1.2.4 1.2.5 1.2.6 1.3 1.3.1 1.3.2 1.3.3 LTE Principle 3 LTE Architecture 6 LTE Network Interfaces 7 LTE Services 11 Circuit‐Switched Fallback 12 Voice over LTE 13 IMS Centralized Services 16 Over the Top Solutions 16 SMS Alternatives over LTE 17 Converged Communication 19 LTE Key Technology Overview 19 Orthogonal Frequency Division Multiplexing 20 MIMO 21 Radio Resource Management 22 3 2 LTE Optimization Principle and Method 24 2.1LTE Wireless Optimization Overview 24 2.1.1 Why LTE Wireless Optimization 24 2.1.2 Characters of LTE Optimization 24 2.1.3 LTE Joint Optimization with 2G/3G 25 2.1.4 Optimization Target 25 2.2 LTE Optimization Procedure 26 2.2.1 Optimization Procedure Overview 26 2.2.2 Collection of Mass Nerwork Measurement Data 28 2.2.3 Measurement Report Data Analysis 30 2.2.4 Signaling Data Analysis 31 2.2.5 UE Positioning 32 2.2.5.1 Timing Advance 33 2.2.5.2 Location Accuracy Evaluation 35 vi Contents 2.2.5.3 2.2.5.4 2.2.6 2.2.7 2.3 2.3.1 2.3.1.1 2.3.1.2 2.3.2 2.3.2.1 2.3.2.2 2.3.2.3 2.3.3 2.3.3.1 2.3.3.2 2.3.3.3 2.3.4 2.3.5 2.3.6 2.3.6.1 2.3.6.2 2.3.7 2.3.8 2.3.8.1 2.3.8.2 2.3.9 2.3.10 2.3.11 2.3.11.1 2.3.11.2 2.3.11.3 2.3.11.4 2.3.11.5 2.3.12 2.3.12.1 2.3.12.2 2.3.12.3 2.3.12.4 2.3.13 2.3.13.1 2.3.13.2 2.3.14 2.3.14.1 2.3.14.2 2.3.14.3 Location Support 36 3D Geolocation 37 Key Performance Indicators Optimization 42 Technology Evolution of Optimization 43 LTE Optimization Key Point 44 RF Optimization 44 RSRP/RSSI/SINR/CINR 44 External Interference 48 CQI versus RSRP and SINR 51 CQI Adjustment 51 SINR Versus Load 54 SINR Versus MCS 56 Channel Power Configuration 58 RE Power 58 CRS Power Boosting 64 Power Allocation Optimization 66 Link Adaption 67 Adaptive Modulation and Coding 69 Scheduler 70 Downlink Scheduler 72 Uplink Scheduler 74 Radio Frame 75 System Information and Timers 76 System Information 76 Timers 81 Random Access 83 Radio Admission Control 85 Paging Control 86 Paging 86 Paging Capacity 92 Paging Message Size 95 Smart Paging 95 Priority Paging 96 MIMO and Beamforming 97 Basic Multi‐Antenna Techniques 100 2D‐Beamforming 101 2D MIMO and Parameters 104 Massive‐MIMO 105 Power Control 107 PUSCH/PUCCH Power Control 107 PRACH Power Control 109 Antenna Adjustment 111 Antenna Position 112 Remote Electrical Tilt 113 Antenna Azimuths and Tilts Optimization 117 2.3.14.4 VSWR Troubleshooting 118 2.3.15 Main Key Performance Indicators 120 Contents Part 2 3 Main Principles of LTE Optimization 123 Coverage Optimization 125 3.1 Traffic Channel Coverage 125 3.1.1 Parameters of Coverage 126 3.1.2 Weak Coverage 128 3.1.2.1 DL Coverage Hole 128 3.1.2.2 UL Weak Coverage 128 3.1.2.3 UL and DL Imbalance 129 3.1.3 Overlapping Coverage 129 3.1.4 Overshooting 130 Tx1/Tx2 RSRP Imbalance 132 3.1.5 3.1.6 Extended Coverage 132 3.1.7 Cell Border Adjustment 135 3.1.8 Vertical Coverage 137 3.1.9 Parameters Impacting Coverage 138 3.2Control Channel Coverage 138 4 Capacity Optimization 140 4.1 RS SINR 140 4.2PDCCH Capacity 141 4.3PUCCH Capacity 144 4.3.1 Factors Affecting PUCCH Capacity 145 4.3.2 PUCCH Dimensioning Example 151 4.4Number of Scheduled UEs 152 4.5 Spectral Efficiency 153 4.6DL Data Rate Optimization 154 4.6.1 Limitation Factor 156 4.6.2 Model of DL Data Throughput 157 4.6.3 UDP/TCP Protocol 158 4.6.4 MIMO 161 4.6.4.1 DL MIMO 161 4.6.4.2 4Tx/4Rx Performance 163 4.6.4.3 Transmission Mode Switch 163 4.6.4.4 UL MU‐MIMO 164 4.6.5 DL PRB Allocation and Utilization Mechanism 4.6.6 DL BLER 167 4.6.7 Impact of UE Velocity 169 4.6.8 Single User Throughput Optimization 170 4.6.8.1 Radio Analysis – Assignable Bits 171 4.6.8.2 Radio Analysis – CFI and Scheduling 171 4.6.8.3 Radio Analysis – HARQ 171 4.6.9 Avarage Cell Throughput Optimization 172 4.6.10 Cell Edge Throughput Optimization 172 4.6.11 Some Issues of DL Throughput 173 4.6.11.1 Antenna Diversity not Balanced 173 4.6.11.2 DL Grant is not Enough 173 4.6.11.3 Unstable Rate 175 165 vii viii Contents 4.7UL Data Rate Optimization 175 4.7.1 Model of UL Data Throughput 176 4.7.2 UL SINR and PUSCH Data Rate 176 4.7.3 PRB Stretching and Throughput 179 4.7.4 Single User Throughput Optimization 180 4.7.4.1 Radio Analysis – Available PRBs 181 4.7.4.2 Radio Analysis—Link Adaptation 181 4.7.4.3 Radio Analysis – PDCCH 182 4.7.5 Cell Avarage and Cell‐edge Throughput Optimization 182 4.7.6 Some Issues of UL Throughput 183 4.8 Parameters Impacting Throughput 185 5 Internal Interference Optimization 188 5.1 Interference Concept 188 5.2DL Interference 190 5.2.1 DL Interference Ratio 191 5.2.2 Balance Between SINR and RSRP 192 5.3UL Interference 192 5.3.1 UL Interference Detection 194 5.3.2 Generation of UL Interference 196 5.3.2.1 Cell Loading Versus Inter‐Cell Interference 196 5.3.2.2 Unreasonable UL Network Structure 197 5.3.2.3 Cross slot interference 199 5.3.3 PUSCH Tx Power Analysis 200 5.3.4 UL Effect of P0 and α 202 5.3.5 PRACH Power Control 204 5.3.6 SRS Power Control 206 5.3.7 Interference Rejection Combinin 209 5.4Inter‐Cell Interference Coordination 210 5.5UL IoT Control 210 5.5.1 UL Interference Issues and Possible Solutions 210 5.5.2 UL IoT Control Mechanism 210 5.5.3 PUSCH UL_SINR Target Calculation 212 5.5.4 UL Interference Criteria 213 6 Drop Call Optimization 216 6.1Drop Call Mechanism 216 6.1.1 Radio Link Failure Detection by UE 217 6.1.2 RadioLink Failure Detection by eNB 220 6.1.2.1 Link Monitors in eNB 220 6.1.2.2 Time Alignment Mechanism 221 6.1.2.3 Maximum RLC Retransmissions Exceeded 224 6.1.3 RadioLink Failure Optimization and Recovery 225 6.2Reasons of Call Drop and Optimization 227 6.2.1 Reasons of E‐RAB Drop 227 6.2.2 S1 Release 230 6.2.3 Retainability Optimization 233 6.3RRC Connection Reestablishment 233 6.4RRC Connection Supervision 239 Contents 7 Latency Optimization 7.1 7.2 7.3 7.4 7.5 7.6 User Plane Latency 244 Control Plane Latency 247 Random Access Latency Optimization Attach Latency Optimization 248 Paging Latency Optimization 250 Parameters Impacting Latency 250 244 8 Mobility Optimization 247 254 8.1Mobility Management 255 8.1.1 RRC Connection Management 256 Measurement and Handover Events 256 8.1.2 8.1.3 Handover Procedure 260 8.1.3.1 X2 Handover 261 8.1.3.2 S1 Handover 267 8.1.3.3 Key point of X2/S1 Handover 267 8.2Mobility Parameter 269 8.2.1 Attach and Dettach 272 8.2.2 UE Measurement Criterion in Idle Mode and Cell Selection 273 8.2.3 Cell Priority 276 8.3Intra‐LTE Cell Reselection 276 8.3.1 Cell Reselection Procedure 278 8.3.2 Inter‐Frequency Cell Reselection 279 8.3.3 Cell Reselection Parameters 282 8.3.4 Inter‐Frequency Reselection Optimization 283 8.4Intra‐LTE Handover Optimization 285 8.4.1 A3 and A5 Handover 285 8.4.2 Data Forwarding 290 8.4.3 Intra‐Frequency Handover Optimization 291 8.4.4 Inter‐Frequency Handover Optimization 292 8.4.5 Timers for Handover Failures 296 8.5Neighbor Cell Optimization 297 8.5.1 Intra‐LTE Neighbor Cell Optimization 297 8.5.1.1 Neighbor Relations Table 297 8.5.1.2 ANR 298 8.5.2 Suitable Neighbors for Load Balancing 299 8.6Measurement Gap 299 8.6.1 Measurement Gap Pattern 299 8.6.2 Measurement Gap Versus Period of CQI Report and DRX 304 8.6.3 Impact of Throughput on Measurement Gap 304 8.7Indoor and Outdoor Mobility 305 8.8Inter‐RAT Mobility 306 8.8.1 Inter‐RAT Mobility Architecture and Key Technology 307 8.8.2 LTE to G/U Strategy 309 8.8.3 Reselection Optimization 314 8.8.3.1 LTE to UTRAN 315 8.8.3.2 UTRAN to LTE 319 8.8.4 Redirection Optimization 320 8.8.4.1 LTE to UTRAN 320 ix x Contents 8.8.4.2 UTRAN to LTE 322 8.8.5 PS Handover Optimization 322 8.8.5.1 LTE to UTRAN 322 8.8.5.2 UTRAN to LTE 324 8.8.6 Reselection and Redirection Latency 325 8.8.7 Optimization Case Study 326 8.9Handover Interruption Time Optimization 326 8.9.1 Control Plane and User Plane Latency 329 8.9.2 Inter‐RAT Mobility Latency 332 8.10Handover Failure and Improvement 332 8.11Mobility Robustness Optimization 335 8.12Carrier Aggregation Mobility Optimization 341 8.13FDD‐TDD Inter‐mode Mobility Optimization 345 8.14Load Balance 346 8.14.1 Inter‐Frequency Load Balance 346 8.14.2 Inter‐RAT Load Balance 348 8.14.3 Load Based Idle Mode Mobility 349 8.15High‐Speed Mobile Optimization 351 8.15.1 High‐Speed Mobile Feature 353 8.15.2 Speed‐Dependent Cell Reselection 354 8.15.3 PRACH Issues 356 8.15.4 Solution for Air to Ground 358 9 Traffic Model of Smartphone and Optimization 360 9.1Traffic Model of Smartphone 360 9.1.1 QoS Mechanism 362 9.1.2 Rate Shaping and Traffic Management 366 9.1.3 Traffic Model 371 9.2Smartphone‐Based Optimization 372 9.3High‐Traffic Scenario Optimization 372 9.3.1 Resource Configuration 374 9.3.2 Capacity Monitoring 375 9.3.3 Special Features and Parameters for High Traffic 377 9.3.4 UL Noise Rise 379 9.3.5 Offload Solution and Parameter Settings 379 Part III 10 Voice Optimization of LTE 383 Circuit Switched Fallback Optimization 385 10.1Voice Evolution 385 10.2CSFB Network Architecture and Configuration 386 10.2.1 CSFB Architecture 386 10.2.2 Combined Register 387 10.2.3 CSFB Call Procedure 392 10.2.3.1 Fallback Options 392 10.2.3.2 RRC Release with Redirection 393 10.2.3.3 CSFB Call Procedure 395 10.2.4 Mismatch Between TA and LA 397 Contents 10.3CSFB Performance Optimization 402 10.3.1 CSFB Optimization 402 10.3.1.1 Main Issues of CSFB 402 10.3.1.2 CSFB Optimization Method 403 10.3.2 CSFB Main KPI 407 10.3.3 Fallback RAT Frequency Configuration Optimization 409 10.3.4 Call Setup Time Latency Optimization 411 10.3.4.1 ESR to Redirection Optimization 416 10.3.4.2 Twice Paging 416 10.3.5 Data Interruption Time 418 10.3.6 Return to LTE After Call Complete 419 10.4Short Message Over CSFB 422 10.5Case Study of CSFB Optimization 423 10.5.1 Combined TA/LA Updating Issue 423 10.5.2 MTRF Issue 425 10.5.3 Track Area Update Reject After CSFB 425 10.5.3.1 No EPS Bearer Context Issue 428 10.5.3.2 Implicitly Detach Issue 428 10.5.3.3 MS Identity Issue 428 10.5.4 Pseudo Base Station 428 11 VoLTE Optimization 434 11.1VoLTE Architecture and Protocol Stack 435 11.1.1 VoLTE Architecture 435 11.1.2 VoLTE Protocol Stack 435 11.1.3 VoLTE Technical Summary 438 11.1.4 VoLTE Capability in UE 439 11.2VoIP/Video QoS and Features 442 11.2.1 VoIP/Video QoS 442 11.2.2 Voice Codec 444 11.2.3 Video Codec 446 11.2.4 Radio Bearer for VoLTE 449 11.2.5 RLC UM 454 11.2.6 Call Procedure 457 11.2.6.1 LTE Attach and IMS Register 458 11.2.6.2 E2E IMS Flow 458 11.2.6.3 Video Phone Session Handling 462 11.2.7 Multiple Bearers Setup and Release 466 11.2.8 VoLTE Call On‐Hold/Call Waiting 467 11.2.9 Differentiated Paging Priority 468 11.2.10 Robust Header Compression 470 11.2.10.1 RoHC Feature 470 11.2.10.2 Gain by RoHC 470 11.2.11 Inter‐eNB Uplink CoMP for VoLTE 475 11.3Semi‐Persistent Scheduling and Other Scheduling Methods 477 11.3.1 SPS Scheduling 477 11.3.2 SPS Link Adaptation 478 11.3.3 Delay Based Scheduling 481 11.3.4 Pre‐scheduling 482 xi xii Contents 11.4PRB and MCS Selection Mechanism 484 11.4.1 Optimized Segmentation 484 11.4.2 PRB and MCS Selection 485 11.5VoLTE Capacity 486 11.5.1 Control Channel for VoLTE 487 11.5.2 Performance of Mixed VoIP and Data 488 11.6VoLTE Coverage 491 11.6.1 VoIP Payload and RoHC 492 11.6.2 RLC Segmentation 492 11.6.3 TTI Bundling 498 11.6.4 TTI Bundling Optimization 502 11.6.5 Coverage Gain with RLC Segmentation and TTI Bundling 507 11.6.6 MCS/TBS/PRB Selection 509 11.6.7 Link Budget 510 11.7VoLTE Delay 513 11.7.1 Call Setup Delay 516 11.7.1.1 Call Setup Time 516 11.7.1.2 Reasons for Long Call Setup Time 516 11.7.2 Conversation Start Delay 519 11.7.3 RTP Delay 521 11.7.4 Handover Delay and Optimization 525 11.8Intra‐LTE Handover and eSRVCC 527 11.8.1 Intra‐Frequency Handover 527 11.8.2 Inter‐Frequency Handover 528 11.8.3 Single Radio Voice Call Continuity Procedure 529 11.8.4 SRVCC Parameters Optimization 539 11.8.4.1 Handover Parameters 539 11.8.4.2 SRVCC–Related Timer 539 11.8.5 aSRVCC and bSRVCC 543 11.8.6 SRVCC Failure 543 11.8.7 Reducing SRVCC Voice Gap and eSRVCC 545 11.8.7.1 Voice Interruption Time during SRVCC 545 11.8.7.2 eSRVCC 549 11.8.8 Fast Return to LTE 552 11.8.9 Roaming Behavior According to Network Capabilities 555 11.9Network Quality and Subjective Speech Quality 555 11.9.1 Bearer Latency 558 11.9.2 MoS 561 11.9.2.1 Voice Quality 561 11.9.2.2 Video Quality 570 11.9.3 Jitter 571 11.9.4 Packet Loss 572 11.9.5 One Way Audio 575 11.9.6 PDCP Discard Timer Operation 576 11.10Optimization 577 11.10.1 Distribution of Main Indicators of Field Test 580 11.10.2 Compression Ratio and GBR Throughput 584 11.10.3 RB Utilization 584 11.10.4 BLER Issue 587 Contents 11.10.5 Quality Due to Handover 589 11.10.6 eSRVCC Handover Issues 589 11.10.7 Packet Loss 592 11.10.7.1 Packet Loss due to Poor RF 592 11.10.7.2 Packet Loss due to Massive users 592 11.10.7.3 Packet Loss Due to Insufficient UL grant 592 11.10.7.4 Packet Loss due to Handover 601 11.10.7.5 Packet Loss Due to Network Issue 601 11.10.8 Call Setup Issues 601 11.10.8.1 Missed Pages 602 11.10.8.2 IMS Issues 604 11.10.8.3 Dedicated Bearer Setup Issues 609 11.10.8.4 CSFB Call Issues 612 11.10.8.5 aSRVCC Failure 612 11.10.8.6 RF Issues 612 11.10.8.7 Frequent TFT Updates 617 11.10.8.8 Encryption Issue 618 11.10.9 Call Drop 619 11.10.9.1 Call Drop 619 11.10.9.2 Radio Link Failure 622 11.10.9.3 RTP‐RTCP Timeout 624 11.10.9.4 RLC/PDCP SN Length Mismatch 626 11.10.9.5 IMS Session Drop 626 11.10.9.6 eNB/MME Initiated Drop 632 11.10.10 Packet Aggregation Level 632 11.10.11 VoIP Padding 633 11.10.12 VoIP Ralated Parameters 635 11.10.13 Video‐Related Optimization 635 11.10.13.1 Video Bit Rate and Frame Rate 637 11.10.13.2 Video MoS and Audio/Video Sync 637 11.10.14 IMS Ralated Timer 637 11.11UE Battery Consumption Optimization for VoLTE 638 11.11.1 Connected Mode DRX Parameter 643 11.11.2 DRX Optimization 644 11.11.2.1 State Estimation 644 11.11.2.2 DRX Optimization and Parameters 644 11.11.2.3 KPI Impacts with DRX 648 11.11.3 Scheduling Request Periodicity and Disabling of Aperiodic CQI 652 11.12Comparation with VoLTE and OTT 654 11.12.1 OTT VoIP User Experience 654 11.12.2 OTT VoIP Codec 657 11.12.3 Signaling Load of OTT VoIP 658 Part IV 12 Advanced Optimization of LTE 663 PRACH Optimization 665 12.1Overview 665 12.2PRACH Configuration Index 669 12.3RACH Root Sequence 673 xiii xiv Contents 12.4PRACH Cyclic Shift 674 12.4.1 PRACH Cyclic Shift Optimization 674 12.4.2 Rrestricted Set 679 12.5Prach Frequency Offset 682 12.6Preamble Collision Probability 683 12.7Preamble Power 684 12.8Random Access Issues 687 12.9RACH Message Optimization 689 12.10Accessibility Optimization 692 12.10.1 Reasons for Poor Accessibility 692 12.10.2 Accessibility 693 12.10.3 Accessibility Analysis Tree 695 12.10.4 Call and Data Session Setup Optimization 697 12.10.5 RACH Estimation for Different Traffic Profile 698 13 Physical Cell ID Optimization 702 13.1Overview 702 13.2PCI Optimization Methodology 703 13.2.1 PCI Group Optimization 705 13.2.2 PCI Code Reuse Distance 705 13.2.3 Mod3/30 Discrepancy Analysis 708 13.2.4 Collision and Confusion 708 13.3PCI Optimization 709 14 Tracking Areas Optimization 711 14.1TA Optimization 712 14.1.1 TA Update Procedure 713 14.1.2 TA Optimization and TAU Failure 715 14.2TA List Optimization 716 14.3TAU Reject Analysis and Optimization 719 15 Uplink Signal Optimization 721 15.1Uplink Reference Signal Optimization 721 15.1.1 Coding Scheme of UL RS 722 15.1.2 Correlation of UL Sequence Group 723 15.1.2.1 UL Sequence Group Hopping 725 15.1.2.2 UL Sequence Hopping 726 15.1.2.3 UL Cyclic Shift Hopping 726 15.1.3 UL Sequence Group Optimization 727 15.2Uplink Sounding Signal Optimization 729 15.2.1 SRS Characters 730 15.2.2 Wideband SRS Coverage 736 15.2.3 Dynamic SRS Adjustment Scheme 736 15.2.4 SRS Selection Dimension and Confliction 737 15.2.5 SRS Conflict and Optimization 739 16 HetNet Optimization 741 16.1UE Geolocation and Identification of Traffic Hot Spots 741 16.2Wave Propagation Characteristics for HetNet 745 Contents 16.3New Features in HetNet 746 16.4Combined Cell Optimization 747 16.5Cell Range Expansion Offset 748 16.6HetNet Cell Reselection and Handover Optimization 17 QoE Evaluation and Optimization Strategy 17.1QoE Modeling 753 17.2Data Collecting and Processing 756 17.3QoE‐Based Traffic Evaluation 757 17.3.1 Online Video QoE 757 17.3.1.1 Video Quality Monitoring Methods 761 17.3.1.2 RATE Adaptive Video Codecs 763 17.3.1.3 Streaming KPI and QoE 764 17.3.1.4 Video Optimization 766 17.3.2 Voice QoE 769 17.3.3 Data Service QoE 770 17.3.3.1 Web browsing 770 17.3.3.2 Online Gaming 774 17.4QoE Based Optimization 776 18 Signaling‐Based Optimization 780 18.1S1‐AP Signaling 780 18.1.1 NAS signaling 782 18.1.2 Inactivity Supervision 783 18.1.3 UE signaling Management 785 18.2Signaling radio bearers 786 18.3Signaling Storm 788 18.4Signaling Troubleshooting Method 18.4.1 Attach Failure 788 18.4.2 Service Request Failure 796 18.4.3 S1/X2‐Based Handover 796 18.4.4 eSRVCC Failure 798 18.4.5 CSFB Failure 800 Appendix 802 Glossary of Acronyms 820 References 823 Index 825 788 752 751 xv xvi About the Author Xincheng Zhang graduated from the Beijing University of Posts and Telecommunications in 1992. He has worked in mobile communication for 25 years as a technical expert with a solid understanding of wireless communication technologies. Starting out in the early days of GSM rollouts, he has many years of planning and optimization experience in 2G, 3G, 4G, and 5G networks, working in operator and vendor environments. He is working as a senior wireless network specialist in the fields of antenna arrays, analog/digital signal processing, radio resource management, and propagation modeling, and so on. He has participated in many large‐scale wireless communication system designs and optimization for a variety of cellular systems using various radio access technologies, including GSM, CDMA, UMTS, and LTE. xvii Preface Mobile communication has become ubiquitous and mobile Internet traffic is continuously growing due to the technology that provides broadband data rates (3G, LTE) and the growing number of mobile dongles and mobile devices like tablets or smartphones that enable the usage of a tremendous number of internet applications through the mobile access. Mobility, broadband, and new device technology have changed the way people connect and communicate. Smartphones have changed the characteristics of the control and user plane, leading to a huge impact on RAN and e2e network capacity, end‐user experience, and perception of the network, which has changed with the advent of new devices and applications. Subscribers want the same internet experience that they have at home, anytime, anywhere, so the long‐term network is under strain and optimization is needed. Many of the new services aim to enhance the experience of a phone conversation by allowing sharing of content other than speech. The quality of all these services needs to be monitored to ensure that users experience a high‐quality service. Low‐bit cost is an essential requirement in a scenario where high volumes of data are being transmitted over the mobile network. To achieve the proposed goals, a very flexible network that aggregates various radio access technologies is created. This network should provide high bandwidth, from 50‐100 Mbps for high mobility users, to 1Gbps for low mobility users, technologies that permit fast handoffs, which is necessary in a QoS framework that enables fair and efficient medium sharing among users. The core of this network should be based on internet protocol version 6—IPv6, the probable convergence platform of future services. The other key factor to the success of the network is that the terminals must be able to provide wireless services anytime, everywhere, and must adapt seamlessly to multiple wireless networks, each with different protocols and technologies. Subscriber loyalty has shifted to devices and applications; quality of experience becomes the fundamental service provider’s differentiation. In this background, the 3GPP long term evolution (LTE) is created and adopted all over the world. High‐speed, high‐capacity data standard for mobile devices is on its way to becoming a globally deployed standard for the fourth generation of mobile networks (4G) supported by all major players in the industry. LTE builds on EUTRAN, a new generation radio access network, and the evolved packet core (EPC), which provides flexible spectrum usage and bandwidths, high data rate, low latency, and optimized resource usage. As LTE has been used as a mobile broadband service, we need to understand the effects of the LTE terminals providing services and how to optimize the network. Actually, for operators, the challenge is not only to optimize 2G, 3G, and 4G but also how to balance the use of those systems, including WiFi. The entire service delivery chain needs to be optimized and the optimization aims to improve network efficiency and the mobile broadband service quality. xviii Preface It is known that LTE doesn’t have basic voice and SMS support. To mitigate this, 3GPP roposes a fallback to circuit‐switched (CS) network for voice and SMS. Although voice has p loosened its weight in the overall user bill with the rise of more and more data services, voice is the dominant source of revenue for operators and is expected to remain so for the foreseeable future. On one hand, 3GPP defines the concept of CS fallback for the EPC, which forces the UE to fall back to the GERAN/UTRAN network where the CS procedures are carried out. On the other hand, voice over LTE will become a mainstream mobile voice technology. VoLTE ecosystem is building up fast due to its strong end‐to‐end VoLTE solution portfolio including the LTE radio, EPC, mobile softswitch, IMS, and its extensive delivery capabilities of complex end‐to‐ end projects. It is the world’s most innovative voice solution for LTE‐based networks and big VoLTE growth is expected since wide‐scale commercial VoLTE started in Korea during 2012. The operators will want to have the best possible observability for this new voice service with fast call setup, low latency, and high speech quality. Actually, VoLTE will be among the most critical and complex technologies mobile operators will ever deploy as VoLTE testing is quite complex due to inherent intricacy of the technology covering the IMS/EPC core, radio network, and UE/IMS client. They expect to be able to monitor how their customers experience accessibility, retainability, as well as the quality of the voice service. Obviously, much of the observability is already in place, but there are reasons to believe that there are missing parts. Under this background, to meet customers’ requirements for high‐quality networks, LTE trial networks must be optimized during and after project implementation. The basis and the main inputs that allowed the creation of this handbook were based on optimization experience, whereas the scope of this book is to provide network engineers with a set of processes and tasks to guide them through the troubleshooting and optimization. For a network optimization engineer, he/she needs to know how good the quality of mobile broadband applications is, and how the network capabilities impact the performance, and how to identify the most critical network KPIs that impact customer experiences. This book is divided into four parts. The first is called “LTE Basics and Optimization Overview,” and proceeds with an introduction to general principles of data transfer of LTE. This chapter is dedicated to the reader who is not acquainted with this area. The second part, titled “Main Principles of LTE Optimization,” and the third part, “Voice Optimization of LTE,” makes up the core of the book, since it describes coverage, capacity, interference, mobility optimization, and includes two chapters that provide step‐by‐step optimization of CFSB and VoLTE. The fourth part “Advanced Optimization of LTE” takes a more applied perspective in PRACH, PCI, TA, QoE, Hetnet, and signaling optimization. Thanks to the many people in China who shared their views acquired from years of experience and valuable insights in wireless optimization, the Optimization Handbook covers the basics of optimization rules, solutions, and methods. It is evident that this book does not cover many other important areas of optimization of LTE networks. Nonetheless, I sincerely hope that readers will find the information presented to be interesting and useful to inspire you to go and do optimization with a renewed vigor in order to help you build a better LTE network. January 1, 2017 Xincheng Zhang 823 References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 GSMA PRD IR.94 ‐ IMS Profile for Conversational Video Service GSMA PRD IR.92 ‐ IMS Profile for Voice and SMS GSMA PRD IR.39 ‐ IMS Profile for High Definition Video Conference (HDVC) Service GSMA PRD RCC.07 ‐ Rich Communication Suite 5.1 Advanced Communications Services and Client Specification. 3GPP TS 23.292 IP Multimedia Subsystem IMS centralized 3GPP TS 23.272, Circuit Switched Fallback in Evolved Packet System 3GPP TS 24.237 IP Multimedia Core Network subsystem IP Multimedia Subsystem (IMS) service continuity. 3GPP TS 24.301: “Non‐Access‐Stratum (NAS) protocol for Evolved Packet System (EPS).” 3GPP TS 26.114 ⟪IMS;Multimedia Telephony;Media handling and interaction⟫ 3GPP TS 36.104: “Evolved Universal Terrestrial Radio Access (E‐UTRA); Base Station (BS) radio transmission and reception.” 3GPP TS 36.211: “Evolved Universal Terrestrial Radio Access (E‐UTRA); Physical Channels and Modulation.” 3GPP TS 36.212: “Evolved Universal Terrestrial Radio Access (E‐UTRA); Multiplexing and channel coding.” 3GPP TS 36.213 “Physical layer procedures.” 3GPP TS 36.331 “Radio Resource Control (RRC).” 3GPP TS 36.321 “Medium Access Control (MAC) protocol specification.” 3GPP TS 36.322 “Radio Link Control Protocol Specification.” 3GPP TS 36.323 – PDCP 3GPP TS 36.401 – E‐UTRA Architecture Description 3GPP TS 36.410 – S1 interface general aspects & principle; 3GPP TS 36.411 – S1 interface Layer 1 3GPP TS 36.412 – S1 interface signaling transport 3GPP TS 36.413 – S1 application protocol S1AP 3GPP TS 36.414 – S1 interface data transport 3GPP TS 36.420 – X2 interface general aspects and principles 3GPP TS 36.421 – X2 interface layer1 3GPP TS 36.422 – X2 interface signaling transport; 3GPP TS 36.423 – X2 interface application part X2AP 3GPP TS 36.442 – UTRAN Implementation Specific O&M Transport 3GPP TS 29.118 – SGs application part 3GPP TS 29.274 – GTP‐C 3GPP TS 29.281 – GTP‐U 3GPP TS 23.280 (Sv Itf ), 3GPP 23.401 (access) LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. 824 References 31 IETF RFC SCTP – RFC 2960, RFC 4960 32 IETF RFC 3095, “RObust Header Compression (ROHC)” 33 IETF RFC 4815, “RObust Header Compression (ROHC):Corrections and Clarifications to 34 35 36 37 38 39 40 41 42 43 RFC 3095” IETF RFC 4995, “The RObust Header Compression (ROHC) Framework” IETF RFC 3843,“RObust Header Compression(ROHC): A Compression Profile for IP” IETF RFC 768, Unreliable Datagram Protocol (UDP) IETF RFC 3550, Real‐Time Protocol (RTP) IMS Architecture, TS23.228, TS23.401, TS24.229, TS33.203 Session Initiation Protocol (SIP), TS24.229, RFC3261 Session Description Protocol (SDP), TS24.229, RFC4566 IMS Authentication and Security, TS23.228, RFC 2406, RFC2451, RFC3602 Media Aspects, TS26.114, TS23.167, TS24.623, RFC4867, RFC3966, RFC4825 ITU‐T Recommendation H.264 (03/2010 or newer): “Advanced video coding for generic audiovisual services.” 1 Part 1 LTE Basics and Optimization Overview 3 1 LTE Basement Mobile networks are rapidly transforming—traffic growth, bit rate increases for the user, increased bit rates per radio site, new delivery schemes (e.g., mobile TV, quadruple play, IMS), and a multiplicity of RANs (2G, 3G, HSPA, WiMAX, LTE)—are the main drivers of the mobile network evolution. The growth in mobile traffic is mainly driven by devices (e.g., smartphone and tablet) and applications (e.g., mainly web browsing and video streaming). To cope with the increasing demand, mobile networks have based their evolution on increasingly IP‐centric solutions. This evolution relies primarily on the introduction of IP transport, and secondly, on a redesign of the core nodes to take advantage of the IP backbones. The first commercial LTE network was opened by Teliasonera in Sweden in December 2009, and marks the new era of high‐speed mobile communications. The incredible growth of LTE network launches boomed between 2012 and 2016 worldwide. It is expected that more than 500 operators in nearly 150 countries will soon be running a commercial LTE network. Mobile data traffic has grown rapidly during the last few years, driven by the new smartphones, large displays, higher data rates, and higher number of mobile broadband subscribers. It is expected that the mobile broadband (MBB) subscriber numbers will double by 2020, reaching over 7 billion subscribers, that MBB data traffic will grow fourfold by 2020, reaching over 19 petabytes/ month. The internet traffic, MBB subscriber, and relative mobile data growth is illustrated in Figure 1.1. 1.1 LTE Principle To provide a fully mature, real‐time–enabled, and MBB network, structural changes are needed in the network. In 2005, the 3GPP LTE project was created to improve the Universal Mobile Telecommunications System (UMTS) mobile phone standard to cope with future requirements, which resulted in the newly evolved Release 8 (Rel 8) of the UMTS standard. The goals include improving efficiency, lowering costs, improving services, making use of new spectrum opportunities, and better integration with other open standards. Long‐term evolution (LTE) is selected as the next generation broadband wireless technology for 3GPP and 3GPP2. The LTE standard supports both FDD (frequency division duplex), where the uplink and downlink channel are separated in frequency, and TDD (time division duplex), where uplink and downlink share the same frequency channel but are separated in time. After Rel 8, Rel 9 was a relatively small update on top of Rel 8, and Rel 10 provided a major step in terms of data rates and capacity with carrier aggregation, higher‐order Multi‐Input‐Multi‐Output (MIMO) up to eight antennas in downlink and four antennas in uplink. The support for heterogeneous network (HetNet) was included in Rel 10, also known as LTE‐Advanced (Figure 1.2). LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Bandwidth demand 140,000.0 Video Conferencing Video Streaming Audio Streaming 120,000.0 P2P 100,000.0 Video Streaming / TV VoIP 80,000.0 Ecommerce Web / Internet Web Surfing 60,000.0 Rich text Email high 40,000.0 VoIP (VoLTE) low Text Email low high 20,000.0 0.0 Delay demand 2011 2012 2013 2014 2015 2016 2017 2018 2019 Internet traffic on LTE Mobile traffic type (Source: ABI Research) Mobile Data Traffic 8,000 7,000 18 000 6,000 16 000 Europe LAT APAC total MEA NAM 14 000 5,000 12 000 4,000 10 000 8 000 3,000 6 000 2,000 4 000 2 000 1,000 MBB subscriber growth Figure 1.1 The internet traffic, MBB subscriber, and relative mobile data growth. MBB data traffic 20 19 20 20 17 18 20 16 20 15 20 14 20 13 20 20 11 12 20 10 20 09 20 08 20 07 20 06 20 05 20 2020 2019 2018 2017 2016 2015 2014 2013 2012 2011 2009 2010 2008 2007 2006 0 2005 0 20 MBB Subscriber in Million 20 000 LTE Basement Phase 2+ (Release 97) Release 99 Release 6 Release 8 GPRS 171.2 kbit/s UMTS 2 Mbit/s HSUPA 5.76 Mbit/s LTE +300 Mbit/s Release 9/10 LTE Advanced GSM 9.6 kbit/s EDGE 473.6 kbit/s HSDPA 14.4 Mbit/s Phase 1 Release 99 Release 5 HSPA+ 28.8 Mbit/s 42 Mbit/s Release 7/8 Figure 1.2 3GPP standard evolution. 5G 10 Gbps Peak Average LTE-A 1 Gbps 5G in 2020 (Ave. ~1Gbps Peak ~5Gbps) LTE 100 Mbps HSDPA, HDR Cat. 11 (Ave. ~240Mbps, Peak ~600Mbps) 10 Mbps 1 Mbps WCDMA, CDMA2000 Cat. 9 (Ave. ~180Mbps, Peak ~450Mbps) Cat. 6 (Ave. ~120Mbps, Peak 300Mbps) Cat. 4 (Ave. ~24Mbps, Peak 150Mbps) Cat. 3 (Ave. ~12Mbps, Peak 100Mbps) 100 Kbps HSDPA (Ave. ~2Mbps, Peak 14Mbps) 2000 2005 2010 2015 2020 Figure 1.3 Downlink data rate evolution. Among the design targets for the first release of the LTE standard are a downlink bit rate of 100 Mbit/s and a bit rate of 50 Mbit/s for the uplink with a 20‐MHz spectrum allocation. Smaller spectrum allocation will of course lead to lower bit rates and the general bit rate can be expressed as 5 bits/s/Hz for the downlink and 2.5 bits/s/Hz for the uplink. Rel 10 (LTE‐Advanced), was completed in June 2011 and the first commercial carrier aggregation network started in June 2013 (Figure 1.3). LTE provides global mobility with a wide range of services that includes voice, data, and video in a mobile environment with lower deployment cost. The main benefits of LTE include (Figure 1.4): ●● Wide spectrum and bandwidth range, increased spectral efficiency and support for higher user data rates 5 LTE Optimization Engineering Handbook 100% CDF 6 “Average”Tput ~0.12bps/Hz 50% “Cell Edge”Tput ~0.06bps/Hz (95% coverage) 5% cell edge cell centre Tput Figure 1.4 Throughput of a user, 10 users evenly distributed in cell. ●● ●● ●● ●● Reduced packet latency and rich multimedia user experience, excellent performance for outstanding quality of experience Improved system capacity and coverage as well as variable bandwidth operation Cost effective with a flat IP architecture and lower deployment cost Smooth interaction with legacy networks LTE air interface uses orthogonal frequency division multiple access (OFDMA) for downlink transmission to achieve high peak data rates in high spectrum bandwidth. LTE uses single carrier frequency division multiple access (SC‐FDMA) for uplink transmission, a technology that provides advantages in power efficiency. LTE supports both FDD and TDD modes, with FDD, DL, and UL transmissions performed simultaneously in two different frequency bands, with TDD, DL, and UL transmissions performed at different time intervals within the same frequency band. LTE supports advanced adaptive MIMO, balance average/peak throughput, and coverage/cell‐edge bit rate. Compared to 3G, significant reduction in delay over air interface can be supported in LTE, and it is suitable for real‐time applications, for example, VoIP, PoC, gaming, and so on. Spectrum is a finite resource and FDD and TDD system will support the future demand, which are shown in Figure 1.5. TDD spectrum can provide 100‐150MHz of additional bandwidth per operator, TD‐LTE spectrum with large bandwidth will be a key to operators future network strategy and one of the way to address capacity growth. 1.1.1 LTE Architecture LTE is predominantly associated with the radio access network (RAN). The eNodeB (eNB) is the component within the LTE RAN network. LTE RAN provides the physical radio link between the user equipment (UE) and the evolved packet core network. The system architecture evolution (SAE) specifications defines a new core network, which is termed as evolved packet core (EPC) including all internet protocol (IP) networking architectures (Figure 1.6). Evolved NodeB (eNB): Provides the LTE air interface to the UEs, the eNB terminates the user plane (PDCP/RLC/MAC/L1) and control plane (RRC) protocols. Among other things, it performs radio resource management and intra‐LTE mobility for the evolved access system. At the S1 interface toward the EPC, the eNB terminates the control plane (S1AP) and the user plane (GTP‐U). LTE Basement B42 (3,5GHz) 250 TDD FDD 200 200 150 B43 (3,7GHz) BW (MHz) B40 (2,3GHz) B3 (1,8GHz) 100 B28 (700MHz) 100 90 75 60 50 B5 (850MHz) 35 25 B8 (900MHz) B2 (1,9GHz) B39 20 (1,9GHz) 200 70 60 B1 B10 (2,1GHz) (1,7/2,1GHz) 50 B7 (2,6GHz) B38 (2,6GHz) 0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 3400 3600 3800 4000 Frequency (MHz) Figure 1.5 Spectrum of LTE. Mobility Management Entity (MME): A control plane node responsible for idle mode UE tracking and paging procedures. The Non‐Access Stratum (NAS) signaling terminates at the MME. Its main function is to manage mobility, UE identities, and security parameters. The MME is involved in the EPS bearer activation, modification, deactivation process, and is also responsible for choosing the SGW for a UE at the initial attach and at time of intra‐ LTE handover involving core network node relocation. PDN GW selection is also performed by the MME. It is responsible for authenticating the user by interacting with the home subscription server (HSS). Serving Gateway (SGW): This node routes and forwards the IP packets, while also acting as the mobility anchor for the user plane flow during inter‐eNB handovers and other 3GPP technologies (2G/3G systems using S4). For idle state UEs, the SGW terminates the DL data path and triggers paging when DL data arrives for the UE. Packet Data Network Gateway (PDN GW): Provides connectivity to the UE to external packet data networks by being the point of exit and entry of traffic for the UEs. The PDN GW performs among other policy enforcement, packet filtering for each user and IP address allocation. Policy and Charging Rules Function (PCRF): The PCRF supports policy control decisions and flow based charging control functionalities. Policy control is the process whereby the PCRF indicates to the PCEF (in PDN GW) how to control the EPS bearer. A policy in this context is the information that is going to be installed in the PCEF to allow the enforcement of the required services. Home Subscription Server (HSS): The HSS is the master database that contains LTE user information and hosts the database of the LTE users. 1.1.2 LTE Network Interfaces LTE network can be considered of two main components: RAN and EPC. RAN includes the LTE radio protocol stack (RRC, PDCP, RLC, MAC, PHY). These entities reside entirely within the UE and the eNB nodes. EPC includes core network interfaces, protocols, and entities. These entities and protocols reside within the SGW, PGW, and MME nodes, and partially within the eNB nodes. 7 HSS • Subscription Profiles • Security information • MME (IP) address for UE S6a UE • IMEI (equipment) • IMSI (SIM card) • Temporary GUTI • User Plane IP MME • Mobility Management • Session Management • Security Management • Selects SGW based on TA • Selects PGW based on APN S11 S1-MME PCRF:Policy & Charging Rules Function DNS • TA to SGW IP query • APN to PGW query PGW: Packet Data Network Gateway HSS:Home Subscriber Server EPC:Evolved packet Core PCRF • QoS rules • Charging rules Gx Rx X2 S1u eNB • Radio control and resource management • Inter eNB communication via X2 Figure 1.6 Nodes and functions in LTE. S5/S8 SGW • Data forwarding • Data buffering SGi PDN (i.e. IMS or internet) PGW • Gateway between the internal EPC network and external PDNs • User IP address allocation • User plane QoS enforcement SGW:Serving Gateway UE:User Equipment EUTRAN:Evolved UTRAN eNodeB:Enhanced Node B VLR:Visitor Location Register MSC:Mobile Switching Centre MME:Mobility Management Entity LTE Uu: LTE UTRAN UE Interface LTE Basement Uu: Uu is the air interface connecting the eNB with the UEs. The protocols used for the control plane are RRC on top of PDCP, RLC, MAC, and L1. The protocols used for the user plane are PDCP, RLC, MAC, and L1. LTE air interface supports high data rates. LTE uses OFDMA for downlink transmission to achieve high peak data rates in high spectrum bandwidth. LTE uses SC‐FDMA for uplink transmission, a technology that provides advantages in power efficiency. S1: The interface S1 is used to connect the MME/S‐GW and the eNB. The S1 is used for both the control plane and the user plane. The control plane part is referred to as S1‐ MME and the user plane S1‐U. The protocol used on S1‐MME is S1‐AP on the radio network layer. The transport network layer is based on IP transport, comprising SCTP on top of IP. The protocol used on S1‐U is based on IP transport with GTP‐U and UDP on top. The X2 interface is a new type of interface between the eNBs introduced by the LTE to perform the following functions: handover, load management, CoMP, and so on. X2‐UP protocol tunnels end‐user packets between the eNBs. The tunneling function supports are identification of packets with the tunnels and packet loss management. X2‐UP uses GTP‐U over UDP/ IP as the transport layer protocol similar to S1‐UP protocol. X2‐CP has SCTP as the transport layer protocol is similar to the S1‐CP protocol. The load management function allows exchange of overload and traffic load information between eNBs, which helps eNBs handle traffic load effectively. The handover function enables one eNB to hand over the UE to another eNB. A handover operation requires transfer of information necessary to maintain the services at the new eNB. It also requires establishment and release of tunnels between source and target eNB to allow data forwarding and informs the already prepared target eNB for handover cancellations. NAS is a control plane protocol that terminates in both the UE and the MME. It is transparently carried over the Uu and S1 interface. S6a: S6a interface enables transfer of subscription and authentication data between the MME and HSS for authenticating/authorizing user access to the EUTRAN. The S6a interface is involved in the following call flows, initial attach, tracking area update, service request, detach, HSS user profile management, and HSS‐initiated QoS modification, and so on. S11: Reference point between MME and SGW. This is a control plane interface for negotiating bearer plane resources with the SGW. The above‐mentioned LTE network interfaces are shown in Figure 1.7. IP connection between a UE and a PDN is called PDN connection or EPS session. Each PDN connection is represented by an IP address of the UE and a PDN ID (APN). As shown in Figure 1.8, there are two different layers of IP networking. The first one is the end‐to‐ end layer, which provides end‐to‐end connectivity to the users. This layers involves the UEs, the PGW, and the remote host, but does not involve the eNB. The second layer of IP networking is the EPC local area network, which involves all eNBs and the SGW/PGW node. The end‐to‐end IP communications is tunneled over the local EPC IP network using GTP/UDP/IP. Moreover, in LTE, IDs are used to identify a different UE, mobile equipment, and network element to make the EPS data session and bearer establishment, which can refer to Annex “LTE identifiers” for reference; the summary of IDs is shown in Table 1.1. 9 SGi Interface Comunicates CPG with external networks. Diameter SCTP IP L2 S6a Interface AAA interface between MME and HSS that enables user access to the EPS L1 IMS/External IP networks HSS IP L2 S11 Interface Control plane for creating, modifying and deleting EPS bearers. MME IP L2 S1-MME Interface Reference point for control plane protocol between E-UTRAN and MME TCP IP L2 L1 L1 S10 S1-AP Gx L2 L1 SCTP Rx Interface Transport policy control, charging and QoS control. IP S10 Interface AAA interface between MME and HSS that enables user access to the EPS L1 Diameter SGi UDP GTPv2-C UDP PCRF GTP-C S6a Rx IP L2 S11 Gx Interface Provides transfer of policy and charging Rules from PCRF to PDN Gw. PDN GW S5/S8 Serv GW S5/S8 Interface Control and user plane tunneling between Serving GW and PDN GW S1-U S1-MME L1 Diameter TCP IP L2 L1 GTP-C/GTP-U UDP IP X2-AP GTP-U SCTP UDP IP L2 L1 Figure 1.7 LTE network interfaces. X2 Interface Connects neigboring eNBs eNB X2 S1-U Interface Reference point for user plane protocol between E-UTRAN and MME GTP-U L2 UDP L1 IP L2 L1 LTE Basement UE Control plane end-to-end layer Uu UE IP address App IP PDCP User plane eNB RRC Signaling MME SGW PGW S1 S11 GTP-C S11 GTP-C Signaling S1 Bearer DRB eNB S5 Bearer SGW S5/S8 PGW SGi IP GTPu UDP IP APN GTPu UDP IP End to End service PDCP GTPu UDP IP EPS Bearer (ID) E-RAB (ID) RLC RLC MAC MAC L2 L2 L2 PHY PHY L1 L1 L1 P D N S1-u Figure 1.8 LTE‐EPC control and data plane protocol stack. Table 1.1 Classification of LTE identification. Classification LTE identification UE ID IMSI, GUTI, S‐TMSI, IP address, C‐RNTI, UE S1AP ID, UE X2AP ID Mobile equipment ID IMEI Network element ID GUMMEI, MMEI, Global eNB ID, eNB ID, ECGI, ECI, P‐GW ID Location ID TAI, TAC Session/bearer ID PDN ID (APN), EPS bearer ID, E‐RAB ID, DRB ID, LBI, TEID 1.2 LTE Services LTE is an all packet‐switched technology. The telephony service on LTE is a packet‐switched mobile broadband service relying on specific support in LTE radio and EPC, which is needed to meet the expectations of telephony. On the other hand, the handling of voice traffic on LTE handsets is evolving as the mobile industry infrastructure evolves toward higher, eventually ubiquitous, and finally, LTE availability. Central to the enablement of LTE smartphones is to meet today’s very high expectation for the mobile user experience and to evolve the entire communications experience by augmenting voice with richer media services. Voice solutions of LTE include VoLTE/SRVCC, RCS, OTT, CSFB, SVLTE, and so on. LTE radio and EPC architecture does not have a circuit‐switched (CS) domain available to handle voice calls as being done in 2G/3G. The voice traffic in the LTE network is handled through different procedures. The first one, which is mainly used, still remains on the circuit switch network (e.g., 2G or 3G) by maintaining either parallel connection and registration on these network or by switching to them whenever a voice call is initiated or terminated. The second one, which is when the voice call stands over LTE, the voice service is named VoLTE or VoIMS when the IP multi‐media system (IMS) service function is included. Video in LTE is one of the most importanr services. The demand for video content continues to grow among data services. Web video traffic growth has accelerated, as the number of internet‐ enabled devices has increased and more people depend on the mobile internet. Recently, a group of key operators, infrastructure, and device vendors announced a joint effort to facilitate the evolution of mobile communication toward RCS (rich communication suite). The core feature set of RCS includes the following services: enhanced phonebook, with service capabilities and presence enhanced contacts information; enhanced messaging, which 11 12 LTE Optimization Engineering Handbook enables a large variety of messaging options including chat and messaging history, and enriched call, which enables multimedia content sharing during a voice call. It is believed that RCS is a promising evolution in LTE, many operators have announced to support the RCS. 1.2.1 Circuit‐Switched Fallback The basic principle of circuit‐switched fallback (CSFB) is that once originating or receiving a CS voice call by the UE connected over LTE, it will move to either GSM or UMTS network (fallback) where the call proceeds. One major requirement for the realization of CSFB is the overlay of LTE with GSM, UMTS, or both. It is the quickest implementation both at terminal and at network sides and is mandatory for international roaming scenarios. With CSFB, UE will attach to the network through LTE, MME will ask MSC to update UE location in its database, when the UE is operating in LTE (data connection) mode and when a call comes in, the LTE network pages the device. The device responds with a special service request message to the network, and the network signals the device to move to 2/3G to accept the incoming call. Similarly for outgoing calls, the same special service request is used to move the device to 2/3G to place the outgoing call. CSFB for operator means very little investment since only few modifications are required in the network, additional interface (SGs) between MME and MSC is required shown in Figure 1.9. With basic CSFB implementation, the additional delay to set up the voice call is less than 1.3s to 3G or about 2.8s to 2G, which is acceptable from an end‐user perspective. This delay is significantly reduced with the activation of PS handovers when falling back to 3G and of RRC release with 3GPP Rel 9 redirections to 2G/3G. The CSFB option offers complete services and feature transparency by enabling mobile service providers to leverage their existing GSM/ UMTS network for the delivery of CS services, including prepaid and postpaid billing. SGs interface is used to carry signaling to move the access network carrying the voice traffic from 2/3G to the LTE and from LTE to 2/3G. This interface maintains a connection between the MSC/VLR and the MME and its main role is to handle signaling and voice by SGsAP application. Gn‐C interface is the interface connecting the MME to the SGSN in the pre‐Rel 8, it is replaced by the S3 for Rel 8 or later. This interface is required when a CSFB call is established to initial the signaling with SGSN. In case CSFB with PS handover the data established over the LTE will be carried over 2/3G network, the interface Gn‐C or S3 is used to establish the signaling sessions with the SGSN to forward pending data over the LTE toward the 2/3G packet core. To forward the data from the PGW, an additional interface named Gn‐U is required between the SGSN and the PGW in pre‐Rel 8 and the S4 interface between the SGSN and the SGW in Rel 8 or later. Uu UTRAN Iu-ps SGSN Gs Gb UE Um Iu-cs GERAN MSC/ VLR A Gn SGs S1-MME MME S11 LTE Uu E-UTRAN S1-U Figure 1.9 Standard architecture for CSFB. SGW S5/S8 PGW LTE Basement Data LTE (eNode B) Voice 2G/3G Base Station Figure 1.10 Dual radio handsets. CSFB is a single radio solution of handset, in order to make or receive calls, the UE must change its radio access technology from LTE to a 2G/3G technology, and uses network signaling to determine when to switch from the PS network to the CS network. The shortcoming is that someone on a voice call will not be able to use the LTE network for browsing or chatting, and so on. Except CSFB, dual‐radio handsets (SVLTE) shown in Figure 1.10 support simultaneous voice and data— voice provided through legacy 2G or 3G network and data services provided by LTE. Dual‐radio solutions use two always‐on radios (and supporting chipsets), one for packet‐switched LTE data and one for circuit‐switched telephony, and as a data fallback where LTE is not available. The dual radio has the benefit in which simultaneous CS voice and LTE data is available; the drawback is the complexity from the device point of view, since more radio components are required increasing the cost, size, and power consumption. Dual‐radio solutions also force the need for double subscriber registration leading to split legacy and LTE records in the subscriber data managers. As a matter of fact, lack of dual‐radio eco‐system for 3GPP markets and the top six main chipset vendors are addressing the 3GPP market with singe‐radio terminal and CSFB, while the top chipset vendors for 3GPP2 markets are supporting dual‐radio solution for the 3GPP2 market. The above considerations have lead to a clear split in the market for early LTE support of voice services with mobile networks based on 3GPP technologies adopting CSFB, while 3GPP2 markets have adopted a dual‐radio solution for early LTE deployments. CSFB addresses the requirements of the first phase of the evolution of mobile voice services, which commercially launched in several regions around the world in 2011. CSFB has become the predominant global solution for voice and SMS inter‐operability in early LTE handsets, primarily due to inherent cost, size, and battery life advantages of single‐radio solutions on the device side. CSFB is the solution to the reality of mixed networks today and throughout the transition to ubiquitous all‐LTE networks in the future phases of LTE voice evolution. 1.2.2 Voice over LTE After CSFB, LTE voice evolution introduces native VoIP on LTE (VoLTE) along with enhanced IP multimedia services such as video telephony, HD voice and rich communication suite (RCS) additions like instant messaging, video share, and enhanced/shared phonebooks. The voiceover LTE solution (VoLTE) is defined in the GSMA1 Permanent Reference Document (PRD) IR.92,2 based on the adopted one‐voice profile (v 1.1.0) from the One Voice Industry Initiative. Video‐related additions are described in GSMA IR.94. 1 At the 2010 GSMA mobile world congress, GSMA announced that they were supporting the one voice solution to provide voice over LTE. After that, industry aligned 3GPP based e2e solution for GSM equivalent voice services over LTE. 2 The VoLTE IR.92 is from October 2010 put in maintenance mode and only corrections of issues that may cause frequent and serious misoperation will be introduced. 13 14 LTE Optimization Engineering Handbook VoLTE specifies the minimum requirements to be fulfilled by network operators and terminal vendors in order to provide a high‐quality and interoperable voice over LTE service. The VoLTE solution is scalable and rapidly deployed, offering rich multimedia and voice services, seamless voice continuity across access networks, and the re‐use of existing network investments including business and operational support assets. In terms of the operators, the deployment of VoLTE means that it is opened to the mobile wideband speech path of evolution. Also, VoLTE can offer a competitive advantage by providing a superior voice service quality with HD voice and video, shortening setup times for the calls and guaranteeing bit rate, and offering simultaneous LTE data together with the voice call. Finally, a richer end‐user experience; to be able to provide end users the benefit of real‐time communications can be another VoLTE attraction. Better multimedia, video‐conferencing, or video chat while still maintaining a voice call, are all possible revenue opportunities of VoLTE. Introducing VoLTE on a standard‐based IMS provides the service provider with a true converged network where services are available regardless of the access type network. Blending services with an IMS service architecture enables an operator to cost‐effectively build integrated service bundles. VoLTE can evolve voice services into rich multimedia offerings, including HD voice, video calling, and other multimedia services (i.e., start a voice session, add and drop media such as video, and add callers, presence) available anywhere on any device, combining mobility with service continuity. VoLTE is an advancement from today’s voice and video telephony to full‐fledged multimedia communication to utilize the full potential of LTE and to improve customer experience. The IP‐based call is always anchored in IMS core network to carry and establish a voice call over an LTE network. Now, in both 3GPP and 3GPP2 markets, there is a clear consensus to adopt the IMS‐based VoLTE solution for the LTE deployments. Two transport modes are also used on the network and determines the quality of the voice call over an IP network. The VoIP’s best effort, mainly over the internet and based on some widely deployed applications, such as Skype, Google talk, and MSN, uses this mode with no guarantee of the quality. Other technology such as LTE propose to carry the VoIP with the guarantee of the quality of this call over the end‐to‐end network. For VoLTE, the installed solution aims at being partially compliant with GSMA PRD IR.92.3 One voice was an effort to use already‐defined standards to specify a mandatory set of functionality for devices, the LTE access network, the evolved packet core network, and the IP multimedia subsystem in order to define a voice and SMS over LTE solution using an IMS architecture. Some VoLTE handsets are already commercial including the features such as emergency call, location based services, and so on. In case VoLTE through IMS is the mode used, two connections are required with the LTE network—Rx interface between the P‐CSCF and the PCRF and the Gx interface between the PCRF and the PGW for dynamic PCC rules. The Gm interface is a virtual interface established between the SIP application on the end user and the P‐CSCF function of the IMS network where it is connected (Figure 1.11). Along with VoLTE introduction, 3GPP also standardized Single Radio Voice Call Continuity (SRVCC) in Rel 8 specifications to provide seamless continuity when an UE handovers from LTE coverage (E‐UTRAN) to UMTS/GSM coverage (UTRAN/GERAN). With SRVCC, which is depicted in Figure 1.12, the calls are anchored in IMS network while UE is capable of transmitting/receiving on only one of those access networks at a given time. SRVCC protocol evolution have different types according to the function. There are bSRVCC (before alerting 3 Complementary scenarios are also beign defined in the VoLTE profile extension (IR.93) to cope with the cases where LTE coverage needs to be complemented with existing WCDMA/GSM CS coverage. LTE Basement A GERAN Gb S4 Gn SGSN IuPS IuCS UTRAN S3/Gn HSS ISUP Sv S6d Gm MSC IMS Gi SGW Gm Rx Sv S6a EUTRAN S5 MME S1-MME S11 PGW Gx SGi PCRF S1-U Figure 1.11 Standard architecture for VoLTE. SRVCC VoLTE CS Core SRVCC SRVCC CS Legacy RAN SRVCC IMS SRVCC SRVCC Evolved Packet Core LTE RAN SRVCC CSFB Semi-Persistent Scheduling TTI Bundling Common IMS SRVCC RCS CSFB Fast Return after CSFB Emergecy call on VoLTE Emergecy call w/SRVCC Rel-8 Rel-9 eSRVCC aSRVCC Rel-10 SRVCC function rSRVCC vSRVCC Rel-11 Figure 1.12 SRVCC and evolution. SRVCC), aSRVCC (alerting phase SRVCC), vSRVCC (video SRVCC), and vSRVCC (reverse SRVCC, HO 3G/2G → LTE). Up to now, VoLTE launches are taking place in Korea, the United States (AT&T, T‐Mobile, Verizon), Russia (MTS), and Asia (NTT Docomo, SingTel, M1, Starhub, HKT). T‐Mobile U.S. launched VoLTE in Seattle on May 22, 2014. AT&T launched in three markets on May 23, 2014 with “crystal clear conversations.” SingTel launched on May 31, 2014 in Singapore using 4G Clear Voice. In 2015 and 2016, more and more countries launched VoLTE, like China, Canada, France, and Denmark. 15 16 LTE Optimization Engineering Handbook 1.2.3 IMS Centralized Services In a CS network telephony, services are provided by the MSC (based on the subscription data in the HLR). In IMS telephony, services are provided by the telephony application server. Multiple service engines introduce synchronization problems and differences in user experience. IMS centralized services avoids these problems by assuming that only one service engine will be used. IMS plays an essential role in IMS centralized services. The UE performs SIP (session initiation protocol) registration with the IMS network. IMS‐AKA (IMS‐authentication and key agreement) procedures are followed for authentication. Integrity protection, whereby integrity of SIP signaling messages is ensured, is mandatory. The use of ISIM (IP multimedia services identity module) or USIM (UMTS subscriber identity module) is required during the IMS authentication. SIP signaling messages are ASCII text messages and could thus be quite large. Hence, signaling compression is mandatory to reduce the bandwidth requirements, especially for over‐the‐air transmission. IMS centralized services (ICS) enable the use of the IMS telephony service engine for originating and terminating services regardless if a UE is connected via a LTE PS access network or connected via a GSM/WCDMA CS access network. For terminating calls, ICS determines the access network currently in use by a UE to deliver the call via the correct access network. ICS requires an IMS service centralization and continuity application server. 1.2.4 Over the Top Solutions At the same time, there are already a number of applications providing over the top (OTT) voice service on smartphones, which can be used over Wi‐Fi connection but also over cellular networks. OTT application is completely transparent to network and also out of operators’ control. OTT services are those provided without special consideration at the network level (i.e., no special treatment with respect to QoS). Examples of these types of services are YouTube, Vimeo, and DailyMotion, which are very popular today. Skype and GoogleTalk have nearly a billion registered users worldwide. Apple has sold countless iPhones and iPads, many of which are capable of FaceTime video calling. These services are provided directly by content providers (and usually over content delivery networks), generally without any arrangement with the network providers sitting between the content and its consumers. Nowadays, some OTT solutions, such as Skype and FaceTime, often come preinstalled on smartphones, and as these devices become much more widespread, the adoption of OTT solutions for video‐calling services will also increase. LTE supports high bandwidth, low latency, always online, all IP and other characteristics, it is convenient for the development of OTT. OTT application providers have delivered very popular voice, video, messaging, and location services that are shifting consumers’ attention and usage. In addition, while OTT players currently generate revenue using the operator’s network for service delivery, the operator itself doesn’t gain any associated increase in revenues. The Figure 1.13 shows MoS performance based on data from the South Korean market’s most OTT‐friendly operator. In the future, the proportion of OTT voice may be more and more high, especially in the area of long distance calls, as these solutions are familiar to subscribers and have driven user expectations. However, a fully satisfactory user experience cannot be provided by OTT solutions, as there are no QoS measures in place, no handover mechanism to the circuit‐switched network, no widespread interoperability of services between different OTT services and devices, and no guaranteed emergency support or security measures. Consequently, the adoption of OTT clients is directly dependent on mobile broadband coverage and the willingness of subscribers to use a service that lacks quality, security, and flexibility. For example, with VoLTE, using LTE Basement Range : <= x < MOS P863(POLQA)(SEOUL_MOS-M10) MOS P863(POLQA)(SEOUL_KAKAO-M10) MOS P863(POLQA)(SEOUL_SKYPE-M10) MOS P863(POLQA)(SEOUL_GTALK-M10) 100.0 Gtalk 3.43 (ave) 90.0 Kakao 2.95 (ave) 80.0 70.0 60.0 50.0 VoLTE 4.08 (ave) 40.0 30.0 Skype 3.38 (ave) 20.0 10.0 0.0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 MOS-LQO (SWB mode) 3.00 3.25 3.50 3.75 4.00 4.25 4.50 4.75 5.00 Typical operator acceptance criteria Figure 1.13 MoS‐LQO (SWB, super wideband AMR mode). SRVCC, the voice communication can be maintained when the user is moving out of LTE coverage versus OTT call might drop in this case. Voice over LTE delivers very high spectral efficiency, OTT solutions that do not benefit from radio performance enhancements implemented specifically for VoLTE, for example, dynamic scheduling, RoHC, TTI bundling, packet segmentation, and admission control. VoLTE also provides better quality than OTT services and narrowband CS calls, as well as improved call setup time, HD voice quality, E2E QoS guaranteed, longer battery life, emergency call availability, and more efficiency than OTT in terms of network resource consumption. Also, rich multimedia voice could be offered directly enabled on top of VoLTE as well, delivering high user experiences beyond voice and SMS, providing to consumers instant messaging, chat, video, and file sharing. 1.2.5 SMS Alternatives over LTE Basically two options are available for SMS, SMS without IMS, and SMS with IMS. When the device registers as an IMS user, then an incoming SMS will be directed to IMS and delivered via IP. For SMS using CSFB (SMS over SGs), UE can send and retrieve SMS using NAS signaling over LTE. This solution requires SGs interface between CS core and EPC to transport SMS to/ from UE. The SGs interface must support mobility management and paging procedures between the EPS and the CS domain. All the procedures are based on the SGs procedures. SMS over IP is a solution to transport SMS over any PS access using IMS and is mandated to be supported by VoLTE terminals (Figure 1.14). If not registered in IMS, and CS attached, then incoming SMS will be delivered to CS domain. Both options support UE‐originated and UE‐terminated SMSs. In the non‐IMS approach, the MME needs to support the SGs interface toward MSC, MSC pages the UE over LTE via this new channel for terminating SMS. The UE includes the SMS inside a NAS signaling message, and the MME forwards the SMS to the MSC. SMS is supported in both idle mode and connected mode. Of course, if the UE is in idle mode, it is needed to first establish connectivity between the UE and the E‐UTRAN and between the UE and the MME. The UE needs to send 17 18 LTE Optimization Engineering Handbook SMSo SGs Existing SMS SMS NNI CS Core (MSS) GSM / WCDMA RAN SMSo SGs SMS IP SGs SMS IP SMSoIP IMS Core SMSo SGs SMSo SGs SMS IP SMS IP Evolved Packet Core LTE RAN Originating Terminating – SMSoSGs – SMSoIP (VoLTE deployed) GSM HLR DTAP MAP DTAP MAP MSC RANAP/NAS MAP SMS-C WCDMA MSC SGsAP/NAS MAP SIP LTE IP-SMGW Figure 1.14 SMSoSGs and SMSoIP. SMS Using Legacy Framework Legacy CS Core Connected Mode Service Request Page Message SMS via NAS Signaling UE Forwarding of SMS MME MSC Server EPC E-UTRAN Figure 1.15 SMS using legacy framework. a service request message to the MME to exit the idle mode. Once the MSC receives the service request message from MME, it will send the SMS via SGs interface to the MME, which will tunnel the short message to the UE. For a UE‐terminated SMS, a page message would be sent to the UE to get the UE out of the mode. If the UE is in connected mode, it already has all the links established. This will remove the extra service request/paging type signaling exchanges. The UE and the MME can directly place an SMS in a NAS signaling message (Figure 1.15). In the case of SMS using IMS, the UE needs to implement the functions of an SMS‐ over‐IP sender and an SMS‐over‐IP receiver. The IMS core network performs functions of an IP short message gateway (IP‐SM‐GW). For a receiver to get the SMS, the receiver needs to do IMS registration and indicate its capability to receive traditional short messages (Figure 1.16). LTE Basement IMS SMS in a SIP Message CSCF IMS SMS IP-SM-GW UE IP Short Message Gateway (IP-SM-GW) SM-over-IP sender SM-over-IP receiver Figure 1.16 SMS using IMS. 1.2.6 Converged Communication Converged communication is network convergence, media convergence, IT and CT convergence, as well as the convergence of communication and social networks. It will leverage cloud computing, mass data, and other emerging technologies to help drive the next paradigm shifts of the RCS initiative. RCS IP call is derived from VoLTE specifications and as such can be offered over any IP connectivity as an extension to VoLTE. Its primary use case is when VoLTE is not available or not wanted by the user for voice and/or video calls, for example, when roaming or while only CS service is available for voice. RCS IP call uses the same control and media plane as VoLTE but adds RCS‐specific policies. Unlike VoLTE, a RCS IP call does not rely on specific support from the access regarding QoS and mobility; hence operators can enable it over Wi‐Fi, LTE, and even 3G without having VoLTE‐specific support in those accesses. Historically, RCS and VoLTE have been standardized separately from each other, meaning that coexistence aspects have not always been considered. However, that has changed since 2011. RCS, and in particular, RCS 5.1, takes coexistence between VoLTE and RCS into consideration. VoLTE and RCS complement each other, where VoLTE is the basis for primary real‐time communication, and RCS provides enrichment and creates added value in the overall communication experience. RCS can also be deployed with CS voice before VoLTE deployment. So, RCS and VoLTE make use of one number for any service. RCS extends reachability outside the LTE access technology and opens up communication for both multi access and multi device usage. Many operators have different plans for introduction of VoLTE and RCS in their networks. In Europe we see RCS first, while in North America and Korea, VoLTE has already been launched. Depending on the strategy, different evolution paths for coexistence will exist, RCS first, VoLTE first, or a combination from day one (Figure 1.17). 1.3 LTE Key Technology Overview For LTE air interface technology, the choice of multiple‐access schemes was made in December 2005, with orthogonal frequency division multiple access (OFDMA) being selected for the downlink, and single carrier‐frequency division multiple access (SC‐FDMA) for the uplink. In LTE downlink, OFDM will be used to schedule the UEs differently in time and frequency. With these techniques the UEs can be scheduled and receive user data in the downlink. In the uplink, SC‐FDMA is similar to downlink and it is possible to schedule the UEs differently in time and frequency. LTE supports FDD and TDD mode. FDD and TDD LTE mode can be combined (depends on UE capabilities) in the same physical layer. 19 20 LTE Optimization Engineering Handbook Interoperable voice and SMS over LTE (IR.92) HD Voice HD video calling CS-coexistence Multi-device and multiaccess Advanced services roaming VoLTE Messaging Cloud RCS v2/3/4 RCSe RCS v5 RCS-e RCSe feature set RCS IP VoIP Consumer Communication Multimedia RCS 5.x RCS IP Video Additional differentiators Alignment to RCS5 VoLTE Figure 1.17 RCS is continuously innovating and service model are always changing. 1.3.1 Orthogonal Frequency Division Multiplexing OFDMA is the multi‐access technology related with OFDM, the combination of TDMA and FDMA essentially. OFDMA has high spectrum utilization efficiency due to orthogonal subcarriers need no protect bandwidth, support frequency link auto adaptation and scheduling, and easy to combine with MIMO, but it needs strict requirement of time‐frequency domain synchronization and high PAPR. At the transmitter the coded and modulated data stream is split up to a number of sub‐streams. The number of sub‐streams can range from typically 12 (one resource block) and up to 1200 (100 resource blocks at 20MHz bandwidth). Each stream is fed into the IFFT block and transformed into a corresponding subcarrier. Each subcarrier in OFDM is 15 kHz. One subcarrier carries one OFDM symbol. With this subcarrier size, symbol time is 66.7μs, which should be much longer than the delay spread in order to keep the ISI (inter‐symbol interference) low. This choice is based on the average radio channel delay spread (a measure of the radio channel time dispersion) and the coherence time (a measure of how slow the radio channel changes), which should fulfill “Delay spread << Symbol time < Coherence time.” Also, the cyclic prefix should be longer than the expected delay spread in order to completely remove ISI. However, if the symbol time is too long (i.e., longer than the coherence time), the radio channel will change considerably during one symbol (Figure 1.18). This would lead to inter carrier interference (ICI). The biggest drawback of OFDMA is that it suffers from high peak‐to‐average power ratio (PAPR). The higher the PAPR, the higher output power back‐off (OBO) is needed to avoid clipping of transmitted signal, which in turn results in higher adjacent carrier leakage ratio (ACLR) and increased error vector magnitude (EVM). High OBO decreases power efficiency of the transmitter. Similar to OFDMA, the SC‐FDMA physical resource can be seen as a time‐frequency grid with the additional constraint that a resource assigned to a UE must always consist of a set of consecutive subcarriers in order to keep the single‐carrier property of the uplink transmission that can release the UE PA limitation caused by high PAPR. The SC‐FDMA subcarrier spacing equals Δf=15 kHz and resource blocks, consisting of 12 subcarriers in the frequency domain, are defined also for the uplink. LTE Basement Δf = 1/Tu OFDM subcarrier spacing. Orthogonal Subcarriers Centre subcarrier Not Orthogonal Frequency Channel Bandwidth Figure 1.18 OFDMA. 1.3.2 MIMO MIMO (multiple input multiple output) technology is a standard feature of LTE networks, and it is a major piece of LTE’s promise to significantly boost data rates and overall system capacity. MIMO systems use more than one transmit antenna (Tx) to send a signal on the same frequency to more than one receive antenna (Rx). MIMO works under rich scattering conditions; signals from different TX take multiple paths to reach the UE at different times. In order to achieve promised throughputs in LTE systems, operators must optimize their networks’ multipath conditions for MIMO, targeting both rich scattering conditions and high SNR for each multipath signal. MIMO builds on Single Input Multiple Output (SIMO), also called receive diversity that have been around for decades, as well as Multiple Input Single Output (MISO), also called transmit diversity, used in most advanced cellular networks today. SIMO and MISO are both techniques seeking to boost signal to noise ratio (SNR) in order to compensate for signal degradation. MIMO can work as a combination of SIMO and MISO techniques, resulting in even greater SNR gains, further boosting coverage and data rates. It is the well‐known techniques named transmit diversity and spatial multiplexing. The other technique in LTE MIMO is called beamforming. It is only within the last decade that there has been significant interest in such antenna arrays for cellular systems. Beamforming generally requires a different antenna configuration from MIMO operations, which all the transmission paths are closely correlated, and each signal is affected by the transmission medium in a similar way. In this case, the multiple antennas can operate as though they were a single, high‐power antenna with its main lobe illuminating a specific area. Beamforming aims to concentrate the available bandwidth in the targeted area, improving the quality of transmission 21 22 LTE Optimization Engineering Handbook or the signal to interference plus noise ratio (SINR). TD‐LTE can make particularly good use of beam‐forming antennas, as the uplink and downlink are duplexed in the time domain and transmitted and received signals at the same frequency. 1.3.3 Radio Resource Management Radio resource management (RRM) includes assignment, re‐assignment, and release of radio resources. Figure 1.19 shows time scale for different RRM mechanism in LTE network. Radio bearer control is responsible for the establishment, maintenance, and release of radio resources associated with specific radio bearers. The radio bearer control function must consider the overall resources situation and the quality of service (QoS) requirements of in progress sessions when admitting new sessions. The radio bearer control function must also maintain the quality of existing sessions when conditions change due to environmental and mobility activity. Admission and congestion control is a cell‐based operation applied to both uplink and downlink. It is one of the key RRM functions. It is responsible for maximizing the radio resource utilization while meeting QoS requirements of new and existing sessions by intelligent admission or rejection of new radio bearer requests or modification of existing bearers. The task of admission control is to admit or to reject the requests for establishment of radio bearers (RB). Admission control is performed when there are new incoming calls or incoming handover attempts. The system resource limitations and QoS satisfaction ratio are the main considerations for admission control. The allocation/retention priority (ARP) can be used to classify gold, silver, and bronze categories with different admission control thresholds. ARP is an attribute of services and is inherited from evolved packet core. Congestion control is critical to maintain the system stability and deliver acceptable QoS when the system is in congestion. The load measures include the power, the available physical resource block at the air interface, and the transmission resource usage at S1‐U interface. Congestion control method including release low‐priority services to alleviate the overloaded system, or GBR downsizing by sacrificing the quality of GBR services slightly but still maintaining acceptable quality. Connection mobility control is responsible for the management of radio resources during active or idle mode mobility of the UEs. It handles the communication of mobility related parameters to the UE, and makes handover decisions based on both radio conditions and network‐ loading conditions. Layer 2 RRM TTI 1 ms Packet scheduling, fast DL AMC, fast ATB Channel fading time 10 ms Outer link quality control Layer 3 RRM Burst duration L3 signaling delay 100 ms Slow UL LA/AM C/ATB Dynamic MIMO Control Slow UL power control UE DRX/DTX control Figure 1.19 Time scale for different RRM mechanism. Upper layer related control Inter-handover time 1s Channel and location variations 10 s Intercell interference coordination Connection mobility control Load balancing, congestion control Radio admission control Time scale LTE Basement In some situation of commercial LTE network, some serving cells have high load but other neighbor cells’ loads are low because of the differentiation of UE services. Under this condition, it can trigger load balancing algorithm. Load balancing can also be triggered during network attach as part of initial network entry or idle exit. However, eNB will only load balance to different carrier on same sector for network attach, idle exit process. In addition, inter‐RAT radio resource management is responsible for resource management related to handovers with or redirection to a cell from a different radio technology. Handovers may be performed due to resource issues, radio conditions, or network preference specified by the operator. 23 24 2 LTE Optimization Principle and Method 2.1 LTE Wireless Optimization Overview 2.1.1 Why LTE Wireless Optimization The optimization of LTE network mainly refers to the pre‐optimization and the continuous optimization before and after the network launched. Network optimization results and the level of network optimization work, directly related to the future performance of the network stability and capacity. A good network optimization can fully reduce interference level of the whole network, improve the network performance and call success rate, reduce service interruption, improve the data throughput, optimize the whole network handover success rate, and improve the network capacity. Network optimization work is a continuous daily work. Optimization is necessary so the network performance satisfies certain thresholds or targets for key performance indicators (KPIs) agreed beforehand with the operator. After a network is built and before is launched on air, it is necessary to perform the pre‐launch optimization where the common process is to divide the network in groups of sites (clusters) and optimize these clusters so agreed KPIs are achieved. The amount of (pre‐launch) optimization needed depends on the quality of the planning. Changes during pre‐launch optimization are mainly physical (e.g., antenna tilts and azimuths) although they may include also some parameter changes with the scope of optimizing the coverage and the quality of the network. As there is no/very little traffic on the network, counters don’t provide statistically reliable information. Therefore, drive testing is the main optimization method during the pre‐launch optimization to achieve certain field KPIs. After the launch, networks are “alive,” always changing (e.g., traffic conditions, addition of new sites, new software upgrades) so optimization is still needed to keep the high level of performance defined by the KPIs. Since there is traffic on the network, counter information is reliable and it is possible to have a centralized view of how the whole network is performing. In general terms, pre‐launch optimization focuses on a “coarse” tuning of the network and counter based optimization focuses on a fine tuning of the network (i.e., parameter based). Drive testing may still be needed to satisfy operator’s demands and to optimize mature networks (post‐launch optimization) although the scope will be smaller than during pre‐launch optimization. 2.1.2 Characters of LTE Optimization There is an obvious difference between LTE and 2G, 3G wireless network optimization. 2G and 3G networks have had their own standard wireless network optimization process, and formed a set of KPI to reflect the overall situation of the network, including the capacity, coverage, LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. LTE Optimization and Principle and Method acces success rate, quality, and handover. The main difference between LTE wireless network and 2/3G optimization includes: ●● ●● ●● ●● LTE optimization is more complex, traffic and control channels need to be considered the various resources such as frequency domain, time domain, code domain, interference domain, and so on. LTE only supports PS domain operations, but supports a variety of different QoS services, the service based optimization is also a challenge, such as VoIP, and so on. When the carriers deploy LTE system with the reuse factor equals 1, eNB will endure strong co‐channel interference. Although some of the system of same frequency interference avoidance algorithm such as frequency domain scheduling, inter‐cell interference coordination, smart antenna beamforming are deployed, actually co‐channel interference cannot be completely avoided, it will inevitably affect the network’s overall performance. This presents a major challenge to the optimization of LTE. From the operator point of view, optimization in 2/3G system and LTE coexistence (share antenna) phase needs to be considered. LTE initial emphasis is foucused on solving coverage problem. The main task is to avoid decreasing of 2G, 3G network performance, maintain the 2/3G services continuity, but also highlight the LTE service experience. In the mature period of business expansion to LTE, it needs to consider the load balance between LTE and 2/3G, improve the utilization rate of network resources. The mobile network ecosystem is nowdays very complicated and OMC statistics only p rovide an average and flat view of the cell that are not enough to have a clear picture of the actual status. There is the need of a new approach to collect and analyse such data and drive properly the network optimization and expansion. 2.1.3 LTE Joint Optimization with 2G/3G Optimization strategies should also consider especially where the LTE network shares antenna and/or feeder cable infrastructure with legacy 3G and/or 2G networks. Due to 2/3G network in speech coverage is quite perfect, and the performance of LTE high data rate services and delay characteristics that 2/3G networks do not have, so if the stragety of joint optimization of 2/3G and LTE network is considered, from the operators point of view, the network performance and investment will reach optimal results. Based on the above purposes, LTE and 2/3G will be one network, which can be used to upgrade the existed network to one core network, one service platform, and one transport network, support interworking with legacy 2/3G networks, to achieve LTE/2G/3G network integration, and ensure that the rapid construction of network. 2G/3G network acts as an extension of LTE network coverage. At the same time, according to the 2/3G and LTE parameters, the operator will set up a flexible deployment network by 2G/3G/LTE hybrid multi‐network strategy. 2.1.4 Optimization Target When doing optimization, the following performance must be considered. Coverage: Good signal needs to be optimized across the whole cell and coverage holes within a cell’s service area must be minimized. In the system’s coverage area, by adjusting the antenna, power, and other means to make more areas’ signal to meet the needs of the minimum level of service, as far as possible the use of limited power to achieve the optimal coverage. Coverage optimization items include cell range, overlapping, overshooting, RSRP and SINR distribution, DL/UL loading, and DL/UL cell‐edge bit rate. 25 26 LTE Optimization Engineering Handbook Interference: A reasonable level of interference must be contained at cell’s service area in order to provide a quality air interface. The power control parameter optimization is especially important for the performance of the same frequency network. In different scenarios, the power control parameters are adjusted to reduce the system interference, including the traffic channel power control and the control channel power control. Mobility: The quality of the air interface in a cell with respect to handover behavior is good, with no unnecessary handovers. Through adjusting the parameter of handovers, enable the handover areas to be more reasonable is the target of mobility optimization. Control handover area with the reference signal level, if it is too high, the interference to other cells will increase; too low, session drop and access failure will easily happened. Capacity: Capacity optimization items include PRB (includes control channels) utilization, power and spectrum utilization, number of simultaneous EPS bearers, UL RSSI, PDCCH format and S1 utilization, traffic distribution, and intra‐LTE/inter‐LTE load balance. Quality: Quality optimization items include CQI, MCS, RSRQ, BLER, MIMO, Tx diversity, UE Rx diversity, eNB power, UE power, packet loss, jitter, and so on. In general, the network optimization should includes access, session drop, latency, mobility, congestion, paging, and other thematic analysis, and so on; it needs to analyze the performance of the network from different ways. Through the network main index trend, spot the poor quality of topN areas and reason of failure reasons, and improve the engineers to solve the problem. At meanwhile, the optimization should realize the perception of VIP users to evaluate and issues tracking, to achieve user awareness. 2.2 LTE Optimization Procedure 2.2.1 Optimization Procedure Overview LTE optimization can be divided into two steps, one is RAN pre‐launch optimization, the other is post launch optimization. Optimization lifecycle is depicted in Figure 2.1. RAN pre‐launch optimization for LTE assists in the evaluation and tuning of the radio access network in order to ensure that the performance and coverage of the RAN is maximized prior to commercial launch. Network Deployment Continuous optimization: network expansion & planning Initial optimization: Initial tuning or prelaunch optimization Network launch Loading reaches designed capacity In-service optimization: Post-launch optimization Figure 2.1 Optimization life cycle. LTE Optimization and Principle and Method Ideally, RAN initial optimization should be performed on single site verification and a cluster basis, as this allows the network to be optimized in a controlled and manageable way. Single site verification involves function verification and aims to ensure that each site is properly installed and the parameters are correctly configured. Cluster optimization starts after all sites in a planned area are installed and verified. Clusters typically consist of between 15 and 25 sites providing contiguous coverage. The “first tier” of sites surrounding the cluster is also defined as these sites will interact with the cluster and should be considered when optimizing a cluster. When optimizing a network it is recommended that a golden cluster is used at the start of a network optimization and performance improvement program. Drive tests on cluster aim at optimizing settings like antenna tilt or azimuth, transmit power, handover parameters, and list of neighboring cells in order to minimize interferences and to get the best from handover. Frequency planning per cluster should be reviewed and fine‐tuning in this step to minimize interference and hence enhance radio network performance. The initial tuning will aim to prepare both the air interface and troubleshoot system issues to achieve the best end‐user experience with the following objectives in mind: resolve hardware installation issues, parameter integrity issues, and improve design and end‐user experience, which includes availability of service (coverage), mobility, throughput, latency, and bottlenecks in the system, reliability (BLER), IRAT handover; its aim is to maximize SINR and highest modulation in order to achieve the best throughput per resource block. The detail optimization procedure is shown in Figure 2.2. Upon completion of the pre‐launch optimization phase, the acceptance phase will be performed. The post‐launch optimization is carried out on counter level, parameter level, cell level, and terrain/area level. Ususlly it is needed to check any type of hardware alarm/VSWR alarm, check UL RSSI issues, check bad CQI distribution, and so on. This phase is performed to verify the system wide performance after completion of cluster optimization. Radio network optimization is a continuation of cluster optimization, to fine‐tune inter‐cluster performance and to address coverage and performance issues that could not be resolved during cluster optimization. It allows operator to validate the compliance to the key performance indicators (KPI) targets and end‐users KPIs, in order to improve radio behavior of the network in terms of call drop, coverage holes, and handover mechanisms. Poor‐quality cell optimization and ideas on how to improve the system resource utilization, requires that the engineers’ analysis of mobile phone traffic statistics, parameter optimization, and physical optimization of the whole network is utilized to achieve the best performance. LTE radio network optimization tasks shown in Figure 2.3 include the following parts: to integrate a variety of existing network measurement data sources for test preparations, data collection, problem analysis, and parameter adjustment. Cell coverage test Frequency scanning External interference survey and cancellation Clutter planning Access test Drive test route planning Meet the target Data rate test No Latency test Mobility test Parameter, antenna and cable check Figure 2.2 Optimization procedure. Yes Data connection Data analysis Yes The whole network optimization Meet the target KPI result No Coverage, interference and parameter optimization Optimization report 27 28 LTE Optimization Engineering Handbook Collect field data Validate data Load data Process data, calculate KPIs, KQIs Present to engineers for analysis and optimization Post Launch Optimization Tasks KPI Reports Parameter Audits & CR Implementations Worst Performer Analysis Alarm Checks Recommendations Figure 2.3 LTE radio network optimization tasks. Daily optimization report Low throughput Low access rate Yes Meet KPI target KPI daily report Wrong parameter configuration cell list No Choose worse cell KPI worst Top N cell eNB alarm cell list High call drop Antenna, Access, Scheduler, Power control, Mobility, Capacity, ViP site, and E2E optimization solution Low handover success rate Optimization Congestion User complaint cell list Figure 2.4 Daily optimization. Perform analysis of daily drive test is a normal way to optimize the network which includes code and power planning verification, reference signal receive power quality, interference analysis, cell coverage overlap, UL/DL data performance, neighbor relation optimization, antenna tilts, re‐orientations, physical changes, IRAT issues, retainability issues, and accessibility issues. The overview of daily optimization is shown in Figure 2.4. Drive test data has been a primary source of optimization, for example, in rollouts and in specific use cases. However, for the majority of mobile users (of whom ~70% are indoors based on recent reports), drive test data does not tell the whole story. Unlike drive test data, call trace data is “real” data and provides visibility into indoor users unlike simulated drive testing. Call trace data is superior, but it has challenges on geolocation and data volume. 2.2.2 Collection of Mass Nerwork Measurement Data Traditional drive test (DT) is a manual method that allows the operator to gather information about network in some particular area to verify call handling functionality (accessibility, retainability, integrity), to detect interference areas, to note unexpected coverage holes, and to identify any hardware/software faults. Parameters like radio signal quality, signaling events, throughput, and so on, are measured from the field and recorded by dedicated equipment. Drive testing will LTE Optimization and Principle and Method collect cell ID, RxLevel, RxQuality, timing advance, transmit power, GPS position data, and time stamps, and form the main part of cluster tuning. Drive test data with UE or scanner measurements are often useful to identify coverage holes, as such drive tests are usually performed at acceptance testing. The indicators. RSRP, RSRQ, and RS CINR, are good indicators of the downlink quality of the network. Once a cluster drive test is complete, it is recommended to use the measurements to start analyzing the coverage of each cell to see if coverage objectives are met. For example, comparing the measurements with a best server plot can give indications of PCI‐ clashes, swapped feeders, cells not transmitting, or overshooting cells. The major measurement from the scanner is to verify the signal strength and the quality of the signal. The second part is to verify the code planning and the handover areas. It is crucial to minimize cell overlap even more than in WCDMA, as in LTE there is no soft handover. This means that inter‐cell interference will occur even from the second best cell. Scanners can be used during drive tests to provide an independent assessment of RSSI, RSRP, and RSRQ. Scanners require one antenna connection and signals from all transmit antennas can be measured as separate RSRP measurements. Differences between RSRP‐0 and RSRP‐1 can be useful for detecting transmit problems such as swapped feeders and radio unit faults. Conventional optimization based on drive testing or scanner data, is extremely laborious and time‐ consuming. Besides drive test and scanner data, the usage of actual network data is prefered for optimization, which is more accurate than drive‐testing or planning data since no assumption is made. Indicators such as measurement report (MR) data, OMC data, signaling data, MDT data, UE traffic recording, cell traffic recording data, and so on, are called mass data for the communication service provider. Mass data is becoming one of the most talked about technology trends these days, as they are all used to identify and solve radio network problems. The mass data is also very important to deploying powerful real‐time analytics and visualization tools and the automatic optimization of the network. Nowadays, soft data collection is a new cost‐efficient way to collect data. The software collection platform can output all original signaling data through the signaling manufacturer’s mirror output port and collects the key information of signaling data through the special adapter server. Compared with the traditional hard data collection mode, the software method provides a more flexible signaling data collection solution, which ensures the integrity and the accuracy of the data (Figure 2.5). Signaling Analysis HSS MME: Mobility Management Entity S6a Evolved node B (eNB) LTE-UE PCRF: Policy & Charging Rule Function MME X2 Gx or S7 S1-MME SGi S5/S8 Cell Serving Gateway PDN Gateway SAE Gateway Figure 2.5 Signaling collection. Rx S11 S1-U LTE-Uu PCRF IMS/PDN 29 30 LTE Optimization Engineering Handbook Network performance management counters and statistics will also be of significant use uring optimization and tuning activities. Counters and performance indicators can also be d useful during initial tuning. The intelligent platform offers eNB total care and collects/manages data automatically online that provides a comprehensive and systematic solution containing valuable know how on operation and optimization. 2.2.3 Measurement Report Data Analysis Mobile phones communicate periodically with the network, for example, sending back a measurement report (MR) about the quality of the signal that they are receiving and the different cells that are surrounding them. Based on the MR data report collected from the network, the operator can monitor the network performance and track all the UE calls’ performance within the network. It allows customers to debug some problems linked to any related mobile users. MR includes physical layer measurement, MAC layer, RLC layer, PDCP layer measurement, and so on. The measurement results are able to trigger events of radio resource control sub‐ layer, such as cell select, cell re‐select and handover, and so on; it is also able to be used for system operation and maintenance and observing system run status. Some important measurements are listed below: RSRP (reference signal received power), the received power on the resource elements that carry cell‐specific reference signals. RSRQ (reference signal received quality), the relation of N times the RSRP divided by the total received power in the channel bandwidth. The MR data of AOA (angle of arrival) is the angle between the UE and the cell antenna that UE is present. AoA is the angle (usually azimuth) from which a signal arrives relative to geographical north of an antenna array. It is able to be used for UE location. The operator can confirm UE position by the distance between UE and eNB and the AOA. Seen in Figure 2.6 (left), the UE is located in cell A, the distance and the AOA give UE’s location, somewhere within the gray sector. TA (timing advance), is used by the UE to adjust transmit timing. eNB estimates the transmission timing of the UE based on the RACH preamble. Direction of UE C Antenna Normal Direction B Geographical Antenna Bearing North for AoA Calculation Absolute Angle of Arrival (AoA) Relative AoA 8Tx/Rx TDD Antenna system A Figure 2.6 Used angles for AoA calculation. LTE Optimization and Principle and Method RIP (received interference power) measurement will be reported through cell trace eriodically. The RIP report type can be selected as either per subframe or PRB based or p per subframe based. UE uplink_SINR, UE based UL_SINR measurement is reported by cell trace periodically. PHR (Power headroom report) is the index reported by the UE to indicate the estimated power headroom, which tells the eNB how much of the maximum UE transmit power that the UE wanted to use for a certain transmission. The power headroom reporting range is from −23 to +40 dB (negative value means the UE was power limited). The eNB can use the power headroom reports to determine how much more uplink bandwidth per subframe a UE is capable of using. 2.2.4 Signaling Data Analysis In LTE, control plane messages and user plane traffic can be traced from various interfaces. To trace all control messages related to call establishment following interfaces should be traced: S1‐MME, S6a, and S11, tools such as Wireshark, and others can be easily connected to monitor, these interfaces through a network hub or router. Meanwhile, LTE UU and X2 interfaces of the wireless equipment need to provide the output capability of the whole quantity network service data, that is, signaling soft collection. Network element should have the output port of signaling and all kinds of traffic data. OMC‐R can control the function of signaling soft collection. Traffic aggregation adapter (SCA) is responsible for gathering wireless network equipment of signaling data flow. The use of network‐generated logs and automated tools can significantly speed up the network optimization process and lead to significant cost savings. Figure 2.7 and Table 2.1 show an example of LTE signaling load on overall network per application. Signaling Analysis Per Application 4000000 Signaling Load per application (%) Attach 19.2 RRC 34.4 TAU 4.8 2.6 Periodic TAU Paging 21.5 11.5 S1 Handover X2 Handover 6 3500000 RRC 3000000 Attach 2500000 S1 Handover TAU 2000000 X2 Handover 1500000 Paging Periodic TAU 1000000 500000 0 eNodeB MME HSS DNS SGW PGW PCRF Figure 2.7 LTE signaling load of different interfaces and nodes. Table 2.1 LTE network element signaling load per application. Nodes eNB MME HSS DNS SGW PGW PCRF Attach 0.4% 7.4% 24.7% 12.3% 18.5% 24.7% 12.3% TAU 0.2% 5.4% 67.4% 0.0% 27.0% 0.0% 0.0% Periodic TAU 0.1% 9.1% 90.8% 0.0% 0.0% 0.0% 0.0% Paging 3.1% 9.9% 0.0% 0.0% 87.0% 0.0% 0.0% S1 Handover 0.2% 11.7% 0.0% 0.0% 58.7% 29.4% 0.0% X2 Handover 0.4% 9.1% 0.0% 0.0% 90.6% 0.0% 0.0% 31 32 LTE Optimization Engineering Handbook 2.2.5 UE Positioning Moving into 4G, it is time that we go beyond existing business models of generating revenues from serving individual users’ location‐based service (LBS), that is, to venture into a new business model of providing geographic user behavior to meet the needs of the users, as well, mass data (MR or signaling data) –based network optimization will need the user’s auucrate position. Geolocation enables important optimization possibilities, helps in the analysis of the traffic and identifies where the data traffic hot spots are. By following the success story of Google search service, it is foreseen that the positioning for users are evidenced by smartphones going into markets at an accelerated rate and providing a higher quality geolocation solution to support commercial LBS to individual users. Especially 3D geolocation allows for rapid identification of quality issues in the network and precisely directed optimization based on actual users. This helps to save time and effort in maintaining the network at the best level. A more detailed UE positioning description from traditional drive test to geo‐based optimization will be presented in the following sections (Figure 2.8). Smartphone technology combining fast processors, GPS and Wi‐Fi assistance has revolutionized the use of mobile location. This can help both consumer and employee facing systems, often making use of the exact same approaches. From wireless technology or air interface point of view, the UE positioning feature provides a method to identify UE’s geographical location by radio signal measurement. The locations of these cells are known, and the speed at which radio waves travel is fixed and does not vary. By calculating the time delay between radio signals arriving at different cells it is possible to triangulate and find the UE’s location. The different positioning methods supported are: cell ID‐ based (the accuracy depends on radio network density), OTDOA (observed time difference of arrival, medium accuracy) and A‐GPS, as illustrated in Table 2.2. On the other hand, the signaling and user plane geolocation becomes popular now where data service is increasingly popular. The concept of the solution is to make LBS entirely independent of wireless technologies, or air interfaces. Since the LBS‐related messages are transported in the user plane there is no scalability issue faced by radio geolocation. For indoor positioning, the situation is different. Clearly, GPS does not work, because there is no satellite visibility, and signal strength is too weak for cell tower triangulation. When the pathloss is used in the estimation of the distance, it is expected lower accuracy for the position of indoor UE. The most common handset‐based technologies are based on the known WiFi locations or Bluetooth beacons, which have been deployed in a few showcase locations, such as airports, shopping malls, and exhibition centers. WiFi/Bluetooth CID (BSSID/UUID) decoding and reporting via RRC, to enable identity reporting of beacons, which also have benefits for radio resource management when handled via RRC. Alternatively, network‐based approaches can be used where the network uses radio frequency fingerprinting to track the UE. When an app needs to know the location of a device, it requests it from the network management system, independent of the handset operating system. Accurately determining the location of a wireless subscriber is difficult because of the nature of wireless networks. The inherent mobility of wireless subscribers means that they are not tethered to a single location. Furthermore, if a subscriber can once be located, there is no assurance that their location has not changed at a later time. Likewise, the varying conditions of wireless networks mean that network information is not static. A solution that works at one moment in time may not work later when network conditions change. All of these issues contribute to the difficulty in determining the location of a wireless subscriber. The actual location of a wireless subscriber can generally not be known with any degree of certainty. Geolocation technology has evolved across the last decade to match industry expectations and also cope with technology evolution and it is achieved by accurately locating UE position, LTE Optimization and Principle and Method PM counters and CM data + Call traces and CM data Accuracy Drive test Smoothing-based Optimization OSS-based Optimization Geo-based Optimization – RSRP < = –115.000000 > – 115.000000 AND RSRP < = –105.000000 > – 105.000000 AND RSRP < = –95.000000 > – 95.000000 AND RSRP < = –80.000000 > – 80.000000 AND RSRP < = 0.000000 Figure 2.8 Optimization method by advanced geolocation algorithms. both indoors and outdoors, three dimensions in horizontal and vertical directions by using MRs from the user equipment, wherever they are, even within buildings. 2.2.5.1 Timing Advance Different UEs in the cell may have different position, and therefore, different propagation delay, thus this may affect uplink synchronization. eNB’s timing will be phase‐synced to GPS within 100ns to support timing advance (TA) to achieve the tight phase‐sync. TA characteristics can be assumed that uplink arrives 0.3us too late compared to downlink. This error needs to be minimized with a correction of the UL timing. Due to the minimum step length of 0.52us 33 34 LTE Optimization Engineering Handbook Table 2.2 Positioning methods. Method Handset impact Accuracy Availability Cell ID No impact From 200 m to 5 Km 200 m in urban to 5 Km in rural area CID/TA No impact From 100 m to >1Km Mainly use CGI/TA (GSM and LTE) or RTT (3G) and UERxTxDiff (LTE) E‐CGI (GSM) No impact 150‐400 m Any environment AECID No impact From 50 m in urban to 500 m in rural area Any environment, but it requires high penetration of A‐GPS in order to reduce maintenance cost A‐GPS HW & SW 10‐30 meter Doesn’t perform well indoor. OTDOA SW 50‐150 meter Only LTE, did not work well in 2/3G U‐TDOA No Impact 50‐100 meter Cost is high for deploying LMUs. 3D RSS fingerprinting No Impact 10‐30 meter Required 3D maps & network configuration Hybrid Positioning between accuracies of the involved methods No Impact 0,52us The mobile is somewhere in the yellow arc. 0,26us 0,26us A DL n*TA Ideal position pos(X,Y,Z) B UL without TA UL useing TA Speed of light: 300*106 m/s s = v*t; 300*106*0.52*10–6 = 156 m; 300*106 *0.26*10–6 = 78 m Two types of TA: TA (type 1) = (eNB Rx – Tx) + (UE Rx – Tx) Continuous TA TA (type 2) = eNB Rx – Tx PRACH TA Figure 2.9 Timing advance. (TA granularity is 16* Ts = 16/(15000*2048) = 0.52 µs) the UE is told to send the next frame 0.52us earlier. After the correction is applied, the frames will arrive 0.22us to early. This is treated as good since the error has decreased. The least measurable error is 0.26us (Figure 2.9). UE distance and UE velocity relative to eNB can be calculated by TA. UE distance can be calculated based on the median TA value in a series of TA reports. UE velocity relative to eNB can be calculated the difference between the first and last TA values in a series of TA reports, and convert the TA difference to distance difference relative to eNB and then UE velocity is based on the duration for the reports. LTE Optimization and Principle and Method In urban areas, UE is usually connected to eNB under a no‐line‐of‐site condition (reflections on building and terrain). This effect results in an over‐estimation of the distance between UE and eNB impacting on the positioning. 2.2.5.2 Location Accuracy Evaluation Location determination algorithms are the backbone of operator‐offered location‐based services. There are legacy techniques listed below to determine an estimate of the location of a wireless user. Other techniques are possible, but are typically less applicable and require much more complicated analysis and computation. Cell‐ID method The cell‐ID method provides the identification of the serving sector for an active wireless subscriber as an estimate of their location. The cell‐ID can be obtained from signaling information that is used to set up or maintain a wireless call. This technique has limited accuracy since it can only identify the location of a user to the area covered by a cell, which can be quite large. Normally cell‐ID and other UE/EUTRAN measurements can be used to calculate UE’s position, it can be got other informations include the coordinates of the radio antenna, number of sectors in an eNB, beam orientations, and transmitting powers. The cell‐ID method can provide estimated accuracy is 100 m 67% and 300 m 95%. TA+AoA method UE distance can be calculated based on the median TA value, AoA (angle‐of‐arrival, the angle (usually azimuth) from which a signal arrives relative to a reference angle (geographical north) of an antenna array) can be used to determine the azimuth of the UE. Assisted GPS (A‐GPS) Assisted GPS makes use of the GPS capabilities of a UE and the availability of a GPS satellite network to provide very accurate location information for an individual user. Assisted GPS is a refinement of GPS where the network stores information that can be provided to the UE to help the UE speed up the process of acquiring the required number of GPS satellites to allow for GPS geolocation. However, not all UEs are equipped with GPS capabilities today. Furthermore, GPS features tend to use a significant amount of power, causing wireless subscribers to sometimes turn off the GPS functions on their GPS‐enabled phones. Also, GPS requires line‐of‐sight access to multiple GPS satellites, which means that it does not work well indoors. An assisted GPS method can provide an estimated accuracy of 50 m 67% and 100 m 95%. OTDOA OTDOA positioning method relies on the UE measuring the Reference Signal Time Difference (RSTD) on Positioning Reference Signals (PRS) sent from the reference cell and a number of neighboring cells as it requires that the UE can detect and measure on at least two neighbors; the two neighbors must also have “decent” geometry relative to the UE (e.g., not in line as seen from the UE). The OTDOA method can provide estimated accuracy is 100 m 67% and 300 m 95%. Triangulation Triangulation is a mathematical technique where the location of a point in a three‐dimensional environment can be determined by knowing the distance between that point and at least three other fixed points in the network. The more accurate the information about these distances, 35 36 LTE Optimization Engineering Handbook Figure 2.10 Choose signals from three sites and estimate the distance from UE to sites. –9 9d Bm 1d Bm –8 –85 dBm –98 dBm –96 dBm Site1 Site2 Site3 the more accurately the calculated position of the point of interest can be obtained. Location estimates of lesser accuracy can be obtained by knowing the distance to two or even one other point in the network (Figure 2.10). In wireless networks, obtaining the distances from the point of the wireless user to the other fixed points is generally accomplished by measuring the time it takes to transmit information from the point of interest to the other fixed points and using that time to estimate the distance based on the propagation speed of the information. Adjustments can be made to the estimate of the distance by the known delays in the propagation. In urban environments, the triangulation method can provide estimated accuracies of 150 meters most of the time. Several tests were made by combined TA+AoA and triangulation method and in the field at different locations in a live network. Some analysis of the test results are provided in Figure 2.11. It was begun with individual tests to see if there is a systematic error in the coordinate detection produced by the tool. Then it will have an assessment of all the tests to see what accuracy can be achieved with an increase in samples. The blue dots are the measured coordinates relative to the mobile location at (0, 0). It is seen that the over all offset are unevenly scattered around (0, 0). On average, the offset in X is 14.9 meters and in Y is −22.5 meters. The standard deviation is 94.4 meters (Figure 2.11). 2.2.5.3 Location Support The control plane of location support provides positioning information of a user in the EPS system. The positioning methods supported by this feature are assisted GPS (A‐GPS), observed time difference of arrival (OTDOA), and enhanced cell ID, and they generate UE positioning data used to position a LTE UE with high accuracy in the network. In A‐GPS, assistance data are retrieved by the network from GPS signal being sent by the GPS satellite constellation. LTE Optimization and Principle and Method Figure 2.11 Scatter plot of the coordinate offset in meters from all the tests. GHData relative to FieldData 150 100 Distance in meters 50 0 –50 –100 –150 –200 –250 –100 –50 0 50 100 150 200 250 300 Distance in meters The UE performs the GPS measurements with the help of assistance data provided by the network. This position can then be used by emergency services or commercial location services. Positioning data between positioning nodes and the UE is transferred on the control/ user plane. Positioning tasks to be performed by eNB include: ●● ●● ●● ●● ●● ●● ●● ●● ●● Configuration of positioning sub‐frames (for OTDOA) Broadcast of position reference symbols Inform UE about positioning sub‐frames, by broadcasting or by dedicated signaling Measure TA (for ECID) Request UE measurements (for ECID) Measure GNSS timing of cell frames Forward measurements to eSMLC Terminate LPPa protocol Forward LPP containers (assistance data and UE measurements) Following RAN events/measurements are located on the map as a function of actual statistically distributed usage (user id unknown) (Figure 2.12). 2.2.5.4 3D Geolocation Drive test and 2D geolocation provide a network view at street level. Traditional optimization doesn’t consider what is happening in buildings. So, how do we identify problems in the buildings, low coverage areas, traffic hot spots, and so on? Imagine you have a multi‐use high‐rise building with 20 floors. Is the majority of the mobile traffic coming from the restaurant on the first floor, or the office on the tenth floor? 3D geolocation can answer this question and help plan the most cost‐efficient and effective solution. As most of the traffic is indoor, network performance in‐building is very important, because accurate indoor positioning of UEs can have a evaluation of indoor cell loading and efficiency by analyzing how much traffic should belong to an indoor cell but is absorbed by surrounding outdoor cells. A 3D method geolocates network events not just on a horizontal two dimensions (2D) but also at different floors in high‐rise buildings, that is, on a vertical z‐axis (3D). A 3D vision of real traffic with outdoor versus indoor traffic and height with massive geolocation brings 37 38 LTE Optimization Engineering Handbook Value value < –110 –110< = value <–103 –103< = value<–96 –96< = value<–85 –85< = value<–75 value> = –75 Figure 2.12 RSRP plot with geolocation. unparallel optimization opportunities, as the optimization output is different depending on the location and height of the demand. For 3D geolocation, real data from users positioned on three dimensions will get higher accuracy. The method is based on a RF prediction database, which is also called RF Frigerprinting. The whole procedure is shown in Figure 2.13. The RF Frigerprinting can be gotton by network simulation using geographical information and network configuration data. Received signal strengths from a large set of cells to each point within the coverage area is calculated and stored as the Frigerprinting and then determine the UE’s position by comparing/matching with its MR. In the first stage, a database of expected signals strengths of all relevant‐based 3D Geolocation Network Data Create GIS Database MR Data Collection Import Network Data Parsing Predict Coverage Network Consistency OK Geolocation NOK Create finger print DB Field Verification Network Alignment Figure 2.13 3D geolocation procedure. Optimization LTE Optimization and Principle and Method Terrain Clutter Streets Buildings Figure 2.14 GIS database is created from terrain, clutters, and buildings. stations in each point within the coverage zone is created. In the second stage, UEs’ MR, which may include RSRP measurement of the serving and neighboring cells as well as time delay information, are correlated with the database, picking the best matching point as the mobile’s estimated location (Figure 2.14). For detail, 3D geolocation will need three steps: Step 1, collect inputs: Geographical data, including terrain, roads, clutter, and buildings. For more accurate monitoring and optimization, 3D maps are needed. Cell and antenna locations and parameters including adjacent (neighbor) lists, and MR such as timing advance and RSRP events, handover events, dropped calls, inter‐RAT handover events, and so on. Such files can be recorded by the RAN’s OMC or by recording the interfaces. Before step 2, preprocessing is needed to clean up the raw data both for removal of bad data from errors in the collection process and cleaning of some multipath errors correctly measured during collection. It is very critical that operator provides as accurate network configuration data as possible. Step 2, generate 3D coverage prediction: Upon collection of geographical data and network information, a 3D propagation model is used to predict received signal strength at various points in the network. A “fingerprint” (predictive fingerprinting algorithms) with RSRP levels is created for each point by ray tracing simulation as shown in Figure 2.15 and Figure 2.16. Upon completion of this phase, the system maintains a 3D coverage database that contains the received signal strength of each relevant cell at each point of the coverage area. This database contains accurate predictions. The coverage database contains separate layers for area, roads, and buildings. Step 3, locating UE: UE’s MR may include RSRP measurement of the serving and neighboring cells as well as time delay information, are correlated with the database, picking the best matching point as the mobile’s estimated location. The location is determined by several algorithms that use time delay measurements and pattern matching of signal strength measurements to the fingerprints as shown in Figure 2.17. The location determination in performed in three dimensions, that is, height above the ground is calculated for indoor calls. Finally, the “real” network performances of the matched UEs’ MRs at different floors of the buildings were present as shown in the Figure 2.18. The set of localized events creates an alternative coverage database, with the advantage of being a result of real measurements of real terminals. As such, it provides a more realistic picture than drive tests or predictions. The heart of the algorithm is the matching of a set of received RSRP measurements to the predicted RSRP values calculated for each viable point. 3D model should take into account the impact of the buildings shape and heights. It worths to note that the fingerprint database needs to be calibrated with real user information before matching. The calibration is based on actual user latitude and longitude data, and signal strength information, which includes indoor/outdoor frequency sweep data, drive test data, signaling, and MR correlation outputs with user latitude and longitude. 39 40 LTE Optimization Engineering Handbook Figure 2.15 Creating a 3D “fingerprint” database. Figure 2.16 Predicted RSRP from up to 40 sectors. Predicted Signal Levels from 3D ray tracing 1 3 2 cell A B C rxlev 1 70 42 25 rxlev 2 68 43 16 rxlev 3 64 45 28 rxlev 4 65 41 23 Delta between predicted and MR signal levels – 4 Receiving Signal Levels in MR MR SQ1 33 SQ2 131 SQ3 3 SQ4 25 cell A B C rxlev 65 44 27 cell A B C delta 1 5 –2 –2 delta 2 3 –1 –11 delta 3 –1 1 1 delta 4 0 –3 –4 Sum of Square Error Σ x2 Square Sum Figure 2.17 Calculate matching factor to 3D prediction grid. BlueLayer 16MB - Max GreenLayer 12MB - 16MB YellowLayer 8MB - 12MB OrangeLayer 4MB - 8MB RedLayer Min - 4MB BlueLayer –70dBm - Max GreenLayer –80dBm - – 70dBm YellowLayer –90dBm - –80dBm OrangeLayer –100dBm - –90dBm RedLayer Min - –100dBm BlueLayer –8dB - Max GreenLayer –10dB - –8dB YellowLayer –12dB - –10dB OrangeLayer –16dB - –12dB RedLayer Min - –16dB Figure 2.18 Network performances at different floors of the buildings. (See insert for color representation of the figure.) 42 LTE Optimization Engineering Handbook 2.2.6 Key Performance Indicators Optimization The key counters are used to generate the key performance indicators (KPIs) of the network, which are defined on the OMC, and these counters are predefined and initialized as soon as the eNB starts. The optimization of KPIs is still important in LTE. There are a few KPIs that will adversely affect end‐user experience if any of these KPIs under‐perform and it is critical to monitor closely the performance of these main KPIs. This troubleshooting guideline aims to present ideas to pinpoint potential areas for troubleshooting if any of these main KPIs, for example, integrity, accessibility, retainability, mobility, and so on, under‐perform. The most interested field KPIs are listed below as shown in Table 2.3. Integrity, which is the KPIs’ defining quality of the service provided: latency, retransmissions, RSSI for PUCCH/PUSCH, SINR for PUCCH/PUSCH, and CQI/CQI offset. Accessibility, which has KPIs indicating the possibility to access to a service: RRC connection establishment. E‐RAB setup, paging records, and discards. It is a combined metric including RRC, S1, and E‐RAB establishment success rate. In the case of poor accessibility, each success rate must be analyzed individually. Reasons for poor accessibility include but are not limited to poor coverage, high UL interference, UE camping in the wrong cell, or admission reject. Retainability, which is defined as the ability of a user to retain the E‐RAB once connected for the desired duration, which has KPIs indicating the ability to hold/sustain the call: call drop, call completion, E‐RAB drop, E‐RAB normal release, RRC connection re‐establishment, and IP incoming traffic error rate. Table 2.3 Some of the field KPIs. KPI name Application services LTE E2E network service PS data services (FTP, HTTP etc) Control plane User plane Radio bearer services User plane KPI category Service accessibility ratio [%] Accessibility Completed session ratio [%] Accessibility Single user throughput [Mbps] Usage Attach time [ms], AttacTime ms t AttachComplete t Attach Request Integrity Attach success rate [%] Accessibility Service request (EPS) time [ms] Integrity Service request (EPS) success rate [%] Accessibility Service (EPS) Drop rate [%] Retainability Handover procedure Time [ms] Mobility Handover success rate [%] Mobility Round trip time (RTT) [ms] Integrity Single user throughput [Mbps] Usage Service interrupt time (HO) [ms] Mobility Single user throughput [Mbps] Usage Cell throughput [Mbps] Usage Note: The attach time is the interval between the RRC connection request (carrying the attach request) and the reception of a positive response by the UE (attach complete). “Single user throughput” KPI can be specified on each protocol layer, such as application layer, IP layer, L2 layer (PDCP/RLC/MAC layer), and L1 layer (physical layer). LTE Optimization and Principle and Method Mobility, which has KPIs indicating performance of handovers (intra/inter eNB, X2 based): handover preparation, handover success rate, and handover failure rate. Reasons for poor mobility include but are not limited to missing neighbor relations, poor radio conditions, or badly tuned handover parameters. Usage, which has KPIs indicating how LTE network is loaded in terms of data volume, throughput, number of users (active and connected), PRB usage, and cell availability. 2.2.7 Technology Evolution of Optimization The trend of technology evolution of optimization is traffic steering, user steering, and smart optimization. The operators might always ask themselves, is the problem caused by poor radio performances? Is the problem caused by specific devices? Is the problem caused by specific applications? How can we make networks more intelligent? All these questions need to be considered for the network optimization evolution as shown in Figure 2.19. Self‐organizing network (SON) as evolution of optimization has been widely deployed in RAN optimization. SON is defined as a set of use cases that covers the entire network lifecycle: planning, deployment, operation, and optimization by network mass data. No more tedious manual engineering work but an in‐network software platform with artificial itelligence algorithm that continuously monitors the network and sets changes to improve its performance. Self‐optimization is an important improvement area due to the fact that current automatic optimization tools focus on small number of radio parameters and a lot of manual effort is required for optimization activities. The aim of self‐optimization is to fine‐tune initial parameters automatically for improving cell/ cluster performance and dynamically recalculate these parameters in case of network and traffic changes, and make following optimization activities automatic: neighbor cell list optimization, interference control, handover parameter optimization, QoS‐related parameter optimization, load balancing, RACH load optimization, optimization of home base stations, and so on (Figure 2.20). Classical network optimization •Basic RF Opt. includes •Coverage •Interference •Neighbor list •PSC/PCI •Frequency •KPI Opt Efficiency improvement: MR, OSS, signaling data application •RF Opt. includes •Coverage (MR, OSS) •Capacity (OSS) •parameters (MR, signaling data) •antenna and interference •Opt. based on GIS (MR/PM) •KPI improvement (MR,OSS) Intelligent and automatic optimization Much focus on traffic and end user •E2E Service KQI •RF/RAN •Bearer/CN/Transport •UE •SP •Voice/Service •Service QOS •MDT •RF Opt (ANR, PCI, CCO) •KPI/Parameter Opt •MRO •MLB •RACH •A-ICIC [Interference] Figure 2.19 Technology evolution of optimization. Figure 2.20 Online automatic optimization. receive power Info Site locations heights, etc KPIs Network simulation Off-line optimization Optimization method On-line optimization Network Parameter Seting Measurements Measurements User location and 43 44 LTE Optimization Engineering Handbook As stated above, network optimization evolution will be based on a suite of algorithms that calculates the optimized configuration settings for the network. Optimization is performed based on mass 3D geo‐located RF events, OMC data, the actual UE behaviors, as well as p rediction models. 2.3 LTE Optimization Key Point 2.3.1 RF Optimization For LTE, the frequency reuse factor is 1 (the same frequency deployment). The co‐channel interference will become one of the most important factors that influence system performance. Therefore, the RSRP and SINR value distribution is to examine the LTE network coverage and interference of the main index, and the level of SINR value is directly dependent upon the cell direction angle, tilt, PCI planning, quality of antenna/feeder, and so on. Basic physical RF optimization is very important, clear cell dominance areas, minimize cell overlapping, and avoid sites shooting over large areas with other cells. The outputs of the RF tuning process include information of recommended changes to antenna tilts, orientations, and heights, neighbor cell changes including additions and deletions, recommendations to remove cells, which may be causing excessive interference in the network. In fact, although antenna tilting and antenna placement has big impact on other cell interference, it can’t fix bad RF by tuning wireless parameters. 2.3.1.1 RSRP/RSSI/SINR/CINR RSRP (Reference signal received power), defined as the linear average over the power contributions of the resource elements that carry cell specific reference signals within the considered measurement frequency bandwidth. The measurement used to indicate the coverage of LTE system. The formular is: DL N RS ns RSRP i 1 PRS DL N RS ns where, for a given E‐UTRA channel bandwidth, PRS = power in a reference signal (RS). DL = number of RS in a downlink slot, n = number of symbols carrying RS. N RS s Typical working values of RSRP are in the range from −130 dBm (distant from site) to −50 dBm (very close to site), which is shown in Table 2.4. Table 2.4 RSRP value. Reported value* Measured quantity value Unit RSRP_00 RSRP < −140 dBm RSRP_01 −140 ≤ RSRP < −139 dBm RSRP_02 −139 ≤ RSRP < −138 dBm … … … RSRP_95 −46 ≤ RSRP < −45 dBm RSRP_96 −45 ≤ RSRP < −44 dBm RSRP_97 −44 ≤ RSRP dBm LTE Optimization and Principle and Method .93 .88 .83 .78 .73 .68 .63 .58 .53 .48 .75 10 MHz 20 MHz 50 100 –30.79 –27.78 .85 RSSP [dBm] BW # RB Scaling:[–10log (12N)] .95 .105 Measurement: 95 dBm –67 dBm = 28 dB → agrees with theory (27.8dB) .115 .125 RSSI [dBm] Figure 2.21 RSRP versus RSSI for full loaded cell (10MHz). RSSI (Received signal strength indicator) comprises the linear average of the total received power observed only in OFDM symbols containing reference symbols for antenna port0, which is measured over the entire bandwidth, includes co‐channel serving and non‐serving cells, adjacent channel interference, and thermal noise everything. In theory, RSSI (wideband power) = noise + serving cell power + interference power. RSSI can be calculated as: ns RSSI i 1 PiTotal ns where = total received wideband power in the ith symbol (Figure 2.21). When 100% downlink PRB is active without noise and interference, RSSI=12*N*RSRP, where N is number of RBs across the RSSI, which is measured and depends on the bandwidth. RSSI can be expressed as function of several terms: PiTotal RSSI f I ,TxPwrRSi ,TxPwrctrl /trf , Pathlossi , RSRQi , N DetectedCell , LevelBS LastDetected , xLoad i Based on the above fomular, under full load (100% PRB utilization) and high SNR: RSRP dBm RSSI dBm 10 * log 12 * N RSRQ (Reference signal received quality), is the ratio of wanted signal to all received power. It is calculated as N*RSRP/RSSI. N is the number of RBs of the EUTRA carrier RSSI measurement bandwidth as shown in Figure 2.22. RSRQ includes the loading of the non‐reference signal subcarriers, so it is a good measurement to indicate the loading. There are five data subcarriers for every one reference signal subcarrier, which means that if a lot of data are transmitted, the RSRQ will be low even if the received signal is of high quality and there is little to no noise. RSSI increases about 5dB when RB activity increases to 100% in a 10MHz cell. Typical working values of RSRQ are in the range from −3 dB (low/no interference) to −18 dB (high load/high interference) shown in Table 2.5. Example: RSRP=−82dB, RSSI=−54dB, N=100 => RSRQ=10lg100 + (−82)−(−54)=−8dB RSRQ tends to drop off rapidly at the cell edge or as the serving cell load increases, which can make designing an appropriate level difficult. Typically, RSRQ down to −11 dB can be strongly influenced by serving or inter‐cell interference, with no indication as to which is the cause. Below −11 dB inter‐cell, external interference or thermal noise become dominant (Figure 2.23). 45 LTE Optimization Engineering Handbook 0.0% IE 0.0% value CRS RSRP –82.0 30.0 –7.0 CRS SINR RSRQ From the measurement result, we can calculate the relation between RSRP and RSRQ: RSRP/RSSI*NRB RSRP-RSSI + 10log(NRB) = RSRQ (–82) – (–67) + 10log(NRB) = –7 –67 –14 14 –20 96 RSSI PRACH Tx Power PUSCH Tx Power PUCCH Tx Power Pathloss –15 + 10log(NBR) = –7 Log(NRB) = 0.8 NRB = 6 PDSCH Rb Num/slot PUSCH Rb Num/slot TM Mode Handover Delay(s) MR→HO Cmd Delay Figure 2.22 Calculation RSRP and RSRQ. Table 2.5 RSRQ values. Reported value* Measured quantity value Unit RSRQ_00 RSRQ < −19.5 dB RSRQ_01 −19.5 ≤ RSRQ < −19 dB RSRQ_02 −19 ≤ RSRQ < −18.5 dB … … … RSRQ_32 −4 ≤ RSRQ < −3.5 dB RSRQ_33 −3.5 ≤ RSRQ < −3 dB RSRQ_34 −3 ≤ RSRQ dB 100 RSRQ Distribution 110% High Traffic 90 100% 80 90% High Traffic cumulative % 80% Low Traffic cumulative % 70 70% 60 60% 50 50% 40 40% 30 30% 20 20% 10 10% 0 0% –14 –13 –12 –11 –10 –9 –8 –7 –6 RSRQ Figure 2.23 RSRQ distribution in a live network. –5 –4 –3 –2 Low Traffic Percentage Number of Samples 46 LTE Optimization and Principle and Method SINR (Signal to interference and noise ratio) is the ratio of the power of all subcarriers that make up a cell‐specific reference signal to the power of the interference (I) plus noise (N) over the same subcarriers. The measurement indicates the RF channel quality, which is measured using the reference signals transmitted in each subframe, the measurement is sampled in every TTI and averaged per second in the logged output. SINR RSRPserv RSRPother I N In a live network, SINR and RSRP are basically presented a linear relationship, in the main range of RSRP, when RSRP upgrades 10dB, SINR will upgrade about 4‐6dB. At the same SINR, the throughput rate is weakly correlated with the RSRP, and the strong RSRP does not imply a high throughput rate, so RSRP’s high design goal is not required. Unfortunately, SINR is not specifically defined in 3GPP specifications, this means that any SINR measurements methods of UE and scanners are not known. SINR measurements can indicate interference areas, it is impacted by network load (traffic in the neighboring cells will reduce serving cell SINR), the measurement method (RS or SCH) and tools, and PCI planning etc. The SINR defines the throughput, coverage, and capacity of the network, and ultimately the user experience. The factors that impact SINR includes UE position in cell (RSRP), interfering cell load, interferer cell geometry, clutter and terrain type, and reference signal configuration. In theory, downlink SINR of a cell can be calculated from RSRQ if the RF utilization of the cell is known. For example, the serving cell should know its own PRB utilization, while for calculation of a neighbor cell SINR the load can be obtained by the X2 Resource status reporting procedure. RSRQ depends on own cell traffic load, but SINR doesn’t depend on own cell load. For RSRQ to SINR mapping, we can use the number of REs/RB in serving cell is an input parameter for RSRQ to SINR mapping. In practice, mapping from RSRQ to SINR seems difficult. SINR = RSRP*12N P1 + Pn_12N Pn_xN = Pn_ RE* xN x = RE / RB_used, N = # RBs RSSI = Pi + RSRP* xN + Pn_12N RSRQ = N* RSRP SINR = RSRP*12 N N* RSRP RSRQ − RSRP* xN = 12 1 RSRQ −x RSSI Where x=RE/RB, 2RE/RB equals to empty cell that only reference signal power is considered from serving cell. 12RE/RB equals to fully loaded serving cell that all resource elements are carrying data (Figure 2.24). The last one is CINR (reference signal carrier to interference and noise ratio), the sum of the RS resource element powers, divided by the sum of the remaining resource element powers, usually SINR=CINR from engineering point of view. This is averaged over reference signal symbols. RS CINR is inversely related to the traffic load and interference. Carrier is conventional naming for a non‐information bearing signal, that is, unmodulated signal. Signal is conventional naming for information‐bearing signals, that is, either phase, frequency, and/or amplitude modulated signals. 47 LTE Optimization Engineering Handbook RSRQ vs SINR 1 30.00 2 3 4 5 6 25.00 20.00 SINR (dB) 48 15.00 10.00 5.00 0.00 –20 –19 –18 –17 –16 –15 –14 –13 –12 –11 –10 –9 –8 –7 –6 –5 –4 1 2 RE/RB 2 4 RE/RB 3 6 RE/RB 4 8 RE/RB 5 10 RE/RB 6 12 RE/RB –3 –5.00 –10.00 RSRQ (dB) Figure 2.24 RSRQ versus SINR. CINR SINR Wanted _ Carrier Noise Interference Wanted _ Signal Noise Interference 2.3.1.2 External Interference Before optimization, the operator needs to verify the radio link permanence and eliminate any source of interferences that can create service degradation. There are two types of interference in LTE—internal and external—and this part is focused on the external interference. For internal interference, it is generated by other UEs within the cells. External interference, it is generated by external source power supply sources, Wi‐Fi, or sharing a site with different technology (CDMA and LTE, GSM and LTE) or even electrical equipment or discharges in poorly installed or low‐quality RF components or passive intermodulation (PIM) that will have a high chance to have interference during high traffic load. Three fundamental external interference issues must be considered: ●● ●● ●● Spurious emissions, the unwanted emissions from a transmitting system degrading the performance of a receiving system. Blocking, a measure of the receiver ability to receive a wanted signal at its assigned channel in the presence of an unwanted interferer. Isolation between systems, the isolation is defined between the antenna reference point (the transmitter connector) in the aggressor base station and the receiver reference point at the victim base station. The interference level at the victim site depends on various factors: rated output power of the aggressor, antenna types, heights, tilts and azimuths, distance between sites, type of environment, for example, urban, suburban, or rural. For each specific site, a unique relation exists between distance to the aggressor base station and isolation under free space propagation conditions, as shown in the formula below. LTE Optimization and Principle and Method Table 2.6 Occurrence of intermodulation and harmonics. 2ndorder 3ndorder th 4 order 5thorder 2f1 2f2 f1 3f1 3f2 2 f1 f2 2 f1 f2 2 f2 f1 2 f2 f1 4f1 4f2 3 f1 f2 3 f1 f2 3 f2 f1 3 f2 f1 2 f1 2 f 2 2 f1 2 f 2 5f1 5f2 4 f1 f2 4 f1 f2 4 f2 f1 4 f2 f1 3 f1 2 f 2 3 f1 2 f 2 3 f 2 2 f1 3 f 2 2 f1 Lisolation f2 32.4 20logd 20logF f1 f2 Ga Gb D Where Ga and Gb are the antenna gains [dB], F is the frequency [MHz], d is distance [km], D is the decoupling factor achieved by changing direction or tilt of the main antenna lobe [dB]. It is needed to carefully evaluate signal levels from co‐located and adjacent transmitter sites then place and point antennas to maximize isolation from those interference sources. For the worst case when antennas point at each other, D = 0, but increases when the main antenna lobe direction or the tilt is changed. Usually directional Yagi (16dBi gain) antennas can help to determine from which direction the interference are coming from. Another usual external interferers are caused by PIM (passive intermodulation), which is generated by the nonlinear mixing of two or more frequencies in a passive circuit. Inter‐ modulation (IM) products are created when two or more frequencies mix in nonlinear devices in the Tx or Rx paths, possible sources include poor connections or damaged cable. Due to non‐linearity of radio component, the high phase harmonic wave might be generated and inter‐modulated by RX signal, thus the new inter‐module signals (spurious signals) will arrive the receiver causing the interference. IM products of order n are the sums and differences in n terms of the original frequencies. The strengths of the IM products decline with higher orders. In general, for two unmodulated signals, f1 and f2, the following IM products are obtained in Table 2.6. IM is generated whenever more than one frequency encounters a non‐linear electrical junction or material. The resulting creation of undesired signals, mathematically related to the original frequencies, can result in decreased system capacity and/or degraded call quality by increasing the receiver noise floor. The probability of interference increases as more carriers are being transmitted and with more power, and inband third‐order IM products (2f1‐f2) and (2f2‐f1) have the largest magnitude and are used for worst case analysis, which is shown in Figure 2.25. In order to minimize PIM, it should use pre‐tested, PIM‐certified cable assemblies, and it needs to test thoroughly for PIM after installation. There are few ways to measure the interferences. Sweep testing: Line and antenna sweep testing, in the wireless telecommunication industry, was once the main type of testing relied upon for performance certification of a new or recently modified RF system being placed into service on an existing network. This test will be done in sections: from RFport output, after feeder, after the jumper cable. Normally, it will give the ground noise, since the power during the test is in the idle state. Therefore, it’s not accurate to measure and the next test will be used. PIM testing: PIM testing is designed to identify the existence of the unwanted signals that create RF interference. When the source of the unwanted signal is identified, corrective steps 49 LTE Optimization Engineering Handbook 3rd Order Desired 5th Order Interference Interference 7th Order Amplitude f2 – f1 f1 2f1 – 1f2 3f1 – 2f2 4f1 – 3f2 f2 2f2 – 1f1 3f2 – 2f1 4f2 – 3f1 f1 + f2 2f1 2f2 Frequency Figure 2.25 Inter‐modulation. PIM test equipment PA Rx UNA TX Low PIM load DUPLEXER PA COMBINER 50 RX PIM source Figure 2.26 PIM measurement setup. can be taken to eliminated the problem. These unwanted signals will show up as excessive frequency noise that can be within the bandwidth of a received radio. This test carries out at the high power output, simulates the site at the high traffic, therefore, it will give more accurate measurement. This is the passive inter‐modulation testing. The IM between each frequency component will form additional signals at frequencies that are not just at harmonic frequencies of either, but also at the sum and difference frequencies of the original frequencies and at multiple of those sum and differences frequencies (Figure 2.26). Spectrum analysis of an antenna system can assist in locating the source of RF interference. A passive inter‐modulation signal can be traced to a specific frequency with the help of a spectrum analyzer and if the test operator is familiar with the possible sources of interference signals, the problem can be identified and possibly resolved. In some cases, the source of the problem is an environmental interference or a signal from another broadcast carrier (Figure 2.27). Here is an example. In an 1880‐1900 frequency domain, three more obvious protrusions were found: 1881.8, 1890.6, and 1897.6 as shown in Table 2.7. 1881.8 and 1890.6 frequencies are the same as A station GSM1800 intermodulation product. When A station was powered off, except for 1897.6 frequency points, the other two obvious protrusions have disappeared. The signal positions of the other small protrusions were all changed. At the same time it was observed the station surrounding wireless environment, the station is located in a high position that can be seen very far away, so the interference signal should be derived from other base stations. LTE Optimization and Principle and Method Figure 2.27 Spectrum analysis. 2.3.2 CQI versus RSRP and SINR CQI (Channel quality indicator) indicating the instantaneous downlink channel quality perceived by the terminal. CQI is used to report the downlink channel quality by the UE after channel estimation calculated using the reference signal. Similar to HSPA, the CQI reports can be used by the network for downlink channel dependent scheduling and rate control. The LTE CQI reports indicate the channel quality in both the time and frequency domain. Usually low throughput is related to poor dominance and low coverage areas, so clear dominance areas with good RSRP together with good enough CQI/RSRQ/SINR are the natural target for optimization. In the same environment, SINR are proportional to CQI, and one step of CQI according to SINR changing is about 2dB. The CQI is used by the link adaptation function to select the transport format matching the channel conditions. Each CQI (wideband or subband) is first converted into SINR thanks to Figure 2.28. Then, the SINR is amended to take into account the UE speed estimated by layer 1, after mapping SINR to symbol info rate per modulation symbol, we can get the throughput. LTE CQI, MCS and throughput relationship is shown in Table 2.8. In live network, as long as RSRP is low (lower than ‐100 dBm), interference has less impact (CQI is anyway low) in such scenario, RSRP is the bottleneck in this region. At higher RSRP (higher than −100dBm) scenario, interference may become the bottleneck. The relation between modulation scheme distribution and CQI can be found in Figure 2.29. 2.3.2.1 CQI Adjustment The eNB adjusts the UE reported CQI value by taking into account ACK/NACK (acknowledged/ not acknowledged) reports from the UE for received downlink data blocks. The adjusted CQI value is used by link adaptation functionality to select the optimum MCS to achieve BLER target. Starting from the initial value, the CQI offset will be adjusted (StepUp or StepDown) in response to the ACK/NACK for the new transmission of a transport block (Figure 2.30). 51 Table 2.7 An example of IM. Frequevcy No. Frequency PIM 741 803 815 820 827 840 749 792 807 817 830 847 751 766 777 823 1851 1863.4 1865.8 1866.8 1868.2 1870.8 1852.6 1861.2 1864.2 1866.2 1868.8 1872.2 1853 1856 1858.2 1867.4 1875.8 1880.6 1882.6 1885.4 1890.6 1869.8 1875.8 1879.8 1885 1891.8 1859 1863.4 1881.8 1868.2 1870.2 1873 1878.2 1867.2 1871.2 1876.4 1883.2 1860.4 1878.8 LTE Optimization and Principle and Method RSRP –140dBm –95dBm –80dBm –40dBm RSRQ –19.5dB –12dB –7dB –3dB SINR 0dB 7dB 17dB 25dB CINR 0dB 5dB QPSK CQI 16 QAM 0 10, 11 16 QAM 0 20dB 64 QAM 6, 7 QPSK MCS 15dB 15 64 QAM 9, 10 16, 17 CQI values PDSCH Target SINR 1 –6.00 2 –4.00 3 –2.75 4 –0.75 5 1.25 6 2.75 7 5.00 8 6.75 9 8.50 10 10.75 11 12.50 12 14.50 13 16.25 14 17.75 15 20.00 28 Figure 2.28 LTE RSRP, RSRQ, CQI, MCS, CINR, and SINR ranges. Table 2.8 LTE CQI, MCS and throughput relationship. CQI Index Modulation 0 No 1 QPSK 2 3 MCS code rate Information bits per symbol Thpt. (per 1 RB) [Kbps] Thpt. @20MHz [Mbps] — — 0.076 0.1523 25.6 2.6 QPSK 0.12 0.2344 39.4 3.9 QPSK 0.19 0.377 63.3 6.3 4 QPSK 0.3 0.6016 101.1 10.1 5 QPSK 0.44 0.877 147.3 14.7 6 QPSK 0.59 1.1758 197.5 19.8 7 16QAM 0.37 1.4766 248.1 24.8 8 16QAM 0.48 1.9141 321.6 32.2 9 16QAM 0.6 2.4063 404.3 40.4 10 64QAM 0.45 2.7305 458.7 45.9 11 64QAM 0.55 3.3223 558.2 55.8 12 64QAM 0.65 3.9023 655.6 65.6 13 64QAM 0.75 4.5234 759.9 76 14 64QAM 0.85 5.1152 859.4 85.9 15 64QAM 0.93 5.5547 933.2 93.3 0 to 9 10 to 16 17 to 28 — For single code word case: CQI t min CQI t 1 CQIstepup , CQImax , for first HARQ feedback max CQI t 1 CQIstepdown , CQImin , for first HARQ feedback NACK, CQI t 1 , for first HARQ feedback N / A ACK, 53 54 LTE Optimization Engineering Handbook 100% Average of percent of QPSK Average of percent of 16 QAM Average of percent of 64 QAM 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 CQI Figure 2.29 Modulation scheme distribution versus CQI. For two code words case: min CQI t 1 CQIstepup , CQImax , for new transmission HARQ feedback ACK ACK max CQI t 1 CQIstepdown , CQImin , for new transmission HARQ feedback NACK NACK CQI t min max CQI t 1 CQIstepup CQIstepdown / 2, CQImin , CQImax , for new transmission HARQ feedback ACK NACK min CQI t 1 CQIstepup , CQImax , for new transmission HARQ feedback ACK N/A max CQI t 1 CQIstepdown , CQImin , for new transmission HARQ feedback NACK N/A CQI t 1 , for new transmission HARQ feedbacks N/A N/A Figure 2.31 presented the reported CQI versus RRM‐adjusted CQI, which are extracted from the actual measurement data collected during the mobile pathloss measurements. In addition, the graph also gives the distribution of spatial multiplexing and transmit diversity with the reported CQI. 2.3.2.2 SINR Versus Load The SNR varied in the field conditions quite a bit, but the impact of adding load is clear. Figure 2.32 illustrates the variability and is from test cases UDP DL MIMO in different interference conditions. FTP DL throughput degrades as DL inter‐cell 0% load increases to 70% DL load, impacts due to the load from neighbor cells, CQI is shown impacted through all subband. Actually, the SNR distribution is very different in different network. Figure 2.33 is a test result and analysis of the impact radio quality by CQI reporting per subband. LTE Optimization and Principle and Method Data block DL link adaptation:selecct MCS Packet schedule (PDSCH) Evaluate reference signals CQI report CQIreport+ΔCQI=CQI (PUCCH/PUSCH) Decode transport blocks ACK ΔCQI(t–1)+CQIstepup = ΔCQI(t) 1st DL transport block transmissions ACK/NACK (PUCCH/PUSCH) ΔCQI(t–1)–CQIstepdown = ΔCQI(t) NACK UE Figure 2.30 CQI adjustment. 20 18 y = 1.1684× –0.6409 16 RRM CQI 14 12 10 8 6 4 2 0 0 2 4 6 8 10 12 14 16 Reported CQI 100.0% Probability 80.0% 60.0% SIMO 40.0% MIMO 20.0% 0.0% 3 4 5 6 7 8 9 10 11 12 13 14 15 Reported CQI Figure 2.31 Reported CQI versus RRM‐adjusted CQI. eNB 55 LTE Optimization Engineering Handbook 70 60 Huge Thruput degradation (same as CQI drop) Thruput 50 Load_0 Load_50 Load_70 Load_90 40 30 20 10 0 Load_0 Load_50 Load_90 Load_70 –60 –62 –64 –66 –68 –70 –72 –74 –76 –78 –80 –82 –84 –86 –88 –90 –92 1 0.9 0.8 CDF 0% Load 0.7 CDF 100% Load 0.6 CDF 56 0.5 100% Load 0% Load 0.4 0.3 0.2 0.1 0 –10 –8 –6 –4 –2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 SNR Figure 2.32 SNR varied with load. 2.3.2.3 SINR Versus MCS The eNB needs knowledge of the SINR conditions of downlink transmission to a UE in order to select the most efficient MCS/PRB combination for a selected UE at any point in time. The MCS contains two parts, modulation constellation (QPSK, 16QAM or 64QAM) and coding rate, which is the ratio between information bits (from upper layers) and coded bits. Figure 2.34 and Figure 2.35 show a simplified view of MCS distribuion (link adaptation is taken into account) based on the SINR reported by the UE. From Figure 2.34 and Figure 2.35, it can be seen that downlink MCS follows closely DL SINR. Mapping of SINR to MCS is performed by DL link adaptation. Improvements over interference conditions (i.e., reducing bad dominance) will result in higher throughputs. A quantized TBS is selected by looking up the largest TBS defined in 36.213, which is shown in Table 2.9 that does not exceed the unquantized TBS. SINR improvements can be achieved by reducing network load, optimizing performance antenna tilts and pans, reducing or removing overshooting cells, and optimizing eNB Tx power levels. From LTE Optimization and Principle and Method Figure 2.33 CQI reporting per subband obtained from field test. 20.0% [–10,–8] [–8,–6] [–6,–4] [–4,–2] [–2,0] 15.0% 10.0% 0.0% 18.0% MC0 MC1 MC2 MC3 MC4 MC5 MC6 MC7 MC8 MC9 MC11 MC12 MC13 MC14 MC15 MC16 MC18 MC19 MC20 MC21 MC22 MC23 MC25 5.0% [0,2] [2,4] [4,6] [6,8] [8,10] 13.0% 8.0% –2.0% MC0 MC1 MC2 MC3 MC4 MC5 MC6 MC7 MC8 MC9 MC11 MC12 MC13 MC14 MC15 MC16 MC18 MC19 MC20 MC21 MC22 MC23 MC24 MC25 MC26 MC27 MC28 3.0% Figure 2.34 SINR Versus MCS in the range of ([–10,–8], [–8,–6],…,[6,8] and [8,10]). Figure 2.36, it can be seen especially at the lower end of the SINR curve, a small change in the SINR will have a significant impact on downlink throughput. For example, when SINR changed from 9 to 10 dB, 8 to 10 dB, 7 to 10 dB, 11%, 24%, and 39% throughput improvement was achieved accordingly from the field test. The relation of DL SINR and bitrate per RB (under LTE‐FDD, SIMO 1x2, EPA5, Okumura‐Hata model, channel bandwidth is 20MHz) is also shown in Figure 2.36. 57 LTE Optimization Engineering Handbook [14,16] MC28 MC27 MC26 MC25 MC24 MC23 MC22 MC21 [24,26] MC20 MC19 MC18 MC16 [22,24] MC15 MC14 MC13 MC12 MC9 MC11 MC8 MC7 [20,22] MC1 24.0% 22.0% 20.0% 18.0% 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% [12,14] MC0 MC1 MC2 MC3 MC4 MC5 MC6 MC7 MC8 MC9 MC11 MC12 MC13 MC14 MC15 MC16 MC18 MC19 MC20 MC21 MC22 MC23 MC24 MC25 MC26 MC27 MC28 [10,12] 18.0% 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% MC0 58 Figure 2.35 SINR Versus MCS in the range of ([10,12], [12,14],…, [26,28] and [28,50]). 2.3.3 Channel Power Configuration 2.3.3.1 RE Power In downlink, eNB is using rate control rather than power control and allocate different power level for different DL channels. To reduce inter‐cell interference and satisfy the quality of received signal of UE at the same time. When configuring the cell, the scheduler verifies that the available power is not exceeded by check the related OFDM symbols, taking into account the RS (reference signal) power, the SCH powers, the PBCH power, the PCFICH power, the PHICH power, the PDCCH power and the PDSCH power. ●● For transmission mode (TM)‐1/2/3/4: TotalPowerRS TotalPowerP SCH TotalPowerS SCH TotalPowerPBCH H TotalPowerPDCCH TotalPowerPCFICH TotalPowerPHICH TotalPoowerPDSCH ●● For TM‐7: TotalPowerRS TotalPowerP SCH TotalPowerS SCH TotalPowerPBCH H TotalPowerPDCCH TotalPowerPCFICH TotalPowerPHICH * PMax / 12.74 0.5 *TotalPowerPDSCH Port 5 PMax PMax Table 2.9 Downlink and uplink MCS Versus ITBS. DL MCSs MCS ITBS MCS_index UL MCSs Mod order Optimize MCS ITBS MCS_index NPRB Mod order Optimize ITBS 1 2 3 99 100 0 16 32 56 2728 2792 0‐QPSK 0 0 2 0‐QPSK 0 0 2 1 24 56 88 3624 3624 1‐QPSK 1 1 2 1‐QPSK 1 1 2 2 32 72 144 4392 4584 2‐QPSK 2 2 2 2‐QPSK 2 2 2 3 40 104 176 5736 5736 3‐QPSK 3 3 2 3‐QPSK 3 3 2 4 56 120 208 6968 7224 4‐QPSK 4 4 2 4‐QPSK 4 4 2 5 72 144 224 8760 8760 5‐QPSK 5 5 2 5‐QPSK 5 5 2 6 328 176 256 10296 10296 6‐QPSK 6 6 2 6‐QPSK 6 6 2 7 104 224 328 12216 12216 7‐QPSK 7 7 2 7‐QPSK 7 7 2 8 120 256 392 14112 14112 8‐QPSK 8 8 2 8‐QPSK 8 8 2 9 136 296 456 15840 15840 9‐QPSK 9 9 2 9‐QPSK 9 9 2 10 144 328 504 17568 17568 10‐16QAM 9 10 4 10‐QPSK 10 10 2 11 176 376 584 19848 19848 11‐16QAM 10 11 4 11‐16QAM 10 11 4 12 208 440 680 22920 22920 12‐16QAM 11 12 4 12‐16QAM 11 12 4 13 224 488 744 25456 25456 13‐16QAM 12 13 4 13‐16QAM 12 13 4 14 256 552 840 28336 28336 14‐16QAM 13 14 4 14‐16QAM 13 14 4 15 280 600 904 30576 30576 15‐16QAM 14 15 4 15‐16QAM 14 15 4 16 328 632 968 31704 32856 (Continued ) Table 2.9 (Continued) DL MCSs UL MCSs NPRB MCS ITBS MCS_index Mod order MCS ITBS MCS_index Mod order ITBS 1 3 99 100 16‐16QAM 15 16 4 16‐16QAM 15 16 4 17 336 2 696 1064 35160 36696 17‐64QAM 15 17 6 17‐16QAM 16 17 4 18 376 776 1160 39232 39232 18‐64QAM 16 18 6 18‐16QAM 17 18 4 19 408 840 1288 42368 43816 19‐64QAM 17 19 6 19‐16QAM 18 19 4 20 440 904 1384 46888 46888 20‐64QAM 18 20 6 20‐16QAM 19 20 4 21 488 1000 1480 48936 51024 21‐64QAM 19 21 6 21‐64QAM 19 21 6 22 520 1064 1608 52752 55056 22‐64QAM 20 22 6 22‐64QAM 20 22 6 23 552 1128 1736 57336 57336 23‐64QAM 21 23 6 23‐64QAM 21 23 6 24 584 1192 1800 61664 61664 24‐64QAM 22 24 6 24‐64QAM 22 24 6 25 616 1256 1864 63776 63776 25‐64QAM 23 25 6 25‐64QAM 23 25 6 26‐64QAM 24 26 6 26‐64QAM 24 26 6 27‐64QAM 25 27 6 27‐64QAM 25 27 6 28‐64QAM 26 28 6 28‐64QAM 26 28 6 LTE Optimization and Principle and Method bitrate/RB (Mbps) 0.700 0.600 0.500 0.400 SINR vs bit rate per RB (16–QAM vs 64–QAM,1×2) 0.300 0.200 0.100 0.000 –5 0 5 10 15 20 25 30 SINR (dB) 70,000 Throughput [kbps] 60,000 50,000 40,000 30,000 Significant change in DL throughput 20,000 Small change in SINR 10,000 0 –10 0 10 20 30 40 SINR [dB] Figure 2.36 SINR versus bitrate per RB. The RS power and the SCH powers are key RF optimization parameters that influences the cell coverage. It is set according to the cell size. The higher the setting, the larger the cell coverage on the downlink, but leaves smaller power headroom available for other downlink signals and channels. The relation between RS power and power on PDSCH symbols (includes type A and type B symbols), has to be defined and reported to the UE, in order for the UE to demodulate 16QAM and 64QAM.1 PDSCH type A symbols has only PDSCH RE’s, PDSCH type B symbols has mixed RS and PDSCH REs. Parameter PA sets the power of PDSCH type A RE relative to cell‐specific RS, PB sets the power relation between PDSCH type B RE, and PDSCH type A RE (ρB/ρA in 3GPP). The ratio of PDSCH EPRE to cell‐specific RS EPRE among PDSCH REs for each OFDM symbol is denoted by either ρA or ρB according to the OFDM symbol index. The mapping between the possible P‐B values and the actual values of the ratio ρB/ρA is in Table 2.10. ●● ●● E(RE_A) (dBm) = 10 Log P (RE_A) (mw), E(RE_B) (dBm) = 10 Log P (RE_B) (mw) E(RS) (dBm) = 10 Log P (RS) (mw) 1 Decoding QPSK/16QAM/64QAM modulated signal at the UE side requires UE to estimate path loss and phase rotation introduced by wireless channel. The process is often called channel estimation. 61 62 LTE Optimization Engineering Handbook Table 2.10 The mapping between the possible P‐B values and the actual values of the ratio ρB/ρA. ρB/ρA P‐B One antenna port Two and four antenna ports 0 1 5/4 1 4/5 1 2 3/5 3/4 3 2/5 1/2 Table 2.11 Reference RS transmit power. Bandwidth and resource blocks Tx mode Config 5MHz (25RB) 10MHz (50RB) 20MHz (100RB) MIMO 2 × 20W 21.2dBm 18.2dBm 15.2dBm 2 × 40W 24.3dBm 21.2dBm 18.2dBm 2 × 60W 26.0dBm 23.0dBm 20.0dBm 2 × 80W 27.3dBm 24.3dBm 21.2dBm ●● ●● ●● ρ A = E (RE_A) – E (RS), ρ B =E (RE_B) – E (RS) ρ A = P‐A (except for MU‐MIMO, if power is distributed to two UE, ρ A = P‐A‐3dB, this is UE specific parameter) ρB/ρA= P‐B (P‐B is cell specific parameter provided by higher layers) The advertised RS power is transmitted on the SIB2 message. This allows UEs to estimate the pathloss to the system reference point. The value set takes RS transmit powers for common configurations and bandwidths in Table 2.11. Power of PDCCHs, PHICHs, PCFICHs, PBCHs, primary synchronization channels, and secondary synchronization channels is set using an offset from RS power. Table 2.12 gives an example of 20MHz DL power setting in a live network. The way to set optimal RS power is not fully known currently. But 3GPP has supplied a possibility to vary the RS power in the range [−3 −2 −1 0 1.77 3 4.77 6] dB relatively PDSCH, which is shown in Table 2.13. A fundamental requirement of the DL system is that full (or near full) allocation on PDSCH and all other channels utilizes the full radio unit max average power (e.g. 40W). p‐A setting: Power offset between the reference signal and PDSCH channel in the symbols without reference signal, P‐A can be found in RRC connection setup message. P‐B setting: Power offset between PDSCH channel in the symbols with reference signal and PDSCH channel in the symbols without reference, P‐B and RS power can be found in SIB2 (Figure 2.37). LTE Optimization and Principle and Method Table 2.12 An example of DL power setting. Inputs Recommended values 20W radio reference Signal Power 12dBm primarySync Signal Power Offset 2.0dB secondary Sync Signal Power Offset 2.0dB PBCH Power Offset 5.5dB PDCCH Power Offset Symbol1 3.5dB PDCCH Power Offset Symbol 2 and 3 3.1dB PCFICH Power Offset 6.4dB PHICH Power Offset 6.1dB paOffset PDSCH 0dB pbOffset PDSCH 0dB (P‐B=1) Cell DL Total Power 43dBm (20W) Table 2.13 Utilization under different combination of P‐A and P‐B. Utilization PB 0 1 2 3 –6 0.67 0.75 0.86 1 –4.77 0.75 0.86 1 0.83 Figure 2.37 P‐A and P‐B. –3 0.86 1 0.83 0.67 PA (dB) –1.77 0.92 0.92 0.75 0.58 0 1 0.83 0.67 0.5 1 0.97 0.8 0.63 0.47 2 0.94 0.77 0.61 0.44 3 0.92 0.75 0.58 0.42 Note: Only four set P-A, P-B=(0,0), (–3,1), (–4.77,2) and (–6,3) can have the maximum utilization of PDSCH power 63 64 LTE Optimization Engineering Handbook PPDSCH,B = pdschTypeBGain*PPDSCH.A pdschTypeBGain = [5/4, 1,3/4, 1/2], linear Overallocation on control region symbol PDSCH, type B boosted 5/4 crsGain Empty RE Pref 6 4.77 3 1.77 0 –1 –2 PDSCH, type A at reference level –3 Figure 2.38 ρA and ρB. Table 2.14 ρA and ρB parameters setting. TC Description PA, PB in L3 trace Baseline No RS boost & no PB boost physicalConfigDedicated pdsch-ConfigDedicated p-a TC1 TC2 RS boost (No PB boost) PB boost (No RS boost) : dB0 physicalConfigDedicated pdsch-ConfigDedicated p-a : dB-3 physicalConfigDedicated pdsch-ConfigDedicated p-a : dB0 pdsch-ConfigCommon referenceSignalPower p-b :1 : 15 pdsch-ConfigCommon referenceSignalPower p-b :1 : 18 pdsch-ConfigCommon referenceSignalPower p-b :0 : 15 The system constant ρA and ρB is used to control reference signal boosting and control PDSCD type PB boosting. Figure 2.38 gives the overview of all the boost levels. The usually recommended ρA and ρB parameters in a live network are given in Table 2.14. Take the example of p‐a=−3dB; p‐b=1, A PA 3 P RE _ A ½P RS A/ B lookup PB 1,2 antenna ports P RE _ B ½P RS 1 B 3 P RE _ A Compared with baseline, TC1 has higher serving RSRP and SINR (3dB), so PDSCH RE can be scheduled higher MCS, downlink throughput/PRB slightly improved. TC2 has the same RSRP and SINR level, but PDSCH RE power is boosted so PSD on receiver side is higher, downlink throughput/PRB improved (Figure 2.39). 2.3.3.2 CRS Power Boosting CRS is transmitted in each physical antenna port. It is used for both demodulation and measurement purpose. Resource element mapping of CRS depends on the number of antenna ports LTE Optimization and Principle and Method DL Throughput vs Serving SINR DL PDCP Throughput (Mbps) 70 60 Linear (TC2) 50 Linear (Baseline) 40 30 20 10 0 Linear (TC1) 0 5 10 15 20 25 30 35 Serving SINR (dB) Baseline TC1(SINR-3dB) TC2 Linear (Baseline) Linear (TC1(SINR-3dB) Linear (TC2) Figure 2.39 DL Throughput/PRB with TC1 and TC2. Type B OFDM symbol (RS & Type B PDSCH) 1 Type A OFDM symbol 1 PB = 0 1 1 1 X 1 1 1 1 1 X 1 1 1 1 1 5/4 5/4 5/4 X 5/4 5/4 1 5/4 5/4 X 5/4 5/4 5/4 1 1 1 1 1 1 1 1 1 1 1 1 𝜌A = 0 dB PB = 1 1 1 1 X 1 1 1 1 1 X 1 1 1 1 1 1 1 X 1 1 1 1 1 X 1 1 1 1 1 1 1 PB = 2 1 0 1 X 1 1 1 0 1 X 1 1 1 1 1 1 1 1 PB = 3 1 0 0 X 1 1 Power 1 1 1 1 1 1 1 1 0 0 X 1 1 1 1 1 1 1 1 1 1 1 1 1 1 𝜌A = –3 dB 3/4 3/4 X 3/4 3/4 3/4 1 1 1 3/4 3/4 X 3/4 3/4 3/4 1 1 1 1 1 1 1 1 1 1 1 1 𝜌A = –4.7 dB 2/4 2/4 X 2/4 2/4 1 1 1 1 2/4 2/4 X 2/4 2/4 2/4 1 1 1 1 1 1 1 1 1 1 1 1 𝜌A = –6 dB Subcarrier 1 RS 0 1 PDSCH (PB is not yet applied) Reduced power for RS power X Nulled RE 1 PDSCH (Type B) 1 Excess power 1 PDSCH (Type A) Figure 2.40 Power boosting for two and four Tx Ant. that the base station is using. CRS power boosting/de‐boosting2 provides means to optimize the power distribution on resource elements to the network geometry and can be used to obtain improved DL throughput in different scenarios. CRS power boost is often used in network optimization. If the power of CRS is same as all other channel power, it would not be easy for UE to detect it. Thus, higher CRS power is 2 CRS power de-boosting can be used to obtain improved DL throughput, for example, in dense networks or interference limited environments. 65 66 LTE Optimization Engineering Handbook referred in order to detect/decode reference signal. Figure 2.40 gives an explanatin of a certain p degree of offset (P‐A) can be configured in order to boost CRS power for two and four Tx antenna cell. For two and four antenna ports, ρA/ρB can be 5/4,1,3/4,1/2 for different P‐B, the four PDSCH below shows P‐A, P‐B=(0,0) (−3,1) (−4.7,2) (−6, 3) respectively. Due to transmit diversity encoding, the PBCH, PCFICH, PDCCH, and PHICH are transmitted at −3 dB compared to the configured value. If SIMO (TM‐1) is used, the PBCH, PCFICH, PDCCH, and PHICH are transmitted at the configured value without transmit diversity. For FDD LTE (10 MHz bandwidth), the method of RS power calculation according to P‐A/ P‐B (−3/1) is: With CRS power boosting: PRS 43 3 3 10 log 50 10 log 12 43 3 3 17 10.8 15.2 dBm For TDD LTE (20 MHz bandwidth), the method of RS power calculation according to P‐A/ P‐B (−3/1) is: DL _ RS _ PowerBCH PsingleAntenna 10 log 12 N RB 37 10 10log 12 100 9.2dbm 10 log 1 PB 10 log 1 1 So we can get: DL_RS_Power + MIMO_gain = 9.2 + 6 =15.2dBm 2.3.3.3 Power Allocation Optimization DL power allocation can impact system performance in the aspects of coverage (coverage extension, eliminate fade zone at inner area), interference, PDSCH power utilization, mobility, and DL throughput. In LTE, UE assume downlink cell‐specific RS EPRE (energy per resource element) is constant across the downlink system bandwidth and constant across all subframes unless different cell‐specific RS power information is received. According to 3GPP, CRS channel gain corresponds to ρA and PDSCH typeB channel gain corresponds to PB or ρB/ρA. Transmission of CRS in each subframe by all the eNBs might cause high interference on both control and data channels of the UE. Adjustable CRS power provides means to optimize the power distribution on resource elements according to the network topology. Absolute mobility borders need to be adjusted for changed CRS power, for example, for inter‐RAT and inter‐frequency handover cell borders, it is recommended that RSRP mobility thresholds are adjusted in the same direction and by the same amount as CRS channel gain, that is, CRS channel gain is lowered by 3 dB, mobility thresholds based on RSRP are also lowered by 3 dB and vice versa. Power on PDSCH can be tuned to control or reduce cell interference. The parameter P‐A can have eight different values of {−6, −4.77, −3, −1.77, 0, 1, 2, 3}, while P‐B can be set to 0, 1, 2, 3 to avoid over‐allocation of RF signal power. Assume there are N TX/RX antennas in each eNB, for each UEj, the power allocation method is described below: For each antenna i, for each RE, LTE can support variable power per PRB and UE, the allocated power per RE |wi,j|2 ≠ 1, where |wi|2, sector is beamformed by appropriate fixed weights wi, if there is power waste exists, the sum of |wi|2 < 1) For each antenna i, the sum of allocated power per antenna is |wi,j|2 = 1 (over antennai), constant power per PRB with flat power spectrum density For each UE, LTE can support variable power per antenna, the sum of allocated power per UE is |wi,j|2 ≠ 1 (over UEj), thus avoid too many weights with high |wi,j|2 for the same antenna (Figure 2.41). In a live network, the refrence total downlink channel power’s calculation is listed in Table 2.15. LTE Optimization and Principle and Method N antennas allocated to UE j wi,j PRBs allocated power per RE ∣wi,j∣2 ≠ 1 allocated power per antenna sum of ∣wi,j∣2 = 1 (over i) allocated power per UE sum of ∣wi,j∣2 ≠ 1 (over j) antenna i Figure 2.41 Power allocation method. Usually there is only a semi‐static downlink power setting, which allows for the attenuation of the maximum cell output power and an extra attenuation when 2x2 MIMO is configured in the cell. During optimization it is recommended to first modify tilts and azimuths before trying to reduce the maximum cell power to reduce the impact of inter‐cell interference. 2.3.4 Link Adaption Link adaptation (LA) is used for adjusting transmission parameters based on differences in the instantaneous channel conditions by the scheduler. LA selects the transport format to ensure that the quality of service requirements are enforced while using resources efficiently. It is also essential for maximizing user throughput over the air interface. The UE measurement of the downlink channel quality is reported to eNB in terms of channel quality indicator (CQI) report. A CQI report assumes that if a certain modulation and coding scheme (MCS) is used, a throughput optimal target block error rate (BLER), usually defined as 10%, will be achieved in a downlink transmission. LA monitors the actual error rate performance of each user (through ACK/NACK events) and dynamically adjusts the SINR thresholds use in the modulation and coding scheme in order to ensure the desired error rate is achieved. In the uplink, the channel quality estimate consists of predicted transmitted PSD (power spectral density), PSD TX, path loss estimate, and noise+interference measurement. The PSD TX is estimated based on UE reported PHR (power headroom report). This is used to estimate uplink gain together with measured PSD RX. In the downlink, gain to interference and noise ratio (GINR=G/(N+I), where G is the path gain) is used as a measure for channel prediction and varies due to fading and interference. The slow fading component is tracked and used in link adaptation. It can be converted to SINR by adding PSD logarithmically (SINR = PSD + GINR). The UE estimates SINR based on the PSD of the downlink reference signals (RS) and PSD offset between PDSCH and RS. The SINR is converted to channel quality indicator (CQI) which indicates the radio quality, and is used by the link adaptation function to select the transport format matching the channel conditions. This leads to improved radio resource utilization. The RAN performs an adaptive adjustment of the SINR derived from CQI to compensate for errors and mismatches, and fulfills the targeted operating point (Figure 2.42 and Figure 2.43). Uplink link adaptation takes care that the UE uses the MCS, which leads to the desired BLER and the appropriate transmission bandwidth for UL data transfer on the PUSCH. These tasks are performed by AMC and the adaptive transmission bandwidth (ATB) components, respectively. Both act on a rather slow time scale. AMC typically reacts within 10 to 100 ms and ATB 67 Table 2.15 Downlink channel power setting. 30 W PA CRS power Slot0‐OFDM symbol# Slot1‐OFDM symbol# 0 1 2 3 4 5 6 0 1 2 3 4 5 6 3981 0 0 0 3981 0 0 3981 0 0 0 3981 0 0 0 primary sync power 0 0 0 0 0 0 4925 0 0 0 0 0 0 secondary sync power 0 0 0 0 0 4925 0 0 0 0 0 0 0 0 PBCH Power Offset 0 0 0 0 0 0 0 1955 1955 2933 2933 0 0 0 0 PDCCH power 15036 29387 29387 0 0 0 0 0 0 0 0 0 0 PCFICH power offset 1271 0 0 0 0 0 0 0 0 0 0 0 0 0 PHICH power offset 6672 0 0 0 0 0 0 0 0 0 0 0 0 0 pbOffset Pdsch 0 0 0 0 15924 0 0 14013 0 0 0 15924 0 0 paOffset Pdsch 0 0 0 23866 0 21020 21020 0 21020 21020 21020 0 23866 23866 26960 29387 29387 23886 19905 25945 25945 19950 22975 23953 23953 19905 23886 23886 Total power LTE Optimization and Principle and Method CQI, PMI, RI RS-PSD PDSCH-PSD UE estimated SINR PSD Offset BS RS PSD conversion to R rted o p Re CQI Used by Link Adaptation to select TBS Figure 2.42 DL link adaptation. Power Control CQI Report Payload PSDTX,target PSDRX,prb Uplink Gain Measurement Transport Format CRC [N+I]prb Outer Loop Gain Adjustment Selection Noise + Interference Filtering Figure 2.43 UL link adaptation. is even slower operating on a time scale which is a multiple (factor 1 to 50) of the AMC time scale. Uplink LA is operated by means of: ●● ●● ●● Slow inner loop LA acting on either first transmission BLER provided by HARQ or on all transport block (TB) transmission BLER Fast outer loop LA acting on first transmission BLER and providing emergency MCS downgrade at high BLER and fast MCS upgrade at low BLER Slow UL adaptive transmission bandwidth based on UE power headroom reports In the conservative LA algorithm 90% weight was put on previous CQI and only 10% weight was put on new CQI measurements. In some new dynamic LA algorithm have shifted 70% of the weight to new CQI measurements and therefore re‐act faster to changing radio environment. Under new dynamic LA algorithm, more samples observed from drive test results in improvement of BLER‐MCS ratio (Figure 2.44). 2.3.5 Adaptive Modulation and Coding Adaptive modulation and coding (AMC) attempts to choose the modulation and coding scheme (per codeword), which results in the best throughput for the scheduled user’s current RF condition as influenced by both signal fading and interference variations. For new transmissions the MCS is decided based on CQI reports from the UE, for retransmissions the same 69 LTE Optimization Engineering Handbook 12 10 8 BLER [%] 70 6 Old LA Improved LA 4 2 0 10 12 14 16 18 20 22 24 26 28 MCS index Figure 2.44 LA algorithm comparison. Table 2.16 Downlink and uplink AMC characteristics. DL AMC UL AMC fast (1ms) slow periodical (30ms) channel aware, CQI based channel aware, SINR based MCS selection: 1 out of 0‐28 MCS adaptation, +/−1MCS correction output: MCS, TBS output: MCS, ATB UE capabilities support: Max.TBS per TTI UE capabilities support: Power headroom, QoS profile MCS as the original transmission is applied. The main characteristics of downlink and uplink AMC is listed in Table 2.16. eNB determins the MCS in a given PRB allocation. According to 3GPP protocol, TB size is selected by the MCS and number of PRBs allocated, such as Table 2.17. 2.3.6 Scheduler To provide efficient resource usage, the LTE concept supports fast scheduling where resources on the shared channels (PDSCH and PUSCH) are assigned to users and radio bearers on subframe basis according to the users momentary traffic demand, QoS requirements and estimated channel quality. The eNB scheduler determines the trade off between end‐user QoS characteristics and aggregate cell throughput. There are many scheduler strategies in LTE: channel unaware scheduler, channel aware scheduler, interference aware scheduler, round robin scheduler, exhaustive scheduler, time domain scheduler, and frequency domain scheduler. Usually it is assumed that non‐GBR bearer solutions typically will be based on proportional fair (PF) scheduling (without minimum rate) and for GBR bearers, voice/video telephony can be based on proprotinal fair scheduler with minimum rate. Chapter No.: 1 Title Name: <TITLENAME> Comp. by: <USER> Date: 06 Sep 2017 Time: 06:33:35 PM Table 2.17 MCS, number of PRB and TBS selection (example). Number of PRBs MCS Index Modulation Order TBS index (informative) 74 75 95 96 97 98 99 100 21 6 19 31704 32856 40576 40576 42368 42368 42368 43816 22 6 20 34008 35160 43816 45352 45352 45352 46888 46888 23 6 21 36696 37888 46888 48936 48936 48936 48936 51024 24 6 22 40576 40576 51024 51024 52752 52752 52752 55056 Stage: <STAGE> 25 6 23 42368 43816 55056 55056 55056 57336 57336 57336 26 6 24 45352 45352 57336 59256 59256 59256 61664 61664 27 6 25 46888 46888 61664 61664 61664 61664 63776 63776 28 6 26 55056 55056 71112 71112 71112 73712 73712 75376 <WORKFLOW> WorkFlow: MCS Index IMCS Modulation Order QM TBS Index ITBS 0 2 0 1 2 1 2 2 2 3 2 3 c02.indd Page Number: 71 4 2 4 ... ... ... 16 4 15 17 6 15 ... ... ... NPRB ITBS 1 2 3 ... 110 0 16 32 56 ... 3112 1 24 56 88 ... 4008 2 32 72 144 ... 4968 3 40 104 176 ... 6456 3) Nprb 16 328 632 968 ... 35160 ... ... ... ... ... ... 26 712 1480 2216 ... 75376 Transport Block Size (TBS) in bits 72 LTE Optimization Engineering Handbook LTE considers several factors to allow for high performance user scheduling and resource allocation algorithms, for example, users and their queues, present delays, radio qualities, priorities, QoS requirements, power restrictions, buffer filling status, and last scheduled TTI, and so on. Normally, the user scheduling and resource allocation problem attempts to maximize at each scheduling instant a weighted sum rate metric, and proportional fair scheduling, which is shown below: max Su u wu u Ru ru Su Su denotes a particular set of PRBs allocated to user u, ru(Su) is the rate achievable by user u when given resources Su. λu is a per‐user grade of service factor to alter priorities based on user service class (e.g., gold/silver/bronze) and bearer type. wu is a per‐user weighting factor that enforces QoS constraints (e.g., min bit rate, etc.) for a given bearer. Ru is the average rate achieved by user u over a prescribed time window. a is a fairness factor: α=0 results in max C/I scheduling, α=1 results in proportional fair (PF) scheduling. 2.3.6.1 Downlink Scheduler The DL scheduling is done per logical channel, each one characterized by a priority queue. The scheduler calculates a weight for each queue in due time for each scheduling occasion. A downlink scheduler is responsible for allocating RB resource for UE and data transfer via PDSCH. The overview of a downlink scheduler is depicted in Figure 2.45. A downlink AMC decides each UE downlink transmission MCS used based on the CQI. A CQI report can be periodic and aperiodic in the uplink PUCCH/PUSCH channel transmission. And the relevant control information for downlink transmission and decoding is transmitted on the PDCCH. The downlink scheduler functions include following items: ●● ●● Pre‐schedule, usually used for resource allocation of common channel and random access. UE is scheduled by listening to the corresponding downlink subframe (data availability, HARQ retransmission, measurement cycle, DRX state, etc.). Time domain scheduling, which is proportional fair method. It is a compromise between system throughput and fairness among users. Using the minimum data bit Link Adaptation OLQC Packet Scheduler BLER measurements Offset AMC QoS parameters MAC PHY DL transmission on PDSCH Figure 2.45 Downlink scheduler. CQI/PMI/RI report (from UE on PUCCH/PUSCH) LTE Optimization and Principle and Method Channel realization Gain-to-interference ratio [dB] 0 5 0 5 0 5 0 100 90 80 70 60 50 Frequency [MHz] 500 40 30 20 10 0 0 100 200 300 600 400 Time [ms] Figure 2.46 Frequency selective scheduling. ●● rate, applicable to the service that rate requirement is low, the quality of service requirements is high, such as VoIP. Frequency selective scheduling, selectively scheduling the active UEs in different sub‐bands where their channel qualities are better, both cell throughput and coverage can be improved. Ideally, an attempt to allocate to each user the time‐frequency resources in the red and orange regions as shown in the Figure 2.46. A downlink scheduler’s decisions should be made every TTI by eNB. eNB decides which UEs to transmit to (which UEs to be scheduled) and which bearer(s) of a UE to schedule, how many PDCCH resources (CCEs, control channel elements) to allocate and where and how many PDSCH resources (PRBs) to allocate, how much data to transmit, and sends new data or retransmits data that failed HARQ, and what HARQ process ID to use, and so on. For each transmission, eNB should decide the following waveform controlling parameters like transmit mode, modulation scheme, transmit scheme, and precoding matrix and so on. In LTE, HARQ schemes can be categorized as either synchronous or asynchronous, with the retransmissions in each case could be adaptive or non‐adaptive. In a synchronous HARQ scheme, the retransmission(s) for each process occur at predefined times relative to the initial transmission. So there is no need to signal information such as HARQ process number, as this can be inferred from the transmission timing. By contrast, in an asynchronous HARQ scheme, the retransmissions can occur at any time relative to the initial transmission, so additional explicit signaling is required to indicate the HARQ process number to the receiver, so that the receiver can correctly associate each retransmission with the corresponding initial transmission. In an adaptive HARQ scheme, transmission attributes such as the MCS, and PRBs can be changed at each retransmission in response to variations in the radio channel conditions. In a non‐adaptive HARQ scheme, the retransmissions are performed without explicit signaling, and new transmission attributes can either use previous transmission scheme, or by changing the attributes according to a predefined rule. 73 74 LTE Optimization Engineering Handbook Table 2.18 UL MSC and TBs. MCS Index Modulation #RBs TBS (bits) Coding Rate TxPwr (dB) MCS Index Modulation #RBs TBS (bits) Coding TxPwr Rate (dB) 0 QPSK 1 104 0.36 −0.6 16 QPSK 25 4968 0.69 17.6 1 QPSK 1 232 0.81 4.8 17 QPSK 30 3112 0.36 14.1 2 QPSK 2 232 0.4 3.1 18 QPSK 48 5992 0.43 17.3 3 QPSK 2 320 0.56 5.1 19 QPSK 40 4968 0.43 16.5 4 QPSK 3 320 0.37 4.3 20 QPSK 48 4968 0.36 16.2 5 QPSK 3 504 0.58 7.2 21 QPSK 30 5992 0.69 18.4 6 QPSK 5 504 0.35 6.2 22 16QAM 1 320 0.56 7.7 7 QPSK 5 776 0.54 8.9 23 16QAM 2 776 0.67 12.7 8 QPSK 5 1000 0.69 10.7 24 16QAM 2 1000 0.87 15.9 9 QPSK 10 1544 0.54 11.9 25 16QAM 5 1544 0.54 14.3 10 QPSK 15 1544 0.36 11.1 26 16QAM 5 2344 0.81 19 11 QPSK 15 2344 0.54 13.7 27 16QAM 10 3112 0.54 17.4 12 QPSK 20 3112 0.54 14.9 28 16QAM 10 4008 0.7 20.1 13 QPSK 20 4008 0.7 16.7 29 16QAM 10 4968 0.86 22.8 14 QPSK 25 4008 0.56 16.1 30 16QAM 15 4968 0.58 19.8 15 QPSK 25 3112 0.43 14.5 31 16QAM 15 5992 0.69 21.8 2.3.6.2 Uplink Scheduler Different from downlink, where the eNB has full knowledge of transmission buffer, in uplink, eNB relys on the buffer status information, which is sent by UE to perform schduling. The buffer status information including SR and buffer status reporting (BSR). SR is one bit, which can only indicate there is new data arrives in the UE buffer. Based on that information, the function “buffer estimation” residing in eNB scheduler estimates the UE buffer status. Besides BSR, the uplink dynamic scheduling also need considers the following factors, QoS priority weight, channel estimates, spectrum efficiency, and so on. During scheduling and resource allocation, the eNB only needs to calculate the UL_SINR using for a given PRBs, UL and downlink pathloss, and then determines MCS levels according to Table 2.18 for UEs with limited TBS. The eNB may configure the UE to transmit a wide‐band sounding reference signal that can be used for estimating the UL channel quality and frequency selective scheduling as shown in Figure 2.47. Additional channel quality estimation can be obtained from other UL transmissions such as, data transmission or control signaling (CQI reports and HARQ ACK/NACK signals). The UL scheduling in contrary to DL instead operates on groups of logical channels, but is otherwise using similar strategies and weight functions to grant resources. The UL scheduler derives UE grants based on the received BSR. The BSR is provided by the UE and occurs per logical channel group (LCG), where LCG#0 is reserved for the logical channels used by the signaling radio bearers and three (LCG#1‐3) provided for data radio bearers. It is important that the delay sensitive GBR traffic such as that arising from voice or conversational video is preferably assigned to a separate logical channel group. The UE needs LCG to report to the eNB, which radio bearers need UL resources and how much resource they need. The UL scheduler is based on channel unaware to select uplink PRBs, and power control headroom will also LTE Optimization and Principle and Method UEs that are valid for scheduling: UL sync, data to transmit and no DRX etc. UE 1 UE 2 UE 3 UE 4 UE 5 UE 6 Scheduling stragety resource fair, proportional fair delay based... Time domain scheduling Hihgest weight UEs: SRS and PDCCH resource allocation UE 1 UE 2 UE 3 UE 4 Resource allocation SRS measurement interference estimation UE 7 Resource fair frequency selective Figure 2.47 Uplink frequency selective scheduling. impact the result. Thr related scheduling parameters includes initial MCS in uplink and UL target BLER, and so on. 2.3.7 Radio Frame In FDD‐LTE every downlink subframe can be associated with an uplink subframe that is different with TDD‐LTE. 3GPP defined seven TDD UL/DL subframe pattern for different deploy conditions leading to different uplink and downlink resources as well as different GP time length; for configuration 0, 1, 2, and 6, there will be two DL→UL and two UL→DL switch point in each 10ms, and for configuration 3, 4, and 5, there will be only one DL→UL and one UL→DL switch point in each 10ms (Table 2.19). TD‐LTE special subframe has nine pattern shown in Table 2.20, different subframe types represent a tradeoff between throughput and coverage. DwPTS is always reserved for downlink transmission. UpPTS is always reserved for uplink transmission. Guard period required to Table 2.19 TD‐LTE radio frame. UL‐DL configuration DL‐to‐UL switch‐point periodicity Subframe number 0 1 2 3 4 5 6 7 8 9 0 5 ms D S U U U D S U U U 1 5 ms D S U U D D S U U D 2 5 ms D S U D D D S U D D 3 10 ms D S U U U D D D D D 4 10 ms D S U U D D D D D D 5 10 ms D S U D D D D D D D 6 5 ms D S U U U D S U U D 75 76 LTE Optimization Engineering Handbook Table 2.20 TD‐LTE special subframe pattern. Normal cyclic prefix Configuration DwPTS Extended cyclic prefix GP UpPTS 1 OFDM symbols 0 3 10 1 9 4 2 10 3 11 4 12 1 5 3 9 6 9 7 10 8 11 DwPTS GP UpPTS 1 OFDM symbols 3 8 8 3 3 9 2 2 10 1 3 7 8 2 3 9 1 2 ‐ ‐ ‐ 1 ‐ ‐ ‐ 2 OFDM symbols 2 OFDM symbols Table 2.21 TD‐LTE DL/UL user peak throughput (special subframe pattern 7, PUCCH resource 4PRB). DL:UL frame configuration Transmission mode DL L1 throughput [Mbps] Con1(DL:UL=2:2)/Con2 (DL:UL=3:1) TM3 59.6/80 Con1(DL:UL=2:2)/Con2 (DL:UL=3:1) TM7 34.0/46.8 Con1(DL:UL=2:2)/Con2 (DL:UL=3:1) ‐ ‐ UL L1 throughput [Mbps] 20.4/10.2 ensure uplink and downlink transmissions do not clash. Large guard period will limit capacity. Larger guard period normally required if distances are increased to accommodate larger propagation times. Assuming special subframe pattern 7, PUCCH resource 4PRB, TD‐LTE DL/UL user peak throughput is shown in Table 2.21. 2.3.8 System Information and Timers 2.3.8.1 System Information The system information broadcast transmits all necessary parameters and information about the cell to all UEs in the cell. LTE has 12 system information messages which is classified into the master information block (MIB) and a number of system information blocks (SIBs). Broadcast of system information (SI) is a function of the RRC sub‐layer between the UE and eNB. After reading the system information, the UEs know how to behave in the cell and can perform the other idle mode tasks, such as PLMN selection, cell selection and reselection, monitoring the paging channel, and performing tracking area update. MIB is transport on BCH, defines the information about the most essential physical layers of the cell required for receiving further system information: downlink system bandwidth, number of transmit antennas, PHICH c onfiguration (duration and resource), and system frame number as shown below. 78 LTE Optimization Engineering Handbook The MIB is broadcast on the physical broadcast channel (PBCH), while SIBs are sent on the physical downlink shared channel (PDSCH) through radio resource control (RRC) messages. SIB1 is carried by “SystemInformationBlocktype1” message. SIB2 and other SIBs are carried by “SystemInformation(SI)” message. SIB1 contains the information for checking whether a UE is allowed to access a cell and for defining the scheduling of other system information blocks: access related parameters (e.g., whether UE is permitted to camp on the cell), cell selection parameter, scheduling block (SB) providing scheduling details for other SIBs, TDD parameters, SI window length, value tag, and so on. LTE SIBs need to map into SI before transmisson, all scheduling and mapping information of SI are included in the SIB1. UE does not always receive SIBs, SIB1 has a value tag, used to display system information whether changes. The different SIBs are then mapped onto different SIs, which corresponds to the normal the actual transport blocks to be transmitted on DL‐SCH. SIB1 is always mapped, by itself, onto the first SI‐1. The remaining SIBs will be group‐wise multiplexed onto the same SI, which have the same transmission period. The total bits that is mapped onto a single SI must not exceed the transport block and the SIB‐to‐SI mapping for SIBs beyond SIB1 is flexible and may be different in one network. Below is a fragment of SIB1, it gives each SI’s periodicity and members of each group. schedulingInfoList { { si-Periodicity rf8 sib-MappingInfo { sibType3} } { si-Periodicity rf64 sib-MappingInfo { sibType4 sibType5 sibType7 } } } ...... si-WindowLength ms10 sibType6 The example of SIBs scheduling period is shown in Figure 2.48. SIB1 80 ms SIB2 160 ms SIB3 320 ms SIB4/5 640 ms SIB6/7/8 1280 ms eNB MIB BCH SIB1 SIB2 SIB3 SI-1 SI-2 DL-SCH --- SIB1 UE SIB2 SIB3 SIB4 SIB5 SIB6 SIB7 SIBN DL-SCH SI-N Period: SI-1 80 ms Figure 2.48 The example of SIBs scheduling period. SI-2 160 ms SI-3 320 ms SI-4 320 ms SI-5 640 ms SIB8 LTE Optimization and Principle and Method Table 2.22 The function of MIB and SIBs. Systerm parameters related to MIB SIB1 SIB2 SIB3 SIB4 SIB5 SIB6 SIB7 SIB8 SIB9 Cell selection info × PLMN‐id × Tracking area code × Cell Id × Cell barraed × Frequency band indicator × SIB scheduling × × UL EARFCN × UL bandwidth DL bandwidth × Common radio resource config × Paging info × Cell reselection Neighboring cell‐intra f Neighboring cell‐inter f Inter RAT reselection(UTRAN) Inter RAT reselection(GRAN) Inter RAT reselection home eNB ETWS notification SIB10 SIB11 × × × × × × × × × MIB and SIB1 are scheduled in a fixed manner. MIB is with a periodicity of 40ms and SIB1 is 80ms which will occupy the fifth subframe. For the remaining SI, the scheduling on DL‐ SCH is more flexible in the sense that each SI can transmit in any subframe within time window (SI‐window) with well‐defined starting point and durations. Also the paging message is used to inform the UEs in RRC_idle and the UEs in RRC_connected state of the change of the system information. The functions of other SIBs is described below as present in Table 2.22: ●● ●● ●● ●● ●● ●● ●● SIB1: Cell access info (PLMN, TAC, CID…) SIB2 contains the information about common and shared channels; frequency information (UL‐carrier frequency, UL‐bandwidth); default paging cycle SIB3 contains cell re‐selection information, mainly related to the serving cell; SIB4 contains the information about the serving frequency and intra‐frequency neighboring cells related to cell re‐selection; SIB5 contains the information about other E‐UTRA frequencies and inter‐frequency neighboring cells related to cell re‐selection; SIB6 contains the information about UTRA frequencies and neighboring cells related to cell re‐selection; SIB7 contains the information about GERAN frequencies related to cell re‐selection; 79 80 LTE Optimization Engineering Handbook Table 2.23 System information parameters. Parameter Range Recommand Description siWindowLen 5ms (2), 10ms (3), 15ms (4), 20ms (5), 40ms (6) 20ms (5) Common SI scheduling window for all SIs si2Mappinginfo SIB2 (0) SIB2 (0) Define which SIB is in SI‐2 si2Periodicity 80ms (0), 160ms (1), 320ms (2), 640ms (3), 1280ms (4), 2560ms (5), 5120ms (6) 160ms(1) Periodicity of SI2‐message in radio frames si2Repetition 1…4, step 1 1 The No. of transmission of SI2 in a SI window si3Mappinginfo SIB3 (1) SIB3 (1) List of the SIBs mapped to this SI message. si3Periodicity 80ms (0), 160ms (1), 320ms (2), 640ms (3), 1280ms (4), 2560ms (5), 5120ms (6) 160ms(1) Periodicity of SI3‐message in radio frames si3Repetition 1…4, step 1 1 The No. of transmission of SI3 in a SI window si4Mappinginfo SIB4 (2), SIB5 (3), SIB6 (4), SIB7 (5), notUsed (18) SIB4 (2) Define which SIB is in SI‐4 si4Periodicity 80ms (0), 160ms (1), 320ms (2), 640ms (3), 1280ms (4), 2560ms (5), 5120ms (6), notUsed (18) 1280ms (4) Periodicity of SI4‐message in radio frames si4Repetition 1…4, step 1 1 The No. of transmission of SI4 in a SI window ●● ●● ●● ●● ●● SIB8 contains the information about CDMA2000 frequencies and CDMA2000 neighboring cells related to cell re‐selection; SIB9 contains a home eNB identifier; SIB10 contains an ETWS3(Earthquake and Tsunami Warning System) primary notification; SIB11 contains an ETWS secondary notification; SIB11 contains CMAS4 (commercial mobile alerting system) notification; The main system information parameters are listed in Table 2.23. SIB scheduling cycle length reflects the SIB transmission time interval length in the air interface, which reflects the transmission frequency of this SIB. The shorter the cycle, the faster the transmission frequency, the less waiting time for the UE to read the SIB, but more resources to be sent of the SIB. The scheduling cycle of each SIB can be configured individually. SIB scheduling cycle is related with its content of the importance, cell bandwidth resources, delay requirements of UE acquired this SIB. The recommended configuration scheme is shown in Table 2.24. 3 ETWS is an LTE system information message that delivers early warning of impending earthquakes or tsunamis, to all UEs in a specified area. Primary and secondary warning types exist and are transmitted as SIB10 and SIB11. 4 CMAS is a system that distributes a range of (mostly commercial) messages to all UEs in a specified area. A CMAS message may carry any information that is of interest to the public. LTE Optimization and Principle and Method Table 2.24 The recommended SIB scheduling cycle scheme. Bandwidth ≥5M (high speed cell) Bandwidth ≥5M (low speed cell) Sib2SchPeriod 160 ms 160 ms Sib3SchPeriod 160 ms 320 ms Sib4SchPeriod 160 ms 320 ms Sib5SchPeriod 160 ms 320 ms Sib6SchPeriod 320 ms 640 ms Sib7SchPeriod 320 ms 640 ms Sib8SchPeriod 640 ms 640 ms 2.3.8.2 Timers Timers optimization are important items in the whole optimization work. Constants and timers in SIB2 track RRC connection establishment and RRC re‐establishment (RLF detection and recovery) procedure. The following timers will be introduced. T300 is a UE timer, which is started when sending RRC connection request and is stopped upon reception of RRC connection setup or RRC connection reject message. Once the T300 expires, if the number of retransmissions is less than N300, then retransmit RRC connection request, otherwise the connection request fails and UE enters the idle mode. Increase the value of T300, access success rate will be improved (Figure 2.49). T301 is used when the UE is unable to successfully RACH when trying to send RRC connection reestablishment. Started after RRC connection reestablishment request message, stopped reception of RRC connection reestablishment or RRC connection reestablishment reject message as well as when the selected cell becomes unsuitable for reestablishment. When T301 is at expiration UE will go to RRC idle. <tx>The higher setting of T301, UE will have more time for cell selection, if the timer set too low, this will reduce the possibility of successful RRC re‐establishment. Typical duration between request and reject messages varied between 25 ms and 70 ms (Figure 2.50). T302 is a UE timer, it is started upon reception of RRC connection reject message, the UE should start timer T302 and is not allowed to send another RRC connection request on the same cell until the expiry of T302. It has 5s as default, in particular cases of congestions it could be advisable to increase it. RRC: RRCConnectionSetupRequest start T300 processing in eNB stop T300 pos: RRC: RRCConnectionSetup or neg: RRC: RRCConnectionReject or Figure 2.49 T300. 81 82 LTE Optimization Engineering Handbook RRC: RRCConnectionReestablishmentRequest start T301 processing in eNB stop T301 pos: RRC: RRCConnectionReestablishment or neg: RRC: RRCConnectionReestablishmentReject Figure 2.50 T301. T310: T310 starts upon detecting physical layer problem, that is, upon receiving N310 consecutive out‐of‐sync indications from lower layers, and stops upon receiving N311 consecutive in‐sync indications from lower layers, upon triggering the handover procedure and upon initiating the connection re‐establishment procedure (Figure 2.51). T311: started when a radio link failure is declared and initiating the RRC connection re‐establishment procedure when T310 expiry, stops when a suitable cell is selected for cell reselection (Figure 2.52). The UE uses timers T310 and T311 to get time to restore the connection with the eNB. During the time T310 + T311 the UE stays in RRC_connected state. If the UE can not reestablish connection to the eNB where it had a context stored during the normal operation, the UE switches from RRC_connected to RRC_idle and initiates the procedure to establish a new RRC connection to a new eNB. T304 indicates handover failure, for the UE is unable to RACH to target cell at handover. With reception of “RRC connection reconfiguration message” T304 starts, if “RRC connection Out of sync indicators Figure 2.51 T310. in sync indicators Resume RRC connection Start T310 Stop T310 send RRC connection reestablishment Request failure Trigger* suitable cell selected RRC connection reestablishment received perform cell (re)selection UE resumes RRC connection start T311 stop T311/ Start T301 Figure 2.52 T311. stop T301 t LTE Optimization and Principle and Method UE S-eNB T-eNB Measurement report (A3) Last RLC PDU T304 starts RRC connection reconfig MSG-1 UE continues to send MSG-1 untill T304 expiry, but never receives MSG-2 MSG-1 T304 expiry RRC connection reconfig completed can not be received Figure 2.53 T304. reconfiguration complete message” is not received by target eNB, before expiry of T304, handover process will be failed with certain reason (Figure 2.53). The above timer default value settings can be found in the annex. 2.3.9 Random Access Random access has an impact on the average call setup time, while the network has a slight capacity impact on the high load (interference) with different RACH parameters. Its processes are involved on the MAC and PHY layers, which are carried through the PRACH and PDCCH channels, respectively. The RACH channel carries a random access preamble and takes up 6 RB of a subframe. RACH process as shown in Figure 2.54. UE chooses a random access preamble from the BCCH broadcast random access preamble list, followed by computing the OLPC (open loop power control) parameters (initial TX power) and verificate the conflict related parameters (such as random access attempt whether has exceeded the maximum number of attempts). After that, UE sends the initial RACH to the eNB and waits for the response to guide the next attempt. OLPC ensures the power of the repeated preamble of the UE, is higher than the previous preamble. After the successful reception of the uplink RACH preamble, eNB will calculate whether the uplink capacity is allowed for the UE access, and at the same time to calculate the power adjustment as well as the timing parameters. Two types of RA procedures are defined in the standard CBRA (contention based random access), and CFRA (contention free random access) which are shown in Figure 2.55. CBRA involves the UE randomly selecting a preamble sequence from a predefined list in order to access the cell. It is a four‐step process. CFRA involves the UE being sent a preamble sequence to use in the RA process, which uniquely identifies the UE to the cell. It is a three‐step process, which is faster than CBRA, usually in this case, the network informs each of the UE of exactly when and which preamble signature it has to use. For CBRA, network would go through additional process, at later step to resolve these contention and this process is called “contention resolution” step. Table 2.25 gives the main random access parameters. 83 BCH information UE sets the initial transmission power of RACH and send preamble signal Preamble (RACH) Preamble (RACH) Preamble (RACH) PDCCH Random access message (UL-SCH) Figure 2.54 Random access. 1 Random Access Preamble PRACH Random Access Response (adressed to RA-RNTI, RAPID, TA, Inital UL grant, Temporary C-RNTI) 2 PDCCH PDSCH Scheduled transmission 3 (L3 msg. incl UE id, Temporary CRNTI/C-RNTI) PUSCH Contention resolution (Incl. L3 msg, incl UE id.) 4 S-eNB PDSCH T-eNB RA Preamble assignment RA Preamble assignment via dedicated signalling in DL HO command or PDCCH in case of DL data arrival RA Preamble Random Access Response (adressed to RA-RNTI, RAPID, TA, Inital UL grant in case of handover) Figure 2.55 Contention‐based (top) and contention‐free random access (bottom). LTE Optimization and Principle and Method Table 2.25 Random access parameters. Parameter Range Recommend Description prachCS 0…15, step 1 11 PRACH cyclic shift prachConfIndex 3…53, step 1 3 PRACH Configuration index rootSeqIndex 0…837, step 1 ulpcIniPrePwr −120 (0), −118 (1), −116 (2), −114 (3),……,−92 (14), −90 dBm (15) −104 dBm (8) Preamble initial received target power RACH Root Sequence prachPwrRamp 0dB (0), 2dB (1), 4dB (2), 6dB (3) 4dB (2) Power increment step preambTxMax 3 (0), 4 (1), 5 (2), 6 (3), 7 (4), 8 (5), 10 (6), 20 (7), 50 (8), 100 (9), 200 (10) 10 (6) Max. RA transmissions raRespWinSize 2 (0), 3 (1), 4 (2), 5 (3), 6 (4), 7 (5), 8 (6), 10 (7) 10 (7) defines the window size for the random access response in TTIs 2.3.10 Radio Admission Control The task of radio admission control is to admit or reject the establishment requests for new radio bearers. In order to do this, it takes into account the overall resource situation, paging, cell loading, handover, incoming a new call, the QoS requirements, the priority levels and QoS of in‐progress sessions as well as the QoS requirement of the new radio bearer request. The purpose of the admission control is to leave a headroom for other traffic not subject to admission control for the specific resource (signaling, non‐GBR), incoming user mobility, varying radio conditions and intra‐cell mobility. The goal is to protect other traffic and to minimize call dropping (service blocking over service dropping). Non‐GBR capacity varies with GBR load, admission threshold should reserves a minimum non‐GBR capacity (Figure 2.56). Parameters related to radio admission control are shown in Table 2.26. Served traffic Non-GBR Partition 2, QCI = 7,8,9 Maintained by Admission control Non-GBR Partition 1, QCI = 5,6 Admission Threshold (AT) AT2 GBR Partition 2, QCI = 3,4 AT1 GBR Partition 1, QCI = 1,2 Bitrate admitted into partition Bitrate admitted into partition time Figure 2.56 Radio admission control. 85 86 LTE Optimization Engineering Handbook Table 2.26 Parameters related to radio admission control. Parameter Description Range maxBitrateDL Maximum bit rate downlink 1000…300000 kbps, step 500 kbps maxBitrateUL Maximum bit rate uplink 1000…75000 kbps, step 500 kbps maxNumActUE Maximum number of active UEs 0…190, step 1 maxNumRrc Maximum number of RRC connections 0…400, step 1 minBitrateDl Minimum bitrate downlink 5…300000 kbps, step 5 kbps minBitrateul Minimum bitrate uplink 5…75000 kbps, step 5 kbps 2.3.11 Paging Control 2.3.11.1 Paging Paging is a basic feature allowing the MME to notify a UE about incoming data connections. There are a number of reasons why the network needs to initiate contact, the most common of which involves downlink data pending in the S‐GW, which needs to be delivered to the UE. Another reason is that the EPC may also want to establish control plane connectivity for other reasons, for example, the network initiated TAU procedure or for location services. When the MME receives a downlink data notification message from SGW, the MME sends an S1‐AP paging message (each message transfers a single paging record) to all eNBs in the TA list. With the MME functionality configurable and adaptive paging, a page can first be sent only to a single eNB or TA before it is sent to all eNBs in the TA list.5 When the S1‐AP paging message arrives at the eNB it is queued until the valid paging occasion (PO) occurs. The message is then transmitted over the air interface using resources on PDCCH and PDSCH. Several UE may be addressed in the same RRC paging message, which encapsulated and multiplexed paging requests sent from eNB to UEs. RRC paging message consists of up to 16 paging records. The DL control information (DCI) containing the scheduling assignment for the paging message is transmitted over PDCCH. The scheduling assignment is common for all UE monitoring a certain PO. UE monitors PDCCH to detect if a paging message is included and checks if there is P‐RNTI (paging radio network temporary identifier, sent on PDCCH when there is a RRC paging message allocated on PDSCH) allocated in common search space. Figure 2.57 illustrates the paging procedure. Table 2.27 presents the RRC paging message IE and semantics description. When the UE is under the mode of idle, the network wants to send the data, the call flow is S_TMSI based paging, if the network happens error and need to recover (S_TMSI can not be used), the network can do the IMSI paging, UE detach the network after receive the signal, then attach (Figure 2.58). When paging messages arrive in the RAN, the RRC layer tries to send the paging message in the first valid PO. If it is impossible to send the paging message in the first PO because of blocking, for example, attempts are made to send the paging message in subsequent POs according to the DRX (discontinuous reception, for reducing power consumption) cycle in idle mode. DRX will enable an efficient paging procedure that allows the UE to sleep with no receiver processing most of the time and to briefly wake up at predefined time intervals to monitor 5 The MME can employ certain smart mechanisms (e.g., tiered paging, last cell paging, sequential paging) to reduce paging overhead, which in turn can allow larger tracking areas. Figure 2.57 Paging strategy. UE eNB MME S1AP paging RRC paging The MME initiates a paging message which is sent to all eNBs in a TA UE uses the random access procedure to initiate access to the serving cell Random access procedure NAS: Service request NAS messaging continues in order to set up the call 1. Downlink Data Notification S1-AP: initial UE message +NAS: Service request +eNB UE signaling connection ID SGW MME TAC=2 2. S1AP Paging message TAC = 1 3. DCI (PDCCH) RRC Paging Message (PDSCH) TAC = 3 4. Service request TA list: TAC1 TAC3 3. DCI (PDCCH) RRC Paging Message (PDSCH) UE being paged Table 2.27 RRC paging message. IE/Group Name source Semantics description pagingRecordList eNB Only included if an UE specific paging has been triggered by S1AP. SEQUENCE (1..maximum No. of pagingrecords): 1 or more elements in the list depending on the number of UEs to be paged at the same paging occasion. (one paging record per UE) MME UE Identity as provided by S1AP: paging message (UE paging identity) systemInfoModification eNB This IE is present if a system information change is notified. eNb can inform UEs about system information changes. It is done via paging. ETWS‐Indication eNB The IE is present if an ETWS primary/secondary notification shall be indicated. When ETWS (earthquake and tsunami warning system) is to be sent, it is done via paging. eNB This IE is present, if a CMAS notification shall be indicated. > UE Identity >> choise S‐TMSI >>> S‐TMSI >> choise IMSI >>> IMSI > CN‐Domain Provided by the S1AP: paging message. Paging‐v890‐IEs > lateNonCriticalExtension > Paging‐v‐920‐IEs >> cmas‐Indication‐r9 >>nonCriticalExtension Omitted 88 LTE Optimization Engineering Handbook UE eNB EPC UE eNB NAS:Paging (S_TMSI) EPC NAS:Paging (IMSI) NAS:Paging (S_TMSI) NAS:Paging (IMSI) Same as the procedure of service request initiated by UE Same as the procedure of attach request initiated by UE Figure 2.58 Paging procedure. Table 2.28 DRX/DTX parameter. Parameter Range Default Default paging cycle 32rf (0), 64rf (1), 128rf (2), 256rf (3) 64rf (1) Inactivity timer 10…65535 s, step 1 s 10 s Paing nB 1/32T, 1/16T, 1/8T,1/4T,1/2T, 1T, 2T, 4T; T is the defaultPagingCycle. 1T (2) aging information from the network. Paging DRX cycle defines time between two POs of the p same UE. The RRC layer tries to send the paging message during a period specified by the parameter pagingDiscardTimer, after which the paging message is discarded. It is recommended that the pagingDiscardTimer should be set equal to or smaller than T3413. To guarantee at least one retransmission attempt by the RRC layer, the pagingDiscardTimer must be set to a larger value than the defaultPagingCycle. If the MME does not receive the service request within T3413 seconds, it resends the S1‐AP paging message according to the paging profile. The DRX/ DTX parameters are present in Table 2.28. The default paging cycle defines the cell specific paging DRX cycle duration (periodicity of the paging). It also determines the maximum paging DRX duration applicable in the cell. Value rf32 corresponds to 32 radio frames, rf64 corresponds to 64 radio frames, and so on. One radio frame is 10ms. Current recommended value is 64rf, since it improves the paging time. Higher parameter values can save battery in idle mode as listening to paging in less frequently but it also means that call setup is getting longer due to longer average paging time. pagingNb defines the number of possible paging occasions per radio frame, that is, the density of paging occasions. It is used to calculate the number and position of paging occasions (PO) and paging frames (PF). By increasing pagingNb the number of paging occasions per second is increased. 3GPP defined there are maximum 16 users per paging occasion. There is one paging record per UE in each paging occasion so it is possible to page a maximum of 16 UEs. Therefore, increasing pagingNb means it is possible to page more UEs, and increasing the paging capacity (Figure 2.59). SFN0 nB = ½T PO nB = T nB = 2T SFN1 SFN2 SFN3 Figure 2.59 pagingNb. LTE Optimization and Principle and Method Possibility to page this terminal UE receiver circuitry UE receiver circuitry switched off switched off Subframe DRX cycle Paging DRX Cycle Paging DRX Cycle Paging DRX Cycle Paging DRX Cycle PF PF PF PF PF PF #0 #4 #5 #9 PO PO PO PO PDCCH PDSCH T_Frequency P-RNTI T_Format Paging message Figure 2.60 Paging frame and occasion. The positions that paging messages are transmitted on the Uu interface are fixed, which are indicated by the paging frames (PFs) and paging occasion (PO) subframes. One PF is one radio frame, which may contains one or multiple POs. All attached UEs are distributed on all paging frames within one paging DRX cycle (based on IMSI). One PO is a subframe where the P‐RNTI is contained. The PO is transmitted over the PDCCH, the P‐RNTI value is fixed, that is, FFFE. UEs read paging messages over PDSCH according to the P‐RNTI. To receive paging messages from E‐UTRAN, UEs in idle mode monitor the PDCCH channel for P‐RNTI value used to indicate paging. If the UE detects a group identity used for paging (the P‐RNTI) when it wakes up, it will process the corresponding downlink paging message transmitted on the PCH (Figure 2.60). LTE frames numbering has two components, one at frame level, that is, system frame number (SFN) and second one at subframe level, that is, subframe number. So the UE has to know both SFN and subframe number to locate exact position of its page. The SFN of a paging frame (PF) is derived from the following formula: PF SFN modT T div N x UE _ IDmod N The subframe number i_s of a PO is derived from the following formula: i_s UE _ ID/N mod Ns T is DRX cycle of the UE. UE can get the T from two different sources, one from the system information (SIB2) and the one from upper layer (NAS). If upper layer send the value, the UE 89 90 LTE Optimization Engineering Handbook Table 2.29 Example of paging offset and sub‐frame, TDD (all UL/DL configurations). Ns PO when i_s=0 PO when i_s=1 PO when i_s=2 PO when i_s=3 1 9 N/A N/A N/A 2 4 9 N/A N/A 4 0 4 5 9 compares the value signaled from the NAS and the value of DefaultPagingCycle and uses the smaller value between them, otherwise UE has to use the value from SIB2. N = min(T, nB), which means the smaller one among T and nB. nB can be any one of 4T, 2T, T/2, T/4, T/8, T/16, T/32, which comes from SIB2 (IE nB). UE_ID is IMSI mod 1024, calculated by MME and UE. If the paging is triggered by the MME, the UE_ID value is the UE identity index value contained in the paging message on the S1 interface without having to signal the IMSI across the S1 interface. It is a 10 bit string value. Thus, UE are divided in “N” groups according to their UE identity, all UE with equal values for “UE_ID mod N” share the same paging frames. Network send paging with S‐TMSI, but if something (e.g., network failure) happens during registration and it fails to allocate TMSI to the UE, network would send paging with IMSI. If UE get the paging with IMSI, it should tear down all existing bearers and delete TAI, TAI List, KSIASMI and get into EMM‐DEREGISTERED status. And then redo “attach request.” Ns = max(1, nB/T), which means that Ns is the larger value between 1 and NB/T Table 2.29 gives an example of paging offset and sub‐frame under TDD (all UL/DL configurations) system. Example: Example: T = 64 nB = 2T = 128 UE_ID = IMSI mod(1024) N = min (T,nB) = 64 Ns = max (1,nB/T) = 2 UE_ID = IMSI mod 1024 = e.g 0 PF : SFN mod T= (T div N)*(UE_ID mod N) = 0, 64, 128… Although 3GPP allows up to four paging occasions per radio frame. Assume only supports up to 1 paging occasion per radio frame (pagingNb = oneT), that translates into 100 paging occasions per second. Since each paging message may contain up to 16 paging records the maximum air interface capacity is 1600 paging records per second per cell, which is more than what eNB can provide in the control plane, that is, it has been observed a CPU load of roughly 4% for 30 S1 pagings per second. As a consequence, 500 S1 paging messages per second would occupy 2/3 of CPU processing power for paging alone. Based on the above, the default pagingNb= quarterT (equivalent to 400 possible paging messages per second) is the maximum recommended value. UE may also be paged by the network when there is data addressed to that particular UE. UE returns to EMM_ACTIVE/RRC_connected mode as soon as packet arrival is detected. The delay depends on the paging DRX cycle, time to acquire UL synchronization, and time to set up the RRC connection with the eNB. MME Initiated paging fow is shown in Figure 2.61. LTE Optimization and Principle and Method UE eNB MME SGW PGW k data Downlin Page Page Response Time Page RACH Preamble RACH Response RRC Connection Request RRCConnection Setup S1-AP Initia l UE Messa ge Attach Requ est eNB UE S1 AP ID IE Total Delay RRCConnectionSetup Complete Attach Request tup req ntext se Initial co Security Mode Command Security Mode Complete RRCConn Reconfiguration Req RRC conn Reconfiguration Complete Initial Cont ext Setup UL DATA Response DL DATA Update Be arer Request arer Modify Be Response Figure 2.61 MME initiated paging flow Here are three paging examples as following. Example 1 > - Paging with s‐TMSI PCCH-Message ::= SEQUENCE +-message ::= CHOICE [c1] +-c1 ::= CHOICE [paging] +-paging ::= SEQUENCE [1000] +-pagingRecordList ::= SEQUENCE OF SIZE(1..maxPageRec[16]) [1] OPTIONAL:Exist | +-PagingRecord ::= SEQUENCE | +-ue-Identity ::= CHOICE [s-TMSI] | | +-s-TMSI ::= SEQUENCE | | +-mmec ::= BIT STRING SIZE(8) [00000001] | | +-m-TMSI ::= BIT STRING SIZE(32) [000000000000000000000 | 00000000001] | +-cn-Domain ::= ENUMERATED [ps] +-systemInfoModification ::= ENUMERATED OPTIONAL:Omit +-etws-Indication ::= ENUMERATED OPTIONAL:Omit +-nonCriticalExtension ::= SEQUENCE OPTIONAL:Omit Example 2 > - Paging with IMSI PCCH-Message ::= SEQUENCE +-message ::= CHOICE [c1] +-c1 ::= CHOICE [paging] +-paging ::= SEQUENCE [1000] 91 92 LTE Optimization Engineering Handbook +-pagingRecordList ::= SEQUENCE OF SIZE(1..maxPageRec[16]) [1] OPTIONAL:Exist | +-PagingRecord ::= SEQUENCE | +-ue-Identity ::= CHOICE [imsi] | | +-imsi ::= SEQUENCE OF SIZE(6..21) [15] | | +-IMSI-Digit ::= INTEGER (0..9) [0] | | +-IMSI-Digit ::= INTEGER (0..9) [0] ...... | | +-IMSI-Digit ::= INTEGER (0..9) [9] | +-cn-Domain ::= ENUMERATED [ps] +-systemInfoModification ::= ENUMERATED OPTIONAL:Omit +-etws-Indication ::= ENUMERATED OPTIONAL:Omit +-nonCriticalExtension ::= SEQUENCE OPTIONAL:Omit Note: IMSI is used as paging identity when reattach from UE is required. It indicates that error occurred. Example 3 > - Paging for system Info Modification RRC_LTE:PCCH-Message PCCH-Message ::= SEQUENCE +-message ::= CHOICE [c1] +-c1 ::= CHOICE [paging] +-paging ::= SEQUENCE [0100] +-pagingRecordList ::= SEQUENCE OF OPTIONAL:Omit +-systemInfoModification ::= ENUMERATED [true] OPTIONAL:Exist +-etws-Indication ::= ENUMERATED OPTIONAL:Omit +-nonCriticalExtension ::= SEQUENCE OPTIONAL:Omit 2.3.11.2 Paging Capacity The higher amount of paging frames the more PDCCH and PDSCH resources may be used for paging. More aggregated paging records in one RRC paging message that is sent in one PO causes less PDSCH occupation comparing to higher number of RRC paging messages with less paging records inside. When average paging load is low, one can reduce value of pagingNb, thus more UEs under one PO will be aggregated and in average more page records per each RRC paging message is expected. What is more, less POs will decrease PDCCH load that less P‐RNTIs will be used to sent the same amount of paging records. On the other hand more paging records per one PO will cause higher blocking probability, it should be balanced. The the number of EPS paging attempts, received/discarded S1AP Paging messages by eNB can be got from PM counter listed in Table 2.30. Paging DRX cycle does not change the maximum paging capacity because the number of resources that can be used for paging is not changed. Paging DRX cycle reduces blocking probability at the cost of call setup time. Table 2.30 The number of EPS paging attempts, received/discarded S1AP Paging messages. EPS paging attempts This counter counts the the number of paging attempts of initial and repeated. PageS1Received This counter counts the number of received S1AP Paging messages in the RBS. PageS1Discarded The number of S1AP Paging messages that are discarded and not routed to any cell LTE Optimization and Principle and Method defaultPagingCycle = rf32 Paging Occasion (per Paging Frame) Paging Frame nB = 4T => 0 16 32 nB = 2T 0 16 32 nB = T 4 9 0 4 9 0 4 9 0 4 9 0 4 9 0 4 9 0 4 9 0 4 9 => 0 16 32 nB = (½)T 0 16 32 0 16 32 nB = (¼)T nB = (1 8)T 0 => => => 0 16 32 => The value of nB determines the rate at which paging frames occur from the cell perspective, i.e. there is an impact upon paging capacity. => nB = (1 16)T 0 16 32 nB = (1 32)T => 0 16 32 Figure 2.62 Paging frame and paging occasion. PagingNB is used to calculate the number of POs within one paging DRX duration, which in turn is used to calculate the PO. PagingNb changes the maximum paging capacity because it increases number of resources that can be used for paging, it reduces blocking probability and increases maximum number of pagings. Paging frames are distributed on all radio frames according to pagingNb parameter value. PagingNb has impact on number of POs per considered time frame, higher value of pagingNb more POs per time period. Paging frame and paging occasion example are shown in Figure 2.62. If eNB KPI indicates more paging discards, it can be considered to increase the maximum number of paging records parameter or to allow more number of paging occasions, that is, increase nB (Table 2.31). Paging capacity has impact on call setup time, reachability of UE, and tracking area size. Excessive paging blocking value leads to delay or failure of the paging procedure. Here are four type of paging DRX cycle and pagingNb settings usually configured in a live network. It can be seen that higher value of pagingNb (½) will result to reduced number of UEs assigned to one PO (20UEs/PO or 10UEs/PO) and longer paging DRX cycle will result more POs within one cycle reducing number of UEs assigned to one PO (Figure 2.63). Assumption1: 320 UEs/cell, pagingNb = ½, 32/64 paging DRX cycle Assumption2: 320 UEs/cell, pagingNb = ¼, 32/64 paging DRX cycle Paging blocking probability is another factot that impacts on eNB paging capacity as it limits maximum number of pagings that can be handled by one eNB. Pagings arrival is poisson distribution Table 2.31 Paged UEs per second per cell. nB T/32 T/16 paging occasions per radio frame 1/32 1/16 Ns T/8 T/4 1/8 T/2 1/4 T 1/2 2T 1 4T 2 4 1 1 1 1 1 1 2 4 Number of paging record per radio frame 16 16 16 16 16 16 32 64 Max No. of paged UEs per second per cell 50 100 200 400 800 1600 3200 6400 93 94 LTE Optimization Engineering Handbook Paging frame (10 ms) 50 PO/s = max 800 pagings/s 32rf Paging DRX Cycle (16POs) 32rf Paging DRX Cycle 1 second 32rf Paging DRX Cycle 32rf Paging DRX Cycle 320/16 = 20UEs/PO 50 PO/s = max 800 pagings/s 320/32 = 10UEs/PO 1 second 64rf Paging DRX Cycle (32POs) 64rf Paging DRX Cycle 25 PO/s = max 400 pagings/s 32rf Paging DRX Cycle (16POs) 32rf Paging DRX Cycle 1 second 32rf Paging DRX Cycle 32rf Paging DRX Cycle 320/8 = 40UEs/PO 25 PO/s = max 400 pagings/s 320/16 = 20UEs/PO 1 second 64rf Paging DRX Cycle (32POs) 64rf Paging DRX Cycle Figure 2.63 Four type of paging DRX cycle and pagingNb. Table 2.32 PDSCH resources occupation for paging. N_PRB I_TBS 1 2 3 0 16 32 1 24 2 32 3 4 4 5 6 7 8 9 10 56 88 120 152 176 208 224 256 56 88 144 176 208 224 256 328 344 72 144 176 208 256 296 328 376 424 40 104 176 208 256 328 392 440 504 568 56 120 208 256 328 408 488 552 632 696 in a live network and blocking probability determines average number of records per one RRC paging message. For the calculation of consumption of PDSCH resources, assume 16 pagings records in one PO, the length of S‐TMSI is 32bits, the length of MME record is 8bits. For each PO, the maximum transmit payload are 16*40+1+1=642 bits, it will occupy 10 PRBs PDSCH resources shown in Table 2.32. It can be seen that when PDSCH resources is 10 PRBs and MCS level is 4, the TB size is 696 bits accordingly. Actually, in a live network, it is normally not to use higher MCS level for the paging channel. Now let’s take an example of how to calculate eNB’s paging capacity. It is simply assumed the traffic model (three cells/eNB) of a network is listed below (variety of applications and smartphone types cause that the real network behavior might be significantly different): ●● ●● ●● ●● ●● ●● ●● Users (RRC connected + RRC idle) per eNB – 450 MTC/MOC for VoIP – 50%/50% MTC/MOC for PS/background – 30%/70% VoIP holding time – 90 sec; PS data mean holding time – 312 sec. PS data sessions per user per busy hour – 1; VoIP sessions per user per busy hour – 1. Background traffic per every user InactivityTimer: 10 sec. So, in this traffic model, eNB’s paging capacity is 4.3 pagings/s/eNB. LTE Optimization and Principle and Method 2.3.11.3 Paging Message Size The physical layer will add a 24 bit CRC to the transport block and then complete channel coding. The default channel coding value of 0.12 means that high quantities of redundancy are added before transmitting the paging message across the air‐interface. The value of 0.12 and QPSK are always used on the PDSCH when transferring a paging message that helps to make paging more reliable. Assuming a resource block includes 132 resource elements per resource block pair within a subframe. Table 2.33 illustrates the requirement for a large number of resource blocks when the paging load is very high. 2.3.11.4 Smart Paging LTE allows the operator to configure a first page is to be distributed, to the area of a single eNB, TA or the whole TA list. The selected area in which the page is sent can be decided based on different criteria: ●● ●● ●● UE‐related criteria, which is used to send a page to a single eNB or a TA for known stationary UEs such as electricity meters UE last reported location criteria, which is used to send a page to a single eNB or a TA when a the location of the UE is unknown Service‐related criteria by APN, QCI, and ARP. These criteria are used for time critical services to guarantee that the page is sent directly to all eNBs in TA list In general, paging strategy is step‐by‐step paging in a live network. When the MME wants to page the UE, first paging the last visit of the eNB, if the UE can not be paged, then last eNB and its neighboring eNBs will be paged, if the UE still can not be paged, then eNBs in last TA will Table 2.33 PDSCH paging load. Number of paging records 1 Paging message size (bits) 56 Transport block size (bits) 56 # Bits after channel coding 667 # RE 334 # RBs 3 2 104 120 1200 600 5 3 144 144 1400 700 6 4 192 208 1934 967 8 5 232 256 2334 1167 9 6 280 280 2534 1267 10 7 320 328 2934 1467 12 8 384 392 3467 1734 14 9 408 488 4267 2134 17 10 456 488 4267 2134 17 11 496 552 4800 2400 19 12 544 552 4800 2400 19 13 584 600 5200 2600 20 14 632 632 5467 2734 21 15 672 696 6000 3000 23 16 720 776 6667 3334 26 95 96 LTE Optimization Engineering Handbook Page strategy IMEI TAC User level IMSI MSISDN The basic paging stragegy is: Trigger 1 Trigger 2 Service level APN ARP Voice QCI SMS last eNB -> last eNB and its neighboring eNBs -> eNBs in last TA -> eNBs in last TA List (3GPP standard) Figure 2.64 Paging strategy. be paged, and finally eNBs in last TA List will be paged. Paging the last visit of the eNB, UE paging success ratio can reach to 70% to 90%, and reduce paging signaling in large quantities. At the same time, different levels or different characteristics of users can be set to different paging attributes. With different attributes of paging, it can be greatly reduced the paging signaling load (amount of paging S1 paging messages decrease 50%). Intelligent intelligent paging strategy can be set different attributes, according to different levels or different characteristics of users based on APN, QoS, IMSI, IMEI, and so on. The paging strategy can be further refined and optimized, for example, accurate paging scheme can only be applied for slow moving UEs, for high mobility users, paging over tracking area, and gradually expand the range of paging shall be applied, thus will greatly reduced the paging signaling load (Figure 2.64). 2.3.11.5 Priority Paging In a system with mixed paging priorities, higher priority paging can preempt existing lower priority paging in the queue, the paging success rate will be higher for higher priority of the paging message compared to lower or no priority. Assuming there are non‐prioritized paging messages and prioritized paging messages per second, the prioritized paging messages shall be equally distributed between priority 1 and 8. All paging messages shall be sent equally distributed in time based on priority. S1AP paging message listed in Table 2.34, which shows a paging payload. 3GPP allows up to 16 S1 pages per modem‐to‐UE page message. In case when more than 16 pagings are considered Table 2.34 S1AP paging message IE and semantics description. IE/Group name Source Semantics description Comments UE identity index value MME IMSI mod 1024 – UE paging identity MME S‐TMSI or IMSI* – Paging DRX MME Paging DRX cycle CN domain MME CN Domain – PS or CS When more than 16 pagings are sent in one PO, they are put in RRC paging message according to descending paging priority value List of TAIs At least one TAI shall be present; up to 16 > TAI list item >> TAI MME >>> PLMN identity MME >>> TAC MME Paging priority MME Used in control plane overload cases LTE Optimization and Principle and Method Token rate–r, short interval (10 ms) or continues excess flow P1 excess flow Tokens? P2 Accepted pagings P3 Tokens? Excess paging Accepted pagings Tokens? prioLevel1 Excess paging prioLevel2 Paging Accepted pagings prioLevel3 Excess paging Figure 2.65 Priority paging. to be sent in one PO, they are put in RRC paging message according to descending paging priority value. Figure 2.65 gives a description of Token based priority paging. 2.3.12 MIMO and Beamforming From 3GPP roadmap of multi‐antenna point of view, the evolution of LTE releases in 3GPP brings significant new capability in the domain of multi‐antenna operation. Advanced multi‐ antenna solutions are key components to achieve the LTE network requirements for high peak data rates, extended coverage, and high capacity. Nowadays, market is driven by advanced antennas and more complex passive antennas. Beamforming and spatial multiplexing illustrated in Table 2.35 have been widely deployed in the live network, while diversity and beamforming aim to improve received signal power of a single information stream, spatial multiplexing aims to share the signal power between multiple parallel streams. Smart antenna is a multiple antenna elements system, which combined with signal processing to dynamically select or form the “optimum” beam pattern for each user. Smart antennas usually categorized as switched beam and adaptive array, and there are four types of smart antennas as shown in Figure 2.66: uniform linear array (a), circular array (b), two‐dimensional grid array (c) and three‐dimensional grid array (d). Now in the LTE live network, usually classical antenna and active antenna are deployed. A classical antenna consists of subelements, two antenna ports per column or smart antenna that has at least eight antenna port. Smart antenna can shape a beam by weighting of subelements so called beamforming (BF). The beam shape could be fixed vertically or adaptable horizontally Table 2.35 Beamforming and spatial multiplexing. 2D Beamforming Digital beamforming with integrated radios and antennas 3D Beamforming 3D beamforming is integrated with radios and antennas, support user specific beamforming. Require phased‐array technology, including analoque, digital and hybrid beamforming technologies MIMO and massive MIMO More antenna elements required for capacity increase. However higher frequency => smaller antennas 97 LTE Optimization Engineering Handbook ∆x y (a) x ∆Φ ∆z y x ∆z ∆y y ∆x (b) x (c) (d) ∆x y z x Figure 2.66 Four types of smart antennas based on element arrangement. one column Subelement weights Figure 2.67 Classical antenna. ments w1 subele antenna port PA w4 PA w5 w8 w1 PA PA antenna ports 98 PA w4 PA w5 PA PA PA w8 PA Figure 2.68 Flexible active antennas. to any direction and shape by changing amplitude and phase. An active antenna, composed of a multiple set of low power active transceivers modules, which are connected to eNB through CPRI. PAs are tightly integrated inside antenna, one PA per subelement. Active antenna offers beamforming more flexibility to tune tilt angles without mechanical actions, more flexibility to change the antenna vertical beam (Figure 2.67 and Figure 2.68). LTE Optimization and Principle and Method For beamforming, two precoding methods were evaluated using a full buffer traffic model in the 3GPP 3D UMi scenario: one is SVD precoding using the uplink long term spatial channel correlation matrix, the other is codebook based precoding using grid‐of‐beams, where the precoder matrix is chosen based on the uplink long term spatial channel correlation matrix. For spatial multiplexing, the eNB is able to transmit and receive multiple parallel information streams within the same spectrum which requires multiple antennas at both ends of the radio link and the maximum number of parallel streams is equal to the minimum of the number of transmit and receive antennas, therefore increasing the spectral efficiency. Most MIMO implementation nowadays is X‐pol (one is 45 degree, the other is ‐45 degree) MIMO and not spatial MIMO. The evolution of DL transmission modes on top of Rel 8 was the enhanced transmit diversity, beamforming and spatial multiplexing, Rel 9 adds dual layer beamforming. Rel 10 extends the dual layer mode of TM8 to TM9 with up to eight layers. Rel 11 adds TM10 with up to eight layers and provide support for DL beamforming on dedicated control channel and optimized DL CoMP operation, and in Rel‐10 UL SU‐MIMO with up to four‐layer, UL reference signal enhancements for improved UL MU‐MIMO (Rel 10) and UL CoMP (Rel 11) will be supported. 3GPP transmission modes (downlink) are present in Table 2.36. It is worth to mention that 3D beamforming is planning introduced in 3GPP Rel 13, it is not only a new beamforming feature but a whole new product and site solution, instead of 8 pipes with fairly high power per pipe, the idea is to use 64 pipes with low power per pipe. Either beamforming or spatial multiplexing, the goal of optimizing a MIMO system is to achieve the highest throughput and connectivity possible in a given environment by leveraging the multipath potential of the environment. Table 2.36 3GPP transmission modes (downlink). 3GPP Rel‐8 Transmission scheme tx mode of PDSCH Antenna Port Feedback Beam‐ Spatial Spatial forming multiplexing diversity 1 Single‐antenna port CRS CQI No No No 2 Transmit diversity CRS CQI No No Yes 3 Open‐loop spatial multiplexing CRS CQI, RI No 1‐4 layer Yes 4 Closed‐loop spatial multiplexing CRS CQI, PMI, RI Yes 1‐4 layer No 5 Multi‐user MIMO CRS CQI, PMI, RI Yes Yes No 6 Closed‐loop Rank=1 precoding CRS CQI, PMI, RI Yes No No 7 Beamforming single‐ antenna port; port 5 DM RS CQI No No Rel‐9 8 Dual layer beamforming DM RS CQI, PMI, RI Yes 1‐2 layer No Rel‐10 9 Closed‐loop spatial multiplexing DM RS CQI, PMI, RI Yes 1‐8 layer No Rel‐11 10 TM9 with DL CoMP and E‐PDCCH DM RS CQI, PMI, RI Yes 1‐8 layer No Yes 99 100 LTE Optimization Engineering Handbook 2.3.12.1 Basic Multi‐Antenna Techniques For LTE 10 transmission modes, only four MIMO technology programs are focused on in this part: beamforming, transmit diversity, SU‐MIMO, and MU‐MIMO. With at least eight antennas at the transmitter and only single antenna at the receiver (referred to as MISO) it is possible to obtain beamforming (BF). With this method the transmission signal is steered in a beneficial direction toward the UE, as the UE moves throughout the cell, the switched beam system detects the signal strength and continually switches the beam as necessary. This is accomplished by adjusting the phase and amplitude of the different antenna elements by multiplying the signal with complex weights. Each element of the array will be fed with the same signal and this enables “smart” antennas to modify their radiation pattern dynamically to alter the direction and shape even a beamforming antenna enables the entire eNB signal to be directed to the single user. It would be possible to dip the beam by causing element B to lag in phase behind element A, thus the line describing points where the radiation arrives in phase is no longer horizontal but instead dips toward point Z as shown in Figure 2.69. BF support to tune shapes horizontally and vertically according to typical UE positions and traffic needs, and even horizontal/vertical sectorization. Transmit diversity aims to increase the robustness of data transmission. When the same data is transmitted redundantly over more than one transmit antenna, this is called TX diversity. This increases the SNR. Space‐time codes are used to generate a redundant signal. Transmit diversity mode includes two transmitting antenna’s SFBC (space‐frequency block code) and four transmitting antenna’s SFBC+FSTD. In LTE, transmit diversity is used as a fallback option for some transmission modes, such as when spatial multiplexing cannot be used. Control channels, such as PBCH and PDCCH, are also transmitted using transmit diversity. MIMO includes single‐user mode SU‐MIMO and multi‐user mode MU‐MIMO. For SU‐ MIMO, data is divided into separate streams for one UE, which are then transmitted simultaneously over the same air interface resources under spatially uncorrelated channels (Figure 2.70). MU‐MIMO, the program will be the same time‐frequency resources through space division with sharp beams, assigned to different users under spatially correlated channels (Figure 2.71). Beamforming and MU‐MIMO prefer correlated antennas, precoding weights are chosen to maximize signal toward a given UE and are related to the instantaneous channel gains and phases of the signal paths through the MIMO channel. With correlated antenna ports as shown in Figure 2.71, the signal paths through the MIMO channel will all have the same instantaneous Y Phase shif A Feed Phase shifters θ X B Z Figure 2.69 Cell shaping and sectorization. e P S- y la De Figure 2.70 SU‐MIMO (left) and transmit diversity (right). LTE Optimization and Principle and Method Figure 2.71 Correlated (left) and decorrelated antennas (right). v v W W ϕ t t Ideal weights have same gain on all ports only phase changes Ideal weights differ in gain and phase across ports fading, so the ideal weights differ only in their phase slope. With decorrelated ports, the fade depth varies across the multiple signal paths through the MIMO channel. Optimal weights would ideally bias power away from the deeply faded paths and are therefore complex, with different gain and phase. It is worth to say beamforming is implicit in multi‐layer SU‐MIMO and MU‐MIMO in LTE, besides improving a UE’s signal strength or quality. 2.3.12.2 2D‐Beamforming Beamforming (BF) algorithm can make wave to any direction and shape by changing amplitude and phase. BF principle is based on channel reciprocity, for TDD system, uplink and downlink has similar channel response and similar covariance matrices.6 HDL HUL R DL R UL R HH H Normally, HUL is estimated by eNB based on sounding reference signal (SRS transmits eriodically), which is used to determine the precoding, for example, grid of beams or p Eigenvalue‐based beamforming, evaluating the required periodicity and calculate DL weights. There are many possible ways of choosing the beamforming vector. Two different algorithms for estimating the precoding are usually used, the simpler grid of beams (GoB) algorithm, and the more complex eigenvalue based beamforming (EBB). The GoB‐based method calculates the DL beamforming weights based on the angle of arrival information. The EBB‐based algorithm calculates the DL beamforming weights based on the SVD decomposition of the spatial channel matrix. GoB based method is effective/efficient in small angle spread (AS) while considered less effective in large AS cases. EBB based methods can work in large AS cases. For FDD system, although uplink and downlink channels on separate frequencies experience independent fading, they should have similar spatial characteristics. Using UL spatial correlation matrix to derive the DL precoder matrix would result in a 5% to 10% loss in average user throughput and a 4% to 50% loss in cell‐edge user throughput compared to using closed‐loop feedback to derive the precoder matrix. In a TDD system with Na receiver and transmitter antennas, let the transmitted signal from a user with a single transmit antenna at subcarrier k of Nsc subcarriers be denoted X(k). Then, the received signal can be written as Y (k ) H (k ) X (k ) D(k ), where Y (k ) [Y1 (k ),Y2 (k ),,YN a (k )]T is the received signal at the eNB, H (k ) [ H1 (k ), H 2 (k ),, H N a (k )]T is the uplink spatial channel vector, and D(k ) [ D1 (k ), D2 (k ),, DN a (k )]T is the interference and noise at the eNB receiver antennas. In the receiver, an estimate H est (k ) of the channel vector H (k ) is calculated from the user’s reference signals. The precoding vector w(k ) [w1 (k ), w2 (k ),, w N a (k )]T is then 6 If the delay is within the channel coherence time, it will approximately hold. 101 102 LTE Optimization Engineering Handbook CW1 u4 u3 u2 u1 Short term weight w1 Long term weight w1 1 5 w5 w2 w2 w6 2 6 CW1 w12 λ/2 w21 3 7 w4 4 8 w3 w7 Short term weight w11 w22 CW2 w8 u4 u3 u2 u1 Long term weight w1 1 5 w5 2 6 3 7 w4 4 8 w2 w6 w3 w7 λ/2 w8 Figure 2.72 Single‐stream (left) and dual‐stream (right) beamforming. c hosen using this estimated channel vector H est k . It can be seen that the gain from the beamforming is related to the factor g H (k )* w(k ) . While 3GPP Rel 8 of the LTE specification defines transmission mode (TM) 7 for beamforming with one layer, Rel 9 defines transmission mode (TM) 8, to support dual‐stream beamforming, in which two different and independently coded data streams are separately transmitted from two logical antenna ports, combines beamforming with 2x2 MIMO spatial multiplexing capabilities. This feature implemented for TDD only, can result in improvements in coverage and capacity. Either of the two data streams is generated by four or eight antennas in beamforming transmission mode. The two data streams both form directional beams toward the target UE, which increases SINR. Dual‐stream beamforming incorporates both spatial multiplexing and beamforming during downlink transmission. This helps provide spatial multiplexing gains, diversity gains, and array gains (Figure 2.72). TM8 is a mode where data is transmitted over two spatial layers as two independent beamformed streams. It has the similar covariance matrices as one layer beamforming: RDL ≈ RUL(R = HHH) H Hermitian operator matrix transposition conjugate complex In general, TM8 is expected to have higher throughput at cell edge, slightly better or similar throughput as TM3 in medium points, lower throughput in good points. Assuming the channel vectors at PRB i estimated from SRS from two polarization antenna elements groups are described by hi,1 and hi,2 respectively, then the instantaneous spatial channel covariance at current sub‐frame n is computed by averaging over two polarization antenna elements groups over all the used PRBs. R inst n i h i ,1 h iH,1 h i ,2 h iH,2 The long‐term BF covariance matrices (weight) averaging over time by a recursive filter of first order, if long term BF covariance matrices is used, it can get better performance in channel models with low AS (dominant eigenvector). Rave n Rave n 1 1 Rinst n where α is forgetting factor, which is inverse proportional to the settling time of filter. As stated above, the performance of beamforming strongly depends on the instantaneous channel information. If the channel condition of an UE varies frequently, then measurement results or reported results on the channel condition do not exactly reflect the current channel LTE Optimization and Principle and Method Table 2.37 TM8 Port7/8 beamforming gain imbalance. No RSRP SINR CQI0 CQI1 MCS CW0 MCS CW1 1 –56.95 23.01 13.78 14.66 25.17 25.19 2 3 4 5 6 7 8 –81.07 –75.57 –99.23 –94.81 –103.51 –91.28 –95.44 14.46 14.70 9.01 10.88 10.00 5.44 0.23 8.82 11.75 6.62 7.23 5.52 4.81 3.65 5.64 6.25 7.62 9.73 8.75 6.74 4.50 19.12 20.69 12.78 14.04 11.58 6.99 4.61 10.21 11.65 16.63 19.30 17.72 9.90 7.37 TM8 Port7/8 BF gain imbalance condition in time, thus the dual‐stream beamforming’s performance will decrease. It is also worth to mention that often TM8 Port7/8 beamforming gain imbalance had been observed in the field test as shown in Table 2.37. The key in beamforming is to generate the weighting vectors. There are two classes of DL beamforming schemes, cell‐specific DL broadcasting beamforming for cell‐specific channel and UE‐adaptive DL beamforming for PDSCH. There are different algorithms for calculating the optimum UE‐adaptive DL beamforming weightings. For example, it is possible to determine from the direction of the received uplink signal (DoA or AoA) if the angular spread is small, or from the uplink sounding reference signals channel estimation to calculate the beamforming weightings. Beamforming weight calculation by channel estimation is described below (Figure 2.73): ●● ●● ●● Frequency (PRB) ●● step 1: collect snapshots of instantaneous H (data matrix) step 2: determine covariance matrix: R = HHH, the format is M x M = 8 x 8, which depends on UE, subcarrier and TTI step 3: averaging of R → Rav, over frequency (over 4 most recently sounded PRBs) and over time (over past SRS receptions (IIR or FIR approach)) step 4: eigenvalue decomposition (EVD): Rav = V Λ VH, Λ is diagonal matrix with eigenvalues, decreasing absolute value => 1. eigenvalue is dominant, V and VH are orthogonal eigenmatrices (M x M) SRS hopping SRS periodicity D S U D D t 1 HH k Hk Rav,f = 4 k ∈ {f with SRS} Figure 2.73 Averaging of R. D t+5 S U D D D S U t + 10 D D time (TTI) (1−α)Rav,t(t−5) + αRav,f(t) IIR Rav,t(t) = , 1 t Rav,f(k) Lk = t–5(L+1) FIR 103 104 LTE Optimization Engineering Handbook ●● step 5: determine dominant eigenvector e1, corresponds to max eigenvalue => e1 = v1 = first column of VH, e1 is the optimum BF vector w (maximizes “ergodic capacity” C, “averaging” over all possible channels H, cuboid!) C = maxE{log 2 det(I w H R w} w/ w 1 ●● ●● step 6: use dominant eigenvector for DL BF: w = e1 implementation: step 4 and 5 merged by very fast “power method” => e1 directly 2.3.12.3 2D MIMO and Parameters This part is foucused on spatial multiplexing of 2D MIMO. Spatial multiplexing allows a radio link composed of M transmit and N receive antennas (MxN) to exchange up to N independent data streams (codewords). Number of codewords is decided based on transmission mode, RI sent from UE, and so on. For UEs with low SINR beamforming should be used and for high SINR UEs spatial multiplexing should be used (Figure 2.74). For a spatial multiplexing system the corresponding generalization of the classic Shannon formula reads as: C log 2 1 SNR1 log 2 1 SNR2 log 2 1 SNRk where SNRk = Sk/N now denotes the SNR of the kth information stream, k = min(n,m). Each MIMO mode can be switched to transmit diversity mode (TM2), and TM3 through TM9. Before pushing to MIMO parameters, a few basic concepts should be reviewed. Codeword (CW) is a transport block that has been processed by the physical layer in terms of CRC addition, channel coding, and rate matching. When two codewords are transferred, they do not need to be of equal size. CQI reporting, link adaptation, and HARQ run independently for each codeword. LTE supports simultaneous transmission resources in the same block by two relatively independent codeword, which is by spatial multiplexing (SM) technology to achieve. Each set of data sent through the antennas in a spatial multiplexing operation is called a layer. Layer mapping is needed for MIMO that maps the modulated symbols belonging to either one or two codewords onto a number of “layers” where the number of layers are less than or equal to the number of antenna ports, and a channel matrix rank is corresponding. PMI, the signal is “pre‐coded” (i.e., multiplied with a precoding matrix) at eNB side before transmission, optimum precoding matrix is selected from predefined “codebook” known at eNB and UE side. Rank, equivalent to the total number of layer. Rank indicator (RI), the number of layers that can be supported under the current channel conditions and modulation scheme. RI indicates M Tx precoding Modulation + coding Select # code words Modulation + coding RI Layer mapping CQI Figure 2.74 Spatial multiplexing procedure. V N Rx H Demod UH Demod PMI H = UΣVH LTE Optimization and Principle and Method Table 2.38 Legacy 2D MIMO parameters. Parameter Range Default CQI threshold for fallback to closed Loop MIMO 1 CW Mode 0…16, step 0.1 11 CQI threshold for activation of closed loop MIMO 2 CW Mode 0…16, step 0.1 13 Rank threshold for fallback to closed loop MIMO 1 CW Mode 1…2, step 0.05 1.4 Rank threshold for activation of closed loop MIMO 2 CW Mode 1…2, step 0.05 1.6 CQI threshold for fallback to MIMO diversity 0…16, step 0.1 9 CQI threshold for Activation of open loop MIMO SM 0…16, step 0.1 11 Rank threshold for fallback to MIMO diversity 1…2, step 0.05 1.4 Rank threshold for Activation of open loop MIMO SM 1…2, step 0.05 1.6 the number of freedom degrees measured by the UE, which represents the maximum capacity of the Tx/Rx channel in terms of independent streams. Antenna port, it is not equal to the number of antennas, but rather a different channel estimation reference signal pattern. For ports 0 to 3, corresponding to RS transmission pattern of the multi‐antenna; for Port 4, corresponding to the PMCH, MBSFN case of RS; for port 5, corresponding to the UE Special RS. MIMO thoughput gains depend on above factors. In reality, maximizing rich scattering conditions within a cell, configuring the eNB to properly match MIMO parameters settings to real‐world conditions, and ensuring that UEs can take full advantage of the multipath conditions. Selection of the correct SU‐MIMO mode depends on factors such as mobility, CQI, and channel correlation (Rank). MIMO optimization process requires accurate measurement of these multipath conditions in order to achieve the best performance for a given environment while avoiding the time and expense of guesswork. Finally, the legacy 2D MIMO parameters are listed in Table 2.38. In a live network, SU‐MIMO optimization is the primary focus of operators attempting to maximize throughput gains. eNB make MIMO decision mainly based on UE report RI that decided by RS SINR and radio environment (the lower channel correlation the better). For low SINR, the two codewords is not easy to distinguish even with the lower channel correlation, if SINR > 12dB, the two codewords is easier to use MIMO, sometimes if the channel correlation is high, SM mode doesn’t increase throughput even with high SINR. From the test data, it is found rank1 CQI reflects the DL SINR fairly well, rank2 CQI have a heavy dependency over channel correlations. Higher correlated channel yield much smaller chances of rank2 reports, and the values are much lower. From Figure 2.75, it can be seen that the proportion of scheduled dual streams and the number of RI=2 are approximate linear relationship, i.e. the more probability of the reported RI=2, more dual streams will be scheduled. In addition, CQI and the dual streams scheduled are also approximately proportional relationship in a live network. 2.3.12.4 Massive‐MIMO Many operators are quite interested in high‐rise building coverage, but legacy 2D beamforming only allowed controlling the beam pointing in azimuth direction. Massive MIMO techniques is introduced that exploit both the azimuth and elevation dimensions that are characterized by focusing the transmit power radiated to a user in the cellular system based on digital beamforming methods such that the peak of a resulting beam can be dynamically controlled in azimuth as well as in elevation direction, so massive MIMO can significantly improve user data 105 LTE Optimization Engineering Handbook 30 SM Throughput (Mbps) TxDiv Throughput (Mbps) 25 For low SNRs, TX Diversity gives slightly better throughput than SM, since the power from both eNB TX antenna is aggregated to decode one code word. This is in line with theoretical simulations 20 15 10 Although the two cases are static TX diversity and SM, switching point could be around this area for dynamic adaptation of MIMO 5 0 –4 –2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 SINR (dB) 2 Average Reported Rank 106 1.9 No correlation 1.8 Med correlation 1.7 High correlation 1.6 1.5 1.4 1.3 1.2 1.1 1 –10 0 10 20 30 DL SINR (dB) RI = 2 reported ratio RI = 2 scheduled ratio 100.00% 80.00% 60.00% 40.00% 20.00% 0.00% 100.00% RI = 2 scheduled ratio 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% 5 7 9 Figure 2.75 MIMO ratio versus RI and CQI. 11 13 CQI LTE Optimization and Principle and Method rate in a high‐rise scenario. Massive MIMO deployed high‐gain adaptive beamforming and high‐order spatial‐multiplexing with a large‐scale array. The antenna type to be used is typically a planar antenna array consisting of multiple cross‐polarized antenna elements arranged in a rectangular two‐dimensional grid. From 3GPP perspective, DMRS and/or CSI‐RS should support more than eight antenna ports, for example, 32 or 64 or even more antennas/ports, beamforming main lobe sharpens as number of antenna elements increases. Massive MIMO, many subelements are controllable from baseband (Figure 2.76). Massive MIMO needs acquiring channel state information at the eNB for downlink MU‐ MIMO, also needs efficient signaling for multiplexing large numbers of UEs. TDD is often viewed as an easier problem than FDD as it leverages DL/UL reciprocity; for FDD, a codebook feedback solution may have difficulty that large number of antenna elements require large codebook size, which will result extremely high UE codebook search complexity. The UE needs receive CSI‐RS and computes azimuth and elevation PMI. User‐specific beams now formed in elevation domain to provide optimal throughput at all levels of the building. 2.3.13 Power Control Power control and proper power configuration will reduce inter‐cell interference and power consumption. This leads to higher cell capacity and the control of the maximum data rate for UE at cell edge and limit the interference that cell‐edge users create to the neigboring cells. In addition, it helps to prolong the battery life of the UE. In LTE, fractional power control (FPC) is introduced to allow a more flexible trade‐off between spectral efficiency and cell‐edge rates. 2.3.13.1 PUSCH/PUCCH Power Control The basic power control of PUSCH aspects are open‐loop power control with slow aperiodic closed loop correction factor, it is based on fractional pathloss power control. With fractional pathloss power control, it will be easier to set power control parameters that will enable higher UL peak rate in the cell without sacrificing cell‐edge performance. It enables a trade‐off between maximized UL cell edge bitrates versus improving the overall UL cell capacity. The formula of PUSCH power control is: PPUSCH i min PMAX , 10.log10 M PUSCH i P0 _ PUSCH . PL TF TF i f i where PMAX is the maximum allowed power that depends on the UE power class. MPUSCH(i) is the bandwidth factor, expressed in number of resource blocks taken from the resource allocation valid for uplink subframe i. P0_PUSCH is a parameter obtained as a sum of a cell‐specific nominal component p0NominalPUSCH signaled from higher layers and a UE‐specific component p0UePUSCH. P0_PUSCH is the basic starting point of open loop power control. α is partial pathloss compensation factor, a cell‐specific parameter signaled from higher layers in order to support fractional power control. α=1 corresponds to classic UL power control, that is, full pathloss compensation. PL is the downlink path loss estimate calculated in the UE. ΔTF(TF(i)) denotes the power offset depending on PUSCH transport format TF(i). f(i) is PUSCH close loop power control adjustment derived from TPC command in subframe i‐4. Both accumulated and non‐accumulated power control rules are used. Compared with PUSCH’s fractional pathloss power control, PUCCH only uses a complete pathloss power control mechanism. The PUCCH power control procedure is used to guarantee the required error rate; it aims at achieving a target SIR the value of which guarantees the 107 Physical Antenna Connectors Virtual Antennas Antenna Ports DMRS or Data Physical Antenna Sub -elements Wprec Wvirt Wfeed CSI-RS Reference - or data signals Precoder Antenna Array Constrained Virtualization Unrestricted Virtualization (per RE) st 21 floor User specific beams Broadened elevation sector beam. 250 13th floor RX TX 63 m Site 8th floor RX 6 RX Average sum rate (Mbps/cell) Highrise 200 150 100 50 Matched filter Zero-forcing Interference free 23 m nd 2 80 m floor 0 0 100 200 300 Number of antennas Figure 2.76 Massive‐MIMO principle. 400 500 LTE Optimization and Principle and Method required error rate. Higher settings of this parameter will improve PUCCH reception, but will also drive higher UE TX power leading to interference to neighboring cells, and vice versa. PPUCCH(i) = min{ PCMAX, P0_nominal_PUCCH+Pathloss+h (nCQI,nHARQ)+∆F_PUCCH(F)+g(i) (UE specific parameter) Unlike PUSCH, the PUCCH power control is based on SINR instead of power spectral density (PSD) as there is no link adaptation on PUCCH; fixed MCS is used over PUCCH. The UE‐received PUCCH SINR will be compared with the target PUCCH SINR and TPC values will be generated based on the difference between the two values. The target PUCCH SINR default value will be estimated based on the minimum required by the received SINR in order to achieve a certain erasure or BER performance. eNB sets a semi‐static nominal power P0 for all UEs in the cells first (P0_PUSCH for PUSCH and P0_PUCCH for PUCCH) and broadcast it to all UEs by SIB2 (UplinkPowerControlCommon: p0NominalPUSCH, p0NominalPUCCH). P0_PUSCH values range is (−126 to 24) dBm. Each UE has a UE‐specific nominal offset power (P0_UE_PUSCH for PUSCH and P0_UE_PUCCH for PUCCH), which is sent to UE by dedicated RRC signal (UplinkPowerControlDedicated: P0_UE_PUSCH, P0_UE_PUCCH). It is worth to note that P0_PUSCH is different for semi‐persistent grant and dynamic scheduled grant (SPS‐ConfigUL: p0NominalPUSCH‐Persistent). Table 2.39 gives the main parameters of power control. Figure 2.77 shows two set of p0NorminalPusch (−106dBm versus −96dBm) comparation. When p0NorminalPusch is set to −106dBm, uplink throughput reduces by approximately 2Mbps in good RF conditions as lower uplink received power target leads to lower UL_SINR and further leading to lower MCS assignment by link adaptation. With −106dBm, uplink received power target is lower by 10dB. Hence, this leads to a much lower uplink Tx_power. 2.3.13.2 PRACH Power Control Open‐loop power control is applied for initial transmission of RACH. The transmit power is determined taking into account the total UL interference level and the required SINR operating point, which can be determined at the UE as: PRACH _ msg1 min PCMAX , PL P0 _ PREAMBLE PREAMBLE N PREAMBLE 1 RAMP _ UP Table 2.39 The related parameters of power control. Parameter Description Range Default P0UEPucch Power offset for UE PUCCH TX power calculation −8…7 dB, step 1 dB 0 dB P0UEPusch Power offset for UE PUSCH TX power calculation −8…7 dB, step 1 dB 0 dB P0NomPucch Nominal power for UE PUCCH TX power calculation −127…‐96 dB, step 1 dB −96 dB P0NomPusch Nominal power for UE PUSCH TX power calculation −127…‐96 dB, step 1 dB −100 dB srsPwrOffset Power Offset For SRS Transmission Power Calculation 0…15, step 1 α α, Indicates the compensation factor for path loss. α0 (0), α0.4 (1), α0.5 (2), α 1 (7) α 0.6 (3), α 0.7 (4), α 0.8 (5), α 0.9 (6), α 1 (7) 7 109 LTE Optimization Engineering Handbook Throughput 25 –96 dBm Mbps 20 –106 dBm 15 10 5 0 Good Medium Poor UL Tx Power 25 20 dBm 110 –96 dBm 15 10 –106 dBm 5 0 Medium Good –5 Poor Figure 2.77 Two set of p0NorminalPusch (–106dBm versus –96dBm) comparation. P0‐PREAMBLE is the preamble received power set point determined at the eNB. This parameter is calculated from the target SINR operating point, and the UL interference‐plus‐noise (IN) power in the PRACH resource. P0 _ PREAMBLE SINRT arg et IN M arg in ΔPREAMBLE is the power offset value dependent on PRACH preamble format, which is given by prach‐ConfigIndex. The preamble format–based power offset values are presented in Table 2.40. According to the estimated received power of RACH preamble, the eNB is able to know the SNR condition of the UE initialized the random access. Thus it will assign a certain power to the UE so that it can send message 3 with reasonable power to enable it to receive the message 3 correctly so that in most cases UE will not have to restart a new random access due to the failure of message 3 transmission. Table 2.40 ΔPREAMBLE value. Preamble format DELTA_PREAMBLE value 0 0 dB 1 0 dB 2 −3 dB 3 −3 dB 4 8 dB LTE Optimization and Principle and Method Tx power to meet the target Rx power UE eNB RACH preamble x x x RACH preamble +∆Ramp RACH preamble +N∆Ramp RAR y y y RRC Connection Req +N∆Ramp + δmsg2 RRC Connection Req +N∆Ramp + δmsg2 PO_Pre PO_Pre PO_Pre (δmsg2) RRC Connection Req +N∆Ramp + δmsg2 Target Rx power PO_Pre + ∆msg3 PO_Pre + ∆msg3 PO_Pre + ∆msg3 RRC Connection Setup Closed loop power control with accumulation RRC Connection Setup complete (+NAS setup) NAS Setup + Authentication P0_nominal_PUSCH RRC Connection Reconfiguration Figure 2.78 Power control algorithm. MSG 3 Tx Power 10 * log 10 Mpusch Last Preamble Power deltaPremableMsg 3 PC _ msg 2 PC_msg2 is determined in the range of −6 to 8 dB (eight values) based on the preamble detection performance. So, the power control algorithm during the whole UE RACH procedure is shown in Figure 2.78. 2.3.14 Antenna Adjustment Besides antenna azimuths and tilting, antenna placement also has big impact on other cell interference in real environment. A poor site design can have a significant impact upon the performance of a potentially good site. Site design involves identifying an appropriate location for each antenna and the eNB cabinet. When there is a requirement to achieve a specific isolation from another radio system then that isolation is easier to achieve if the antennas are separated vertically rather than horizontally. The most important requirement is that antennas should be mounted such that their main beams are not obstructed. This should include both the horizontal and vertical half power beamwdiths, that is, the beamwidths at which the antenna gain has decreased by 3 dB. In the case of a roof‐top site, obstructions could be other antennas or cabins located on the same or a neighboring roof. In the case of mast or pole 111 112 LTE Optimization Engineering Handbook Main beam is not obstructed Poor Position Good Position Figure 2.79 Examples of poor and good roof‐top antenna locations. mounted antennas, obstructions could be trees or nearby buildings. Figure 2.79 illustrates examples of poor and good roof‐top antenna positions. 2.3.14.1 Antenna Position In the example of the poor antanna installation position, the antenna is located behind and slightly higher than some existing antennas. In this case the main beam of the antenna is obstructed and its performance is likely to be deteriorated. In the example of the good position, the antenna is located in front of and slightly lower than some existing antennas. In this case the main beam of the antenna is not obstructed, although care should be taken that the rear lobe of the antenna does not cause interference to the other radio systems. There is also a requirement to ensure that the edge of the roof‐top does not cause shadowing of the antenna. If an antenna is positioned on the edge of a roof‐top then it is unlikely to incur any shadowing from the roof‐ top itself. However, as the antenna position is moved away from the edge then the antenna is more likely to incur shadowing. Antennas that are located away from the edge should be mounted with an increased height. A general rule is that if you can walk in front of the antenna then it should be mounted 3 m above the roof‐top. Figure 2.80 illustrates the principle of shadowing from a roof‐top and suggests a range of heights that could be used to avoid shadowing. In most cases an antenna would be mounted less than 10 m from the edge of the building and its suggested height would be obtained by dividing the distance to the edge by two. Wherever possible, antenna mountings should allow the height and azimuth of the antenna to be adjusted. In the case of antennas mounted on walls then the azimuth should be configured to ensure that the horizontal beamwidth of the antenna is not compromised. In general, a 15‐degree safety margin should be added to each side of the half‐power horizontal beamwidth and then a check made to ensure that the composite beamwidth is free from obstruction. Figure 2.81 illustrates the principle of avoiding shadowing from the walls upon which antennas are mounted. General rule Clearance angle h d d < 10 m h > d/2 10 < d < 20 m h > d/3 d > 20 m h > d/4 Figure 2.80 Principle of avoiding shadowing from a roof‐top. LTE Optimization and Principle and Method Figure 2.81 Principle of avoiding shadowing from walls. Direction of main beam Poor Position 15° safety margin 15° safety margin Direction of main beam Half power beam width Good Position When there are other antennas on the same mast, the same roof‐top or the same wall, then the isolation from those antennas should be maximized without compromising the position of the LTE antennas. The isolation requirement will depend upon the systems to which the antennas belong. The isolation requirement can be translated into a physical separation using curves that plot the measured isolation as a function of physical separation. These curves depend upon the gain patterns of the antennas being used and whether or not the antennas are cross‐polar. As an example, the LTE system requires 40 dB of isolation from the UMTS system. If the antennas have a vertical separation then there should be at least 0.2 m between the base of one antenna and the top of the other antenna. If the antennas have a horizontal physical separation and a horizontal beamwidth of 65° then there should be at least 0.5 m between them. Whenever possible, a vertical separation should be combined with a horizontal separation to increase the achieved isolation. In cases where these separations cannot be achieved, then the isolation requirement should be solved in other ways. Alternatively, the isolation requirement can be achieved using a diplexor and allowing the two radio systems to share the same feeders. A diplexor typically offers 40 dB of isolation. Radio systems may also share the same antennas. In general, this has the drawback of restricting both radio systems to using the same antenna downtilts; that is, downtilts cannot be configured separately for each system. Antennas that have remotely controllable tilts are generally more expensive, but tilt changes can be made with relative ease. 2.3.14.2 Remote Electrical Tilt Antenna tilting is a very powerful method to control network capacity and performance optimization. With the tilt, it directs irradiation further down (or higher), concentrating the energy in the new desired direction. The remote electrical tilt (RET) function enables the operator to control and optimize the coverage area by modifying the inclination of installed antennas, without the need for climbing masts. The RET provides electrical tilt for tuning and optimizing the network by adjusting the vertical lobe‐angle (adjusting the phase‐shifter on the antenna) of the antenna. The RET unit can be mounted on any antenna with tilting capability, regardless of height, gain, or band. The RET unit communicates over an interface by the open specifications defined 113 114 LTE Optimization Engineering Handbook Triple Band Dual Band Single Band Manual Adjustment Knobs Electrical Tilt Indicators iRET Tilt indicator - reading a value of 4° (+/– 1°) RetSubunit RET AuPort RfBranch TMA RfBranch Rx TmaSubunit RfPort Tx/Rx & DC RBS Figure 2.82 Remote electrical tilt. by the antenna interface standards group (AISG) to ensure basic interoperability of antennas and to control infrastructure (Figure 2.82). Antenna tilt is defined as the angle of the main beam of the antenna below the horizontal plane. Positive and negative angles are also referred to as downtilt and up‐tilt respectively. Antenna downtilt can be adjusted mechanically or electrically. Electrical tilt is realized by wire LTE Optimization and Principle and Method + Antenna Axis Mechanical downtilt, the pattern in the front goes down, and behind goes up. Mechanical Tilt Mechanical Tilt Electrical Tilt Total Tilt Figure 2.83 Mechanical downtilt, electrical tilt, and total downtilt. feed phase shifting. This phase shifting is realized by modification of feed cable length in antenna factory. In electrical downtilt, phases of antenna elements are adjusted so that desired tilt angle is achieved by tilting main, side, and back lobes uniformly contrary to mechanical downtilt. The total tilt is the inclination of the maximum of the antenna’s main beam with respect to the horizontal plane. The performances of vertical sectorization depend on antenna tilt, vertical beam width, and front‐to‐back ratio, and so on (Figure 2.83). Downtilt calculation: Downtilt=arctan(h/D)+ (Beamwidth/2). The optimal amount of tilt is a trade‐off between coverage and interference reduction, it depends on the real‐time traffic situation with a varying degree of user clustering or hotspots. In a realistic network, the traffic characteristics are dynamically changing and the optimum tilt to the current traffic conditions, which referred to as automatic tilt control. It is possible to pre‐define tilt settings for different times of the day (such as rush hour, midday, evening, night) and different times of the week based on historic data (Figure 2.84). The antenna effects are combined as a sum of antenna gain, horizontal pattern, and elevation pattern. The sum of horizontal and vertical patterns is limited for a common front‐to‐back attenuation Am and SLAv. The antenna gain of horizontal and vertical radiation patterns are shown below: A min 12 A min 12 3 dB 3 dB , Am , Am 25dB , SLAv , SLAv 20dB A , min{ AH AV , Am Assume that the antenna vertical (3dB) beamwidth is 120 and antenna height is 30m. The relation of downtilt and Dmin, Dmax is shown in Table 2.41. The algorithm for tilt control is based on relative load between different cells covering the same area/cluster. First, find Max load and Min load in the cluster, then if the Max load minus Min load is greater than a predefined margin, find downtilt for sector with Max load and do up‐tilt for the sector with Min load. The load for cell number m is defined as: Lm = Im/N + Im, Where Im is the total interference experienced and N is thermal noise. 115 116 LTE Optimization Engineering Handbook 0° Downtilt Angle θ Side Lobe 3dB Beamwidth Main Lobe Rapidly Decreasing Signal Strength Region h Usable Signal Area Shadow Area Dmax Vertical Beamwidth Antenna Height Dmin (m) Dmax (m) Figure 2.84 Downtilt calculation. Table 2.41 The relation of downtilt and Dmin, Dmax. Downtilt (°) Dmin (m) Dmax (m) Downtilt (°) Dmin (m) Dmax (m) 0 285 infinite 10 105 429 1 244 infinite 11 98 343 2 213 infinite 12 92 285 3 189 infinite 13 87 244 4 170 infinite 14 82 213 5 154 infinite 15 78 189 6 141 infinite 16 74 170 7 130 1719 17 71 154 8 120 859 18 67 141 9 112 572 19 64 130 LTE Optimization and Principle and Method 2.3.14.3 Antenna Azimuths and Tilts Optimization The antenna system plays an important role in mobile communications. Antenna height, tilts, and azimuths (to certain extent) are strong primary RF shaping factors. The performance of the entire network is affected by improper type, location, or configured parameters of the antenna system. Figure 2.85 gives an example of how reduced antenna height improved application coverage. The antenna tilt is mainly affected by the coverage radius of the cell and the average SINR value in the coverage area, and the optimization of the tilt must take into account the balance between RF coverage and SINR. Figure 2.86 is the driving test results of the impact of the tilt on the SINR value. Usually in the tilt adjustment will encounter the following three typical scenarios: uplink coverage limited, overshooting, and coverage holes, and so on. Uplink coverage limited: When a cell is in a continuous coverage zone, due to the large tilt, worse propagation conditions or big penetration loss, and so on, the terminals may encounter rdaio link failure, although RSRP is still higher, it is not able to make the handover. This situation may be uplink coverage limited, which can be considered to reduce the tilt to improve coverage. Tilt 35% cells with >40 m, >15° 45% cells with <40 m 40 Low antenna (<40 m) gives 60% calls CQI >10; 80% user throughput >10 Mbps High antenna (>40 m) gives 35% calls CQI >10; 40% user throughput >10 Mbps 30 20 cells potentially serving high rise Indoor Traffic 10 0 0 20 40 60 80 Figure 2.85 Antenna height versus tilt. 20 18 16 Average SINR 14 12 10 8 6 4 2 0 0 2 4 6 8 10 12 Antenna Downtilt, degrees Figure 2.86 Average SINR versus tilt. 14 16 Height (m) 117 LTE Optimization Engineering Handbook Overshooting: When a cell is in a continuous coverage zone, due to the small tilt, good propagation conditions, and so on, overshooting will happen. When a cell has coverage beyond the intended coverage area, it leads to interference problems especially if the signal strength of the overshooting cell is high. This issue can be determined by observing the handover area. Usually the solution is to change the antenna configuration of the overshooting cell, for example, tilting down the antenna, redirecting the antenna orientation, or reducing the antenna height. Coverage holes: The serving RSRP within this area is below the minimum required signal level (qRxLevMin) to set up and maintain LTE service. There are two causes for coverage holes: lack of signal power and high interference. The coverage holes may cause handover failure or session drop. At this time it should adjust tilts of the cells to make up the hole based on the location of coverage hole and the surrounding RF conditions. 2.3.14.4 VSWR Troubleshooting VSWR stands for voltage standing wave ratio, it is a measure of how efficiently RF power is transmitted from a power source, through a transmission line, into a antenna. VSWR is the ratio of the peak amplitude to the minimum amplitude of a standing wave, calculated by the ratio of the highest voltage to the lowest voltage along the transmission line. This ratio is a function of the reflection coefficient, which in simple terms is just a measurement of power reflected from the antenna (Figure 2.87). Back to Transmitter Voltage Amplitude Reverse Power Vmax Transmission Line Vmin Forward Power To Antenna 2 1.5 Voltage (normalized) 118 1 0.5 0 –0.5 –1 –1.5 –2 0 5 10 Forward wave Composite wave Figure 2.87 VSWR concept. 15 20 25 Reflected wave Detected standing wave 30 LTE Optimization and Principle and Method VSWR V max / V min Return loss (RL) is the loss of power in the signal returned/reflected by a discontinuity in a transmission line. It is usually expressed as a ratio in decibels (dB). RL dB 10 log10 Pi Pr Where RL is the return loss in dB, Pi is the incident/forward power and Pr is the reflected/ reverse power. The relation between VSWR and RL is present as below: Example: by substituting VSWR = 1.5 and RL/20 = X, it can get: 1 10 x / 1 10 x 1.5 x 2.5 0.5 10 x 1.5 10 1 RL 20 * 0.6989 13.979 ~14 1.5 10 x 1 / 10 x 1 10 x 1 5 10 x X log10 5 0.6989, A high value enables detection of a low reflected power (high return loss). A low value requires a high reflected power (low return loss) to generate an alarm. The normal VSWR range is 1 to1.5. If current VSWR is more than a specified threshold, then eNB will generate relevant alarm. The VSWR antenna supervision enables supervision of the feeder cables. Measurements are made on the reflected radio power, making it possible to detect breaks or loose connections in the cables connected to the radio unit. An alarm is raised if return loss is below the configured VSWR sensitivity value. It is possible to activate VSWR antenna supervision on all RF ports supporting measurements of reflected radio power in the radio unit. Supervision can be made on RF ports used for downlink transmitter branch where power is transmitted. Possible causes of VSWR issues include improperly installed antenna, damaged or defective antenna hardware, damaged feeder cable or jumper cable (i.e., excessive bend radius), dirty or loose cable connection, improperly sealed/weatherproofed cable connection, water in the antenna of antenna feeder, and snow on the antenna, and so on. 119 120 LTE Optimization Engineering Handbook 2.3.15 Main Key Performance Indicators It is necessary for the LTE to provide (periodically or on demand) a range of key performance indicators (KPI). The aim of KPI monitoring is in order to provide global information of network behaviour, detect the different problems, analyzing and correlating them with the rest of statistics in order to fix problems encountered. These KPIs ensure adequate analysis of LTE performance in commercial operation environments. The operator‐associated target performance guidelines and the resultant measured KPIs give a means to measure actual performance against target performance. The KPIs will provide the ability monitor network performance at the user plane, control plane, and network level ensuring a full understanding on the overall LTE and eNB performance. These KPIs will enable the monitoring and subsequent troubleshooting of network performance, and will support performance analysis at the user plane, control plane, and network level. Take TDD_LTE, a 2300 MHz operation, or a 20 MHz bandwidth system, for example, LTE RAN KPIs are proposed in Table 2.42. Table 2.42 TDD_LTE RAN KPIs. Drive test based Recommended KPI Drive test OSS based based Session Setup Success >= 99.3 % Rate (%) >= 98.5 % DL RLC Throughput (Mbps) >= 23.2 ‐ E‐RAB establishment successrate (%) >= 99.15 % UL RLC Throughput (Mbps) >= 9.2 ‐ Minutes Per >=55min/drop >= 45 min/drop DL PDCP Abnormal Release (%) Throughput (Mbps) >= 20 ‐ Uplink User Throughput (Mbps) >= 12 >= 15 UL PDCP Throughput (Mbps) >= 8 ‐ Downlink User Throughput (Mbps) >= 30 >= 35 DL Application Throughput (Mbps) >= 21 >= 24.5 Uplink Peak User Throughput (Mbps) >= 30 >= 33 Latency Downlink (ms) <= 15 ‐ Downlink Peak User Throughput (Mbps) >= 65 >= 70 Packet Loss Rate ‐ Uplink <= 0.3% ‐ Latency ‐ Attach Time (ms) <= 550 <= 650 Packet Loss Rate ‐ Downlink <= 0.3% ‐ Latency Round Trip Time ‐TDD (ms) ‐ <= 45 Packet Jitter ‐ <= 5 ms Latency control plane idle to active ‐ <= 375ms Average CQI >= 10 >= 11 Packet Loss Rate ‐ Round Trip (ms) ‐ <= 0.5 % Handover Preparation Success Rate ‐ Intra LTE >= 99.89 % ‐ Handover Success Rate ‐ Intra LTE >= 99.4 % >‐= 98.0 % Handover Execution Success Rate ‐ Intra LTE >= 99.1 % >‐= 98.0 % Handover Interruption Time ‐ User Plane ‐ <= 250 ms Handover Interruption ‐ Time ‐ Control Plane Recommended KPI OSS based >= 99.4 % <= 90 ms LTE Optimization and Principle and Method Recommended KPI OSS based Drive test based Recommended KPI Drive test OSS based based Cell Availability (%) >= 99.99 % ‐ Radireceived interference power PUSCH <= ‐117 ‐ dBm/PRB RRC Connection Success Rate (%) >= 99.4 % >= 99.2 % Radio received interference power PUCCH <= ‐117 ‐ dBm/PRB S1 Establish Success Rate (%) >= 99.1 % ‐ Voice Telephony ‐ Jitter ‐ Added ERAB establishment success Rate >= 99.4 % ‐ Voice Telephony(VT) ‐ <= 0.3 % Packet Loss <= 0.4 % PRACH Success Rate >= 95.0 % >= 92.0 % VT ‐ Speech Quality ‐ >= 4.1 [MOS] DL PDSCH Throughput (Mbps) >= 30 >= 35 VT ‐ Session Setup Success Rate >= 99.0% >= 98.7 % UL PDSCH Throughput (Mbps) >= 12 >= 15 VT ‐ Session Setup Time <= 1.9 sec <= 2.8 sec DL MAC Throughput (Mbps) >= 26.4 ‐ Voice Telephony ‐ Speech Delay ‐ <= 80 ms UL MAC Throughput (Mbps) >= 10.5 ‐ Voice Telephony ‐ Call Drop Rate) <= 0.35 % <= 0.8% KPI analysis Tuning Activity Found Issue Fundamental Analysis User data rate Site audit Accessibility Antenna tuning Weak coverage Call drop rate Parameter Optimization Abnormal coverage PCI optimization Overshoot Coverage analysis Pilot pollution SINR analysis <= 6 ms Handover succ rate Neighbor relation check No handover Handover failure Handover analysis Figure 2.88 KPI analysis and LTE network tuning process. Performance monitoring and reporting are based on a complete set of counters and indicators and are performed at different levels (network level, MME level, cell level, etc.), for different time intervals (busy hour, daily, weekly, etc.) and with different level of details. Some generic reports will be executed daily or weekly at network to give a global view of the network, and some reports will be more focus on a specific monitoring domain and will be use at cell zone level for detailed monitoring and investigations when an issue has been encountered (Figure 2.88). Figure 2.88 gives an overview of KPI analysis and LTE network tuning process. In the following part of the book, we will diccussed them more deeply. 121 123 Part 2 Main Principles of LTE Optimization 125 3 Coverage Optimization 3.1 Traffic Channel Coverage The coverage of a functional network is given by the downlink (DL) and the uplink (UL) coverage. Coverage optimization is the key for LTE network optimization. The purpose is to highlight the domain serving cell and reduce the coverage overlapping. However, UL coverage will normally be the limiting factor. UL coverage limitation can result in radio link failures in case the RSRP levels are still good and do not trigger any intra or inter‐RAT handover. DL coverage can be decreased by antenna tilting. Caution should be used when using antenna tilt to decrease DL coverage since there needs to be a balance between UL and DL. A cell located in an area of continuous LTE coverage that, because of, for example, actual propagation being better than assumed in planning or an error in setting the tilt value provides coverage farther than expected. It can be detected by looking at the average inter‐site distance calculated over the target cells for outgoing handovers or by looking into drive test or scanner results plot marked with serving cells IDs along with handovers. The coverage area should be reduced to a certain degree by downtilting. The definitions of some coverage issues are listed below. ●● ●● ●● Coverage holes definition: RSRP < −120dBm Weak coverage definition: RSRP < −105dBm Overshooting definition: Beyond the expected coverage, the victim cell’s RSRP is good, but RSRQ/SINR is poor as shown in Figure 3.1. One of the highest costs in RAN optimization is coverage measurement and optimization. Traditionally, this requires intensive GPS‐based drive testing followed by offline analysis and correction. Today, the better solution uses data (e.g., measurement report) collected from real subscribers to create a finer granularity geographic distribution than is possible with cell level stats. This data is used to analyze the coverage and interference relationships in each area to provide new tilt values to optimize the balance between coverage and interference. Find out the cross‐boundary area and the weak coverage area and investigate if there is any RAN hardware issue, misconfigured parameters issues, RF issues, or eNB site location issues; the proposed LTE coverage optimization process is depicted in Figure 3.2. The main method of the coverage optimization includes adjustment of the antenna azimuth, adjustment of the antenna downtilt, adjustment of the antenna height, adjustment of the location of the site, adding new sites or RRU for the poor coverage area, adjustment of the RS power, and so on. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. 126 LTE Optimization Engineering Handbook RSRQ (dB) –0 –10 –12 –14 –16 –18 –20 –22 –84 DL interference zone –80 –76 –72 –68 –64 –60 –56 –52 RSRP (dBm) Figure 3.1 DL interference estimator based on RSRQ. Coverage test preparation Site location problem Coverage test including drive test or MR with GIS distribution HW problem RF problem Coverage test result analysis Is the result meet requirement? Test report Parameter problem No Coverage test result analysis Yes Operational test Figure 3.2 LTE coverage optimization process. 3.1.1 Parameters of Coverage For LTE, most often the coverage is determined by UL link. The parameters impacting DL and UL coverage are shown in Table 3.1. In DL, the eNB’s transmit power of all kinds of channels are influencing the cell range. qRxLevMin, is the minimum DL required level of received RSRP in a cell, which is broadcasted in SIB1, in object CellSelectionReselectionConf. The parameter services the same purpose in LTE what parameter ACCMIN services in GSM, impacts the cell size in idle, for cell selection and re‐selection, and used in computation of selection criterion Srxlev. A very low Coverage Optimization Table 3.1 Main parameters impacting DL and UL coverage. DL coverage UL coverage qRxLevMin a factor RS power p0NominalPUSCH/p0NominalPUCCH SCH power offset, PBCH power offset, PCFICH power offset, PHICH power offset, PDCCH power offset Initial UL_SINR target for PUSCH pboffsetPDSCH, pboffsetPDSCH Max UL_SINR target for PUSCH Cell DL total power Min UL_SINR target for PUSCH value (more negative) of qRxLevMin will have more UEs connected to LTE network, but that will impact access success rate and paging success rate. A very high‐value (less negative) of qRxLevMin will result in less number of UEs connected to LTE network. The UL cell coverage is defined as a target service that UE must satisfy at cell‐edge conditions. The cell coverage is independent of channel bandwidth and is strictly dependent of UE maximum Tx power. For UL PUSCH coverage, different kinds of UL_SINR according to services are chosen to ensure that the target SINR at the cell edge does not go too low so as to cause problems with synchronization. α factor is intended to allow partial compensation of the path loss, the value of this parameter represents a trade‐off between minimizing interference and maximizing throughput, its value must be set according to the client’s desired network behavior. p0NominalPUSCH represents the necessary signal level per radio bearer (RB), and of the received signal for correct decoding, which is sent over BCH and not updated with interference. The parameter impacts UE power and interference, before any power control commands is being received from the eNB. Higher settings of the parameter will improve PUSCH reception, but will also drive higher UE Tx power leading to interference to neighboring cells, and vice versa. Besides PDSCH/PUSCH coverage, it is necessary to study PDCCH coverage, since PDSCH may have beamforming gain to improve coverage, while PDCCH cannot be beamformed. That implies PDCCH cannot reach the same coverage as PDSCH in some scenarios. In this case, the SINR of the PDCCH could become negative, which requires a power boost to retain the desired BLER. Automatic boost of PDCCHs using CCE aggregation levels less than the maximum CCE aggregation level is an important feature in a live network (Figure 3.3). Cell center Cell edge 1-CCE 2-CCE 4-CCE 8-CCE CCE aggregation level PDSCH (beam forming) Attenuate Boost Boost Boost Attenuate PDCCH with boost Attenuate PDCCH Attenuate Cell coverage: Optional boost Max boost level set Figure 3.3 PDCCH boost. Selected power setting 127 128 LTE Optimization Engineering Handbook 3.1.2 Weak Coverage 3.1.2.1 DL Coverage Hole Independent of the cause, coverage holes can be detected by looking at the inter‐RAT handover statistics per each adjacency. More specifically, in case of a hole, a number of terminals served by a set of neighboring LTE cells, will perform inter‐RAT handover toward the same set of 3G/2G cells. Inter‐RAT handover statistics may detect DL coverage holes from various causes: worse than expected propagation, blocking by a building, and indoor penetration losses. The factors affecting DL coverage include equivalent isotropic radiated power (EIRP), combining loss, path loss (PL), frequency band, distance between a receive point and an eNB, scenarios (urban and suburban areas) and terrains (plains, mountains, and hills) of electric wave propagation, and antenna related parameters, and so on. To resolving weak coverage problems, it can analyze the EIRP of each cell and ensure it can reach maximum values if possible, if so, increase RS power, adjust antenna, and deploy new eNBs if that cannot be resolved, even uses RRUs, indoor distribution systems to resolve indoor weak coverage. 3.1.2.2 UL Weak Coverage Because of, for example, excessive downtilt, attenuation due to worse than expected propagation, or high indoor penetration, suffers from UL coverage limitation. UL coverage limitation can result in radio link failures, in case the RSRP levels are still good and do not trigger any sort of intra or inter‐RAT handover. The eNB can use PHR (power headroom report) to determine how much more UL bandwidth per subframe a UE is capable of using. The index reported by the UE to indicate the estimated how much transmission power left for a UE to use in addition to the power being used by current transmission. The power headroom reporting range is from −23 to +40 dB. If the power headroom value is (+), it indicates “I still have some space under the maximum power” implying “I can transmit more data if you allow.” If the power headroom value is (−), it indicates “I am already transmitting the power greater than what I am allowed to transmit.” PHR measurements are piggy backed with other RRC messages. In a live network, the UE Tx_power range is usually between −15 dBm up to the maximum value of 23 dBm. The high concentration of samples for 23 dBm indicates that the coverage is not optimal at some parts of the area. Normalized power headroom is an efficient way to estimate UL weak coverage, as UE may send to eNB the power headroom report at the moment of PUSCH transmission, which indicates the distance between UE max Tx power versus the required Tx power of the specific PUSCH transmission within which the power headroom report is sent. Based on the report, eNB is able to estimate the maximum UL grant size (w.r.t. number of PRBs) the UE can support with the current PUSCH transmit power density. The range of normalized power headroom is typically between −5 and 20 dB. When the value approaches 0 dB or even lower, it shows UE has trouble meeting the required UL Tx power due to max Tx power. PUSCH RSSI/PUSCH SINR measurement is an another way to detect UL coverage and UL interference problems. The range of the UL SINR report is typically between −10 to 15 dB. When the report shows UL SINR lower than −5 dB, UE is likely in UL coverage challenge location. Figure 3.4 points the bad UL coverage and interference area based on PUSCH RSSI/ PUSCH SINR. The factors affecting UL coverage include eNB receiver sensitivity, antenna diversity gain, UE transmit power, propagation loss of UL radio signals, tower‐mounted amplifiers (TMAs) on UL, and so on. Coverage Optimization SINR UL close loop power control lower RSSI threshold UL close loop power control upper RSSI threshold UL close loop power control upper SINR threshold Ideal area Bad UL coverage region Noise floor per RB UL close loop power control lower SINR threshold UL interference region RSSI Figure 3.4 Bad UL coverage and interference area. 3.1.2.3 UL and DL Imbalance Sometimes, the eNB cannot receive UL signals because of limited power when UEs perform random access or upload data. In this situation, the UL coverage distance is less than the DL coverage distance. Imbalance between UL and DL involves limited UL or DL coverage. In limited UL coverage, UE transmit power reaches its maximum but still cannot meet the requirement for UL BLERs. In limited DL coverage, the DL DCH transmit code power reaches its maximum but still cannot meet the requirement for the DL BLER. Imbalance between UL and DL leads to service drops. The most common cause is limited UL coverage in a live network. For the optimization of UL and DL imbalance, the UL measurement reports on the Uu interface can be used to analyze and solve problems. The criteria of UL and DL imbalance (without external interference) can be presented as: ●● ●● UL limited: PHR < 3 dB, and RIP < −100dBm, and UL_SINR < 0 dB, and RSRP > −105dBm DL limited:RSRP < −110dBm, and UL_SINR > 0 dB, and PHR > 3 dB, and RIP < −100dBm If UL interference leads to imbalance between UL and DL, it is needed to monitor eNB alarms to check for the interference and check whether the equipment works properly and whether alarms are generated if imbalance between UL and DL is caused by other factors, for example, UL and DL gains of repeaters and trunk amplifiers are set incorrectly, the antenna system for receive diversity is faulty when reception and transmission are separated, or power amplifiers are faulty. 3.1.3 Overlapping Coverage LTE network use the same frequency, if the overlap area is too big between two cells, it will cause a lot of interference with each other. As discussed previously, interference reduces the throughput in a strong way, so the identification of the strongest interferers is needed. From drive test data, an area of strong interference shall be identified as an area with good RSRP but with low SINR or RSRQ. However, SINR and RSRQ measurements are dependent on network load and measurement methods. In an unloaded network, SINR measured from RS reflects the PCI planning quality. RS SINR has also a random component depending on the radio frame synchronization state between neighboring sites (no controlled synchronization by the network). Thus, it is possible to have locations with good RS SINR in unloaded network, but which suffer from heavy interference when there is traffic load in a neighboring (interfering) cell. 129 LTE Optimization Engineering Handbook Polluter 100 80 60 40 20 LTE19510_3 LTE19520_2 LTE19970_2 LTE19973_1 LTE19520_1 LTE19514_2 LTE19917_3 LTE19510_1 LTE19511_1 LTE19514_3 LTE19973_2 LTE19514_1 LTE19509_1 0 LTE19511_2 In noise limited area, multiple signals with adequate power are present in a given area, resulting in high RSSI without a dominant server. SINR value from any of these signals will be impeded resulting in poor service. LTE19511_3 130 Figure 3.5 Pilot pollution. Scanner RSRP measurements provide the most accurate method to identify interfering sites. By looking best N cells RSRP values, it is possible to identify places that don’t have dominant server or where there are many cells visible inside a small power window. Those locations could still have fairly good performance in unloaded network, but they would suffer from interference when traffic load increases in surrounding cells. When an area having detected too many high power pilots as compared to the best serving pilot, the area is defined as pilot pollution as shown in Figure 3.5. Pilot pollution can be dominated by number of signals is more than three, and RSRP is more than −105dBm, and difference with the serving pilot is less than 6 dB. The harm of pilot pollution includes frequent cell reselection in idle state, ping‐pong handover, low SNR and high BLER, drop call, and low throughput. So reducing the pilot pollution (overlap area) will be an important step forward to secure the high network performance and give the end user a good apperception of the LTE network. At a first stage, corrective actions to improve the situation are changes of antenna tilt, azimuth, and eventually, eNB power of the interfering cells. Antenna patterns should have low sector power ratios to minimize overlap, and high upper sidelobe suppression to minimize coverage in unwanted directions. A second stage might be the change of antenna type and height or add new site/RRU to create dominance. 3.1.4 Overshooting Overshooting happens when a cell is giving coverage out of its designed coverage area into a neighbor cell causing interference. Because of, for example, actual propagation being better than assumed in planning or an error in setting the tilt value provides coverage farther than expected, overshooting is happening. It can be detected by looking at the average inter‐site distance calculated over the target cells for outgoing handovers or by looking into drive test results plot marked with serving cells IDs along with handovers. Overshooting cell usually contributes to low throughput and drop call. The coverage area should be reduced to a certain degree by downtilting or power setting. Overshooting can be detected with propagation delay counters. Propagation delay distribution shows that the sector is shooting as far as 10 km away. Analysis of traffic versus distance for cell/user is necessary for tilt optimization and interference reduction (Figure 3.6). Cells ranked in order of network impact for a structured approach to address interference issues. For engineering, timing advance distribution and handover relations can be analyzed for boomer coverage as shown in Figure 3.7 and Figure 3.8. Coverage Optimization Overshooting Cell ‘B Cell ‘A Figure 3.6 Unwanted island areas in cell “A” and optimizes cell “B” to remove them. Figure 3.7 TA distribution. Figure 3.8 Handover relations. 131 132 LTE Optimization Engineering Handbook There are different ways to identify overshooting. Based on PCI information of serving sector and target sector, calculating RSRP delta and distance between these sectors, and checking the random access response messages for the timing advance value, overshooting is identified. Overshooting can be minimized using physical optimization (mechanical/ electrical tilt, azimuth optimization, reducing the antenna height) or parameter optimization like power of nominal cell and/or overshooting cells, handover thresholds, maximum cell range, and so on, as shown below where boomers are identified with distance measured. Here is an example of a downtilt overshooting cell to remove interference, which is shown in Figure 3.9. 3.1.5 Tx1/Tx2 RSRP Imbalance The antenna system is one of the major components contributing to the performance of a wireless communication system. When using multiple antennas, the antennas and feeders must align well, when at least one antenna works well, but the others may be disconnected or have RF loss, there is almost no gain from the disconnected antennas in such scenarios. Before launching the network, it will have to make it possible to discover both slowly failing and wrongly mounted antenna systems timely and remotely. In live network, some scanners and data collection/post processing tools can distinguish between transmit (Tx) branches of each sector by utilizing Layer1 measurements of received RSRP/SINR per antenna branch to monitor the antenna status and detect if there are problems with the antenna. In the case below, UE‐received RSRPs of Tx1/Tx2 are imbalanced as shown in Figure 3.10, in areas served by this sector could still maintain a connection to the sector, but the performance would be impacted. Remote measurements on the radio unit, such as return loss and VSWR and associated alarms, can identify these issues. Typically, this points out to hardware issues at the cell. High RF delta between the UE receive ports can have huge impact on the throughput especially using MIMO, it should be verified that the delta is within 0 to 10 dB. With practical integrated UE antennas power imbalance is difficult to control since the UE antenna patterns are not identical. In live network, the antenna problems are usually in the antenna branch, which is the receiver or transmitter chain. A fault in the antenna could be related to the TMA, or the cable/ feeders. 3.1.6 Extended Coverage Large cells are suitable to obtain coverage in sparsely populated areas where the requirement for capacity is low. Examples of such areas are deserts, coastal areas, or sea environments. For 3GPP standard, LTE enable cell sizes up to 100 km. The limit corresponds to the maximum timing advance value that can be sent to a UE. This will give the UE a minimum amount of processing time after receiving data in DL and before transmitting a response in UL. For each cell, the maximum desired cell range can be defined from 1 up to 100 km by 3GPP. First, it is important to carefully select the correct combination of PRACH format type and special subframe type. Either PRACH format or special subframe type will be the upper bound for maximum cell radius, independent of link budgets. For LTE FDD, only PRACH format 0, 1, 2, 3 set an upper bound for maximum cell radius (Figure 3.11). There are many factors that limit cell size: ●● RACH delay: An unsynchronized (no timing advance) UE’s ability to send a RACH preamble and have it arrive at the eNB within the RACH window is dependent on the size of the preamble and the round trip delay of the cell. Coverage Optimization 3000 2500 Number of samples 2000 1500 1000 Tilt needed 500 0 0 156 312 468 624 779 935 1091 1247 1403 1559 1715 1871 2027 2183 2338 2572 Distance (m) Figure 3.9 Downtilt and related propagation delay. ●● Cyclic shift orthogonality: Each UE will utilize a different cyclic shift of a root ZC sequence. However, the delay (and delay spread) of the channel will make the ZC sequence of a far UE look just the same as the ZC sequence of a close UE that has been cyclically shifted. In order to prevent this, the cyclic shift of the UEs must be kept further apart than the round trip delay time + delay spread. 133 LTE Optimization Engineering Handbook Sc RSRP Tx1 (dBm): EARFCN_Cl CN-5230_Cl-352 [Min, –120) (271) (20.67%) [–120, –115) (207) (15.79%) [–115, –110) (309) (23.57%) [–110, –105) (255) (19.45%) [–105, –100) (157) (11.98%) [–100, –95) (63) (4.81%) [–95, –90) (19) (1.45%) (20) (1.53%) [–90, –80) (6) (0.46%) [–80, –70) (4) (0.31%) [–70, Max) Sc RSRP Tx2 (dBm): EARFCN_Cl CN-5230_Cl-352 [Min, –120) (7) (0.53%) [–120, –115) (28) (2.14%) [–115, –110) (184) (14.04%) [–110, –105) (215) (16.4%) [–105, –100) (260) (19.83%) [–100, –95) (222) (16.93%) [–95, –90) (188) (14.34%) (193) (14.72%) [–90, –80) (12) (0.92%) [–80, –70) [–70, Max] (2) (0.15%) PCl 352 TX1 RSRP PCl 352 TX1 RSRP Figure 3.10 Tx1/Tx2 RSRP imbalance. SSF8 PRACH format 4 PRACH format 3 SSF7 PRACH format 2 PRACH format 1 SSF6 Special sub-frame type 134 PRACH format 0 SSF5 SSF4 SSF3 LTE supports 4 preamble formats for both FDD and TDD and one format specific for TDD in order to provide reasonable RACH configurations to various environments. The preamble structure for the 4 formats suitable for both FDD and TDD. SSF2 SSF1 SSF0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 Maximum cell radius (km) Figure 3.11 Cell range according to PRACH format. ●● SNR will determine the eNBs probability of detecting the RACH sequence. For larger cells, users at the cell edge require more signal power (hence a longer preamble) to meet the required probability of detection. UE was able to successfully attach to the network after 15 km with preamble Format 1 being used in the PRACH. After a successful attach at 15.5 km from the site, UL and DL throughput tests were conducted. Table 3.2 summarizes the results observed. An example of extended coverage with parameter configured is shown in Figure 3.12. If the maximum cell range feature is activated and if the cell range exceeds 15 km, extra UL resource blocks is used for random access, which reduces UL capacity. For some cases, repeater can be deployed to extend coverage. It worth to mention that when the UE have a signal from both the macro base station and the repeater, the normal cyclic prefix Coverage Optimization Table 3.2 UL and DL throughput tests at 15.5 km from the site. RSRP (dbm) CINR (db) Throughput (Mbps) DL −118.31 4.3 8.23 UL −118.31 4.3 0.35 will in some locations not cover the delay difference between the direct path and the repeater path that will result to the performance degradation. An example is shown in Figure 3.13. 3.1.7 Cell Border Adjustment In previous chapters, it can be seen that the cell size is impacted by a number of factors, output power and antenna tilt are two examples. In a live network, it has three coverage borders according to different parameters. ●● ●● ●● “Real” coverage area is defined by output power, radio conditions, and so on; “Idle mode UEs” cell size is defined by qRxLevMin. The qRxLevMin parameter defines the minimum received signal strength an idle mode UE shall receive in order to stay in the cell. If received power is less that this level the UE will try to make a reselection from the cell. “Connected UEs” cell size toward other RANs defined by a2Thresholds; The a2Threshold parameters define a mobility threshold before a RRC connected UE tries to move, using for example, handover, release with redirect. During troubleshooting activities in a live network it has been seen that when the idle mode parameters are not tuned correctly, the eNB will experience an abnormal number of time outs of the RRC connection set up procedure. Similarly, if the bad coverage parameters are tuned incorrectly the eNB will experience an increase in radio link failures where, if tuned correctly, a redirection to other RAT could have been possible. The KPI measurement gives the smallest cell size that should determine the cell size in both idle and RRC connected mode. The optimization of the cell size should be automatically tuned to achieve an optimal accessibility and retainability KPI in a live network (Figure 3.14). When bad KPIs are observed, cell size is decreased (if possible), when too good KPIs are observed, cell size is eventually (a hysteresis is used) increased. The accessibility and retainability KPIs are checked periodically. The cell size (determined by parameters Qrxlevmin and a2Threshold or b2Threshold1) is decreased when statistical base is sufficient AND the percentage of failed RRC connection set up procedures due to time out is higher than the acceptable level OR the percentage of dropped connections due to radio link failure is higher than the acceptable level. The cell size is increased when statistical base is sufficient AND the percentage of failed RRC connection set up procedures due to time out is lower than the acceptable level minus a hysteresis AND the percentage of dropped connections due to radio link failure is lower than the acceptable level minus a hysteresis. Otherwise the cell size is maintained. The operator can set different qRxLevMin for different cells respectively according to different scenarios and RF condition (e.g., tuning cluster cell‐edge, idle and active mobility cell‐edge matching, inter‐cells gap, inter‐cells overlap, cells load balance). The tuning purpose of qRxLevMin is to set suitable cell size, includes decreasing the gap between cells, controlling overlap of cells, matching the idle mode cell size and connected mode cell size, and so on. The qRxLevMin and a2Threshold shall be adjusted together. So if, for example, qRxLevMin is decreased a2Threshold will also be decreased. qRxLevMin and a2Threshold will not be enforced 135 Distance Total: 36.610 Km RSRP = –129 dBm, SINR = –4 dB Maximum cell range is 36.61 Km Figure 3.12 Example of extended coverage. (See insert for color representation of the figure.) Coverage Optimization 70 m high source eNB Received power from both path about the same and delay difference larger than Cyclic Prefix (4.7 µs) gives degraded performance 2.7 km 2.7 km Repeater adds ~5 µs delay. 1.4 km corresponds to 4.7 µs extra delay Figure 3.13 Repeater in LTE. RRC Connected Mode cell boarder towards CDMA is defined by parameter: or t eventA2Threshold : CH :R CH DC g ra ove dC RA CH CC ep eR :R CH Ba DC :R RC RC RC Co n ne Co n ne Co n ne cti on Idle Mode cell boarder is cti on cti on Re q Se st tup Se defined by parameter: ue tup Co m Qrxlevmin ple te 2G Coverage Figure 3.14 Cell size in idle and RRC connected mode. to have similar values. An offset parameter can exist in LTE that defines the difference between the a2Thresholds and qRxLevMin. Default value for this offset will be 0 dB. When the algorithm starts the a2Thresholds will be set using this offset. For cell border optimization, there are still some weight factors to be considered: how much is the RRC connection set up time out failure ratio, how much is the RRC reestablishment failure ratio, how much is the initial RRC context set up failure ratio, how much is the abnormal release ratio, and how fast is the hysteresis shall go down to allow cell size increases, and so on. 3.1.8 Vertical Coverage In most scenarios, users are predominantly distributed in the azimuth plane, so in the narrowly spaced, horizontal configuration, antenna is clearly the best all‐around antenna. The vertically configured antenna may show benefits for scenarios, such as the urban canyon, coverage of high‐rise buildings and small cells. The vertical configuration requires UEs to be distributed in elevation. 137 138 LTE Optimization Engineering Handbook Outer beam Inner beam z UE elevation UE x y Figure 3.15 Vertical sectorization. In a typical macro cell environment, the horizontal configuration requires more than 4λ spacing for decorrelation and less than 0.5λ for high correlation. The vertical configuration was found to require wider spacing of more than 12λ for low correlation and a wider acceptable spacing of less than 8λ for high correlation. Vertical sectorization is made possible through the vertically active antenna array, which can be used to form multiple, narrow vertical beams pointed in different directions in order to create multiple sectors, this has the advantage in allowing reuse of resources (PRBs) between beams, but also introduces additional UL interference due to the overlap of the beams. Each vertical sector has a separate physical layer cell ID (Figure 3.15). 3.1.9 Parameters Impacting Coverage The parameters of coverage have an impact on coverage are presented in Table 3.3. Most often the coverage is determined by UL link, thus for optimizing coverage it is needed to optimize levels of UL power. 3.2 Control Channel Coverage For control channel coverage optimization process, the primary set of parameters are for power settings that can have a big influence on coverage, for example, reference signal power, primary/ secondary sync signal power offset, and so on. The higher the power the better the coverage but also the higher interference in the neighboring cells. The values must be tuned so that cells coverage with same frequency does not overlap. Based on field and analytical studies, it is recommended that adjusting CCCHs and reference power offsets to maintain current amplifiers power utilization as well as to maintain PDSCH and CCCH’s SINR levels better than the current level. Coverage Optimization Table 3.3 Parameters impacting UL/DL coverage. Parameters impacting UL coverage Parameters impacting DL coverage pUSCHPowerControlAlphaFactor qRxLevMin p0NominalPUSCH cellDLTotalPower/referenceSignalPower uplinkSIRtargetValueForDynamic PUSCHscheduling primary/secondary SyncSignalPowerOffset sEcorrInit pBCHPowerOffset ulSyncSINRsyncToOOSTreshold pDCCHPowerOffsetSymbol ulSyncSINROOStoSyncTreshold pCFICHPowerOffset qRxLevMin pHICHPowerOffset deltaFPUCCHFormat1 pbOffsetPdsch/paOffsetPdsch sIRTargetforReferencePUCCHFormat phichResource minSIRtargetForFractionalPowerCtrl dlTargetSINRTableForPDCCH maxSIRtargetForFractionalPowerCtrl pdcchAggregationLevelForCRNTIGrantsIn CommonSearchSpace pathLossNominal pdcchAggregationLevelForUESearchSpace p0NominalPUCCH n310/t310 139 140 4 Capacity Optimization Capacity optimization focuses on the resource utilities and high traffic solution. It is not only about user plane dimensioning but also interference control, signaling solutions, and user connectivity dimensioning. LTE cell and user throughput is an important KPI of capacity optimization. Many high‐resource utilities are usually UL limited while the networks in general are DL limited, which adds further complexity on the high capacity optimization. Figure 4.1 gives the general principle of high capacity optimization strategy. The optimization engineer can collect OMC and trace data and site latitude, longitude and azimuth from operator network, and calculate the average network utilization, traffic volume, and DL spectrum efficiency over the heaviest busy hour (BH), and make necessary update on capacity assumptions based on the better understanding of operator requirements. Then, based on predicted traffic increase and offload from network rollout plan, estimate the near‐ term traffic volume, resource utilization, and license requirement. Usually, the network total traffic can be calculated from the product of forecasted subscriber number and the forecasted traffic per subscriber. Figure 4.2 shows the service work flows from a LTE capacity planning service. Capacity optimization triggers are based on network resources, radio spectrum limitations, and the forecasted traffic. When performing root cause analysis of LTE low throughput investigations, it is important to analyze the radio conditions, signaling flows, logging messages, and cell loading to understand low throughput reasons. In order to improve the SIR, it may be necessary to perform antenna tilting to reduce over shooting cells and minimize coverage overlap as this may cause excessive inter‐cell interference. 4.1 RS SINR RS SINR is the most important factor impacted the network capacity. RS SINR is determined by position in cell (RSRP), interfering cell load, interferer cell geometry, clutter and terrain type, and reference signal configuration, and so on. A high‐speed LTE data network requires excellent SINR distribution planning besides parameter optimization and troubleshooting. Various internal and external assessments have shown that the network has room for improving the SINR and CQI distribution which would be a precursor to any additional optimization under static and mobility scenarios, and directly impacts the accessibility, retainability, throughput, and quality. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Capacity Optimization Parameters Optimization RF resource not enough Traffic distribution BBU resource not enough Resource not enough Hot spot Backhaul resource not enough Traffic prediction Frequency expansion Channel dimensioning License expansion Add more cell Cell rehoming Core network resource not enough Load balance Core Expansion Figure 4.1 High‐capacity optimization strategy. DL/UL volume per cell average N busy hours DL/UL volume forecast eNB resource utilization average N busy hours (license, CP, UP etc.) eNB resource utilization forecast Active DL/UL UEs per cell average N busy hours Active DL/UL UEs forecast Spectral efficiency theoretical limit per cell Max DL/UL cell throughput VoLTE capacity monitoring and planning VoLTE resources forecast Max active UEs for acceptable DL/UL user throughput DL/UL volume triggers eNB resource triggers User throughput triggers Key capacity indicator Limit Trigger # of pages/s/eNB 100 70 # of Peak RRC connected users/sector 167 80% of limit # of Peak Radio bearers/sector 500 80% of limit 100% 80% Avg PRB utilization%/sector VoLTE resources triggers Capacity trigger Figure 4.2 Capacity optimization trigger. 4.2 PDCCH Capacity PDCCH (Physical DL control channel) is used to transfer DCI (DL control information) for the scheduling of DL resources on PDSCH and UL resources on PUSCH. PDCCH symbols in the beginning of each subframe can be dynamically adjusted based on the load and radio conditions. UE in poor radio coverage will require 8 control channel elements (CCE, PDCCH can be aggregated in groups of 1, 2, 4, and 8 CCEs) for each PDCCH transmission 141 142 LTE Optimization Engineering Handbook Control format indicator 2 1.5 MHz NrSymPdcch = 1 NrSymPdcch = 2 NrSymPdcch = 3 1 0.5 0 System Bandwidth 5 10 5 10 13 27 44 21 20 21 55 88 10MHz bandwidth –9 –5 –3 –1 1 3 5 7 9 11 13 15 17 19 21 23 25 SINR Figure 4.3 Number of PDCCH symbols versus DL SINR (10 MHz bandwidth). where as UE in good radio conditions may need 1 CCE. Therefore, higher aggregation levels should be employed at cell‐edge whereas lower aggregation levels are meant for users close to the cell. PDCCH capacity is measured in terms of CCEs, which are 9 sets of REGs (resource element groups), that is, 36 REs as each REG contains 4 RE. PDCCH coding is restricted to QPSK and coding scheme (convolutional 1/3) in order to enhance the decoding in low SINR conditions at cell‐edge with BLER below 1%. Figure 4.3 shows the number of PDCCH symbols used in the different cells versus the DL SINR conditions measured by the UE and the maximum number of CCEs for different band width configurations and different values of PDCCH symbols. As expected, under good radio conditions, low aggregation levels are employed for UL and DL. As radio conditions worsen the aggregation levels will tend to increase and so will the average number of required PDCCH symbols. In extremely low DL SINR the average PDCCH symbols is just below 2, since drives were performed with a single user and 10 MHz bandwidth configurations for all the cells. The relationship between code rate and the number of symbols occupied by PDCCH, the number of antennas and the size of TB (10 MHz bandwidth) is shown in Table 4.1. Additionally, the PDCCH capacity for user data scheduling is reduced by PDCCH CCEs used for system information scheduling over PDSCH, paging, preamble assignment and random access message 2 scheduling carried by different DCI formats. The aggregation level used for broadcast (SIB), paging preamble assignment and RA response is specified by parameters and is limited to aggregation levels 4 and 8 to ensure reliable decoding across the cell coverage area. The first phase of PDCCH capacity calculation procedure is the calculation of the total resources occupation of control channel. Assuming conventional CP, TDD‐LTE (DL:UL = 2:2) special subframe ratio 10:2:2, the total resources of control channel (REG) = total RB resources × control channel symbol number × 3. Control domain OFDM symbol number relates with CP configuration, special subframe configuration, PCFICH channel, and other related information. In the calculation of each subframe number of scheduled users, high bandwidth under conventional subframe is generally considered 3 OFDM symbols, then 20 MHz bandwidth under conventional subframe control domain resource is 100 * 3 * 3 = 900 REG (Figure 4.4). Figure 4.4 gives the PDCCH capacity analysis procedure. It is needed to determine the total REGs in the control domain with system bandwidth and other parameters, and calculate the REGs occupied by other control channels except PDCCH. Then, the CCES used for PDCCH can be retrieved, and the allowed scheduled users per TTI according to the cell conditions will be determined. Based on the above analysis, it is concluded that the parameters that affect the PDCCH channel capacity are: system bandwidth, CP length, subframe configuration, special subframe configuration, the number of antenna ports (two ports in the example), PHICH parameter like Capacity Optimization Table 4.1 PDCCH symbols versus TBS. 3 PDCCH Sym TBS 2 PDCCH Sym 1 PDCCH Sym 2 Ant. 1 Ant. 2 Ant. 1 Ant. 2 Ant. 1 Ant. 1384 0.115 0.11 0.105 0.1 0.096 0.092 1800 0.15 0.143 0.136 0.13 0.125 0.12 2216 0.185 0.176 0.168 0.161 0.154 0.148 2856 0.238 0.227 0.216 0.207 0.198 0.19 3624 0.302 0.288 0.275 0.263 0.252 0.242 4392 0.366 0.349 0.333 0.318 0.305 0.293 5160 0.43 0.41 0.391 0.374 0.358 0.344 6200 0.517 0.492 0.47 0.449 0.431 0.413 6968 0.581 0.553 0.528 0.505 0.484 0.465 7992 0.666 0.634 0.605 0.579 0.555 0.533 7992 0.333 0.317 0.303 0.29 0.278 0.266 8760 0.365 0.348 0.332 0.317 0.304 0.292 9912 0.413 0.393 0.375 0.359 0.344 0.33 11448 0.477 0.454 0.434 0.415 0.398 0.382 12960 0.54 0.514 0.491 0.47 0.45 0.432 14112 0.588 0.56 0.535 0.511 0.49 0.47 15264 0.636 0.606 0.578 0.553 0.53 0.509 15264 0.424 0.404 0.385 0.369 0.353 0.339 16416 0.456 0.434 0.415 0.397 0.38 0.365 18336 0.509 0.485 0.463 0.443 0.424 0.407 19848 0.551 0.525 0.501 0.479 0.459 0.441 21384 0.594 0.566 0.54 0.517 0.495 0.475 22920 0.637 0.606 0.579 0.554 0.531 0.509 25456 0.707 0.673 0.643 0.615 0.589 0.566 27376 0.76 0.724 0.691 0.661 0.634 0.608 28336 0.787 0.75 0.716 0.684 0.656 0.63 30576 0.849 0.809 0.772 0.739 0.708 0.679 31704 0.881 0.839 0.801 0.766 0.734 0.705 36696 1.019 0.971 0.927 0.886 0.849 0.815 Ng = 1/6, assuming the normal subframe control domain occupies three OFDM symbols, special subframe control domain occupies one OFDM symbol, the allowed scheduled users according to different PDCCH aggregation level is shown in Table 4.2. In a live network, PDCCH CCEs utilization above 75% indicates high PDCCH CCEs utilization. High PDCCH utilization can be caused by large number of active users, poor coverage, high interference, parameters and traffic model, and so on. RRC connection set up failure/incoming handover preparation failure under high‐capacity conditions usually have a 143 144 LTE Optimization Engineering Handbook System bandwidth, CP, DL and UL subframe configuration, special subframe configuration, and antenna configuration The symbols occupied by CFI (PCFICH control domain) The REs occupied by RS in control domain The REGs occupied by PHICH The REGs occupied by PCFICH (4REG) The REGs occupied by other channels except PDCCH The total REGs in control domain The CCEs can be used for PDCCH Determin the allowed scheduled users per TTI according to PDCCH aggregation level Figure 4.4 PDCCH capacity analysis procedure. Table 4.2 The example of allowed scheduled users. subframe subframe configuration PDCCH aggregation level 0 1 2:2 0 88 1 44 3:1 2 3 4 5 6 7 8 9 20 87 88 20 87 10 43 44 10 43 2 22 5 21 22 5 21 3 11 2 10 11 2 10 0 88 20 87 88 88 20 87 88 1 44 10 43 44 44 10 43 44 2 22 5 21 22 22 5 21 22 3 11 2 10 11 11 2 10 11 high PDCCH utilization behind them. High PDCCH utilization usually indicates the SRB traffic is using the majority of the scheduling opportunities and leaves very little opportuni ties for DRB traffic under high load conditions. The eNB is probably limited on CCEs during low throughput. If CCE has nearly reached the maximum CCE aggregation level, the feature of PDCCH power boost can further improve PDCCH coverage, but decrease PDCCH CCEs utilization. 4.3 PUCCH Capacity The physical UL control channel (PUCCH) is used for the transmission of signaling (scheduling requests, HARQ acknowledgments, channel state information) when no simultaneous UL data is being sent. The PUCCH resources are always allocated on the extreme ends of the band width in order to maximize the number of contiguous PRBs that can be allocated for the Capacity Optimization Different UEs are separated on PUCCH by means of FDM and CDM. FDM is used between the RB whereas CDM is used inside the PUCCH RB. PUCCH utilizes two separate and LTE-specific ways to realize CDM: by means of cyclic shifts of sequences with suitable zero-autocorrelation properties (i.e., CDM inside the SC-FDMA block) and by means of block-wise spreading with the orthogonal cover sequences (i.e., CDM between multiple SC-FDMA blocks) subframe0 slot0 slot1 subframe1 slot2 slot3 subframe9 slot18 slot19 RB0 m=1 m=0 m=1 m=0 m=1 m=0 RB1 m=3 m=2 m=3 m=2 m=3 m=2 m=2 m=3 m=2 m=3 m=2 m=3 m=0 m=1 m=0 m=1 m=0 m=1 RBn ZC_root1 CSsymbols OC1 RS_OC1 ZC_root2 CSsymbols OC2 RS_OC2 ZC_rootn CSsymbols OCn RS_OCn Figure 4.5 PUCCH multi‐user. PUSCH. Additionally, users will hop between the bandwidth edges (intra subframe hopping) in order to provide frequency diversity gain for PUCCH transmissions. PUCCH are by default located at the lower and upper band edges of the UL system band width. The PUCCH region needs planning in order to get an appropriate balance between control and data traffic resources. It is important to correctly dimension the PUCCH area, not only to avoid unnecessary overheads that reduce the throughput (if more resources than needed are reserved) but to avoid poor RRC set up success rates (in case insufficient PUCCH resources are reserved to handle the maximum number of active users in the cell). PUCCH is code‐multiplexed of multiple UEs in one RB pair, which is spreading in frequency domain within symbol by ZC sequence, spreading in time domain symbols by orthogonal covers (format 1/1a/1b). PUCCH capacity typically limited by inter‐cell interference, so hop ping technique is used for avoiding interference for neighboring cells that including group hopping (ZC root sequences hopping on slot basis) based on cell ID, cyclic shift hopping on symbol basis, scrambling sequence per radio frame, and orthogonal covers hopping on slot basis (format 1/1a/1b only) (Figure 4.5). 4.3.1 Factors Affecting PUCCH Capacity The PUCCH supports multiple formats. Table 4.3 provides a list of these formats and the type of information that is carried with the different PUCCH formats. Formats 2a and 2b are sup ported for normal cyclic prefix only. The outer part of the PUCCH region will be employed to transmit Format 2 type information. This is mainly CQI reports with or without HARQ Table 4.3 Info carried in the different PUSCH formats. PUCCH format Information Modulation scheme 1 Scheduling request (SR) N/A 1a 1‐bit ACK/NACK with/without SR BPSK 1b 2‐bit ACK/NACK with/without SR QPSK 2 20‐bit CSI (CQI + PMI + RI) QPSK 3 Up to 10‐bit(FDD)/20‐bit(TDD) ACK/NACK QPSK 2a 20‐bit CSI + 1‐bit ACK/NACK QPSK + BPSK 2b 20‐bit CSI + 2‐bit ACK/NACK QPSK + QPSK 145 146 LTE Optimization Engineering Handbook Network and PUCCH parameter PUCCH resource allocation PUCCH capacity Bandwidth and DL/UL configuration PUCCH RB config during report period PUCCH RB config per subframe Report period of UL control information UL control information capacity PUCCH transport format PUCCH resources supported per RB PUCCH resource config UL control channel resource allocation Figure 4.6 PUCCH channel estimation procedure. acknowledgments. This region will be followed by another section in which it is possible to have a mix of Formats 1.x and 2.x. Formats 1 are employed for the transmission of scheduling request (SR) and Ack/Nacks. Finally the inner part of the PUCCH region allocates resources dedicated for Formats 1. UE is allocated SR and CQI resources during set up procedure, and the resources are kept as long as UE is UL synchronized. The amount of CQI, SR, and ACK/NACK will affect PUCCH loading. CQI is periodic based, it can be moved to PUSCH in case of concurrent data transmission. CQI will not transmit if UE is in DRX sleep mode. SR is transmitted only when SR (and related buffer status reports) is triggered. Traffic enters the buffer corresponding to a logical channel group with higher priority than what exists in the buffer, or at expiry of the retxBsrTimer. ACK/ NACK is transmitted per UE‐specific DL PDCCH allocation. The PDCCH allocation controls the PUCCH allocation, that is, there is some freedom in the number of ACK/NACKs per RB pair. In case of concurrent data transmission, ACK/NACK can be moved to PUSCH. The capacity of PUCCH channel estimation scheme is shown in Figure 4.6. In order to improve the usage of RB, one UE occupy one RB waste too much, so consider CDM (cyclic shifts in frequency and orthogonal sequence in time) to make multiple UE shared one RB in one TTI in the same PUCCH Format. PUCCH resource calculation can be found in Figure 4.7 and Figure 4.8. For PUCCH format 1/1a/1b, 3 symbol used for DMRS, 4 symbols left used for PUCCH for mat 1/1a/1b. At most 12 × 3 = 36 UE can use in the same RB. For PUCCH format 2/2a/2b, only used for period CQI reporting, non‐period CQI reporting used in PUSCH. PUCCH format 3 used for CA condition, each carrier needs to transmit DL data in the same TTI, and the UE can only send one PUCCH on the Pcell in each subframe, including all the serving cell’s ACK/NACK information. In LTE, one UE can use 5 serving cell at most, if each cell use MIMO, it needs 10 ACK/NACK for FDD. For TDD, if it likes DL:DL:UL, then in the UL, the UE needs to response 10 × 2 DL = 20 ACK/NACK, if it likes DL:DL:DL:DL:UL, then the UE needs to response 10 × 4 = 40 ACK/NACK. And even sometimes need to add one more bit for SR. The number of RB‐pairs for format 1 can be shared between SR and HARQ resources. The number of pairs for these resources can be calculated by: nRB , Format 1 nPUCCH ,SR nPUCCH , HARQ 36 Capacity Optimization Figure 4.7 Max number of resources per PRB. Max resources per PRB m=0 format 2/2a/2b: RB NSC = 12 m=3 m=2 m=2 m=3 Max resources per PRB format 1/1a/1b: RB c· NSC PUCCH PUCCH = {c = 3, Δshift = 1} = 36 Δshift m=0 Figure 4.8 RB Calculation for format 1/1a/1b. (2) NRB (1) NCS = 0 PUCCH Δshift PUCCH δoffset (1) nPUCCH slot _ nr f(Cell_id,...) symbol _ nr ZCroot (slot), common for all users CS (symbol, slot) RBnr (slot) OCindex (slot) The number of resource blocks per slot allocated for format 2 is calculated by: nCQI ,res nRB , Format 2 ncapTp ,CQI * 10 nSF , PUCCH ncqi,res is the wanted number of CQI resources on the PUCCH, specified by parameter noOfPucchCqiUsers. ncap is the CQI resources per RB‐pair, equal to 4, Tp,cqi is the periodicity for CQI reporting. The number of RB‐pairs for PUCCH is given by: nRB , PUCCH 2 nRB , Format 1 nRB , Format 2 2 Table 4.4 gives the parameters deployed for PUCCH physical resources configuration (10 MHz). 147 148 LTE Optimization Engineering Handbook Table 4.4 Parameters deployed for PUCCH. Parameters Range Description default nCqiRb 1…98 Number of PRBs dedicated to Formats 2.x 1 Tperiod,cqi 2 ms (0), 5 ms (1), 10 ms (2), 20 ms (3) Periodicity of periodic CQI/PMI feedback on PUCCH or PUSCH 10 ms n1pucchAn 10…2047 Offset to decide the number of resources reserved for SRI (and A/N from persistent PDSCH scheduling in later releases). 18 1…3 Maximum number of cyclic shifts allowed for Formats 1/1a/1b 2 Tperiod,SR 5 ms (0), 10 ms (1), 20 ms (2), 40 ms (3), 80 ms (4) Scheduling request periodicity in the cell. The recommendation is to have one scheduling request (SR) configured per frame. 10 ms prachFreqOff 0…94, step 1 First physical resource block available for PRACH in the UL system frequency band. Roundup [PUCCH resources/2] 3 puschHopOffset 0…98 The PUSCH hopping offset parameter defines the offset used for PUSCH hopping, expressed in a number of resource blocks 5 PUCCH shift 1 subframe with PRACH occasion 1 PRB for SR 6 PRBs for PRACH prachFreqOff = 3 1 PRB for CQI (RI & PMI). 49 48 47 46 . . 8 7 6 5 4 3 2 1 0 PRACH 1-3 PRBs for ACK/NACKS. DMRS Unused PUCCH PUSCH Figure 4.9 RB allocation for PUCCH (example). Within an RB pair, 12CQI, 36SR, and 36ACK/NACK resources can be used when code division multiplexing is applied. The number of PRBs dedicated to Formats 2.x is defined by the param eter nCqiRb. Control information from multiple users can be scheduled on the same PUCCH timeslot by means of using Zadoff‐Chu CAZAC codes, which will preserve the orthogonality. Since a PUCCH RB spans over 12 sub‐carriers there will be 12 cyclic shifts available per PUCCH RB, it is not recommended to employ all 12 cyclic shifts since this leads to excessive interference and therefore degradation in performance (Figure 4.9). Every UE with a DRB in the cell is required to send CQI reports periodically. For this purpose every UE will require a Format 2.x PUCCH resource assignment. The periodicity of the CQI Capacity Optimization reports is set by the parameter Tp,cqi. The total capacity provided exclusively for Format 2.x in a cell will be defined by: nCqiRb * 12 * Tp ,cqi 1 * 6 * 10 ms 120users For the transmission of Formats 1.x a similar approach as for Format 2.x is employed. Cyclic shifts for orthogonality between users are employed along with code division multiplexing (CDM) increasing the number of users, multiplexed orthogonally on a single PRB. The CDM is limited by the number of reference symbols that are available for the time domain spreading. This leads to a spreading factor (SF) =3 for normal cyclic prefix (CP) and a SF = 2 for extended cyclic prefix. The number of resources for PUCCH format 1.x per RB is defined by: SF * 12 / deltaPucchShift 3 * 12 / 2 18 users per PRB Where deltaPucchShift is a parameter that defines the maximum number of cyclic shifts (range: 1, 2 and 3) allowed for Formats 1.x. Considering normal cyclic prefix the maximum number of Formats 1.x that can be multiplexed per PRB is 36. In a live network, it does not recommend the usage of all 12 cyclic shifts due to the high interference generated, therefore, it is recommended to set deltaPucchShift to 2, reducing the maximum number of Formats 1.x per PRB to 18. For scheduling requests, semi‐persistent scheduling Ack/Nacks and Ack/Nack repetition, a fixed number of resources is reserved per TTI. This is defined via the parameter n1pucchAn. The exact resource to be employed inside this reserved region for the above use cases is communicated to the UE via explicit signaling. For HARQ acknowledgments of dynamic scheduling, the resource to be used by the UE is a function of the first CCE used for the PDCCH scheduling. This means that the resources for Formats 1.x can vary between the ranges speci fied below: 0 Format 1. xResource range n1 pucchAn Max CCE The number of Format 1.x resources required has to be dimensioned considering the num ber RRC connected users per cell and the amount of users scheduled per TTI in the DL Apart from having dedicated PRBs for Formats 1.x and Formats 2.x there is a possibility of having a mixed format region. In this region the resources per PRB are divided between the 2 Formats. For formats 1.x, it can be reserved up to eight cyclic shifts per PRB via the parameter pucchNAnCs. The remaining cyclic shifts can be employed for Formats 2.x considering there are two guard cyclic shifts as shown in Table 4.5. Table 4.5 Mixed region resource allocation for pucchNAnCs =6. Cyclic shifts RB xx Cyclic shifts 0 Format 1.x RB xx 6 Guard 1 2 Format 1.x 7 Format 2.x Format 1.x 8 Format 2.x 3 Format 1.x 9 Format 2.x 4 Format 1.x 10 Format 2.x 5 Format 1.x 11 Guard 149 150 LTE Optimization Engineering Handbook If the mixed format region is employed the total capacity for Formats 1.x and Formats 2.x, it can be calculated as per the equations below: Formats 1.x capacity per PRB = {(n1pucchAn − pucchNAnCs * SF/deltaPucchShift) * deltaPucchShift]/(SF*12)} + roundup (pucchNAnCs/8) ●● Formats 2.x capacity per PRB = nCqiRb * 12 * cqiPerNp + (12‐2‐ pucchNAnCs)*Roundup(puc chNAnCs/8) So the maximum PUCCH capacity reserved in a cell for the normal cyclic prefix case (i.e., SF = 3) is defined by the PUCCH resources for Formats 1.x and Formats 2.x as: ●● MaxPucchResourceSize nCqiRb roundup maxNumOfCce n1PucchAn pucchNAnCs * * deltaPucchShift / 3 * 12 3 / deltaPucchShift roundup pucchNAnCs / 8 Each UE will know the physical resource blocks to be used for transmission of PUCCH in slot ns from the following equations obtained from 3GPP 36.211 nPRB m 2 if m ns mod 2 mod 2 0 UL 1 N RB m 2 if m ns mod 2 mod2 1 Where the variable m depends on the PUCCH format. For formats 1, 1a, and 1b: 2 1 N RB m 1 if nPUCCH nPUCCH sf N cs1 / sf N scRB / PUCCH shift PUCCH shift 2 N RB N cs1 8 c N cs1 / PUCCH shift otherwise Where : N cs1 : Number of cyclic shifts reserved for Formats1. x in the mixed region pucchnanCS 1 nPUCCH : Cyclic Shift assigned to the UE for Formats1. x PUCCH shift 2 N RB : Number of Cyclic Shifts used per RB deltapucchshift : number of PUCCH resources reserved for formats 2 x nCqiRb Whereas for formats 2, 2a, and 2b: m = round down (cyclic shift assigned for Formats 2.x/12) In addition, the number of dynamic ACK/NACK channels can be correlated with the number of scheduled users at the same time. In the actual system, CQI/PMI/RI, SRI information of various reporting period may occur and PUCCH channel RB allocation can also according to the change of number of users, and therefore, PUCCH capacity will change with the adjustment of the resources allocation. Capacity Optimization 4.3.2 PUCCH Dimensioning Example The algorithm to assign slots for Formats2.x always assigns the slot with lower load first as a way of controlling the interference when there is low number of users per slot. Considering cqiPerNp =10 ms/1RB for Formats 2.x would be able to handle up to 120 UEs with 12 cyclic shifts. In a live network, the recommendation is not to use more than six cyclic shifts (i.e., 60 users in this case). This means if it set nCqiRb =1, the PUCCH would not have enough capacity. The solution is to increase nCqiRb =2. In case the interference was still high in Format 2.x, it should be possible to continue increasing nCqiRb so number of users per slot is reduced further. Every user in RRC connected mode requires to be assigned a Format1.x resource for scheduling resource requests. The maximum number of RRC connected users per cell is no less than 100 for the case of 10 MHz bandwidth. In order to accommodate 100 resources for SR, considering a 10 ms SR periodicity (cellSrPeriod =10 ms), there would need to be 10 resources reserved per TTI, that is, n1pucchAn =10. Dimensioning of the number of Format 1.x resources required for acknowledgments from dynamic scheduling is dependent on the maximum number of CCEs that are available for the PDCCH scheduling. For this example, a 10 MHz bandwidth with 1 PDCCH symbol per subframe is assumed. This leads to a maximum of 10 CCEs available. The maximum number of acknowledgments that will be received per TTI depends on the number of UEs that can be scheduled per TTI. This is set by the parameter maxNumUeDl. The maximum value of this parameter depends on the bandwidth and for 10 MHz is 10 UEs. Considering normal cyclic prefix, maxNumUeDl =10 and deltaPucchShift =2 Max Resources for Format 1.x per PRB c * 12 /deltaPucchShift 18 0 Forrmat 1. x Resource range n1 pucchAn Max CCE 10 10 20 Number of PRB required for Formats 1.x Roundup 20/18 2 With two PRBs for Formats 1.x, we would have in total 36 resources per TTI out of which the first 10 are always reserved for Scheduling Requests according to the parameter n1pucchAn and the remaining 26 will be employed for acknowledgments of dynamic Ack/Nack scheduling depending on the position of the first CCE used in the PDCCH scheduling. As it can be seen with the current configuration chosen a total of 10 CCEs are available, which means in theory we could schedule up to 10 users in the DL per TTI with aggregation level 1. In reality, PDCCH will schedule users in UL and DL and with different aggregation levels so with the current configuration it would be difficult to reach the maximum number of simultaneous users per TTI specified by parameter (maxNumUeDl =10). Even if this number would be achieved, the capacity reserved for dynamic Ack/Nacks supports up to 26 users so there would be no problem. In the above configuration the mixed format region was not employed since no further capacity was required for Formats 2.x, therefore, pucchNaCs =0. The total PUCCH region for the above given example would be 4 PRBs: 2PRB for Formats 2.x and 2 PRB for Formats 1.x. With above given configuration: ●● ●● For formats 2.x: m = round down (cyclic shift assigned for Formats 2.x/12) =2 Since nCqiRb = 2, the assigned cyclic shift format by the eNB will always be in the range 0 to 23, which means for that for formats 2.x m = 0 and m = 1 For formats 1.x: 1 1 m if nPUCCH 1 nPUCCH 18 1 otherwise 18 151 152 LTE Optimization Engineering Handbook Knowing that the Format 1.x range is delimited to a maximum value of 20, this means that Formats 1.x will use m = 2 and m = 3. 4.4 Number of Scheduled UEs The user throughput is impacted by the number of users in the cell. If an excessively large number of users have accessed the cell and eNB are exhausted when a UE accesses the cell, the user throughput will be low. The maximum number of active UEs is dependent on the RAN license. In fact, when the maximum number of users of the cell began to restricted, it can be improved by increasing inactivityTimer. Inactivity timer is a system parameter controlling the transition from RRC connected state to idle state. ●● ●● ●● Actively scheduled user: the definition of the number of actively scheduled UEs are the scheduled UEs per TTI by eNB, there is a license feature on the eNB capacity. The No. of UL actively scheduled UEs is dependent on the channel of PRACH, SRS, and PUCCH, the No. of DL actively scheduled UEs is dependent on the channel of PHICH and PDCCH. Connected user: connected user is 3GPP defined concept, a connected user is defined as UE in the state of RRC_connected. UE is considered connected UE, if it has at least one DRB established. When a user is in RRC_connected state, it does not necessarily need to transfer any data. The maximum range for the amount of simultaneous connected users depends on the digital unit hardware, the number of RRC connected users is relevant for the radio net work dimensioning. Usually a single cell can offer no less than 1200 connected users. Attached user: the definition of a connected user is different from a simultaneously attached user. Simultaneously attached users in the EPC include users in both RRC_idle and RRC_ connected states. Another critical key distinction is that connected users is not equal to subscribers in the cell (including detached users). The number of connected users is initially determined by considering the number of subscribers per site and the traffic profiles. Obviously, the more the subscribers the more licenses are required. The type of traffic and the amount of traffic also impacts the dimension ing. User plane activity statistics describe user plane traffic patterns and activity: when and how frequently subscribers are sending and receiving data packets. Figure 4.10 shows that the ratio between number of connected user and number of attached subscribers. 35,0% % of Subs active last week, 61 sec timer Attached RRC connected 30,0% 25,0% 20,0% 15,0% 10,0% 5,0% 0,0% Day 1 Day 2 Day 3 Day 4 Day 5 Figure 4.10 Connected users not more than 15% of attached subscribers. Day 6 Day 7 ratio pf active subscribers during busy hour [%] Capacity Optimization 60% 50% M2M (op1) M2M (op2) PC (op1) PC (op2) iPhone (op1) iPhone (op2) Android (op1) Android (op2) feature phone (op1) feature phone (op2) 40% 30% 20% 10% 0% 2 sec 10 sec 60 sec 3600 sec inactivity timer [sec] Figure 4.11 Connected users vs. RRC inactivity timer. (See insert for color representation of the figure.) The calculation of the number of connected users should consider the below factors: inactivity timer,traffic profile, and other traffic related parameters. The share of connected users depends heavily on the RRC inactivity timer, the lower the value (inactivityTimer,range 10…65535 s, step 1 s, default: 30s), the more connected users can be supported, the UE can switch from the active state to the idle state earlier. However, this will affect the user experience, when the new data service transmission occurs, the UE will fre quently to establish a session. The eNB triggers a RRC connection release procedure upon the expiry of tInactivityTimer. The UE context release request S1AP message is sent to the MME. The eNB sends the RRC connection release message to the UE and then clears all resources related to the UE. Whereas lower values of this parameter results in longer battery time, higher values of this parameter improves throughput and latency performance. Therefore, a good bal ance should be maintained between a higher value and a lower value. Setting parameter tInactivityTimer to values too low results in too frequent releases of UEs, which result in an increased number of RRC set up attempts. The increased number of RRC set ups result in an over all increased signaling load on the network as traffic builds up in a network. Setting parameter tInactivityTimer to values too high results in some UEs using more system resources than needed. This could result in congestion in case of high traffic loads in the network. Figure 4.11 shows the ratio of active subscribers for different terminal types during busy hour depending on the value of the inactivity timer. 4.5 Spectral Efficiency Spectral efficiency is the user data transmission rate over a given radio bandwidth (1Hz). It is determined by multiple factors, including coverage, quality of signal, MIMO, and so on. From eNB available data, it can be estimated the spectral efficiency by modulation and coding scheme used (MCS), network layers overhead, control and signaling overhead and cyclic prefix over head. Modulation and coding scheme are related to the radio signal quality, that is, SINR. Overheads are the radio resources that cannot be used by the user application payload. Control and signaling overhead represents the resources cost for common and dedicated signaling. Cyclic prefix overhead is determined by the format of OFDM symbol transmission. 153 154 LTE Optimization Engineering Handbook According to the spectral efficiency and the spectrum bandwidth, maximum cell throughput is defined with the assumption that there is enough data in the DL buffer to fill up all TTI. User throughput is calculated from the cell throughput and number of active UEs in scheduler. When there is less than one average active UE, the user throughput is equal to the spectral efficiency with the DL bandwidth. CellThpmax Thpuser Spectrum _ Bandwidth Spectral _ Efficiency Spectral _ Efficiency BWDL max 1, Active _ UE DL spectral efficiency can be computed as: Spectral _ Efficiency Tu Ts 1 DOH MIMOMode 15 1 PDCPOH 1 RLCOH . 1 MACOH Ts S BW 1 RTX RLC & MAC Wi Nbits persymboli CRi i 1 Tu is useful symbol time excluding cyclic prefix time. It is the part of the symbol that can be effectively used for data transmission. Ts is 0.5 ms/7 for normal cyclic prefix. DOH is DL over head of RE used for control and signaling information. CR is coding rate estimated using CQI reported from UE and CQI to MCS mapping table. SBW is equal to 15000 Hz. 4.6 DL Data Rate Optimization In LTE, DL throughput is directly correlated with SINR. Typically, the UE has an algorithm to report CQI to the eNB depending on the SINR measurement. DL UE throughout increases as CQI increases or MIMO usage increases, and as DL iBLER decreases or UE‐eNB distance decreases. The general troubleshooting strategy is described in the following along with different factors responsible for poor throughput. ●● ●● ●● ●● ●● ●● ●● Excessively high BLER (bad coverage) DL interference (bad CQI) MIMO parameters Scheduling algorithm Low demand CQI reporting frequency Other (VSWR, backhaul capacity) Analysis flow for DL throughput investigation if backhaul or other physical issues work well. ●● ●● ●● ●● CQI and 64/16QAM ratio: average CQI and the ratio of 64QAM samples and RI (rank indi cator, decides how many codewords is used for the data transmission) indicates DL SINR status. Average CQI should be high (>10%), the ratio of 64QAM sample should be high (>10%). TM modes: MIMO (tm3) versus TxD (tm2) versus SIMO (tm1). UE scheduling percentage of TTIs (how often is the UE scheduled). PRB(DL) and PDCCH utilization: High PRB and PDCCH utilization would impact the DL throughput. Capacity Optimization ●● ●● ●● DL latency and RLC retransmission: High value of DL latency (>9 ms) and RLC retransmis sion (>1%) would impact DL throughput. Power limited UE and No. of A2 events: High occurrence of transport block PWR restricted and the counter of bad coverage report indicates poor DL coverage. RRC connected users: DL throughput would reduce with increase in number of con nected users. Steps to troubleshoot low DL throughput ●● ●● ●● ●● ●● ●● ●● ●● ●● ●● Check to see the RF condition Check to see power setting is correct (40 watt; 100% parameter discrepancies) Check to see dl64QAM license is activated Check to see No of Tx antenna and No of Rx antenna, both of which should be 2 Check to see if vswrSupervisionActive = true for both branches A and B Check to see the drive tester is using the right application server/port# when doing driving test Check to see the actual TX/RX powers (check RRU and connections) Check counters from vendor OSS Check whether the number of users in the cell is excessively large. If an excessively large number of users have accessed the cell and eNB are exhausted when a UE accesses the cell, the user throughput will be low. Check HSS profile that there is no APN‐AMBR or UE‐AMBR limitation, which can be checked in the message of S1AP_INITIAL_CONTEXT_SETUP_REQUEST. The default value of AMBR is more than 150 MB. Still the default QCI should be set correctly, QCI should be set „Non‐GBR,“ l the value can be 6, 8, 9, should not be 5 (QCI5 is used for IMS signaling, QPSK) or 7 (QCI7 is used for UM) as it can’t achieves the peak rate throughput. IMSI ........................ CTXID ....................... PDN TYPE .................... AP NAME ..................... QOS CLASS IDENTIFIER ........ PRIORITY LEVEL .............. PRE-EMPTION CAPABILITY ...... PRE-EMPTION VULNERABILITY ... VPLMN DYNAMIC ADDRE ALLOWED . PGW ALLOCATION TYPE ......... CHARGING CHARACTERISTICS .... AMBR DOWNLINK ............... AMBR UPLINK ................. ●● ●● ●● ●● xxxxxxxxxxxxxxx 10 BOTH test-ap-2 6 2 DISABLED DISABLED YES DYNAMIC NORM 20000 65535 Check if there is packet loss in transport network Change to Linux FTP server and optimize UE TCP settings Check whether relative license information is incorrect and if the license has not expired Check whether the traffic volume to the eNB is insufficient. A common reason for the insuf ficient input traffic volume is a bottleneck transmission bandwidth at an intermediate node. One of the efficient way of poor throughput troubleshooting is by OMC counters as listed below. ●● ●● ●● UL Interference on PUCCH/PUSCH: pmRadioRecInterferencePwrPUCCH/pmRadio RecInterferencePwr, for good throughput in UL_RSSI value should be low ‐105 dBm SINR of PUCCH/PUSCH: pmSinrPucchDistr/pmSinrPuschDistr PRB Utilization in DL/UL 155 LTE Optimization Engineering Handbook ●● ●● ●● ●● ●● ●● RLC ACK/NACK: pmRlcArqUlack/pmRlcArqUlNack, the total number of successful RLC PDU transmissions (ACKs/NACKs) in the UL direction, for good throughput in UL ‐UL RLC NACK ratio should be low CQI distribution in DL: for good throughput in DL‐ average CQI should be high (>10) Transport block PWR restricted Scheduling activity per cell in UL and DL Rank distribution MIMO/SIMO: for good throughput in DL‐ high samples of MIMO Rank 2 are needed Number of A2 events (UE in poor coverage) :pmBadCovEvalReport 4.6.1 Limitation Factor The focus of this kind of problem is the identification of bottleneck of the data rate. That is to say, whether it is limited because of the air interface, core network, or transmission? Usually, low data rate region is a weak coverage, no dominant cell area, so for the optimization of a region, the main coverage area, good RSRP, RSRQ, SINR is the optimization goal that must be achieved, the average user rate and the changes of number of users and RF environment are closely related. Actually, the analysis of low DL throughput is very complex, an example from Figure 4.12 that can be seen that low throughput is uncorrelated with DL SINR measured. CQI and RI provides the SINR/antenna layer reception reports from the UE point of view. Understanding the relationship between chosen MCS, assigned PRBs, and assignable bits in the scheduler are important for sorting core network issues/UE issues from air interface issues. Scheduling percentage means the amount of TTIs (typically measured per second) that the UE was scheduled. This is also related to the resources allocated for PDCCH (control channels). In some networks, average UE DL rate target is 1Mbps, and the vast majority of users should achieve the rate. Low backhaul capacity likely will have an impact on the rate, especially for high load cell. In addition, the settings on the server and tests computer must check to ensure that it will not become the bottleneck of the rate limiting. PDSCH phy Throughput [kbps] 156 80000 70000 60000 50000 40000 30000 20000 10000 Low throughout uncorrelated with DL SINR measured 0 –6 –3 0 3 6 9 Figure 4.12 DL throught versus RS SINR (2.6GHz). 12 15 18 21 24 UE RS SINR (dB) Capacity Optimization In conclude, the impacted factors of low throughput includes UL/DL bandwidth, signaling quality (CQI, SINR, MCSs), number of DL bearers, QoS (QCIs, GBR), transmission mode (SISO, MIMO, TxDiv), backhaul delay (RTT‐affecting on ACK‐ed traffic), UE capacity, UL power control (TPC command), scheduler, and so on. Before investigating any throughput issues, it is best to rule out the most obvious and basic issues that might affect end‐user throughput. Some essential checks are required, for example, network basic troubleshooting, radio network parameters, PC/server settings, UE categories, UE subscriber profile,1 eNB parameters, and network enabled features, and so on. Sometimes laptop specification, FTP server configuration, or Linux TCP setting can impact throughput (processors, memory, USB bus, HDD speed, plugged into AC power, etc.). MTU settings in PC (can cause fragmentation on S1, especially with IPSEC on GTP‐U) can also impact throughput. 4.6.2 Model of DL Data Throughput The model of LTE DL throughput and its influencing factors can be categorized as time domain, radio domain, and efficiency domain. Time domain, the main factors includes PDCCH DL grant, it needs to check the number of scheduled per TTI if it gets enough scheduling. 1) Cell loading and number of active UEs. 2) FTP server performance. 3) Check data amount to see if it has been injected enough DL data to eNB to schedule, if there’s insufficient input traffic volume, throughput is definitely lower than the max value. 4) Compress mode for inter‐frequency measurement. Radio domain, the main factors includes number of scheduled RB per subframe, which can be got from OMC counters. 1) 2) 3) 4) 5) Cell loading and number of active UEs. FTP server performance. eNB alarm, for example, VSWR alarm, RF alarm or RRU low power alarm. Parameter of PA/PB is not reasonable; power amplifier efficiency is not 100%. Interference of some PRB. The factors of efficiency domain include RSRP, CRS_SINR, CQI/RI, DL MCS, TBS, coding rate, transmission mode, and re‐transmission rate. Figure 4.13 shows the relation of PDSCH throughput versus RSRP (left) and SNR (right). Figure 4.13 PDSCH throughput versus RSRP and SNR. 1 UE subscriber profile consists of MSISDN number, aggregate maximum bit rate (AMBR), max requested bandwidth in downlink/uplink, RAT frequency selection priority, APN configuration profile. 157 Probability LTE Optimization Engineering Handbook 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% QPSK 16QAM 64QAM 3 4 5 6 7 8 9 10 11 12 13 14 15 Reported CQI 45 40 35 % of samples 158 30 25 20 15 10 5 0 QPSK 16QAM 64QAM Modulation format Figure 4.14 Modulation probability versus CQI in a live network. The eNB needs knowledge of the CQI conditions of DL transmission to a UE in order to select the most efficient MCS/PRB combination for a selected UE at any point in time. Figure 4.14 shows the probability of different modulation schemes in DL versus different reported CQI values. Table 4.6 shows the ideal layer1 DL bits per scheduling block for different antenna/radio configurations. In conclude, the following radio network parameters directly impact end‐user throughput: channel bandwidth, number of used Tx/Rx antennas, transmission mode, number of OFDM symbols used for PDCCH, maximum transmission power, p0NominalPucch (some UEs need this to be increased or ACK/NACKs are not received successfully on PUCCH), and p0NominalPusch (some UEs need this to be increased from default or lots of errors seen on PUSCH), and so on. 4.6.3 UDP/TCP Protocol For LTE throughput optimization, if FTP throughput is bad, it is often difficult to pinpoint if this is caused by radio, transport, or TCP setting problem, as the engineer can test with UDP to check as UDP shows higher throughput because it is not as affected as TCP. UDP tests were subsequently performed to validate results and better represent achievable throughput perfor mance. If UDP throughput is good, then it is a transport or TCP setting problem. When trans port had problems (discarded packets observed), it would cause retransmissions and TCP congestion control to reduce throughput. If UDP throughput is also bad, then it’s probably a radio problem. It could be either interference or parameters settings are incorrectly, also might be a problem with HSS profile (Figure 4.15). UDP is a simple data transmission protocol with 8 byte header based on IP that does not pro vide guarantees for transmission reliability nor for data integrity. As it maintains a connectionless Table 4.6 DL bits per scheduling block. dlCyclicPrefix = 15 KHz => 7 OFDM symbols T× Diversity 2×2 MIMO Resource Elements per Resource Block 84 168 RE per Scheduling Block (2 × RB) 168 336 RS RE (per RB)/RS RE (per Scheduling Block) Control Region Size in OFDM symbols 8/16 1 RE per CRS (OFDM*12 − 4 RS Tx), (OFDM*12 − 8 RS MIMO) 2 16/32 3 1 2 3 8 20 32 16 40 64 Tot Num RE per Scheduling Block available for PDSCH (best case w/o SCH/BCH) 144 132 120 288 264 240 Bits per Scheduling Block ‐ QPSK (2) 288 264 240 576 528 480 Bits per Scheduling Block ‐ 16QAM (4) 576 528 480 1152 1056 960 Bits per Scheduling Block ‐ 64QAM (6) 864 792 720 1728 1584 1440 20 MHz Max Theoretical L1 Thrpt (Mbps) 86.4 79.2 72 172.8 158.4 144 Tot Num RE per Scheduling Block available for PDSCH (worst case with SCH/BCH in SB) SCH = 24, BCH = 4 × 12 − 4 per CW 76 64 52 152 128 104 Bits per Scheduling Block ‐ (QPSK) 152 128 104 304 256 208 Bits per Scheduling Block ‐ (16QAM) 304 256 208 608 512 416 Bits per Scheduling Block ‐ (64QAM) 456 384 312 912 768 624 DL throughput, UDP 100 –75 –80 80 70 –85 60 –90 50 –95 40 30 20 Average of Phy DL TP[Mbps] RSRP [dBm] CINR [dB], DL PHY tput [Mbps] 90 –100 Average of Serv Cell CINR –105 Average of Serv Cell RSRP 10 –110 11 :0 11 7 :0 11 7 :0 11 7 :0 11 7 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 11 8 :0 8 0 DL throughput, FTP –75 90 Average of Phy DL TP[Mbps] 80 Average of Serv Cell CINR 70 Average of Serv Cell RSRP 60 –80 –85 –90 50 40 30 20 10 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 10 :5 9 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 11 :0 0 0 Figure 4.15 UDP versus TCP, same drive route for both. –95 –100 –105 –110 RSRP [dBm] CINR [dB], PHY DL tput [Mbps] 100 160 LTE Optimization Engineering Handbook state it does not provide any reliability in terms of ordered delivery and duplicate protection and does not also provide any means of congestion control. Thus, for UDP traffic, no feedback channel is necessary. This means that UDP data transfer in DL in principle does not need UL traffic (of course, some UL traffic is necessary for lower layer functions). Hence, measuring the DL data throughput using UDP may be possible even with a bad UL quality. TCP is a much more complex protocol than UDP, since it provides a low delay and reliable byte stream based on IP. TCP is a connection oriented, point‐to‐point, reliable transport pro tocol. TCP has its built in congestion management algorithm such as fast retransmission and fast recovery. The minimum TCP header is 20 bytes. This is achieved by sending an acknowl edgment from the receiver to the sender for each TCP segment that has been delivered on the receiver side to the higher protocol layer. In the current context this means that a TCP data stream in DL requires a UL connection with sufficiently good quality, and using TCP reduces the throughput due to acknowledgments and retransmissions. The maximum throughput that can be reached by TCP is given by the “TCP window size” (wnd, iperf parameter –w) divided by the round trip time (RTT). This is because the TCP sender stops sending data if the acknowl edgment for the last sent segments is missing. Assuming an RTT of 20 ms (=1/50 s), the TCP window size must be set to at least x/50 bytes, if x is the expected TCP throughput in bytes/s. Example: For reaching 24 Mbps (= ~3 Mbyte/s) throughput on a connection with 20 ms RTT the TCP window size must be at least (3 Mbyte/s) * 0.02 s = 60 Kbyte. In order to avoid bandwidth limitation because of an insufficient TCP window size, it is rec ommended to set the TCP window size to 256 Kbyte or more. Besides, MTU (maximum transmission unit) is another factor that can cause fragmenta tion on S1 interface and influence the throughput. If there is a datagram needs to be trans mitted at the IP layer, and the datagram is larger than MTU, then the IP layer needs to segment the datagram into smaller fragments that are smaller than MTU. To improve effi ciency, it is necessary to avoid IP segmentation and reassembly during PDCP <‐> GTP translation, as well as to set MTU to a value as large as possible. Usually, MTU should not be greater than 1450 bytes. For example shown in Figure 4.16, compared with MTU=1400B (right), for MTU=1500B settings (left), the same UE cannot reach max FTP throughput, the FTP throughput = ~41 Mbps, MTU = 1500Bytes Figure 4.16 FTP throughput with different MTU size. FTP throughput = ~87 Mbps, MTU = 1400 Bytes Capacity Optimization throughput fluctuates unpredictably. The reason is probably fragmentation/reassembly somewhere in the network, when sets MTU=1360Bytes by default, there is usually no IP fragmentation even with S1 IPsec. In summary, TCP has shown to be non‐optimum in wireless networks as it was primarily designed in the 1990 for wired network, thus wireless network have however evolved to adapt slowly to TCP. It is worth mentioning, because the UL data sessions are mostly very small and due to a TCP slow start, they do not reach the maximum achievable throughput levels before session end. 4.6.4 MIMO With multiple antennas at both transmitter and receiver, it is possible to achieve spatial mul tiplexing or beamforming. eNB automatically selects the most appropriate diversity, beam forming or depending on radio conditions. The maximum number of layers that can be created depends on the radio channel characteristics and the number of transmit and receive antennas and the channel rank. Dynamic open/close loop MIMO, open/close loop TxDiv, and open/close loop spatial multiplexing (SM) switching decision is based on RI and CQI. The parameters of RI and CQI are the thresholds that are used by DL scheduler to decide the transmission scheme switching decision between Rank1 beam‐forming, TxD, and spatial multiplexing mode. 4.6.4.1 DL MIMO This section focuses on the DL throughput optimization via adjusting the MIMO mode switching parameters. In the adaptive MIMO mode, the switching between TxDiv and spatial multiplexing (SM) is governed by the CQI and RI feedback from the UE. In order to maximize cell capacity it is desirable for the UE to employ SM when it is located under good radio conditions (typically SINR> 20 dB) and to switch to TxDiv when radio conditions are not so favorable. Figure 4.17 plots the expected layer 2 throughput for different SINR values derived from link level simulations for EPA5 for a 10 MHz bandwidth. For low SNRs, TX diversity gives slightly better throughput than SM, since the power from both eNB TX antenna is aggregated to decode one code word. This is in line with theoretical simulations. For high SNR, SM gives much higher throughput than TX diversity, since two DL Throughput vs DL C(I+N) Thresholds 70000 Throughput [kbps] 60000 50000 TxDiv => SM 40000 30000 16QAM => 64QAM 20000 QPSK => 16QAM 10000 0 –6 –4 –2 –1 1 2 4 6 7 9 11 13 15 16 19 26 28 31 32 DL C(I+N) - dB Figure 4.17 DL throughput versus SINR and switching SINRs based on simulations. 161 LTE Optimization Engineering Handbook code words are sent at the same time. Although the two cases are static TX diversity and SM, the switching point could be around SNIR = 10 to 13 dB for dynamic adaptation of MIMO mode (open loop or closed loop MIMO) as shown in Figure 4.18. Optimum switching point is suggested to be at around SNIR = 18 dB for UEs at medium speed of 5 to 30 km/h. Antenna correlation and antenna power imbalance are the key factors affect the MIMO per formance. Both of these factors increase the SNR difference between streams. In order to mini mize antenna correlation needed to consider practical deployment, power imbalance between antenna branches were discussed in Chapter 3. Figure 4.19 shows a measurement example with a real UE and antenna correlation artificially altered with a fading simulation, which are the terms of different degrees of correlation. In order to switch from SM to TxDiv, it is required for the CQI‐ or RI‐averaged values to be below certain thresholds (RiThreshold2 and CqiThreshold2) defined for these two metrics. In order to switch from TxDiv to SM it is required for the CQI and RI averaged values to be above 25 Throughput (Mbps) SM Throughput (Mbps) TxDiv Throughput (Mbps) 20 SM Throughput (Mbps) 15 TxDiv Throughput (Mbps) 10 5 SINR (dB) 0 –4 –2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 Figure 4.18 DL throughput versus SINR. MIMO Usage Probability 100% 80% 60% 40% 20% 27 25 23 21 19 17 15 11 13 9 7 5 3 1 –1 –3 –5 0% –7 162 SNR CQI SM SM CqiThreshold1 CqiThreshold2 Time RI Filtered cqi, ri RiThreshold1 RiThreshold2 Time Figure 4.19 MIMO mode versus measured SNR in unloaded network. 28 30 32 34 Capacity Optimization a certain thresholds (RiThreshold1 and CqiThreshold1) defined for these two metrics. The recommended values for these parameters involved in the open loop dynamic MIMO are: ●● RiThreshold1 = 1.8, RiThreshold2 = 1.4, CqiThreshold1 = 11, CqiThreshold12 = 9 The thresholds defined for the switching are based on the CQI/RI, allow for hysteresis. Depending on the difference between these up and down thresholds the switching between the modes will be faster or slower. 4.6.4.2 4Tx/4Rx Performance It is known that four‐antenna cell application has been widely used in networks, which can be used to extend the cell coverage and increase user data rates. The main motivation for 4TX/4RX is enhanced coverage and enhanced cell‐edge data rates. Compare with 2TX/2RX and 4TX/4RX, the average improvement in the user data rate was +50% in UL and +25% in DL. The data volumes increased even more because the cell coverage area also increased. The typical antenna solution for 4TX/4RX is XX‐polarized antenna in a single radome. The number of antennas is not increased compared to 2TX/2RX but the antenna just gets slightly wider, which is shown in Figure 4.20. Another antenna option would be two separate X‐polar ized antennas, which would be more difficult for the site solution. The simulations and the field measurements indicate that XX‐polarized antenna gives better performance. 4Rx can improve UL coverage for all channels, the cell average bit rate and primarily the cell‐edge bitrate. With 4Rx, interference rejection combining (IRC) generally improves the per formance when 4‐way Rx diversity is used. Figure 4.21 shows that due to 4RX, cell‐edge is extended with MIMO 4x2 versus 2x2, and cell‐edge UL throughput gets a better performance due to the antenna spatial diversity and up to 3 dB of coverage improvement. With the IRC feature activated, the system dynamically switches between the combining methods IRC and maximum ratio combining (MRC) depending on interference situation. 4.6.4.3 Transmission Mode Switch 3GPP Rel 9 defines multiple transmission mode, which is controlled by RRC; mode switching/ selection is higher layer configured and on a slow basis (several seconds) in order to ensure optimum UE DL throughput and performance. The transmission mode is selected based on wideband SINR and UE speed. The CQI report is firstly mapped to SINR, then sum over streams (in case of rank 2) and finally average over frequency. The beamforming weighting factor is based on periodic SRS (10 ms), SRS Rx_SINR impacts to beamforming weighting factor is considered. Beamforming performs good enough for high speed even with 350 km/h coverage due to intra‐polarization beamforming. This still can give higher gain than broadcasting and inter‐cell interference of beamforming is more Figure 4.20 Antenna options for 4TX4RX. Single XX-pol antenna Two X-pol antennas 163 LTE Optimization Engineering Handbook DL Tput (Mbps): 2x20 W DL Tput (Mbps): 4x10 W DL Tput DL Tput (Mbps): 4x20 W 60 50 40 30 20 10 0 Figure 4.21 DL and UL throughput versus pathloss. DL Tput (Mbps): 4*20 W DL Tput (Mbps): 2*20 W DL Tput (Mbps): 4*10 W –75 –80 –85 –90 –95 –100 –105 –110 –115 –120 –125 Path Loss (dB) UL Tput (Mbps): 2RX 25 UL Tput 164 UL Tput (Mbps): 4RX DL Tput (Mbps): 4RX 20 15 DL Tput (Mbps): 2RX 10 5 0 –75 –80 –85 –90 –95 –100 –105 –110 –115 –120 –125 Path Loss (dB) robust to FSS performance. But TM3 provides higher peak rate due to more robust to SRS channel estimation error and SRS capacity problem. TM3/TM8 inter‐mode switch is based on the DL SINR estimation in eNB. TM3 is used in good RF condition to get better DL throughput. TM8 is used in poor and middle RF condition to get better throughput by utilizing beamforming gain. But in rapid fading channel, TM8 may have worse performance than TM3 as shown in Figure 4.22. From Figure 4.22, it can be seen that TM3/8 switch has lower throughput than TM3 and TM8 below SINR 23 dB and high DL BLER in high SINR environment in TM8. 4.6.4.4 UL MU‐MIMO Contrary to DL, 3GPP Rel 8 and 9 do not standardize single user MIMO in UL, UE is able to transmit only one stream of data, instead multi‐user MIMO (or virtual MIMO, V‐MIMO) is supported, which allows pairs of UEs with appropriate radio conditions to be scheduled on the same time and frequency radio resources. Usually with SINR threshold and DoA threshold, UEs can be filtered to be paired. The pairing criterions are based on radio conditions of individual UEs and potential pairs (othogonality), pairing candidates are chosen from among the UEs that are to be scheduled in the same TTI. Advanced eNB receiver is needed to be able to separate data streams from MU‐MIMO UEs. Othogonality between the UEs got from the SRS soundings is obtained from the channel coefficients of UE pairs. Orthogonality is a wideband value. Calculation of the channel vec tor is done in the same way as long term beamforming vector (hi, hi are also wideband values). Oij 1 * hUEi hUEj hUEi 2 hUEj Only the UE pairs with wideband orthogonality metric above the threshold can be considered for pairing. UE pairs with low orthogonality will interfere one another too much, as MU‐MIMO receiver implementation does not cancel the interfering UE. Capacity Optimization Figure 4.22 DL throughput versus transmission mode. DL PHY throughput (kb/s) 80000 70000 60000 50000 40000 30000 20000 10000 0 –17 –14 –11 –8 –5 –2 1 4 7 TM38 10 13 16 19 22 25 28 TM3 TM8 DL BLER (%) 20 Fast fading 15 10 5 0 –17 –14 –11 –8 –5 –2 1 4 7 10 13 16 19 22 25 28 In single‐user mode preliminary MCS is calculated based on the wideband SINR measure ment provided by physical layer. UEs that are scheduled in MU‐MIMO mode will be assigned with MCS lower than in single user mode. Assuming SINRSIMO(i) and SINRSIMO(j) as the UL wideband SINR of UE i and UE j before pairing respectively, the UL wideband SINR of UE i after pairing SINR MU‐MIMO(i) and SINR MU‐MIMO(j) are: SINRMU MIMO i SINRMU MIMO j 1 Oij * SINRSIMO j 1 SINRSIMO j 1 Oij * SINRSIMO i 1 SINRSIMO i * SINRSIMO i ; * SINRSIMO j ; The scheduling metric relative SINR for the whole UE pair is calculated by averaging over two relative SINR of UEs in the UE pair. It notes that when both of the MU‐MIMO multiplexed TB shall be re‐transmitted, they shall not be multiplexed instead of with non‐overlapped RBs allocation. 4.6.5 DL PRB Allocation and Utilization Mechanism In order to reduce the inter‐cell interference in low load scenarios, an enhancement of the resource utilization optimized strategy in live network is introduced. The start index of resource 165 LTE Optimization Engineering Handbook allocation is based on ICIC feature for decreasing the PRB collisions between cells as shown in Figure 4.23. In this way, cell‐edge users (cell‐edge/non‐edge classification is done using the event A3) are prioritized for allocation close to the start index. Usually, an operator starts the PRB allocation at the start index with “cell‐edge VoIP” UEs, continuing with “cell‐edge best effort” UEs, and finally continuing with “non‐edge” UEs. After change PRB start position offset, the expected gain in non peak hour is obvious, but not in busy hour. From Figure 4.24, it can be seen that in a non‐peak hour, a 194 kbps improvement (11:00‐16:00) for DL user throughput that there is no changes in a BH. In live network, trying to avoid high PRB utilization is necessary. There are very limited means of reducing of PRB utilization. However, some of these methods can be tried on case‐by‐ case basis. ●● ●● ●● ●● Traffic offload to less utilized neighboring cells Reduce control channel resources (Before that check PDCCH utilization) Add bandwidth as bandwidth increase would increase number of PRBs Reduce inactivity timer value so that inactive user can be released early. PRB Collision 0 0 Parameter Current dlConfigurableFrequencyStart 0 0 0 0 0 0 Starting point for allocations Configured 0 0 0 0 0 Proposal S1 = 0, S2 = 34, S3 = 67 7000 400 6000 350 5000 300 250 4000 200 3000 150 2000 100 1000 50 No Change RRC User Active User DL 0 3/22 11:00 3/22 12:00 3/22 13:00 3/22 14:00 3/22 15:00 3/22 16:00 3/22 17:00 3/22 18:00 3/22 19:00 3/22 20:00 3/22 21:00 3/22 22:00 3/22 23:00 3/15 11:00 3/15 12:00 3/15 13:00 3/15 14:00 3/15 15:00 3/15 16:00 3/15 17:00 3/15 18:00 3/15 19:00 3/15 20:00 3/15 21:00 3/15 22:00 3/15 23:00 0 DL User Throughput DL Cell Throughput Figure 4.24 Throughput gain after PRB start position offset changed. Throughput (kbps) 450 67 34 0 34 67 0 Figure 4.23 PRB start position offset. User 166 67 34 Change parameter-PRB start position offset Capacity Optimization UL average PRB utilization 100 90 80 DL average PRB utilization 70 60 50 40 30 20 10 0 0 200 400 600 800 1000 1200 1400 1600 Figure 4.25 Average PRB utilization. It is worth noting that UL PRB utilization grows very rapidly compared to the DL because there is no MIMO or 64QAM in UL, therefore, the UL efficiency is much less compared to the DL. From Figure 4.25 it can be seen that for 400Kbps throughput, ~20% PRB utilization in DL and ~90% PRB utilization in UL. 4.6.6 DL BLER There is a strong correlation between high BLER and low throughput, so it is clear that the issue is purely in RF, not an issue for any backhaul/transport problems. In most cases, ratio of retransmission is the major impact factor for throughput, BLER (block error rate) is needed to check in case of problem. BLER loop convergence algorithm provides the link adaptation in order to meet the target BLER. DL and UL link adaption are targeting to sustain certain BLER for the first transmission. iBLER (initial BLER) can be expressed as: iBLER f TBS , coding rate , modulation, SINR BLER target denotes the initial BLER target that is used to determine MCS together with CQI. BLER target is defined by parameters dlTargetBler and ulTargetBler. Default value for both parameters is 10%. BLER target could be changed and optimized for different radio envi ronments. If the BLER is higher than 10%, the channel condition is poor and will result in low throughput. BLER is a metric that eNB controls within a configurable target (DL initial BLER target) to optimize the RF operation. Too‐high BLER performance beyond the target often indicates some aspect of RF problems. In live network, DL iBLER is below 10% when the average CQI is above 4. DL UE throughput is decreasing as DL iBLER increases, DL iBLER is decreasing as average CQI increases (i.e., RF improves), which is shown in Figure 4.26. The expected range of iBLER is between 5% and 15%. If the performance is significantly (e.g., more than double) poorer, it may be usually caused by several reasons: ●● UE RF estimate accuracy issue with CQI report – for example, CQI reports are much inflated comparing with UE’s actual RF condition 167 LTE Optimization Engineering Handbook 80 70 Average of DL iBLER DL iBLER distribution 60 DL iBLER (%) 50 40 30 20 10 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 CQI Index 16.00 80 14.63 14.00 DL UE Throughput (Mbps) 168 70 Average UE DL throughput UE DL throughput distribution 12.00 60 10.54 10.00 50 8.70 7.51 8.00 40 6.21 6.00 30 5.01 4.16 3.50 4.00 3.05 20 2.60 2.18 2.00 1.78 1.42 10 1.01 0.72 0.58 0.34 0.18 0.00 0.12 0.00 0.00 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 0 100 DL iBLER (%) Figure 4.26 DL iBLER versus CQI and DL UE throughput versus DL iBLER. ●● ●● UE’s tx power or RF condition is poor such that the ACK/NAK it sends over UL is too weak for eNB to detect UE’s RF is very poor such that it misses frequently the DL grants forcing eNB to retransmit the packets. High BLER caused by inter‐cell interference In this scenario, it have really good CINR value and really good RSRP value, but inter‐cell interference where in the reference signal (RS) from an adjacent sector interferes with the traf fic resource blocks in the serving sector can be seen. The signature to look for is high RSRP, high CINR, and very high BLER, greater than 10%. As a result of the link adaptation deployed Capacity Optimization Figure 4.27 BLER target issue example. in LTE, the eNB would start reducing the MCS (modulation and coding scheme) given to the UE. DL residual BLER One of the possible issues that could lead to the decrease of throughput is related with the excess of residual BLER. In order to evaluate this possible issue, it is advised to check the DL residual BLER distribution that it is based on measurement of number of MAC PDUs that are not acknowledged by the UE after the maximum allowed number of transmissions. The DL residual MAC BLER is distributed by different intervals as it follows: ●● ●● ●● ●● ●● DL BLER low (from 0% to 1%) DL BLER low‐medium (from 1% to 2%) DL BLER high‐low (from 2% to 5%) DL BLER high‐medium (from 5% to 10%) DL BLER high (above 10%) With the intervals above, means that it must have the samples as much as possible on the low BLER classification. An example of incorrect setting of BLER target is shown in Figure 4.27. In the site A base station, the cell download rate is too low (20Mbps), MCS is also low, but RSRP and DL SINR is pretty good, and 16QAM ratio is high, while BLER is low. By preliminary analysis, the phenomenon is not caused by interference. It is found that DL BLER target value is configured at 1%, so that scheduler will decrease MCS to guarantee BLER target. For FTP service, it does not require such a high BLER requirements, and this will lead to not be able to use higher‐order MCS, which leads to low download rate. After modified DL BLER target from 1% to 10%, DL data rate can achieve above 35Mbps. 4.6.7 Impact of UE Velocity The DL throughput is impacted by UE velocity. The network performance influence by UE velocity relative to eNB is shown in Figure 4.28, negative velocity means that UE is moving closer toward eNBs, and positive velocity means thet UE is moving farther away from eNBs. The UE velocity group X stands for group [X, X+10], UE mobility may help or harm RF condition (i.e., CQI), the UE experiences and thus DL UE throughput. 169 LTE Optimization Engineering Handbook DL UE Throughput, CQI, DL iBLER v.s UE Velocity Reative to eNB 600000 16.00 547349 14.00 500000 11.40 12.00 10.59 11.97 11.11 10.20 10.00 8.43 10.91 10.50 10.59 10.72 10.67 10.57 10.61 9.95 9.74 400000 8.80 300000 8.00 6.00 200000 Sample Count DL UE Throughput(Mbps), CQI, DL iBLER (%) 170 4.00 100000 2.00 254 739 1268 2614 4317 8690 –80 –70 –60 –50 –40 –30 21891 29756 –20 –10 46431 12438 5890 3816 2207 1438 479 20 30 40 50 60 70 0 0.00 0 10 UE Velocity group (MPH) Average of UE_DL_Thp(Mbps) Average of Avg_CQI_Reported(Overall) Average of DL iBLER(%) Count of UE_DL_Thp(Mbps) Figure 4.28 Impact of UE velocity (relative to eNB). From Figure 4.28, it can be seen that as UE velocity increases, average UE DL throughput and CQI decreases a bit, while DL iBLER increases. 4.6.8 Single User Throughput Optimization This part is focused on the radio analysis to improve the DL single user throughput. Single‐user throughput optimization needs UE trace logs to spot the problem. Here we pre sent an example of trace parses. This example is a modified output of DL, which summarizes a number of traces onto one line, like transmission mode, MCS (modulation and coding scheme), PRB occupied, TBS (transport block size), assignable bits, HARQ, CQI/RI, DL BLER, and so on. Note how RI=2 is reported, then transmission changes from TxDiversity to MIMO. HARQ ACK/NACK refers to the transmission four subframes earlier. sfn | sf |mode |dlModul | mcs1 |mcs2 280 | 4 |TxDi | 64QAM | 16 | 0 280 | 5 | | | | 280 | 6 |TxDi | 64QAM | 18 | 0 280 | 7 |TxDi | 64QAM | 18 | 0 280 | 8 |MIMO | 16QAM | 13 | 13 280 | 9 |MIMO | 16QAM | 13 | 13 281 | 0 |MIMO | 16QAM | 12 | 12 281 | 1 |MIMO | 16QAM | 13 | 13 281 | 2 |MIMO | 16QAM | 13 | 13 281 | 3 |MIMO | 16QAM | 13 | 13 281 | 4 |MIMO | 16QAM | 13 | 13 281 | 5 |TxDi | 16QAM | 30 | 0 281 | 6 |MIMO | 16QAM | 30 | 30 281 | 7 |MIMO | 16QAM | 30 | 30 |prb |Ndf |Tbs1 |Tbs2 |AssBits | Harq | dlBler | cqi |ri | 25 | Y |7736 | 0 |8771784 | | | 11 | 2 | | | | | |A | 0% | | | 25 | Y |7992 | 0 |8764088 | A | 0% | | | 25 | Y |7992 | 0 |8756144 | A | 0% | | | 25 |Y Y |5736 | 5736 |8748192 | A | 0% | | | 25 |Y Y |5736 | 5736 |8736760 | | | | | 25 |Y Y |4968 | 4968 |8737384 | A | 0% | | | 25 |Y Y |5736 | 5736 |8763568 | N | 0% | | | 25 |Y Y |5736 | 5736 |8776208 | N N | 2% | | | 25 |Y Y |5736 | 5736 |8800856 | N N | 4% | | | 25 |Y Y |5736 | 5736 |8825504 | N N | 6% | | | 25 | N |7992 | 0 |8862160 | N N | 8% | | | 25 |N N |5736 | 5736 |8862200 | N N | 10% | | | 25 |N N |5736 | 5736 |8862200 | A A | 10% | | | | | | | | | | | | | | | | | Capacity Optimization 4.6.8.1 Radio Analysis – Assignable Bits If the UE is sending with high CQI (in the range 10 to 15) and RI=2 but throughput is still very low, then the next check should be assignable bits. Assignable bits means the amount of data in the DL buffer available for the scheduler to schedule for this UE. A classic symptom of low assignable bits is that the UE is scheduled with a high MCS but a low number of PRBs. This scheduler strategy maybe attempts to send with the highest possible MCS and least number of PRBs so that leftover PRBs could be assigned to another UE, or, another symptom is that the UE is not scheduled every TTI (and nothing else is available to schedule). Possible causes for low assignable bits include data received from core network is not enough to fill the RLC buffers in eNB and RLC status messages are not being received fast enough and RLC buffers are full. RLC status messages are sent between eNB and UE to inform about lost RLC packets. It needs to check that non‐TCP based traffic is not being sent with too large packet size. For iperf‐based traffic, the recommended MTU size is 1360 bytes (default is 1470). For RLC status messages not being received fast enough, it needs to check that RLC discards which will trigger TCP congestion control and lower throughput (discards on UDP traffic will not affect throughput). 4.6.8.2 Radio Analysis – CFI and Scheduling SIBs require PDCCH resources, typically SIBs consume four or eight CCEs of PDCCH resources. If a UE is in good SINR conditions, the scheduler may allocate only one CCE for that UE. In that case, because of limited positions in PDCCH, it is quite likely that a PDCCH colli sion occurs especially in low system bandwidths. If a UE is in bad SINR conditions, the scheduler may allocate a large number of CCEs for that UE (two or four or eight CCEs) depending on the configured CFI there may only be common search space available or it may still collide with other PDCCH users when other DL users are scheduled. 4.6.8.3 Radio Analysis – HARQ Each transport block transmission is represented as a HARQ process. Each HARQ process data is held in memory until NDI is toggled (NDI – new data indicator2 (physical layer bit tog gled for new data, i.e., new data is to be sent). This allows fast retransmission of erroneously received data. The scheduler’s representation of an HARQ process is as follows: feedback sta tus (ACK, NAK, DTX, Pending), TBS, MCS, and RV (redundancy version, HARQ has 4 redun dancy versions, rv0, rv2, rv3, rv1). Increasing the default number of transmissions means that RLC parameters also need to be modified and will require larger RLC buffers. The parameter tPollRetransmit also impacts the DL throughput; as the RLC send data by AM transmit mode, receiver will feedback ACK/ NACK to transmitter, and if time is out of the value of tPollRetransmit, the transmitter will retransmit related PDU, so the value of tPoll is important and the recommended value is tPollRetransmit=40 ms. If tPollRetransmit set too low, it will result to premature initiated retransmission rather than received ACK/NACK, this is due to a time out in advance. This will affect the normal transmis sion and reduce data rate. If tPollRetransmit set too high, it will lead to delay in the launch, can not quickly complete the normal transmission, resulting in a decline in data rate. Table 4.7 gives the field test of the responding UE PDCP throughput, RB Num/slot and percentage of lower than 2Mbps with different tPollRetransmit settings of signaling and data. 2 Do not confuse with newDataFlag, which is scheduler internal flag where 1 means new data and 0 means retransmission. 171 172 LTE Optimization Engineering Handbook Table 4.7 UE throughput with different tPollRetransmit settings. Test (ms) RSRP SINR PDCP throughput RB Num/slot percentage of lower than 2Mbps signaling 80, data 500 −79.2 18.5 33.5 94.811 1.14% signaling 160, data 160 −78.5 19.3 29.4 74.122 3.97% signaling 80, data 80 −78.3 18.5 27.0 86.974 8.00% signaling 40, data 40 −79.5 18.3 31.9 93.672 1.71% signaling 20, data 20 −78.4 19.3 30.6 93.577 0.54% In case of rank 2 spatial multiplexing there are 16 HARQ process per UE instead of 8, but there are two processes that share the same ID; scheduler sees them as separate processes that are coupled to each other. 4.6.9 Avarage Cell Throughput Optimization Average cell throughput is the average number of successfully transmitted data bits in one second per DL bandwidth of all active users in one data frame. In a live network, if the RSRP/SINR does not show any special issues, with normal maximum values and distribution according to the specific location, coverage of the cells and distribution of traffic in the cells, that means trying to increase RSRP close to the site and avoiding serving users far from the node. This fact will lead to an increase in the average throughput of the cells as well. The parameter of crsGain is another factor impacted throughput. The configuration of crsGain to values higher than 0 will increase the area of coverage, increasing the power allocated to reference signal, but decreasing power in other resources, therefore, the average throughput in the cell can be affected. Due to this fact, RS boost must be used only in specific cases, and it is not a recommended configuration to be implemented in the whole network. The other impact factors include bandwidth, control format indicator (CFI) format, UE category, MIMO transmission mode, and loading. For TDD LTE system, UL‐DL configuration and special subframe configuration also impact the average cell throughput. The DL through put calculation flow is shown in Figure 4.29. The target average DL throughput obtained for the LTE sectors of the capacity layer (high‐ frequency band) and coverage layer (low‐frequency band) is different. Unlike the coverage layer, the objective of capacity layer is to provide a throughput as high as possible. The mean frequency efficiency can be achieved to 2 bits per hertz in the capacity layer. 4.6.10 Cell Edge Throughput Optimization Reasonable DL coverage threshold should also reflect the throughput requirements of DL and UL. In most cases LTE UL is always limited coverage due to the mobile terminal is power lim ited. The problem is how much Tx_power will be limited. It depends on the DL power and DL/ UL rate requirements. A typical case is that volume of DL traffic and data rate are always higher than the UL. Therefore, RF DL/UL imbalance is normal, and that will match the data rates. If the imbalance is very high, which means that the DL‐received power is either too high (may interfere with), or too low (because the DL power is insufficient, can not fully play the UL power of the potential). A very important fact is that the UL performance and coverage do not Capacity Optimization BW UL/DL config Transmission mode Determine PRB number of DL subframe MCS index PRB number PRB number CFI Determine total RE of DL sbframe Transmission mode Deduct common channel/signal overhead Determine TB size UE category Physical bits number Determine the physical bits number N TBS valid? Y throughput Figure 4.29 DL throughput calculation flow. depend entirely on the RSRP and RSRQ. Therefore, higher DL transmit power may improve the DL cell average and cell‐edge data rate, but it is not able to improve the UL coverage. Generally speaking, the diversity of MIMO can improve the cell‐edge throughput peak. Therefore, in order to improve the UL coverage performance, through higher levels of diversity (using 4RX diversity), they can keep a low noise in LNA application. On the other hand, LTE uses universal frequency reuse (N=1) without soft hand off. Consequently, high levels of interference and low SINR can be expected near the cell‐edge. Traffic channel performance at the cell edge can also be enhanced via ICIC feature. 4.6.11 Some Issues of DL Throughput Throughput issues can be anywhere in the network. A series of questions are needed to ask first: How many eNBs are affected? How many subscribers/UEs are affected? Was the test done using UDP or TCP? 4.6.11.1 Antenna Diversity not Balanced During field testing, it is found RSRP and RSRQ display good (SINR is around 27 dB, RSRP is around ‐80dBm) but DL throughput can not achieve 35Mbps, which is shown in Figure 4.30. It is found that the imbalance of the two antenna diversity reception is the cause of the prob lem. RSRP of antenna0 and antenna1 display huge power gap, RANK2 SINR is low shown in Figure 4.30. Antenna 0’s RSRP is around ‐76dBm, antenna 1 is around ‐95dBm; the gap is more than 20 dB. Play back all the test data, and it can be seen that the antenna1’s RSRP continued at a low level. It is suspected that the antenna interface issues or problems with the antenna. After adjusting the eNB transmit antenna (antenna TX power gap, check RRU) and test again, the test antenna is found to be received normally, and throughput increased by about 40Mbps. 4.6.11.2 DL Grant is not Enough From Figure 4.31, it can be seen that under good SINR (17.5 dB), DL throughput was only able to reach to around 1Mbps. The reason is too low number of DL grant allocation, the average is around 200 for a period of time. 173 174 LTE Optimization Engineering Handbook Figure 4.30 Antenna diversity not balanced. Figure 4.31 Low DL grant allocation. The reasons of low number of DL grants are generally include: UE is in DRX state, the DL is not scheduled or rarely scheduled, PDCCH false alarms or missed detection, and lack of RB resources, and so on. In this case, DRX was off and PDCCH misdetection ratio is nor mal, it is found finally that there were other users occupied the PDSCH resources after OMC counter‐investigation. When locking to a single‐user test, the DL Grant can reach to 600, and DL throughput resumed normal. Capacity Optimization 4.6.11.3 Unstable Rate In this case, the cell radio coverage was good, RSRP was about −75dBm, the average SINR was about 25 dB; in addition, BLER and double MCS are normal, but the DL data rate is often below 10Mbps or no rate. The issue is DL user data rate was unstable, sometimes reduced to 10Mbps, even decreased to 0Mbps. The problematic cell coverage seemed to be well, and MCS was normal, but the PDCP data rate was not qualified. The investigation work was following: first, check if there is any equip ment (including UE, FTP Sever, etc.) alarm. Next, check the radio parameters configuration, like RRU power, bandwidth, time slot allocation, and so on, were right or not. If the rate was still so unstable after changed, someone still needs to check the cable between RRU and antenna and whether the line sequence was correct or not. Sometimes, mapping of the RRU port and antenna mouth (line sequence) does not meet the requirements during installation and causes an unstable data rate. 4.7 UL Data Rate Optimization UL throughput optimization is not a trivial task in since there are different features that affect the UL throughput. UL scheduler assignments will decide how many PRBs are allocated to each UE, UL adaptive modulation and coding will decide the MCS to be used by each UE every time it is granted UL resources, and adaptive transmission bandwidth will reduces the number of PRBs assigned to the UE in the UL based on the UEs power headroom reports in order to favor retainability of the call. The general troubleshooting strategy is described in the following along with different factors responsible for poor UL throughput. ●● ●● ●● ●● ●● ●● High BLER (bad coverage) UL interference (high RSSI) Low power headroom Scheduling algorithm Low demand Other (VSWR, backhaul capacity) Analysis flow for UL throughput investigation: ●● ●● ●● ●● ●● Alarm and parameter/feature check: make baseline audit for parameter and feature. RSSI: High UL RSSI would impact the UL throughput. Percentage of 16 QAM samples: Low usage of 16 QAM modulations scheme in UL would impact the UL throughput. PUCCH and PUSCH SINR: Poor UL_SINR conditions would impact UL throughput. Power limited UE: High number of power limited UE indicates poor UL coverage. For UL, the above mentioned is the fundamental areas of analysis for UL. It will begin this sec tion with an overview of UL scheduling and link adaptation. BSR is the mechanism the UE uses to inform the eNB about the amount of data waiting in its RLC buffers. PHR is the mechanism the UE uses to inform the eNB about remaining power at the transmitter (or power limitations). The num ber of PRBs available for UL scheduling has some 3GPP specified limitations, which is different from DL. This means that, for example, the maximum number of PRBs for a single UE able to be scheduled in 5 MHz is 20 and not 23 (with two reserved for PUCCH). The areas of analysis for UL: ●● ●● UL scheduling strategy. BSR (buffer status report), Values ranges from 0 up to >15000 bytes using 64 index values. for example, index 0 for BSR=0, index 1 for 0 < BSR <= 10 and so forth. 175 176 LTE Optimization Engineering Handbook ●● ●● ●● ●● ●● ●● PHR (power headroom report), PHR reports a index value similar to BSR with values between −23 up to 40 dB, values are close to (or less than) 0 means the UE is power limited. When a UE is power limited, the eNB may schedule fewer PRBs in order to reduce the required out put power of the UE, this can in turn reduce throughput. Cell bandwidth versus maximum allowable PRBs, the number of PRBs available for UL scheduling. Link adaptation, SINR of PUSCH is the inputs for UL. MCS available and 16QAM PDCCH SIB scheduling colliding with UL grant, PDCCH collisions can occur with SIB/ DL transmissions as DL and UL grants are both scheduled using the same PDCCH resources. HARQ for UL. 4.7.1 Model of UL Data Throughput The data rate that can be achieved on the air interface under different SINR conditions var ies according to the applied modulation and coding scheme (MCS) as selected by the UL link adaptation function. Channel sounding DMRS and SRS are used so that eNB can under stand channel response across PUSCH and schedule UE accordingly. Also the amount of retransmissions (both on HARQ and RLC layer) is a key factor for the UL data throughput. UL link adaptation tries to choose the MCS, which provides the maximum data throughput at a given maximum rate of retransmissions (BLER‐based link adaptation). For a given UE location and a given MCS, the power that the UE uses per RB in UL finally determines the UL SINR. The open‐loop UL power control (UL‐PC) functionality manages the UE transmit power per RB based on DL pathloss estimation. Hence, the main focus of the UL through put is on verifying the UL link adaptation and UL power control RRM functions and on the maximum throughput achievable for a single UE in terms of layer1 physical throughput and application throughput. For LTE, there is very good correlation between throughput and RSRP on the UL, which is shown in Figure 4.32, as UL throughput depends mainly on propagation loss, while for the DL it is usually expressed as DL throughput with SINR. UL data amount comes from BSR reported by UE, if the engineer doubts the UE doesn’t report enough data, he needs to check the configuration of UE. Another possibility is that in the poor air condition UE really sends BSR but due to CRC fail, eNB misses to handle it. 4.7.2 UL SINR and PUSCH Data Rate UL_SINR is defind as UE’s PUSCH channel SINR received by eNB. The ratio of UL_SINR more than threshold (−3 dB in network) and the mean value can be used to analysis UL service qual ity and interference. The very good point of UL_SINR is more than 22 dB, good point of UL_ SINR is around 15~22 dB, middle point of UL_SINR is arround 5 to 15 dB, bad point of UL_SINR is around −5 to 5 dB, the worst point of UL_SINR is less than −5 dB. The issues can be spotted by combined RSRP and UL_SINR analysis. UL_SINR could also easily be extracted from eNB traces and it was simply estimated from UE measurement data according to UL_SINR = Prx,RB – (noise floor). The relation between UL_SINR from measurement report and pathloss (left) and combinded RSRP and UL_SINR analysis are shown in Figure 4.33. From the right figure, it can be seen that the first quadrant, the samples of RSRPand UL_ SINR are good, this is normally a good cell. The second quadrant, RSRP is bad, UL_SINR is Capacity Optimization PUSCH throughput (Mbps) 25 20 15 10 5 1 RSRP (dBm) –120 –110 –100 –90 –80 PUSCH throughput (Mbps) 25 20 15 10 5 1 RS SINR (dB) –5 0 5 10 15 19 Figure 4.32 PUSCH throughput versus RSRP and SINR. good—this usually happens in an indoor scenario (outdoor cell coverages indoor). The third quadrant, RSRP and UL_SINR are both bad—this is usually happens in a weak coverage scenario. The fourth quadrant, RSRP is good, UL_SINR is bad—this probably happens in a UL interference scenario. In a live network, it needs specially focus on the poor UL_SINR cells that affected throughput in UL. For engineering, UL_SINR can be got from OMC counter as Table 4.8 described. 177 178 LTE Optimization Engineering Handbook 15 SINR (dB) 10 5 0 90 100 110 130 120 140 150 Pathloss –5 SINR calculated SINR simulated –10 110% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 110% Figure 4.33 UL_SINR versus pathloss (left) and combinded RSRP and UL_SINR analysis (right). Table 4.8 UL_SINR investigation from OMC counter. SINR of PUCCH: pmSinrPucchDistr Distribution of the SINR values calculated for PUCCH. PDF ranges: Unit: dB [0]: SINR <= −15; [1]: −15 < SINR <= −12 [2]: −12 < SINR <= −9; [3]: −9 < SINR <= −6 [4]: −6 < SINR <= −3; [5]: ‐3 < SINR <= 0 [6]: 0 < SINR <= 3; [7]: 3 < SINR Condition: each SINR value for PUCCH per UE calculated on a TTI basis yields one sample in the distribution. SINR for PUSCH: pmSinrPuschDistr Distribution of the SINR values calculated for PUSCH. PDF ranges: Unit: dB [0]: SINR <= −5; [1]: −5 < SINR <= −2 [2]: −2 < SINR <= 2; [3]: 2 < SINR <= 6 [4]: 6 < SINR <= 10; [5]: 10 < SINR <= 14 [6]: 14 < SINR <= 17; [7]: 17 < SINR Condition: each SINR value for PUSCH per UE calculated on a TTI basis yields one sample in the distribution. Capacity Optimization UDP UL 45 EVA LC EVA MC EPA HC ETU MC COV EVA MC 40 35 Mb/s 30 25 20 15 10 5 0 –8 –3 2 7 12 17 22 UL SINR Figure 4.34 PUSCH data rate versus UL_SINR. 100.00% 90.00% 80.00% Operator 1 Operator 2 CDF 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Uplink SINR Figure 4.35 UL SINR CDF. In live networks, average cell throughput or cell‐edge throughput are the main optimization purpose. The throughput as function of the SINR in UL is increasing with increased SINR values (see Figure 4.34). UL SINR distribution in a live network refers to Figure 4.35. 4.7.3 PRB Stretching and Throughput Different from DL, UL PRB stretching can’t always bring throughput gain. Assume UE is trans mitting at its maximum power, if the number of PRBs are doubled, the bandwidth (B) doubles, as well as the noise (N), but the signal (S) remains the same since the UE can not increase its transmit power anymore. As it keeps doubling the number of PRBs, it obtains some channel capacity gain, as shown in Figure 4.36. 179 180 LTE Optimization Engineering Handbook Figure 4.36 PRB stretching and channel capacity. It can be seen that doubling the bandwidth does not double the channel capacity, so as it increases the number of PRBs, the channel capacity (throughput) is increased, but at the cost of operating with a lower spectral efficiency. Therefore, PRB stretching is only viable when a small number of UEs are competing for the UL PRBs. When a large number of UEs are competing for the UL PRBs, then it makes more sense to grant a smaller number of PRBs to each UE as a way to preserve the spectral efficiency and so the UL cell capacity. 4.7.4 Single User Throughput Optimization When trying to reach the max theoretical UL throughput in the field in static tests, UL radio is the main factor and the related analysis is necessary. Similar as DL, single user throughput optimization needs UE trace logs to spot the problem. Here we present an example of trace parses of UL. This example is a modified output of UL, which summarizes a number of traces onto one line, like UL_SINR, MCS (modulation and coding scheme), PRB occupied, TBS (transport block size), BSR, PHR, HARQ, UL_TPC, UL BLER, and so on. BSR and PHR are decoded into human readable formats along with PRBs. sfn|sf|rxPwrPus|prb|ulTpc|sinr|ulModul|mcs|ndf|ul bsr 266| 6| -95.6 | 48| 0:1 | 22 | 16QAM | 23| Y | 266| 7| -95.6 | 48| 0:1 | 22 | 16QAM | 24| Y | 266| 8| -95.6 | 48| 0:1 | 22 | 16QAM | 24| N | 266| 9| -95.6 | 48| 0:1 | 23 | 16QAM | 24| Y | 267| 0| -95.7 | 48| 0:1 | 22 | 16QAM | 24| Y |>150000 267| 1| -95.8 | 40| 0:1 | 22 | 16QAM | 24| Y | 267| 2| -95.6 | 48| | 23 | 16QAM | 24| Y | 267| 3| -95.6 | 48| 0:1 | 22 | 16QAM | 24| Y | 267| 4| -95.6 | 48| 0:1 | 22 | 16QAM | 23| Y | 267| 5| -95.6 | 48| 0:1 | 22 | 16QAM | 23| Y |>150000 267| 6| -95.6 | 48| 0:1 | 23 | 16QAM | 24| N |>150000 267| 7| -95.6 | 48| 0:1 | 23 | 16QAM | 24| Y | 267| 8| -95.6 | 48| 0:1 | 22 | 16QAM | 24| Y | |phr | | | | | | | | | | | | | 32 |ul tbs| ul crc |har|ulBler| | 25456| | A | 2% | | 25456| | A | 2% | | 25456| ERR 3182| N | 5% | | 24496| | A | 5% | | 24496| | A | 5% | | 21384| | A | 5% | | 25456| | A | 5% | | 25456| | A | 4% | | 25456| | A | 4% | | 25456| | A | 4% | | 25456| | A | 4% | | 25456| | A | 4% | | 24496| | A | 4% | Capacity Optimization 50 45 40 35 nrb 30 25 20 15 10 Number of RBs measured 5 Number of RBs simulated 0 90 100 110 120 130 140 150 Lsa [dB] Figure 4.37 Number of UL RBs in a live network. 4.7.4.1 Radio Analysis – Available PRBs PUCCH takes a minimum 1 PRB on each side of the UL band for UL control signaling, r educing the size of PUSCH, for example, 5 MHz bandwidth, 25 PRBs available. If only 2 PRBs are reserved for PUCCH, thus 23 PRBs are available for PUSCH transmission. Due to 3GPP specified design limitations in the UL it is not always possible to utilize all free PRBs for UL transmissions. 3GPP TS36.211 Ch 5.3.3 defines the following formula for the number of PRBs on PUSCH for a single transmission: 2 a 3b 5c. Here a, b, and c are non‐negative integers. For 5 MHz, auusming only 23 PRBs are available for PUSCH, but the max number of PRBs for a single PUSCH transmission is only 20 PRBs, this corresponds to a = 2, b = 0, and c = 1 (i.e., three PRBs are unavailable to be used). The result of the measured number of UL RBs as function of pathloss in a live network, which follows the trend of the ideal result (not take effects from fading; see Figure 4.37). The difference may also be explained by a low number of measurement samples. 4.7.4.2 Radio Analysis—Link Adaptation The selected MCS for a certain allocation size is needed to maintain the target BLER (10%) for the first transmission. The inputs of UL link adaptation include UL interference power, received power of UE (across traffical PUSCH PRBs), PHR reports, and HARQ CRC (BLER). eNB can use received power/UL_SINR, PHR, UL interference, and UL HARQ BLER measurements to control UL MCS. For these, it needs to check further based on the parameters mentioned in Chapter 4.7.4.1: ●● ●● ●● ●● High values of UL interference, could there be some external interferer? Are the values of p0NominalPusch in neighbor cells too high? In a multi‐UE environment, too high of values for this parameter may generate more UL interference and eventually lower total throughput on the UL. rxPower is too low, check if p0NominalPusch is set too low. If PHR shows UE at maximum Tx_power, is p0NominalPusch too high causing UE to exceed maximum transmit power? Closed‐loop power control TPC is ignored by the UE? Low values of SINR, is p0NominalPusch too low? Closed‐loop power control TPC is ignored by the UE? 181 182 LTE Optimization Engineering Handbook Figure 4.38 PDCCH assignment. PDCCH PDSCH DL subframe (current) PUSCH UL subframe (4 TTI later) 4.7.4.3 Radio Analysis – PDCCH PDCCH carries both the UL (PUSCH) assignment and DL (PDSCH) assignment indication. In case many PDCCH CCEs are used for DL transmission (e.g., SIB with 8 CCEs. In case of a DL SIB transmission, 8 CCEs of PDCCH may be used for DL grant), it may be that UL grant is not possible to be scheduled in this TTI for a single UE (Figure 4.38)! To reduce processing load when decoding PDCCH, 3GPP (TS36.213 Ch 9.1.1) defines par ticular search spaces within PDCCH depending on number of CCEs for grant, number of CCEs for PDCCH, and RNTI of the UE. Depending on these parameters, it may not be possible to allocate a PDCCH UL grant resource, and therefore, the UE may not be able to be scheduled every TTI even if there are unused PUSCH resources. 4.7.5 Cell Avarage and Cell‐edge Throughput Optimization The analysis method of UL cell avarage and cell‐edge throughput is quite different with DL. The following parameters will deeply affect the performance like pathloss distribution, mapping from SNR to throughput, PUSCH P0 target, α, upper/lower target SINR threshold, upper/ lower target RSSI threshold, and so on. Cell coverage is based on power budget. However, UL is the limiting link in most cases. Reasons are in the natural power limitation on mobile side. UE Tx_power increases slower with FPC when moving from cell center to cell edge. Without UL fractional power control, the SINR target needs to be set conservatively (at a lower value) for cell‐edge coverage, which will sacri fice the cell center throughput. As UL fractional power control allows better trade‐off between spectral efficiency and cell‐edge rate by compensating gradually in open loop power control as a function of the pathloss. This allows a higher SINR target toward cell center. The SINR target automatically drops toward cell edge as the pathloss increases, which will be suitable for cell‐ edge UE’s power profile (less power headroom available, lower data rate, lower interference to neighbor cells). The analysis in Figure 4.39 shows the trade‐off between spectral efficiency (average SE) and cell‐edge rate (edge SE). How much the UL limits the power budget depends on the DL power along with DL and UL throughput requirements. Typically, the DL traffic and throughput (both cell and user) is b igger than in UL. Thus, certain RF unbalance is normal and ideally it should reflect the proportion of UL and DL throughput. For cell‐edge users, when reaching a certain pathloss, the higher the SIR target, the lower the throughput: when the max UE Tx_power is reached (23dBm), the UL scheduler is forced to Capacity Optimization 4.5 improvement/degradation Figure 4.39 Cell average SE and cell‐edge SE. average SE 4 edge SE 3.5 3 2.5 2 1.5 1 0.5 0 1 0.9 0.8 0.7 0.6 0.5 0.4 alpha schedule/grant less RBs in order to be able to transmit more power on each RB and thus reach the SIR target. Overall, the UL SIR target varies between 7.5 dB in cell center and 0 dB in cell edge. 4.7.6 Some Issues of UL Throughput One case shows that while the DL channel quality is excellent in TDD LTE field test, the UL iBLER rate is less than 5%, and the UL throughput can reach to 10Mbps or so. Occasionally iBLER will rise to 20%, resulting in the throughput down to about 7Mbps. Preliminary analysis found that when the iBLER increased, the UE transmit power does not ascend, instead of reducing the MCS to reduce error bit rate. According to the UL power con trol design, there is no exception in this process, as for MCS28, demodulation threshold is 18 dB under 10% packet error, the UL channel quality was constaant between the 22 to 24 dB. So the UE UL TX_power was almost without change, instead of lower the MCS to reduce the error, the lower the MCS results in a lower data rate of UL. For burst errors happened in UL, it needs to capture eNB trace and UL_RSSI to analyze. It is found that the UL_RSSI of antenna 4 is higher than other antennas and its SINR is lower 5 dB compared to other antennas as shown in Figure 4.40, so RRU channel 4 maybe the problem, leading to the demodulation performance degradation. The other case is UL/DL TCP service simulteneous issues and analysis. Also in TD‐LTE field trial, when doing UL/DL TCP service at the same time, peak throughput of UL/DL can not reach peak value. The same UE for concurrent upload, download throughput is significantly lower than the throughput of a single download, as shown in Figure 4.41. In this situation, it can be seen that in single UE scenario, peak throughput can be reached in both DL and UL as shown in Table 4.9. But when simultaneously, throughput will be drop down badly. The reason is caused by TCP mechanism. TCP transmissions will stall when ACKs are delayed or not received, TCP sender is “ACK clocked.” A data packet is only sent upon arrival of an ACK from receiver, which can be shown in Figure 4.42. This is also need to pay attention to that in excessive pathloss scenario that UL ACKs can’t be received at the eNB at a rate sufficient to avoid TCP slow start being initiated. 183 184 LTE Optimization Engineering Handbook RSSI (dBm) –100 RSSI from antenna 0 RSSI from antenna 1 RSSI from antenna 2 –105 RSSI from antenna 3 RSSI from antenna 4 –110 RSSI from antenna 5 RSSI from antenna 6 RSSI from antenna 7 –115 Figure 4.40 Antenna issue. (See insert for color representation of the figure.) only download concurrent upload and download only concurrent upload and download only download download Figure 4.41 Concurrent upload and download issue. Table 4.9 Throughput of simulteneous UL/DL TCP transmission. Comment Traffic RSRP SINR CQI CFI: 3, TDD configure: 2 UL −83.02 29.87 14.56 DL_thp 0.16 RB_num 5.5 UL_thp UL_RB 8.13 80.02 DL −83.79 26.35 14.34 79.85 98.37 1.22 23.56 DL&UL −82.6 27.43 13.36 42.56 84.7 8.03 79.73 Capacity Optimization Application Space “File” XYZ Not Yet Sent by Application Received Reliably (Correct & In-Order) Memory shared Between Application and TCP Receiver Buffer Operating System TCP Receiver Data ACK Data ACK IP Client “File” XYZ Send Buffer Data ACK TCP Sender IP Server Figure 4.42 TCP transmission mechanism. When run DL/UL TCP simultaneously, in both direction, there are data and TCP ack, which are needed to be send. So, there will be congestion between data and TCP ack. Data and TCP ack will be in one queue (buffer) and wait for TCP window swapping. It will bring the issues as: (eNB side for example, the same situation in UE side) ●● ●● ●● DL TCP data package cannot be sent in time because there is delay for DL TCP ACK from UE. TCP ACK for UL TCP window cannot be sent in time because it is in long queue and waiting. Package will be forced to loss during congestion if the buffer is full. To upload and download at the same time, if you do not set the TCP ack priority higher, download data TCP ack will row in the back of the upload data, that is delayed while sending, resulting in the download throughput lower than a download only throughput. AQM (active queue management) algorithm is used for TCP throughput optimization to discard packets before the buffer is full and thus providing rapid feedback to the sender. 4.8 Parameters Impacting Throughput The parameters that can impact the throughput, both on UL and on DL, are listed in Table 4.10 and Table 4.11. In DL, the throughput is most influenced by the type of antenna system that is being used/selected while in UL by the required quality of the received signal, which forces higher powers of PUSCH channel. Note that some parameters can increase the DL throughput while decreasing the UL throughput in the meantime due to the asymmetry of TDD LTE. 185 Table 4.10 Parameters impacting DL throughput. Name Recommended dlMCSTransitionTable Name Recommended dlSINRThresholdbetweenRank1BeamformingAndTM3 dlSinrThresholdBetweenCLMIMOOneLayerAndTxDiv −10 deltaSINRforIntermodeSwitch 3 dlSinrThresholdBetweenCLMIMOTwoLayersAndOneLayer 12 beamformingAlgoRank1 COM‐EBB beamformingAlgoRank2 SU‐BF‐RANK2‐ COMEBB dlSinrThresholdBetweenOLMIMOAndTxDiv subframeAssignment/specialSubframePatterns dlSinrThresholdBetweenRank1BeamformingAndRank2BF 0 α Fairnessfactor 0.8 dlSinrThresholdBetweenTxDivAndRank1Beamforming 0 Dynamic CFI enabled ture rIThresholdBetweenRank1AndRank2 0.6 cFI 3 sinrOffsetForBeamformingPMICQI 0 cFI1/2/3 Allowed TRUE sinrOffsetForBeamformingTxDivCQI 0 cFIThreshold1 2 sinrOffsetForRank1AndRank2CW0 0 cFIThreshold2 6 uLCESINRThresholdBetweenRank1BeamformingAndRank2BF −51.2 cFIIncreaseTimer 5 uLCESINRThresholdBetweenTxDivAndRank1Beamforming −51.2 dlBasicSchedulingMode PF blerThresholdBetweenRank1BeamformingAndRank2BF 1 blerThresholdBetweenTxDivAndRank1Beamforming 0.8 transmissionMode beamFormingAlgo COM‐EBB pmiRIReportR9 FALSE uLCESINRThresholdBetweenTxDivAndBeamFormingIntraTm7 −17 cqiReportingModeAperiodic rm30 sinrOffsetForBeamformingCQICompensation 3 tddAckNackFeedbackMode multiplexing Capacity Optimization Table 4.11 Parameters impacting UL throughput. Name Recommended Value uplinkSIRtargetValueForDynamicPUSCHscheduling According to the real scenario pUSCHPowerControlAlphaFactor 0.8 ulSchedPropFairAlphaFactor 0.5 ulMCSTransitionTable ‐ mCScorrectionForIRC 0 minSIRtargetForFractionalPowerCtrl 0.0 [dB] maxSIRtargetForFractionalPowerCtrl 19.0 [dB] pathLossNominal According to the real scenario p0NominalPUSCH According to the real scenario 187 188 5 Internal Interference Optimization LTE systems deployed with 1/1 frequency reuse are shown to be interference limited. Providing good coverage and improving sector capacity are two critical objectives of most operators. High interference leads to bad impact on several KPIs. For voice‐centric service, this trans lates to reducing the probability of bad coverage zones or zones with low SINR. For data‐ centric service, improving coverage refers to improvement in the lowest 5% achievable data rates. In a cellular system, since the zones with poor coverage are usually areas that see maximum interference, a desirable goal is important to design mechanisms to mitigate the intra‐LTE interference. 5.1 Interference Concept Because optimal LTE performance requires a higher SNR than any previous technology, noise can present a major obstacle to the smooth, efficient, and profitable operation of an LTE net work. High interference in UL and DL can cause total traffic reduction in UL and DL. Figure 5.1 plots the DL (RS) and UL (PUSCH) interference map in a LTE system. UL The UL thermal noise power within the UL system bandwidth consisting of N RB resource blocks as defined in 3GPP standard. It is defined as (No × W), where No denotes the white noise UL power spectral density on the UL carrier frequency and W N RB N scRB f denotes the UL system bandwidth. The measurement is optionally reported together with the received inter ference power (RIP) measurement, it shall be determined over the same time period as the RIP measurement. The reference point for the measurement shall be the RX antenna connector. In case of receiver diversity, the reported value shall be linear average of the power in the diversity branches. The noise floor on the UL is estimated using the formula as following, UL Noise Floor Thermal noise Noise Figure Where thermal noise = k(Boltzmann) * T(290 K)* Bandwidth. For nominal cell values in a network with RRU deployments, when the noise floor without loading is not more than −118 dBm, and no more than −108 dBm with full loading is accepted. From a field test, it can be seen that the high interefence issued cell is 17 dB loss compared with the nominal cell in Figure 5.2. The noise floor on the DL is estimated using the formula as following, Noise Floor RSRP RS SINR LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Internal Interference Optimization Figure 5.1 DL (RS) and UL (PUSCH) interference map. (See insert for color representation of the figure.) Noise rise is a quantity that describes the increase in interference that comes with increased traffic. It is used in dimensioning in coverage calculations as well as in capacity calculations. In coverage calculations it is used as a margin on top of the link budget, to accommodate for the interference created by traffic in adjoining cells, the so‐called interference margin. In capacity calculations, it is used to estimate the SINR, which in turn determines the bitrates that can be achieved by the users (Figure 5.3). A bad SINR is based on bad high noise or high inteference, SINR/SNR can be expressed as: SINR/SNR 1/ 1 inteference/noise power For SINR/SNR = 1, means that noise is a lot higher than interference, and this is a noise‐limited scenario (typically SINR/SNR > 0.5). For SINR/SNR = 0, means that interference is a lot higher 189 LTE Optimization Engineering Handbook PUSCH througput (kbps) 10000 8000 Nominal cell 6000 The drive test shows about a 17 dB loss in coverage due to interference 4000 17 dB 2000 Issued cell 0 85 95 105 115 125 135 145 Pathloss (dB) Figure 5.2 Interefence issued cell versus nominal cell. 6000 A free space pathloss model was used to convert maximum pathloss into cell range 5000 Cell range (m) 4000 3000 2000 1000 0 –119.4 –119.0 –118.6 –118.2 –117.8 –117.4 –117.0 –116.6 –116.2 –115.8 –115.4 –115.0 –114.6 –114.2 –113.8 –113.4 –113.0 –112.6 –112.2 –111.8 –111.4 –111.0 –110.6 –110.2 –109.8 –109.4 –109.0 –108.6 –108.2 –107.8 –107.4 –107.0 –106.6 –106.2 –105.8 –105.4 –105.0 –104.6 190 Noise + Interference / PRB Figure 5.3 UL N + I will reduce cell range in UL limited scenario. than noise, this is a interference‐limited scenario (typically SINR/SNR < 0.5). For SINR/ SNR = 0.5, means that noise is equal to the interference. Usually, there seems to be an interference issue when densifying the network. 5.2 DL Interference RSRP measurement with a scanner is the most reliable way to detect areas with possible interference problems and bad dominance (not impacted by network load, absolute SINR measurement values can’t be used as a reliable performance indicator). Figure 5.4 presents RS received power (dBm) Internal Interference Optimization –65 –70 –75 –80 –85 –90 –95 –100 –105 –110 –115 –120 –125 –130 –135 –140 No dominant cell in the area Figure 5.4 No dominant cell. what no dominant cell is, and the number of PCIs in for example, 10 dB power window is a useful indicator. The interference in an LTE system comes from the surrounding cells, called inter‐cell inter ference. Reduce neighbor cell interference is the key method to improve DL SINR. The case of normal RSRP and lower SINR is generally caused by interference from neighbor cells. For the same frequency network in LTE, co‐channel interference is inevitable, with the increase of the cell loading, SINR value decreases, but the RSRP is basically unchanged. Use overlapping coverage index to evaluate the possible impacts between cells is a very good method to evaluate and optimize the LTE network. SINR = Pwanted Pnoise_ue + Iexternal + Iother_cells SINR ≈ RSRP1 N15kHz + FNF,UE + QL,k ∑ RSRPk k>1 But in the case of no load around the cell, the following two cases will also produce the interference that we do not expect. First, the surrounding area PCI setting is incorrect, the second is neighbor cell overshooting. The border between the adjacent cells need to avoid the value of PCI mode 3 equal, and then by the method of the Chapter 3 to solve the problem of overshooting coverage. 5.2.1 DL Interference Ratio DL interference ratio shown in Figure 5.5 is a cell isolation measure defined as a ratio of inter ference from all cells except the best over the strongest cell signal. It is independent of access technology, bandwidth, receiver performance, and capability. It gives a measure and thus gbest_cell g1 g2 g3 gi Figure 5.5 DL interference ratio. ∑ gi F= i ≠ best_cell gbest_cell 191 192 LTE Optimization Engineering Handbook rovides excellent measure of inter‐cell interference and can be calculated on path gains (or p attenuations) according to the following formula: F i best _ cell gi g best _ cell Geometry describes the user received pathgain in relation to all other cells, which is propor tional to SNR, as seen in the fomular below: G RSRPBestServer RSRPothers 5.2.2 Balance Between SINR and RSRP In LTE, it should more strictly optimizing the coverage among cells. Interference control aims to find the balance between SINR and RSRP. As Figure 5.6 shows, with too little overlap, hand over may fail, whereas with too much cell overlap, higher interference occurs and cell‐edge throughput can be reduced. Again, a balance must be achieved by adjusting overlap margins and cell sizes. This can be achieved with parameters and physical changes. Cell overlapping area is a sensitive area in LTE network optimization. In the area, the SINR and RSRP is refers to the lowest 5% samples, cell reselection, handover, dropped calls, and other failure events will probably happen frequently. Figure 5.7 shows the DL reference signal SINR distribution during handover with different UE velocity from the field test data. DL interference can be shaped according to the degree of the interference, which also defines the preferred area where the PDSCH scheduling should consider. The decision of applying the interference shaping requires the knowledge of the general load situation in its own cell as well as the neighboring cells. The functionality relies on resource status reporting between neigh boring cells using X2 interface. This provides a possibility to consider neighbor cells load in the decision whether to apply interference shaping in a certain cell. The area can expand or shrink to adapt to the load. If the cell adjacent to the cell with interference shaping has high load, with accurate CQI reports, it could allocate the UEs experiencing neighbor cell interference to less interfered resources, improving frequency selective scheduling. When interference shaping is active, scheduling is limited to the preferred resources. The interference‐shaping function uses the principle of a preferred area of the occupied PRBs in frequency domain where the actual allocations are allowed. Criteria for applying interference shaping in a cell is dependent on the loading, a neighbor cell is considered to be under high load if the average PRB utilization as reported by this cell is above a threshold. 5.3 UL Interference In the UL, received interference power (RIP, equivalent UL RSSI) and thermal noise power at eNB reflect the experienced interference at eNB and they indicate the network load. In the UL, the UEs near the cell border will cause most of the interference to adjacent cells, the eNB scheduler can adjust the transmit PSD of mobiles near the cell edge in a frequency selective manner in order to shape the inter‐cell interference that is generated. Internal Interference Optimization SINR ↑ RSRP ↑ SINR ↔ RSRP ↑ SINR ↔ RSRP ↑ SINR ↑ RSRP ↑ SINR ↓ RSRP ↑ U N T SINR ↑ RSRP ↓ SINR ↑ RSRP ↑ SINR ↑ RSRP ↓ U SINR ↑ RSRP ↑ N SINR ↔ RSRP ↓ E D SINR ↑ RSRP ↔ SINR ↑ RSRP ↑ SINR ↑ RSRP ↔ SINR ↑ RSRP ↑ T U SINR ↔ RSRP ↔ N E D UNTUNED TUNED 70,000 More samples with a high SINR 50,000 60,000 DL Throughput [kbps] DL Throughput [kbps] 60,000 40,000 30,000 20,000 Less samples with a low SINR 10,000 0 –10.0 70,000 0.0 10.0 20.0 30.0 SINR [dB] Figure 5.6 Balance between SINR and RSRP. 50,000 40,000 30,000 20,000 10,000 0 –10.0 0.0 10.0 SINR [dB] 20.0 30.0 193 LTE Optimization Engineering Handbook DL SINR during Handover at SeNB and TeNB (dB) 1 0.9 1 2 3 4 120km_win100msec_TeNB: avg = –2.5834 120km_win100msec_SeNB: avg = –4.9875 30km_win100msec_TeNB: avg = –0.077 30km_win100msec_SeNB: avg = –1.6895 3 4 1 2 0.8 0.7 0.6 CDF 194 0.5 0.4 0.3 0.2 0.1 0 –20 –15 –10 –5 0 SINR (dB) 5 10 15 Figure 5.7 DL reference signal SINR during handover. UL wideband noise power calculation: Wide band Noise PowerdB 10 log10 BandwidthPRBNum 1 p 0 Noise Power Linear p /Bandwidth PRBNum More UEs with bad RF conditions trying to excess eNB transmitting high power without suc cess causes high UL noise floor and thus causes access failures/drop calls, which further reduces the spectrum efficiency (capacity) since MCS determines the spectrum efficiency (capacity). With the increase of simultaneous UEs, after the critical point, all of major performance stats worsen dramatically. Interference sources include high traffic in the UL, external source of interference (WiFi, power generator, GSM interference), too high in values of P0NominalPUCCH and P0NominalPUSCH, incorrect installation (wrong feeder type), incorrect parameters setting (UL attenuation and VSWR) and sharing a site with different technology (CDMA and LTE, GSM and LTE) that will have high chance to have interference during high traffic load. 5.3.1 UL Interference Detection RSSI is a measurement of all of the power contained in the used bandwidth. This could be signals, background noise, anything. High UL RSSI issue is an common issue that will impact the UL performance of the LTE network. In a normal case, the UL RSSI on each resource block is about −119 to 120 dBm when the cell is unloaded. If the RSSI is 3 to 5 dBm higher than the normal value at unloaded, UL interference exists. The causes of high UL RSSI are coming from LTE internal caused by inter‐cell UL interference at cell edge, and external system inter ference includes hardware issue, such as antenna, cable, RU module, and wrong parameter configuration. Internal Interference Optimization The UL received interference power, including thermal noise, within one physical resource block’s bandwidth of N scRB resource elements, the reported value shall contain a set of received UL interference powers of physical resource blocks nPRB 0, , N RB 1. UL interference level can be spot by UL RSSI simply. When UL RSSI’s range is −121 to −110 dBm, that means no UL interference. When UL RSSI’s range is −110 to −100 dBm, it means that UL is under medium interference. When UL RSSI is higher than −100 dBm, it means that UL is under higher interference. There are also two counters that are used to measure the interferences for PUSCH and PUCCH channels, pmRadioRecInterferencePwrPusch and pmRadioRecInterference PwrPucch. From these counters at the cell level in which the measurements are averaged over receive antennas, is it possible to identify antenna branches with external interference, as shown in Figure 5.8. pmRadioRecInterferencePwrBranchPrb1~100 Counter 1,000,000 900,000 Branch 0 Branch 1 800,000 700,000 Branch 1 600,000 500,000 Branch 0 400,000 300,000 External interference on branch 1 only 200,000 100,000 0 1 11 21 31 41 51 61 71 Figure 5.8 pmRadioRecInterferencePwr reflect UL interference. 81 91 PRB 195 LTE Optimization Engineering Handbook Figure 5.9 UL_SINR reflect UL interference. Usually RACH failures can give another indication of bad RF environment, and therefore, a ossible indication of high RSSI. The counters for UL_SINR (pmSinrPucchDistr and pmSinr p PuschDistr) can also help identifying bad RF conditions for the UE. Although not conclusive for interference problems, it gives more indication of the RF environment. Figure 5.9 shows each SINR value for PUSCH/PUCCH per UE calculated on a TTI basis yields one sample in the distribution. For UL interference analysis, TDD LTE is more complex than that in FDD system. It also needs additional analysis of GPS out of sync, subframe configuration, special subframe pattern, the surrounding cells’ UE transmit power, and the investigation of external or inter‐RAT interference, and so on. 5.3.2 Generation of UL Interference 5.3.2.1 Cell Loading Versus Inter‐Cell Interference UL RSSI vs NR Cell RRC Connected User –110 –112 –114 –116 –118 –120 –122 –124 –126 –128 –130 250 200 150 100 50 NR1 USER_RRCCONN NR3 USER_RRCCONN Current Cell UL_RSSI_Pwr_dBm NR2 USER_RRCCONN NR4 USER_RRCCONN Figure 5.10 Cell loading versus inter‐cell interference. 6/29/2014 0:00 6/27/2014 0:00 6/25/2014 0:00 6/23/2014 0:00 6/21/2014 0:00 6/19/2014 0:00 6/17/2014 0:00 6/15/2014 0:00 6/13/2014 0:00 0 RSSI (dBm) It is shown in Figure 5.10, UL RSSI of current cells rise with user increasing in neighbor cells. There should be minor inter‐cell interference caused by neighbor cell users. Normally, as the loading increases, PRACH, PUCCH, PUSCH interference is increased by 20 to 40 dB. In case of PUSCH, mainly from inter‐cell interference and for PRACH, PUCCH, both intra and inter‐cell interference exist. Figure 5.11 shows PRACH, PUCCH, PUSCH received level under light loading (20 calls) and heavy loading (200 calls). NR Cell RRC Connected User 196 Internal Interference Optimization Figure 5.11 Light loading (20 calls) and heavy loading (200 calls). 5.3.2.2 Unreasonable UL Network Structure The below scenario is typical in a live network in suburban, urban, and dense urban deploy ment. The UE is at the border of the two cells, its pathloss (PL) to cell 1 is 78 dB, pathloss to cell 2 is 82 dB. In this case fractional power control will assign the UE a high DL data rate in cell 1, but with assigning a high data rate in this case as the IoT from the UE1 to cell 2 is significant (Figure 5.12). This is typical case that neighbor’s interference caused by UE if UL network structure is unreasonable. UL network structure is an important factor affecting the IoT. And, it can cause a cascading effect as follows: if UE1 has 17 dB SINR target, but that causes an IoT of almost 13 dB at cell 2. Now from the point of UE2, which has a slightly better pathloss to cell 2 compare to cell 1, and if you assign a target UL_SINR of 17 dB to UE2, that UE2 will create an IoT of 13 dB at cell 1, and one can see that there is no solution for this issue and very high noise rises, even it does see 40 dB IoT on some busy cells in live network. 197 198 LTE Optimization Engineering Handbook Cell 1 Cell 2 UE1 PL = 78 dB Figure 5.12 Neighbor’s UL interference caused by UE. PL = 82 dB AN UE 2 Usually, the possible solutions of such issues can, if in a high UL interference scenarios, stop PUSCH power control avalanche effects when two adjacent cells are strongly interfering with each other. This can be achieved by blocking the PUSCH power‐up commands and forcing all PUSCH SINR targets to 0 dB on the cell when the level of IoT exceeds a threshold level. If in a medium interference scenarios, backs off the PUSCH SINR target proportionally to the measured level of interference as a typical strategy. Under certain assumptions DL CQI can be indicative of the effect of the UL inter‐cell interference, so that the interference contribution of every UE to the neighbor cells can be calculated. The interference effect of the UE on a neighbor cell also depends on the path loss between the UE and that neighbor eNB. Assumption: PGi ,m ispath gain from UE m to cell i PGs ,m is path gain from UE m to serverr cell s ●● ●● ●● Is and Ii are the total interference received at cell‐s and cell‐i, respectively. Ps and Pi are the transmission powers by cell‐s and cell‐i, respectively. Qs,m and Qi,m are the received DL powers of cell‐s and cell‐i at UE‐m, respectively. Then, the fraction of interference contributed by UE‐m at cell‐i, ri,m, is as follows, assuming that the UL pathloss and the DL path loss are the same/similar. PGi ,mUL _ SINRm The UL IoT rise contribution of a UE on a neighbor cell is ri ,m , also can be PGs ,m presented as: ri,m = UL_ SINRm × Ps / Ii Pi / Is × Qi,m Qs,m Cell-s Ue-m Cell-i If the DL and UL traffic loadings of the cell is the same/similar, or if the DL and UL traffic loading ratio per cell is the same/similar, then we have Ps /I i Pi /I s 1 ri ,m UL _ SINRm Qi ,m Qs, m Then, the fraction of interference contributed by UE‐m at all cells, Rm, can be represented as the ratio of UL SINR and DL SINR, is as follows. Rm i s ri ,m UL _ SINRm DL _ SINRm Internal Interference Optimization GP #0 #2 DwPTS #3 no cross slot interf. DwPTS #2 #0 40 km delay #4 Interfering BS #4 #3 cross slot interf #0 80 km delay DwPTS #2 #4 #3 cross slot interf. #0 DwPTS 200 km delay 300 km delay UpPTS #2 #4 #3 Target BS DP GP DwPTS #2 #3 Target BS cross slot interf. cross slot interf. #0 UP #4 delay DP GP UP Interfering BS Figure 5.13 Cross slot interference. Thus, operators can control Rm (i.e., the fraction of interference contribution by UE‐m to all cells) by setting the UL SINR target for UL power control according to UEm’s RF condition in UL as follows. UL _ SINRm Rm DL _ SINRm 5.3.2.3 Cross slot interference Cross slot interference mainly means a DL signal from “faraway.” eNB arrive into the UL slot of target eNB by a long propagation in the TDD LTE system. In this case, the atmos phere has refraction effect to radio wave propagation with very low propagation loss, just like in free space, especially in the specific climate and environment. Most interference is observed from about 200 km away eNB, sometimes even 300 km, and GP may be not long enough to cover this kind of interference due to long distance propagation. SRS or PRACH in UpPTS and even UL data in UL normal subframe will be interfered, which is shown in Figure 5.13. Cross slot interference mainly happened in likely scenarios include: eNBs on top of hills at different ends of a city separate by a large distance and eNBs in different cities, or separated by a large water body. Cross slot interference should affect mainly subframes 1 and 6 (special subframes), sites are not far enough for other subframes to be affected as shown in Figure 5.14. Interference on one site should correlate with DL traffic on the other site and the effect should be mutual. Here is an example of LTE‐TDD UL interference due to propagation delay. Assuming special sub‐ frame structure 7 used (10:2:2), guard period and UpPTS length are both 0.14 ms (2 sym bols*1 ms/14 symbols). The safe and greenmount are 41.8 km (0.139 ms) apart, likely that there is additional delay not considered here, and so DwPTS from one site overlaps UpPTS on the other due to propagation delay. The maximum “visible” eNB to eNB distance is based on GP, which is depicted in Table 5.1. If two eNB’s are separated greater than GP, they can see each other, then DL and UL sub‐ frames will collide between each other, even though their TDD duty cycles are synchro nized. Separation greater than this limitation is required if there is geographic isolation between the eNBs. If DL of one eNB overlaps with the UL of other eNBs due to propagation delay is more than GAP, which results in IoT increasment at eNBs. In order to spot cross slot interference, it is needed to monitor I + N measurements for each PRB in GP, UpPTS and UL normal subframe respectively. If cross slot interference is founded, reconfiguring of PRACH (format 4) and SRS 199 LTE Optimization Engineering Handbook Obvious interference at GP and UpPTS symbols –60 –70 –80 –90 –100 –110 –120 –130 –140 –150 SF1 SF2 SF6 SF7 –75 Power (dBm) Amplitude 200 31280 8220 0 10,000 20,000 30,000 39480 40,000 62520 50,000 60,000 –100 –125 –150 –175 0 58368 10,000 20,000 Number of samples 30,000 40,000 50,000 60,000 Number of subcarriers Figure 5.14 Time (left) and frequency (right) domain power of all symbols. Table 5.1 Maximum “visible” eNB to eNB distance based on GP. Normal cyclic prefix Configuation 0 DwPTS GP UpPTS Max eNB to eNB (km) 3 10 1 214 1 9 4 1 86 2 10 3 1 64 3 11 2 1 43 4 12 1 1 22 5 3 9 2 193 6 9 3 2 64 7 10 2 2 43 8 11 1 2 22 in UpPTS into UL normal subframe is necessary. Sometimes the worst thing is only a few OFDM symbols interfered, in this case, data symbols may suffer cross slot interference, but RS symbols do not, IRC doesn’t work and the interference can not be canceled in this scenario. Antenna orientation and tilt changes at both ends can introduce 50 dB or more isolation in addition to pathloss due to distance separation. With proper RF planning, it can have eNBs at LOS as close as 34 km using special subframe 7 and not interfere with each other. If sufficient pathloss isolation cannot be achieved then consider using a special subframe with a larger GAP. It notes that when using FDD, the interference between neighboring eNBs is much lower than when using TDD. 5.3.3 PUSCH Tx Power Analysis The eNB received PUSCH power spectral density (PSD) will be compared with the target PUSCH PSD and TPC values will be generated based on the difference between the two values. The target PUSCH PSD default value will be estimated based on α used by UE in its pathloss calculation and the corresponding theoretical optimum operating point based on spectral efficiency. To support network interference control, the target PUSCH PSD will have to be on a per UE basis. In the UL, the power control target for PUSCH and PUCCH can be tuned by the operator. The power control target for PUSCH is specified by p0NominalPusch that given by SIB2 and RRC connection reconfigration message. Depending on the difference between the SINR measured Internal Interference Optimization Power [dBm] Power [dBm] PUSCH TX power PUCCH TX power UE power limited pZeroNominalPusch –103 PUSCH RX power Path loss [dB] Path loss [dB] pZeroNominalPucch –117 SINR [dB] Thermal noise + interference PUCCH RX power SINR [dB] Thermal noise + interference Figure 5.15 p0NominalPucch and p0NominalPusch settings (example). and the SIR target, the eNB will send TPC (transmit power control) commands to ask the UE to increase or decrease the PPUSCH of its next message, this is the closed loop (slow inner‐loop) power control. The parameter is the target PSDrx that can be tuned to find a trade‐off between UL coverage and UL capacity as well as UL user bit rates experienced close to the site. Tests show that in good channel conditions p0NominalPusch is around −106dBm is enough to reach peak rate in a two receive antennas configuration, whereas poor channels require higher values. In a four‐receiver antenna configuration, p0NominalPusch can be further lowered by 3 dB. p0NominalPusch can be adjusted to 2 dB above noise and interference level, to overcome the interference. Noise and interference level on PUSCH can be measured by the OMC counter (Figure 5.15). Assuming p0NominalPusch = −106dBm, the thermal noise is −118 dBm/RB, and no interference from other UEs since there was no load in the network. The SINR in UL was estimated from the UE Tx_power using the following formula: UL _ SINR min UETx _ power 10 log nRB , 106 RS _ power RSRP 118 RS Power RSRP Increasing p0NominalPucch and p0NominalPusch can help to mitigate the adverse impact on perfor mance due to interference. It should be highlighted that increasing the values of p0Nominal would imply in a trade‐off between UL coverage and UL capacity. Setting p0Nominals well above noise and interference level will lead to high UL interference and UE battery consumption. Besides, the UE specific parameters of p0UE‐PUSCH’s initial setting is according to the UE’s category and RF conditions, the function of the parameter is for compensating the inac curate RF link pathloss that works together with p0Nominal, to guarentee the proper UE Tx_power. Thus, P 0 _ PUSCH p0 NominalPUSCH p0UE PUSCH For the closed loop power control adjustment, two power control methods shown in Table 5.2 are tested in the field; the absolute type is used in area1 and the accumulated type is used in area2. The result shows that there has similar trend on the PUSCH power consumption per byte in two areas. In low pathloss range, averagely, it’s shown that area2 uses less power per byte data, in the range of pathloss between 90 dB and 120dBm, it’s shown that area2 also uses less power per byte on the high power efficiency side, which is shown in Figure 5.16. 201 LTE Optimization Engineering Handbook Table 5.2 Parameters used in PUSCH power control test. Cell‐specific (SIB2) UE‐specific (RRC connection set up) Parameters used in PUSCH power control Area1 Area2 p0‐NominalPUSCH(dBm) −75 −67 α 0.8 0.7 p0‐UE‐PUSCH(dB) 1 0 deltaMCS‐enabled PUSCH Power Per Byte (dBm) Close loop power control mode – accumulation enabled 0 0 FALSE TRUE Figure 5.16 PUSCH power consumption per byte of different P0 settings. 15 5 –5 –15 –25 –35 –45 –55 70 80 70 80 90 100 110 120 Pathloss (dB) 130 140 150 25 PUSCH Power Per Byte (dBm) 202 15 5 –5 –15 –25 –35 –45 –55 90 100 110 120 Pathloss (dB) 130 140 150 Higher values of parameters p0NominalPucch and p0NominalPusch shall be deployed in areas that suffer from low UL throughput and that average UL RSSI load is not that high (i.e., below −110 dBm). On the other hand, lower values of parameters p0NominalPucch and p0NominalPusch shall be used during special events like concerts, football games and highly loaded areas, for example, commercial centers and big squares. 5.3.4 UL Effect of P0 and α LTE fractional power control allows the operator to adjust the power control target using p0NominalPUSCH/PUCCH and α parameters. By setting α less than 1, a “fractional path loss Internal Interference Optimization 50 Power [p0 = –103] Power [p0 = –100] 40 Power [p0 = –106] 30 20 Power (p0 = –100) 10 Power (p0 = –103) Power (p0 = –106) –20 –99 dBm –92 dBm –10 –96 dBm –6 0 –6 3 –6 6 –6 9 –7 2 –7 5 –7 8 –8 1 –8 4 –8 7 –9 0 –9 3 –9 6 –9 –1 9 0 –1 2 05 –1 0 –1 8 11 –1 14 0 50 Power [A = 1] 40 Power [A = 0.9] 20 Power [A = 0.8] 10 Power (A = 1) Power (A = 0.8) –10 SR –6P 2 –6 5 –6 8 –7 1 –7 4 –7 7 –8 0 –8 3 –8 6 –8 9 –9 2 –9 5 –9 –1 8 0 –1 1 0 –1 4 07 –1 1 –1 0 13 0 Power (A = 0.9) R UE TX POWER 30 –20 –30 –40 ~–93 dBm ~–106 dBm RSRP Figure 5.17 Impact of p0 component (α = 1) and compensation factor (P0 = −100). c ompensation” is achieved. This has the effect that eNB RX received power gradually drops with the cell signal attenuation even though the UE is not power limited. This means that eNB RX received power will not be equal to P0_PUSCH. In a live network, usually tuning these parameters could be a temporary solution for UL RSSI. These two parameters can be tuned to find a trade‐off between UL coverage and UL capacity and UL user bit rates experienced close to the site. The setting of P0, that is the power received per RB, affects both coverage and capacity. A higher setting leads to higher cell throughput but also higher noise rise. A lower setting should reduce peak throughput in low load scenarios but could improve capacity at high load. Reduction of p0nominal helps reduce the number of UEs reaching maximum power, but the impact is not same as α. Reduction of α helps reduce the number of UEs reaching maximum power. The example shown in Figure 5.17 is based on UL PRB utilization = 10%. Figure 5.18 described the UL effect of P0 and α. ●● ●● ●● power offset P0: –– P0 ↑ = > P(i) ↑, potential limitation for cell‐edge users Pathloss compensation: –– α ↓ = > P(i) ↓, especially for cell‐edge users cell‐edge users discriminated with P0 ↑ and α ↓ (“capacity setting”) 203 204 LTE Optimization Engineering Handbook cell edge P(i) Macro case 1 Macro case 3 Pmax α<1 α=1 P0,2 “capacity” “coverage” P0,1 PL Figure 5.18 UL effect of P0 and α. Figure 5.19 shows the relation between PUSCH RSSI versus PUSCH SINR both from daily average statistics, under two operatos with different power control strategy. For operator A, open‐loop power control is used, P0_PUSCH = −106dBm, α = 1, and each dot presents one eNB. For operator B, closed‐loop power control is used, RSSI target ~ −100dBm, SINR target ~ 20 dB, and each dot presents one eNB. UL power needs to be set by UL interference, when UL interference is larger/lower than a certain value (e.g., −95dBm). Simulations indicate that an α setting around 0.8 may be an opti mized setting for the UL peak rates and the UL cell throughput, it is suggested to use the parameter settings below as present in Table 5.3. PUSCH/PUCCH target values will increase if cells suffer from external interference. The cell had a very high service drop rate, also DL cell throughput was very low due to high UL/DL packet error rate. The reason is the cell suffers from very high UL RSSI (−100dBm) that external interference is suspected. One example shown in Figure 5.20 presents the performance impacted with the changes of P0. After applying the p0NominalPusch/Pucch changes, p0NominalPusch changed to −92dBm from −106dBm, p0NominalPucch changed to −96dBm from −120dBm, several KPIs of the cell significantly improved, especially retainability and throughput. 5.3.5 PRACH Power Control High initial Message3 power means high initial PUSCH transmission power, and will take the closed loop power control many steps to bring the PUSCH power to normal, which results in further UL interference in the system. A live example of parameter setting is in Figure 5.21. According the assumptions, Message3’s power is calculated as: MSG3 Tx_power = 10*log8 dB (RB number) + 32dBm (PRACH power) + 12 dB (deltaPreambleMsg3) + 6 dB (TPC) = 9 + 32 + 12 + 6 = 59dBm Actually, the above Message3’s power is not suitable. Table 5.4 gives four sets of PRACH typi cal parameters configuration in two areas, which were tested in field to find out the proper PRACH parameters settings. As shown in Figure 5.22, it can have following observations. For Area 1, F(0) is almost fixed at 6 dB, Message3 is almost transmitted at maximum power (23dBm), for Area 2, parameter setting is not too aggressive, most of F(0) is −6 dB, Message3 is transmitted at variable power. It is recommended that deltaPremableMsg3 setting to 2 dB to 4 dB; PC_msg2 setting to 0 dB; Mpusch setting to1RB to 3RB. Internal Interference Optimization –125 noise floor –119dBm PUSCH SINR [dBm] 35 30 25 20 ce en L hU er erf 15 int it sw 10 S BT –120 –115 –110 –105 –100 5 –95 –90 –85 0 –80 –5 –10 –15 PUSCH RSSI per PRB [dBm] (open-loop power control) 30 25 PUSCH SINR [dB] 20 –120 15 10 BTSs with UL interference 5 0 –115 –110 –105 –100 –95 –90 –85 –80 –5 –10 –15 PUSCH RSSI per PRB [dBm] (closed-loop power control) Figure 5.19 PUSCH RSSI versus PUSCH SINR. Table 5.3 Parameters settings of PUSCH/PUCCH power control. Recommended value Parameter UL interference larger than ‐95dBm UL interference lower than ‐95dBm PzeroPUCCH −103 −103 PzeroPUSCH −70 −97 α 0.8 1 205 LTE Optimization Engineering Handbook Service_Drop_Rate_SR 3.50 3.25 Service_Drop_Rate_SR (%) 3.00 2.75 2.50 2.25 2.00 1.75 1.50 1.25 1.00 0.75 0.50 3-Mar 1-Mar 27-Feb 25-Feb 23-Feb 21-Feb 19-Feb 17-Feb 15-Feb 13-Feb 9-Feb 11-Feb 7-Feb 5-Feb 3-Feb 1-Feb 30-Jan 28-Jan 26-Jan 24-Jan 22-Jan 20-Jan 18-Jan 16-Jan 14-Jan 12-Jan 0.25 Time (timezone: Europe/Madrid) Service_Drop_Rate_SR (Total) Downlink_Throughput_Kbps 3-Mar 1-Mar 27-Feb 25-Feb 23-Feb 21-Feb 19-Feb 17-Feb 15-Feb 13-Feb 11-Feb 9-Feb 7-Feb 5-Feb 3-Feb 1-Feb 30-Jan 28-Jan 26-Jan 24-Jan 22-Jan 20-Jan 18-Jan 16-Jan 14-Jan 30,000 29,000 28,000 27,000 26,000 25,000 24,000 23,000 22,000 21,000 20,000 19,000 18,000 17,000 16,000 15,000 14,000 13,000 12,000 11,000 10,000 12-Jan Downlink_Throughput_Kbps 206 Time (timezone: Europe/Madrid) Downlink_Throughput_Kbps (Total) Figure 5.20 p0NominalPusch/Pucch affected the throughput. 5.3.6 SRS Power Control Every UE in the UL uses SRS (sounding reference signal) and UE used same TPC commands for both SRS and PUSCH transmissions. It is used for the following reasons: UL radio link supervision, UL channel estimation for UL FSS scheduling, and UL timing alignment. Internal Interference Optimization Figure 5.21 A live example of UE Tx_power. Table 5.4 PRACH typical parameter setting. Parameter (HO RACH) Mpusch deltaPremableMsg3 (dB) PC_msg2 (dB) Area 1 Area 2 Set 1 Set 2 Set 1 Set 2 (19:43:14.753) (19:38:48.689) (20:11:10.96) (20:11:12.23) 8 8 3 3 12 8 0 0 6 8 −6 4 UL parameter pSRSOffset is an offset applied in relative to the PUSCH power that PSRS_OFFSET is a power offset configured by parameter pSRSoffset. It had been discovered the pSRSOffset values that are being used in the network are more optimistic than needed. This has resulted in optimizing the SRS power control parameter, which in turn improved the UL noise and significantly improved every metric in the network (retainability, throughput, BLER, UL noise, data volume, etc.) (Figure 5.23). 207 LTE Optimization Engineering Handbook PUSCH Actual Tx Power F(i) 24 35 21 Area1 30 Area2 25 18 20 15 15 12 10 f(0) = ΔPrampup + δmsg2 9 5 6 0 3 –5 0 –10 19:27:44.44 19:28:47.47 19:31:04.04 19:33:11.11 19:35:29.29 19:40:09.09 19:42:46.46 19:44:25.25 19:45:51.51 19:49:02.02 19:51:58.58 19:54:01.01 19:55:21.21 19:57:49.49 19:59:59.59 20:03:07.07 20:07:21.21 20:11:12.12 20:13:40.40 20:17:48.48 20:21:13.13 20:22:53.53 20:25:46.46 20:28:31.31 208 Figure 5.22 F(0) and PRACH Tx_power. Ks = 0 (MCS) pSRSPoweroffset (eff) The example values for pSRSOffset is 7, some networks proposed value is 3 –10.5 pSRSPoweroffset (config) 0 –9 1 –7.5 2 –6 3 –4.5 4 –3 5 –1.5 6 0 7 1.5 8 3 9 4.5 10 6 11 7.5 12 9 13 10.5 14 12 15 uplinkPowerControlDedicated { p0-UE-PUSCH 0, deltaMCS-Enabled en0 accumulationEnabled TRUE, p0-UE-PUCCH 0, pSRS-Offset 7 } soundingRS-UL-ConfigDedicated setup : { srs-Bandwidth bw0, srs-HoppingBandwidth hbw3, freqDomainPosition 0, duration TRUE, srs-ConfigIndex 17, transmissionComb 0, cyclicShift cs4 } Figure 5.23 pSRSOffset from RRC connection set up message. The transmit power of the SRS signal is set in dBm at the UE as: PSRS i min PMAX , PSRS _ OFFSET 10 log10 MSRS PO _ PUSCH j PL f i where MSRS is the bandwidth of the SRS transmission expressed in number of resource blocks, f (i) is the current power control adjustment state for the PUSCH. In addition to pSRSOffset, there is room for optimizing other SRS related parameters like SRS bandwidth. srsBandwidthConfiguration is set to bw0, which corresponds to 48 RBs in the UL SRS. The operator needs to know if it really need 48RBs for UL SRS or it can be configured with 24RBs every alternate RB and still accomplish UL FSS. Sometimes by reducing the SRS bandwidth in half, it can not only improve UL noise but also UL throughput now that symbol can be used for PUSCH(data). Internal Interference Optimization 5.3.7 Interference Rejection Combinin Interference rejection combining (IRC) is a method to enhance the capacity in UL by mitigating the undesirable inter‐cell interference. IRC is a receiver technique using multiple RX antennas (up to 4) inputs to suppress interference and achieve better performance in term of SNR and BLER. IRC utilizes (unlike MRC) correlation in the spatial domain between antennas and in the time domain to suppress interfering signals from other cells. The IRC feature uses a linear MMSE (minimum mean square error) estimator algorithm to combine Rx‐signals from two or more antennas to sup press interference efficiently. The suppression can be seen as weighting down the signal in the direction of an interferer, so that it does not corrupt the signal from the desired user. IRC improves the BLER on the PUSCH and PUCCH compared to MRC and improves UL through put in UL interference limited systems. The IRC feature uses an “receive signal combining algorithm” that can be viewed as an extension of the MRC algorithm (Figure 5.24). Figure 5.24 IRC gain. 2 epa5_16qam_6rb_1rb, SNR=10dB x 106 IRC MRC 1.8 1.6 IRC gain at 90% throughput Throughput [bps] 1.4 1.2 1.0 0.8 IRC gain at 50% throughput 0.6 0.4 0.2 0 –20 –15 –10 –5 0 5 C/I [dB] 10 15 20 25 209 210 LTE Optimization Engineering Handbook IRC will maximize SINR of the wanted user by applying complex weights to the antenna e lements, IRC is performed per subcarrier, it cannot cancel white noise, just (spatially correlated) interferers. IRC can cancel up to N−1 interferers where N is number of antenna elements. In LTE network, IRC is applicable to PUSCH only, control channels are received using MRC or incoherent combining. 5.4 Inter‐Cell Interference Coordination If the adjacent cell‐edge users use the same resources to transmit/receive in UL/DL, then it will be caused by UL/DL inter‐cell interference. Inter‐cell interference coordination (ICIC) func tionality can reduce the impact of both UL/DL inter‐cell interference to improve throughput performance, particularly important for users with low or medium load. When using dynamic ICIC, eNB selects the starting point of UL/DL PRB in frequency randomly, either the lowest or highest possible frequency. ICIC limited the use of radio resources among the various cells in a inter‐cell coordinated manner, including limit the use of frequency resources or in a certain time‐frequency resources limit its transmission power and so on. UEs in different cells will interfere due to using the same frequencies, ICIC distributes the interference between cells. According to E‐UTRAN network architecture, UL dynamic ICIC program needs information interaction between the inter‐cell. There is two interactive information: HII (high Interference indicator) and OI (overload indicator), wired interface X2 between neighboring eNB can be used to send the HII and OI. 5.5 UL IoT Control Excessively high IoT is observed in certain deployment cases due to the massive UL traffic increase or improper parameters settings. For example, in mixed metro/macro‐cell case where there’s a huge difference in cell coverage between the neighbors. High IoT leads to high UE Tx power (battery drain) as well as degraded UL throughput performance. 5.5.1 UL Interference Issues and Possible Solutions UL interference problems are likely to increase as networks evolve due to phenomena like traf fic increase, Hetnet and network densification. The challenge is to address UL performance issues of traffic and external interference, DL/UL unbalanced cells in such evolved networks. The solutions are short term actions as mentioned before, link adaptation, power control, inter‐cell interference coordination and handover, and admission control that is needed to work hand‐in‐hand. For handover, the currently existing mobility triggers for bad coverage are only based on DL UE measurements are released with redirect, and IRAT and inter‐frequency handover. 5.5.2 UL IoT Control Mechanism Currently UL IoT management is indirectly controlled by FPC (fractional power control) set tings in LTE. FPC parameter settings are depicted in Table 5.5. FPC settings determine the UL SINR targets, which decides the UE Tx_power allowance based on UL pathloss condition. Without FPC, eNB maintains a fixed UL SINR target for PUSCH, so it allows for comparable average UL throughput across cell until UE begins to hit the maximum Tx_power. Internal Interference Optimization Table 5.5 FPC parameter settings. α factor 0.8 uplink SIR_target value for dynamic PUSCH scheduling 19 dB Max. uplink SIR_target value for fractional power control 19 dB Min. uplink SIR_target value for fractional power control 0 dB PLnominal 60 dB Notes: ●● ●● ●● ●● α = 0.8 has been shown to provide the best overall performance in the field, particularly when considering 3D antenna patterns with application of downtilt. min target SINR = 0 dB, which is chosen to ensure no loss of synchronization of UEs at cell edge. max target SINR = 19 dB, which is chosen to ensure highest MCS level can be achieved for UEs in near‐cell conditions. nominal target SINR and nominal path loss (PLnominal) are chosen for particular scenario to give best performance, depends on system loading and desired tradeoff between average user throughput and edge user throughput. In case of limited interference (unloaded neighbor cells) scenario, the higher the setting of UL SIR_target for dynamic PUSCH scheduling, the higher the throughput, but also, the higher the setting the higher the interference generated in the neighboring cells. In loaded network scenarios, the higher the SINR target, the higher the near cell throughput but the higher the interference generated in the different cells of the network and thus the lower the cell‐edge throughput and at some point the lower overall cell throughput too. Consequently, the IoT is based on the relative UL pathloss to the neighboring cells. High IoT drives UL power control to compensate, which in turn creates more interference to neighbors, leading to power control race condition similar to CDMA heavy‐loaded case. Excessively high IoT could even deafen the eNB radio, distort the received signal, and severely impact the UL operability. Thus, connection/handover success and call drop KPI will be impacted. eNB uses the local measurement of interference plus noise (I + N) level as the indication of IoT. To avoid generating excessively high IoT, the UE Tx power needs to be lowered when high IoT is observed. IoT control mechanism in LTE is shown in Figure 5.25. loT? Handover, Call drop rates need to meet KPI targets Feedback: 8–10 dB might be appropriate Cell edge rate: Would like to maximize cell capacity and cell edge rate maxSIRtarget = 19dB Target SINR minSIRtarget = 0dB Cell capacity Higher UL_SINR target allows higher cell throughput, but cell edge users create significant interference to neighboring cells (IoT), and eventually lower total throughput on the UL. If high SINR target was set close to cell, then linear adjustment (in dB terms) beyond a certain pathloss should be done according to fractional power compensation control. Figure 5.25 IoT control mechanism. 211 LTE Optimization Engineering Handbook To avoid generating excessively high IoT, the UE Tx_power needs to be lowered when high IoT is observed. Cell adjusts UL _SINR target for a given user based on estimated UL pathloss. This corresponds to lowering the UL_SINR target at the cell level when elevated IoT level beyond acceptance is observed. 5.5.3 PUSCH UL_SINR Target Calculation PUSCH UL_SINR target can vary from initial target according to experienced path loss. eNB computes closed loop target UL SINR for a given UE based on estimated UL pathloss, which is as follows: SNRCL _ T arg et min max PL PLnominal , SIRTarget _ nominal 1 min SIRTargetFPC ,max SIRTargetFPC where PLnominal = 60dB, it is usually set to the value of the UL pathloss for a UE at cell centre. Parameter PLnominal configures the nominal path loss and corresponds to the path loss at which it wants the SINR target to be UL SIR_target value for dynamic PUSCH scheduling, according to fixed UL SINR target in Figure 5.27, which is also the nominal SINR target. The slpoe is 1−α. PL is an estimate of the average path loss based on the power headroom reports of the UE and the average SRS power. eNB broadcasts p0NominalPusch to assist UE in computing initial UE Tx power: P0 _ PUSCH _ nominal TxPSD dBm/ PRB SINRTarget _ nominal 1 P0 _ PUSCH _ nominal PLnominal ITOT PL UETx _ power TxPSD dBm/ PRB 10 x log10 # of PUSCH PRBs So, PUSCH SINR target can be represented as the function of UL path loss for a selection of IoT levels, as shown in Figure 5.26. 20 18 TargetUL PUSCH SINR (dB) 212 maxSIRtarget 16 Slope 14 12 Fixed UL SINR Target 10 8 6 4 PLNominal 2 minSIRtarget 0 40 50 60 70 80 90 100 110 120 Path Loss (dB) Figure 5.26 PUSCH SINR target versus UL path loss. 130 140 150 160 PUSCH SINR target (dB) Internal Interference Optimization 22 20 18 16 14 12 10 8 6 4 2 0 –2 1 FPC - no IoT correction 2 PUSCH SINR target For IoT=0dB leading to PUSCH SINR target correction –3dB 3 PUSCH SINR target For IoT=4dB leading to PUSCH SINR target correction 0.5dB 4 PUSCH SINR target For IoT=9dB leading to PUSCH SINR target correction 3dB 5 PUSCH SINR target For IoT=15dB leading to PUSCH SINR target correction 6dB 5 0 20 40 60 80 100 120 4 140 31 2 160 180 200 UL path loss (dB) Figure 5.27 PUSCH SINR target versus UL pathloss (example). Note that when the full path loss compensation is used, that is, when α is set to 1.0, the target SINR is always equal to UL SIR_target for dynamic PUSCH scheduling and other parameters are ignored. Following equation is also used to compute the p0NominalPusch power based on PUSCH SINR target: P0 _ PUSCH _ nominal SINRTarget _ nominal 1 PLnominal ITOT Assume α = 1, and SINR_target_nominal = 1 dB, p0NominalPUSCH = 1 + 0 −112 = −111 dBm Assume α = 0.7, and SINR_target_nominal = 15 dB, p0NominalPUSCH = 15 + (1−0.7)*140 – 112 = −55 dBm So, if aggressive FPC mechanism is adopted in no‐load scenarios, cell throughput is higher, if FPC behavior is toned down in high‐load scenarios, call set up and call drop KPIs will be kept at an acceptable level. Figure 5.27 shows an example of PUSCH SINR target versus UL pathloss. Figure 5.28 further illustrates the power control behavior under different pathloss condi tions, it can be divided into four regions with different characteristics. For region 1, UE power control is working within the dynamic range. The power control target p0NominalPusch is met and SINR of PUSCH is constant. For region 2, the UE is transmitting at its maximum power. Received power level per RB and SINR for PUSCH decreases with a higher signal attenuation until SINR reaches the minimum required SINR for PUSCH. For region 3, the received power level per RB and SINR for PUSCH are constant even though the signal attenuation keeps increasing, hence the number of allocated RBs decreases in order to meet the minimum required SINR. For region 4, the number of allocated RBs has reached its minimum, which is two at the moment and SINR for PUSCH decreases with a higher signal attenuation. By increasing p0NominalPusch, the region 1 will be shrunk, the region 2 will be extended further into region 1 and the rest of the region unchanged due to limitation on maximum UE transmit power, which is shown in Figure 5.29. 5.5.4 UL Interference Criteria Two types of quantities were considered as candidates for the UL triggers mechanism: GINR (gain to interference plus noise ratio) and SINR. The main difference between the two quantities 213 RB dBm 60 Pue_TX power used per resource block SINR 50 RB 40 30 20 SINR 0 –10 74 77 80 83 86 89 92 95 98 101 104 107 110 113 116 119 122 125 128 131 134 137 140 143 146 149 152 10 –20 –30 1 2 Pue_TX power used per resource block 3 path loss Figure 5.28 Power control behavior under different pathloss conditions. Uplink SINR/UE Tx/RBs vs RSRP 35 Total UE Tx Power UL SINR UL RBs (PUSCH) 80 15 60 5 –5 40 –15 20 –25 80 100 120 Pathloss [dB] 140 UL Resource Blocks UL SINR [dB], UE TX power [dBm] 100 25 0 160 p0NominalPusch = –106 dBm Total UE Tx Power UL SINR UL RBs (PUSCH) 100 25 80 15 60 5 –5 40 –15 20 –25 80 100 120 Pathloss [dB] 140 p0NominalPusch = –97 dBm Figure 5.29 Different p0NominalPusch versus UE behavior. 0 160 UL Resource Blocks UL SINR [dB], UE TX power [dBm] Uplink SINR/UE Tx/RBs vs RSRP 35 4 Internal Interference Optimization RSRP GINR RSRQ SINR Cell center Cell edge Figure 5.30 Expected GINR and SINR average behavior. is that the gain quantity includes the estimated transmitted power density, not only the received power density. GINR PSDRX PSDTX N I ; SINR PSDRX N I where, PSD ‐ power spectral density; N ‐ noise; received noise PSD; I ‐ interference; received interference PSD; PSDRX ‐ received signal PSD; PSDTX ‐ estimated UL transmit PSD based on power headroom reports from the UE; In a live network, GINR can be chosen as the quantity for the UL interference criteria for the following reasons. This quantity is already calculated and used in link adaptation and the vari ation of the value of GINR with time and distance from cell is much more significant than the variation of SINR. As shown in the picture in Figure 5.30, the expected RSRP, RSRQ, GINR, and SINR average behavior, moving from cell center toward border. The circled lines, it is visible that for GINR a modification of the distance/time quantity gives a much larger variation of the measured quantity than for SINR. In fact, the link adap tation algorithm tends to maintain a constant SINR. SINR is algorithm‐independent, while GINR is not. The trigger shall be based on in eNB calculated UL GINR. It shall be possible to trigger bad coverage based on UL when GINR value is below the configured GINR bad coverage threshold. 215 216 6 Drop Call Optimization 6.1 Drop Call Mechanism Session errors are used to define the event when the FTP session can not be completed for whatever reasons. RRC drops can be considered as radio link failure. After a “RRC drop,” UE will maintain its context in the MME and SGW and the PGW will continue to anchor the IP address. So for internet surfing, the RRC drop may not be considered a user perceived drop as voice. Generally call drop during drive test is counted when RRC connection reestablishment occurs. Drop call will cause abnormal release. The definition of an abnormal session release can be presented by that the release of the E‐RAB had a negative impact on the end‐user. In a packet domain system as LTE it is natural to establish and release E‐RABs. E‐RABs don’t have to be released just because they are inactive, they can be kept to have a fast access once new data arrives. Drop call usually caused by radio link failure (RLF). To check this there are two criteria, if the E‐RAB was considered active at the time of release and the cause value of the release. There are several radio connection supervision schemes in LTE to monitor the air interface link. ●● ●● ●● ●● UE detected DL sync by decoding PDCCH/PHICH and the BLER performance UE detected radio link failures (RLF) by T310 expiry, maximum number of RLC retransmissions, integrity check failure, handover failure (T304 expiry) or non‐handover related random access problem eNB detected radio link failures (RLF) by PUSCH RLF, CQI RLF, Ack/Nack RLF, RLC failure SRB/DRB, also by vendor specific, like UE capability enquiry time out, security mode complete time out, RRC connection reconfiguration complete time out, RRC connection reestablishment complete time out, tRelocOverall time out (e.g. handover preparation or handover execution takes too long time) etc. eNB initiated release: TA timer expiry and maximum RLC (radio link control) retransmissions exceeded Radio link failure (RLF) means the radio link between the eNB and the UE is lost. Once the eNB has detected the loss of the radio link and the radio link failure condition is met, the eNB requests the MME to release the UE context in the eNB, by sending the UE context release request message, with the cause “Release due to E‐UTRAN generated reason,” to the MME. Finally the UE switches to idle state, and the call is dropprd. Once the UE re‐enter the coverage area it may initiate the UE triggered connection re‐activation at any point in time. The main radio link failure and call drop reasons are listed in Figure 6.1. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Drop Call Optimization Physical layer failure RLC layer failure PHY Layer Failure Handover Failure Out of Sync Coverage or Interference Problems Data RLC Unrecoverable Error Mac layer failure PDCP layer failure RRC layer failure SRB RLC Unrecoverable Error RACH Problems Parameter Settings RACH Issues integrity check failure Security Issues Re-configuration failure Radio Link Failure RLF Recovery Call Drop eNB Misconfiguration Figure 6.1 Main possible radio link failure and call drop reasons. Drop call measurements can be based on drive tests or statistics. The main reasons for poor retainability include but are not limited to missing neighbor relations, poor radio environment, badly tuned handover parameters, admission reject (due to lack of licenses). For UL, the radio link supervision is performed in the eNB, for DL, it is carried out by similar supervision function located in the UE. 6.1.1 Radio Link Failure Detection by UE Usually, radiolink failure will trigger RRC connection reestablishment by UE upon T310 expiry, reaching the maximum number of RLC retransmissions and handover failure (T304 expiry), or trigger RLF if UEs fail to read MIB/SIB1/SIB2. UE monitors the DL by decoding PDCCH/PCFICH. Related timers are T310 and T311; counters are N310 and N311. UE judges if there is still keeping DL sync1 by decoding PDCCH/ PCFICH. If the percentage of BLER of PDCCH/PCFICH is greater than or equals to 10%, UE will consider that it is an out‐of‐sync indication, if the percentage of BLER is less than or equals to 2%, UE will consider that it is an in‐sync indication as shown in Figure 6.2. If there are N310 consecutive out‐of‐sync indications, UE starts T310. If there are N311 consecutive in‐sync indications, UE stops T310. If T310 expires, UE will consider a radio link failure, it means that UE has lost the sync with eNB. Then UE starts T311 and goes to reestablishment procedure (Refer to 3GPP 36331, 5.3.11; see Figure 6.3). In case of N310 consecutive out‐of‐sync indications on the UE side as shown in Figure 6.4, the UE will not initiate reestablishment until T310 has expired. If T311 expires (meaning no cell found), then UE goes to IDLE and starts cell reselection. The UE uses timers T310 and T311 to get time to restore the connection with the eNB. During the time T310 + T311 the UE stays in RRC_connected state. If the UE can not reestablish in-sync in-sync in-sync out-sync out-sync out-sync out-sync out-sync out-sync out-sync out-sync BLER of reference PCFICH + PDCCH 10% 2% time Figure 6.2 DL synchronization. 1 The UE is expected to monitor the RS in the DL, based on the RSRP. The UE will determine if it can decode the PDCCH according to the specs. If the reference signals have enough strength such that the UE can decode consistently the PDCCH, then the link is in-sync. 217 218 LTE Optimization Engineering Handbook First Phase normal operation radio problem detection no recovery during T310 no recover during T311 RRC_CONNECTED n311 consecutive outof-sync indications n311 consecutive in-sync indications during t310 Cell reselection and tracking area update if RRC reestablishment fails Second Phase goes back to idle RRC_IDLE radio link failure RRC connection re-establishment attempted during t311 Figure 6.3 RLF due to T310 expiry at UE. UE N310 times out of sync is sent from lower layers. e.g. not able to decode 20 consecutive subframes (PDCCH) N310 times UE detects radio problems and starts T310 T311, default value is 3s. Start upon initiating the RRC connection re-establishment procedure, stop at selection of a suitable cell. When T311 is at expiration, the UE enter RRC_idle state. 1st OOS occurrence N310 times OOS If one of the following conditions is met: 1. T310 expires 2. Random access problem indication from MAC layer 3. Max RLC retransmission reached eNB Default RRC timers: T300 ms 1000, T301 ms 400, T310 ms 2000, T311 ms 3000, T304 ms 1000, T301, default value is 400ms. Start at the transmission of RRC connection reestabilshment request, stop at reception of RRC connection reestablishment or reject message as well as when the selected cell becomes unsuitable. When T301 is at expiration, the UE enter RRC_idle state. Radio link failure detected Start RRC reestablishment procedure and start T311 Suitable cell found Send RRC reestablishment request and start T301 Figure 6.4 UE out‐of‐sync and re‐starts cell reselection. the connection to the eNB, the UE switches from RRC_onnected to RRC_idle and initiates the procedure to establish a new RRC connection. It may then establish a connection to a new eNB. The example displays the DL sync situation in Table 6.1 (refer to Figure 6.5), which shows that consecutive out_of_sync indications are up to 18 times. The setting of N310 is 20, N311 is 1, T310 is 2 s, and T311 is 3 s. Some random access procedures are triggered by scheduling request when UE needs to obtain UL grant to send data. During random access procedure, if UE reaches the RACH maximum attempts (preampleTransMax + 1), the random access does not bring success, such as after consecutive preambleTransMax message 1, and UE does not receive message 2, or after UE sends message 3, and does not receive message 4 (contention failed). In this situation UE will consider Table 6.1 DL sync situation. SFN Sub_fn Out_of_sync_BLER In_sync_BLER Out_of_sync_count In_sync_count 144 1 4.4% 53.0% 0 0 145 1 6.3% 58.4% 0 0 146 1 10.5% 60.8% 1 0 147 1 12.7% 65.0% 2 0 ….. ….. …… …… …… …… 161 1 42.0% 83.6% 16 0 162 1 46.0% 83.7% 17 0 163 1 51.5% 84.4% 18 0 Drop Call Optimization Figure 6.5 The example of DL out of sync. 200 ms 100 ms If timer expires, Radio Link Failed In-Sync In-Sync Out-of-Sync Out-of-Sync Out-of-Sync Out-of-Sync If n311, then stop timer n311 1 t310 2s n310 20 a radio link failure. Then UE starts T311 and goes to reestablishment procedure, or directly goes back idle mode. RLF due to RACH triggered by UL data arrival is shown in Figure 6.6. UE will send to request an UL grant before UE had sent the maximum number of scheduling requests. UE cannot receive UL grants could be due to DL issues (weak DL SINR) or UE running out of UL coverage. The parameter of max number of scheduling requests is recommended as 64 in such scenario. In some random access procedures, the RACH trigger reason is handover. When UE receives handover command, it starts timer T304. If RACH procedure to the target cell does not s ucceed by the time T304 expires, handover failure occurs, UE tries RRC connnection reestablishment with reestablishment cause “handoverFailure.” RLF due to handover failure is shown in Figure 6.7. Random access triggered due to missing PUCCH SR resources, or PDCCH order normal operation First Phase Second Phase Attempting PRACH to serving cell no recovery during T311 RRC_CONNECTED goes back to idle RRC_IDLE RACH failure UE attempting random access to serving cell. Cell reselection and tracking area update if RRC Reestablishment fails radio link failure RRC connection re-establishment attempted to serving cell during t311 Figure 6.6 Non‐handover–related random access issue result RLF. normal operation First Phase Second Phase Attempting PRACH to target cell no recovery during T311 RRC_CONNECTED Handover command T304 running while UE attempting access to target cell. RLF detected T304 Figure 6.7 Handover failure. goes back to idle RRC_IDLE RRC connection reradio link failure establishment attempted to source or target cell during t311 T304 expires RACH preamble RACH Handover start required Cell reselection and tracking area update if RRC Reestablishment fails T304 is strated by the UE following handover command. The UE detaches from the source cell and following acquisition of the target cell, use RA procedure to attempt access. If the UE is unable to access the target cell before the expiration of T304, a RLF is detected. 219 220 LTE Optimization Engineering Handbook RLC retransmission until max RLC retx threshold is reached normal operation First Phase Second Phase RLC retransmissions until max value no recovery during T311 RRC_CONNECTED RLC Failure Radio Bearer Failure UE Release Cell reselection and tracking area update if RRC reestablishment fails goes back to idle RRC_IDLE radio link failure RRC connection re-establishment attempted during t311 dlMaxRetxThreshold Figure 6.8 Maximum UL RLC retransmissions reached. In addition to RLF and inactivity time out functions, there is a RLC algorithm managing generation of RLC failures. The RLC function is responsible for error correction through ARQ, protocol error detection, and recovery. UE receives the RLC status report from eNB after transmitting data. If UE do not receive the ACK message for relevant RLC block, UE needs to re‐transmit the block. If UE reaches the maximum re‐transmitting times, UE considers a radio link failure. Then UE starts T311 and goes to reestablishment procedure, or directly goes back idle mode (Refer to 3GPP 36322 5.2.1, 36331 5.3.11; see Figure 6.8). Finally, all the radio link failure situations mentioned above can be detected by UE. The parameters related to RLF detection and recovery can be refered to Annex LTE timers. 6.1.2 RadioLink Failure Detection by eNB 6.1.2.1 Link Monitors in eNB eNB radio link problem detection mechanisms are vendor internally specified. 3GPP does not specify eNB radio link failures, but some vendors eNB mimics the behaviour of the UE RLF specified in 3GPP. Multiple‐link monitors shown below are defined to detect a radio link problem in the popular eNB. ●● ●● ●● ●● ●● UL PUSCH DTX detection for scheduled UL data. When UE is scheduled for PUSCH transmission, eNB expects to receive UL transmission on the scheduled PRBs. If signal from UE cannot be detected, PUSCH DTX is declared, the detection of the radio link problem by the UL scheduler is based on the comparison of UL grant assignment and the DTX detection on PUSCH for the assigned PRBs. The detection result is received by LTE MAC. No allocated preamble received as a response to PDCCH order. CQI DTX detection for periodic CQI reports in PUCCH and PUSCH. If MAC layer receives CQI DTX consecutive reports from UL PHY layer, the eNB MAC declares a radio link problem. UL Ack/Nack DTX detection for transmitted DL data. If for consecutive DL resource allocations to the same UE the HARQ feedback from the UE is always “DTX,” either for a configurable period of time or for a configurable number of consecutive DL resource allocation attempts, DL scheduler declares radio link problem. SRS DTX detection for radio link problem detection. If MAC layer receives multiple SRS DTXs consecutive reports from UL PHY layer, the MAC declares SRS radio link problem. Some vendors defined UL MAC RLF by SRS SINR as shown in Figure 6.9. Each link monitor has its internal criteria and filtering with internal counters to decide when there is a radio link problem occurring and when there is a radio link problem recovery occurring. When a radio link problem is detected, an eNB internal radio link recovery timer is started. Drop Call Optimization SRS SINR > Threshold1 UL IN_SYNC UL OUT_SYNC SRS SINR > Threshold2 MAC RLF TA Failure UE RACH RRC_CONNECTED Expiry of Inactivity Timer TA UL in-sync TA UL out-of-sync RA success, PDCCH order success RRC Connection Reestablishment RRC Connection Release Radio link failure RRC Connection Setup RRC Connection Release RRC_IDLE Figure 6.9 UL synchronization based on the SRS SINR. The timer is stopped when in case of radio link failure recovery. In case the radio link recovery timer expires the RRC connection is released as well as the S1 released by the eNB using eNB initiated S1 release + RRC connection release, S1 release signaling will cause ECM connected state change to ECM idle state, so that the timer should be longer than T310 + T311 that the UE can perform RRC connection reestablishment to the source cell or to any other cell (Figure 6.10). 6.1.2.2 Time Alignment Mechanism UL time alignment (TA) update is done periodically on per‐need basis. UE sends to eNB TA measurements based on PUSCH, PUCCH signals. eNB calculates the timing advance of every UE and deliver it to UE by TAC (timing alignment command) before the expiry of this timer so as to keep the UE in sync. UE uses this timing advance to adjust the UL transmission to maintain UL sync. As UE detects out‐of‐sync status using a time alignment timer (taTimer, in SIB2), the timer shall be started or restarted whenever an initial TA or a TA update command is 221 222 LTE Optimization Engineering Handbook UE eNB MME S-GW Detect radio link problem, T_RLF expires UE in RRC-CONNECTED S1AP: UE Context Release Request S1AP: UE Context Release Command RRC: RRC Connection Release S1AP: UE Context Release Complete Release all UE related resources, remove UE context Set UE to RRC-IDLE S11 interaction to inform S-GW about connection release Set UE to ECM-IDLE Figure 6.10 RLF triggered by eNB procedure. Traffic Inactivity Stop TA TA Timer UL dedicated resource = OFF eNB TA Ack TA Ack TA Ack DRX Inactivity Timer RRC connection release UE Enter Long-DRX state Figure 6.11 Active to idle state. received (see [3GPP‐36.321], section 5.2). eNB shall send timing advance command in order to avoid expiration of time alignment timer as shown in Figure 6.11. For UE, if time alignment timer expires, UE knows the UL sync is lost, and the UE will not be scheduled. In this situation, UE will clear HARQ buffer, release PUCCH/SRS and clear the received UL grant. After that, if UE wants to transmit data, it needs to re‐sync UL with eNB by random access procedure. The example shown in Figure 6.12 is multiple TA_expire and MAC_ RA_problem, in areas where there are numerous servers of low RSRP. For eNB, eNB will find out the out of sync of UL when it can not measure the exact TA of UE. In a live network, the setting of Timing Alignment Timer can be infinity (0). It means that UE will consider that UL is always sync with eNB. With this setting, if eNB find there is UL problem, such as eNB has lost sync with UE (eNB can not measure the TA of UE), eNB will mark this UE as no_sync. With this mark it will not send any signal to this UE. In this situation, if there is data which eNB needs to send to this UE, eNB can notice UE by “PDCCH ordered random access function” to let UE re‐sync the UL through random access procedure (Figure 6.13). The interval between periodic TA update commands is based on the timing alignment timer reduced by a configurable offset taTimerMargin. TA command period = taTimer – taTimerMargin Figure 6.12 Example of multiple TA_expire and MAC_RA_problem. 224 LTE Optimization Engineering Handbook UE eNB UE in RRC-CONNECTED MME S-GW TA timer expires S1AP: UE Context Release Request S1AP: UE Context Release Command RRC: RRC Connection Release S1AP: UE Context Release Complete Release all UE related resources, remove UE context Set UE to RRC-IDLE S11 interaction to inform S-GW about connection release Set UE to ECM-IDLE Figure 6.13 TA timer expires at eNB. Table 6.2 UL time alignment configuration. Parameters Range Description Setting taTimer 500 (0), 750 (1), 1280 (2), 1920 (3), 2560 (4), 5120 (5), 10240 (6) 10240 Determines the number of subframes after which a UE (6) assumes it is out‐of‐sync in UL if no time alignment command was received. The increased TA timer saves resources due to reduced number of TA command transmissions and also allows UEs more time for answering TA command and thus, probability for RLF might be less in case of UE is shadowed for short time only or not responding for short time. taTimerMargin 0…2560 The parameter defines lead with respect to the taTimer expiration time for starting to send the periodic timing advance command. The actual time interval between updates will be TimeAlignTimer – taTimerMargin. 2000 taMaxOffset Determines the maximum allowed time alignment offset. If the value is exceeded, TA command is sent to the UE to adjust UL timing. 52 (0.52us) 0…5, step 0.01 A configurable parameter taMaxOffset is used to determine the maximum allowed timing alignment offset before a per‐need timing alignment update is required. The following parameters are recommended for UL time alignment configuration shown in Table 6.2. 6.1.2.3 Maximum RLC Retransmissions Exceeded If UE/eNB reaches the maximum re‐transmitting times (ulMaxRetxThreshold/dlMaxRetxThreshold), that’s RLC max. In this situation, UE considers a radio link failure. Then UE starts T311 and goes to reestablishment procedure, or directly goes back idle mode. Example screenshots in the Figure 6‐14 highlight the occurrence of the RLC reset – RF conditions are poor (low RSRP with several servers, PDSCH BLER runs beyond target 10%, and Drop Call Optimization eNB UE AMD PDU (SN=23) ReTX AMD PDU (SN=23) #1 AMD PDU (SN=23) #2 AMD PDU (SN=23) #8 t-PollRetransmit RL Failure Cause: RLC reset Figure 6.14 Example of RLC reset. RLC retransmissions are necessary) leading the UE to reset. This will be followed by initiation of RRC connection reestablishment procedure (Figure 6.14). After max RLC retransmissions has been reached, eNB starts a timer to wait for an UE triggered RRC connection reestablishment with cause “otherFailure.” If the timer expires, eNB releases the UE (S1 + RRC release), otherwise UE has triggered a RRC connection reestablishment procedure and eNB performs the RRC connection reestablishment procedure (as for a RLF) (Figure 6.15). For LTE optimization, proper signaling radio bearer and data radio beare’s RLC retransmission parameter settings can improve mobility and retainability performance. The recommended RLC robustness parameters tuning is shown in Table 6.3. 6.1.3 RadioLink Failure Optimization and Recovery If RRC drop is due to random access (RA) failure at handover, when UE random access attempt to re‐sync, there is no response after the maximum retransmissions, and UE is not able to synch to target eNB. It needs to check message1, message2, message3, and message4 informations and UL interference or timing advance alignment issues with target cell. After message1, UE has not received message2 until to the maximum RACH attempts is due to preamble issue. In this situation, eNB can not decode message1 correctly, or UE can not decode message2 correctly. It is usually due to bad DL/UL quality caused by low coverage or 225 226 LTE Optimization Engineering Handbook UE eNB MME eNB detects RLC maximum retransmission event eNB waits for RRC reestablishment by UE (timer controlled) Case Timer expiration Timer expires S1AP: UE CONTEXT RELEASE REQUEST UE CONTEXT RELEASE procedure UE starts Reestablishment UE detects RLF or RLC maximum retransmission event RRC Connection Reestablishment procedure Figure 6.15 eNB‐triggered release due to maximum number of RLC retransmissions. Table 6.3 The recommended RLC robustness parameters (SRB and DRB). MO Attribute Parameter Description dlMaxRetxThreshold Max. number RLC re‐transmissions in DL/UL before ulMaxRetxThreshold stopping and indicating to RRC that max. number of RLC re‐txs have been reached. Range Recommended 1, 2, 3, 4, 6, 8, 16, 32 16 high interference. If UE does not receive message4 after sending message3, until RACH maximum attempts is due to no TA information in message2, it is contention failed. If RRC drop is due to RLC maximum re‐transmission, it needs to check PUSCH/PDSCH BLER and SINR to see the radio environment. The eNB/UE can not decode (or receive) normally UL/DL RLC data due to UL/DL quality, which usually caused by interference, low signal level, overlapping cell, overshooting cell, or UE terminal. For very low RSRP in serving cell, or sudden low received level in corner, or missing neighbor, or overshooting cell, its solution is to optimize coverage and neighbor relation. For RSRP that not too low in the serving cell, but appears very bad quality due to very high interference, its solution is to deal with the interference. Usually, it will expected 3 to 5 sec to recover the RRC connection after RLF detection that is described below. ●● ●● PDCCH BLER monitoring time: N310 * 200 ms = 2 sec (assume N310 = 10, without DRX) Max Re‐Tx time: Max Re‐Tx Threshold*50 ms = 1.6 sec (assume RLC Max retransmission = 32, RLC retransmission time = 50 ms) Drop Call Optimization ●● ●● ●● ●● ●● RLF wait time: T310 = 2 sec (waiting time for radio link failure) Random access time: 2*RA‐response size * preamble trans Max = 80 ms (assume RA‐response size = 4 ms, preamble trans Max = 10) Time supervision of successful handover completion: T304 = 1 sec Cell reselection time: T311 = 3 sec (RRC connection reestablishment timer) RRC reestablishment time: T301 = 200 ms (RRC reestablishment procedure timer) 6.2 Reasons of Call Drop and Optimization Reasons for poor retainability include but are not limited to coverage issues, handover issues, neighbor issues, interference, and other abnormal events. ●● ●● ●● ●● ●● ●● ●● ●● ●● Poor signal quality in the UL and DL due to poor coverage or path imbalance, fault RRU and RF issues, Tx power limited or others, in the field, as the UE approaches an RSRP of ‐110 dBm or SINR approaches ‐5 dB, the UE may not have sufficient signal strength to maintain the session. Further, it needs to check if the radio parameter settings are optimal like power setting, PCI collision, and so on. Interference, for DL quality, this can be evaluated by measuring the scanner SIR and F factor, to find if there is PCI pollution or overshooting cell. For UL quality, monitor PUSCH BLER and PHICH NACK. Handover failure in target EPC/eNB or system, that is, no neighbor, no measurement report, no handover command, RA failure at handover. Sometimes, Ts1reloc_overall_expiry, or Tx2reloc_overall_expiry maybe the reason. Admission reject, due to lack of licenses Release due to EUTRAN_generated reason RLC_failure_DRB, RLC_failure_SRB Load_balancing_TAU_required RRC_re‐configuration_time out, RRC_reestablishment_reject, due to parameter misconfiguration S1_reset, due to S1 link issues The parameters related to the areas listed in the previous section, the E‐RAB drop rate can be easily improved by tuning the timers T310 and T311 as well as out‐of‐sync, N310, and in‐ sync, N311, along with the currently recommended values for maximising retainability performance. If the drop event is found, the whole procedure should be investigated, for analyzing poor retainability, below parameters in Figure 6.16 are especially needed to pay attention to. 6.2.1 Reasons of E‐RAB Drop The data radio bearer (DRB) and the S1 bearer carry user plane data between the UE and the SGW. S1 bearer carries user plane data between the RAN and the SGW, S1 application protocol Investigate UE MR, RSRP/RSRQ of serving and neighbor cell...... Drop E-RAB established Time elapsed Investigate the new (same) cell performance after call drop, performance and radio quality. Time Investigate UE RLC/PDCP Thp, BLER, Harq Nack ratio, RLC Nack ratio, CQI, RSRP/RSRQ, TA, RANK, and UE Tx_power limitation ratio..... Figure 6.16 Analyzing poor retainability. If HO happened, best ceel measurement should be investigated..... 227 228 LTE Optimization Engineering Handbook (AP) bearer carries control plane data between the RAN and the MME. S1 application protocol is responsible for setting up, modifying, and releasing E‐RABs. Table 6.4 gives a list of the different S1AP procedures. An E‐RAB drop is counted each time the eNB sends an E‐RAB release indication to the MME, and whenever the MME sends the E‐RAB release command to the eNB with a release cause other than normal release, detach, user inactivity and CSFB triggered. E‐RAB can be released by either of E‐RAB release procedure or UE context release procedure, E‐RAB release reasons can be shown in Figure 6.17. E‐RAB drop due to poor radio Actually it can be found that UE loss is the most cause for E‐RAB abnormal releases in the whole network, although poor radio condition in LTE might be the reason. Abnormal E‐RAB releases due to radio faults are caused by faults such as the number of RLC retransmissions reaching the maximum, UE UL out of synchronization, or signaling procedure failures that are resulted from weak coverage, UL interference. If E‐RAB drop is due to poor radio, it needs to check whether UEs are mostly located in weak coverage areas (timing advance) and check the values of the counters related to CQI, bad coverage report, and UE_power restricted, and so on. Table 6.4 S1AP procedures. Elementary procedure Initiating message Successful response message Unsuccessful response message Handover preparation Handover required Handover command Handover preparation failure Handover resource allocation Handover request Handover request acknowledge Handover failure Path switch request Path switch request Path switch request acknowledge Path switch request failure Handover cancellation Handover cancel Handover cancel acknowledge E‐RAB set up E‐RAB set up request E‐RAB set up response E‐RAB modify E‐RAB modify request E‐RAB modify response E‐RAB release E‐RAB release command E‐RAB release responseE Initial context set up Initial context set up request Initial context set up response Reset Reset Reset acknowledge S1 set up S1 set up request S1 set up response UE context release UE context release command UE context release complete Initial context set up failure S1 set up failure UE context modification UE context modification request UE context modification response UE context modification failure eNB configuration update eNB configuration update eNB configuration update acknowledge eNB configuration update failure MME configuration update MME configuration update MME configuration update acknowledge MME configuration update failure Write‐replace warning Write‐replace warning request Write‐replace warning response RRC MME S1AP RRC S1AP MME E-RAB RELEASE COMMAND Includes a list of E-RABs to be released All resources for the E-RAB are released (DRB and S1 Bearer) All resources for the E-RAB are released (DRB and S1 Bearer) E-RAB RELEASE RESPONSE MME initiated eNB initiated Includes a list of released E-RABs RRC S1AP MME E-RAB RELEASE INDICATION Includes a list of released E-RABs RRC S1AP MME UE CONTEXT RELEASE REQUEST Includes release cause UE CONTEXT RELEASE COMMAND UE CONTEXT RELEASE COMMAND All resources for the UE context are released (DRB and S1 Bearer) All resources for the UE context are released (DRB and S1 Bearer) UE CONTEXT RELEASE COMPLETE MME initiated eNB initiated UE CONTEXT RELEASE COMPLETE Drop call reason Other reason UE lost › tS1relocoverall expiry Probable cause: › Poor radio › RRU issue › High UL interference › Transport issue › Improper neighbor relation › IFHO and IRAT feature › Lost due to UE Abnormal Release capability Figure 6.17 E‐RAB release procedure and reasons. HO fail drop call HO analysis HO preparation fail Normal release Probable cause: › IFHO and IRAT feature Probable cause: › TAC/IP configuration issue - CSFB release › Admission control in target cell › RLC mode for ERAB › Target cell downtime HO fail - User Inactivity - IRATredirection 230 LTE Optimization Engineering Handbook Figure 6.18 E‐RAB drop due to transmission fault (example). E‐RAB drop due to handover failures Abnormal E‐RAB releases due to handover failures are caused by congestion on the target cell, intermittent downtimes on target cell, PCI collision, overshooters, and target cell is beyond the second‐tier neighbors. E‐RAB drop due to transmission fault Abnormal E‐RAB releases due to transmission faults are caused by transmission exceptions between the eNB and the MME. For example, the S1 interface experiences intermittent disconnections as shown in Figure 6.18. E‐RAB drop due to congestion Abnormal E‐RAB releases due to congestion are caused by congestion of radio resources on the eNB side. For example, the radio sources are insufficient if the number of UEs reaches the upper limit. If service drops due to congestion occurs in a cell for a long time, high load cell specific parameters setting and load balancing procedures (idle and connected modes) can be enabled to reduce the cell load. In the long term, the cell requires capacity expansion. 6.2.2 S1 Release Any abnormal events happened in S1/Uu will trigger S1 release procedure. S1 release can move the UE from ECM‐connected to ECM‐idle in both the UE and MME, and all UE‐related context information is deleted in the eNB. The initiation of S1 release procedure is either eNB initiated (e.g., radio, user inactivity, release due to UE generated signaling connection release) or MME initiated with cause, for example, authentication failure, detach, NAS, other protocol, and so on. The reasons are list in Table 6.5. Drop Call Optimization Table 6.5 The reasons of S1 release. UE release cause (3GPP) UE release cause (3GPP) Abnormal ‐ cell_not_available (S1_reset) Normal ‐ CSFB_triggered Abnormal ‐ failure_in_the_radio_interface_procedure Normal – detach Abnormal ‐ radio_conection_with_UE_lost (RLC_failure_DRB) Normal ‐ Handover_desirable_for_ radio_reasons Abnormal ‐ radio_conection_with_UE_lost (RLC_failure_SRB) Normal ‐successful_handover Abnormal ‐ release due to EUTRAN_generated reason (RLC_failure_SRB) Normal ‐ UE_not_available_for_ PS_services Abnormal ‐ release due to EUTRAN_generated reason Normal ‐ user_inactivity Abnormal ‐ TS1relocoverall expiry Normal ‐ Load balancing TAU required Abnormal ‐ TX2relocoverall_expiry Normal_release Abnormal ‐ radio resources not available Abnormal ‐ authentication failure Abnormal – control overload In a live network, the value of E‐RAB drop rate is more than UE context drop rate. The below items will trigger UE context release request. For same cases, the source eNB sends the UE context release request with cause “tS1relocoverall‐expiry” to MME followed by the UE context release command from MME indicating the unsuccessful handover. ●● ●● ●● Radio link failure and no RRC reestablishment S1 (TS1relocoverall‐expiry)/X2 release time out (tx2relocoverall‐expiry) UE RRC time out (release‐due‐to‐eutran‐generated‐reason) Usually the operator can extend the parameter tS1relocoverall a little longer to solve the problem, which is shown in Figure 6.19, although this is not the best way to resolve this problem, it must wait for longer to release the source eNB radio resource, which is very wasteful. The final solution is finding out why there is so long of a delay between handover notify and UE context release command in MME and to resolve it. An example of “radio_conection_with_ UE_lost” is shown in Figure 6.20. 100 90 80 70 60 50 40 30 20 10 0 Sum of HoPrepOutS1AttInterEnb Sum of S1 HOSR(%) Day1 Day2 Day3 Day4 Day5 Day6 Day7 Figure 6.19 S1 handover success rate (change tS1relocoverall from 5 s to 8 s). Day8 Day9 1000 900 800 700 600 500 400 300 200 100 0 231 Figure 6.20 Radio_conection_with_UE_lost. Drop Call Optimization Upon reception of the handover command message, the source eNB shall stop the timer TS1RELOCprep and start the timer TS1RELOCOverall. The parameter TRelocOverall is defined as the maximum time for the protection of the overall handover procedure in the source eNB. Upon reception of the UE context release command message, the eNB shall release all related signaling and user data transport resources and reply with the UE context release complete message. If the UE context release procedure is not initiated toward the eNB before the expiry of the timer TS1RELOCOverall, the handover will be failed with cause “TRelocOverall_EXPIRED,” the eNB shall request the MME to release the UE context. Two examples are shown in Figure 6.21 and Figure 6.22. 6.2.3 Retainability Optimization Retainability is defined as the ability of a user to retain its requested service once connected for the desired duration. It can be measured by the number of dropped calls per second or by percentage. As have mentioned before, there could be many reasons when a call drops on LTE. It is important to check layer 3 messages specially during last a few seconds of drive in order to understand cause of drop. The drop location should also be seen with respect to RSRP and SINR surrounding drop. Neighboring sites should also be confirmed for issues near drop location. One important thing to remember is terrain. Use Google Earth or any other software that can show you terrain for location of drop and elevation profile from serving sector. This will help you in understanding the cause of drop. 6.3 RRC Connection Reestablishment The RRC connection reestablishment procedure is defined in 3GPP 36.331 and gives the UE the option to try to re‐connect if a radio link failure is detected. In case of lost connection (detection of radio link failure or handover failure) to the UE, eNB will keep the context for a period of time. During this period it is possible for the UE to do a RRC connection reestablishment request in a cell found after a cell search. This cell can, in principle, be in any eNB. The feature allows the UE to reestablish RRC connection in serving cell in case of radio link failure. The eNB, instead of dropping the call (releasing the UE S1 + RRC release) immediately in case of failure, starts a timer for the reestablishment of the RRC connection; the UE attempts a connection reestablishment in a cell. UE sends identification information in reestablishment message such as C‐RNTI used in the cell where the failure occurred, the PCI of the cell where the failure occurred and a message authentication code (shortMAC‐I). The procedure is initiated by the UE in RRC connection reestablishment in case of detecting radio link failure. For inter‐eNB cases the RLF report is relayed in X2 RLF Indication based on PCI received in reestablishment request. In a number of cases, reestablishment success does not occur (e.g., no UE context on the target) and RLF report cannot be delivered (Figure 6.23). Radio link failure detection is due to for example: ●● ●● ●● ●● ●● ●● Expiry of the timer T310 (started after detection of physical layer problem) Random access problem indication from MAC Indication from RLC that the maximum number of retransmissions has been reached Handover failure (T304 expiry) Integrity check failure indication from lower layers RRC connection reconfiguration failure 233 Figure 6.21 An example of expiry of TS1RELOCOverall. Figure 6.22 An example of expiry of TX2RELOCOverall. 236 LTE Optimization Engineering Handbook UE eNB MME eNB detects RLC maximum retransmission event eNB waits for RRC reestablishment by UE (timer controlled) Case Timer expiration Timer expires S1AP: UE CONTEXT RELEASE REQUEST UE CONTEXT RELEASE procedure UE starts Reestablishment UE detects RLF or RLC maximum retransmission event RRC Connection Reestablishment procedure Figure 6.23 eNB release due to max number of RLC retransmissions exceeded. eNB UE UE RRCConnectionReestablishmentRequest (SRBO,CCCH) eNB HO Preparation RRCConnectionReconfiguration X T301 RRCConnectionReestablishment (SRBO,CCCH) eNB T301 RRCConnectionReestablishmentRequest RRCConnectionReestablishment RRCConnectionReestablishment Complete (SRB1, DCCH) RRCConnectionReestablishmentComplete RRCConnectionReconfiguration procedure to setup SRB2 and DRB RRC Connection Reconfiguration Procedure (re-establishes SRB2 and DRB(s) Figure 6.24 RRC connection reestablishment at the same and at a different cell. At the RRC connection reestablishment request, the UE informs eNB about temporary identity and from which cell it originates. Based on this information it is possible for eNB to activate the correct context in eNB and with the, so far, not released bearers to MME/SGW. The UE sends RRC connection reestablishment request message to the eNB in order to start the procedure and the eNB keeps the UE context alive while timer T311 doesn’t expire. If the UE sends an RRC connection reestablishment request inside this period, the call is reestablished, that is, first the eNB sends DL RRCConnectionReestablishment message for the UE to resume SRB1 and security and after reception of RRCConnectionReestablismentComplete message the eNB reestablishes the SRB2 and the DRB, which is shown in Figure 6.24. If the RRC connection reestablishment request is outside the T311 period, the eNB would have already released these resources and will reply with an RRC connection reestablishment reject, UE will perform cell reselection + TAU. Here is a drop call example shown in Table 6.6. The related timers of RRC connection reestablishment procedure is shown in Figure 6.25. Drop Call Optimization Table 6.6 An example of drop call. Time Channel Direction Message 10:36:32 10:36:32 DCCH UL RRCConnectionReconfigurationComplete BCCH‐SCH DL SystemInformationBlockType1 10:36:32 BCCH‐SCH DL SystemInformation 10:36:33 DCCH UL MeasurementReport 10:36:34 DCCH UL MeasurementReport 10:36:34 DCCH DL RRCConnectionReconfiguration (handover command) 10:36:34 DCCH UL RRCConnectionReconfigurationComplete 10:36:35 BCCH‐BCH DL MasterInformationBlock (In a new cell) 10:36:35 BCCH‐SCH DL SystemInformationBlockType1 10:36:35 BCCH‐SCH DL SystemInformation 10:36:35 BCCH‐SCH DL SystemInformation 10:36:35 BCCH‐SCH DL SystemInformation 10:36:35 CCCH UL RRCConnectionReestablishmentRequest 10:36:36 CCCH DL RRCConnectionReestablishmentReject 10:36:36 BCCH‐BCH DL MasterInformationBlock 10:36:36 BCCH‐SCH DL SystemInformationBlockType1 10:36:36 UL TRACKING_AREA_UPDATE_REQUEST 10:36:36 CCCH UL RRCConnectionRequest 10:36:36 CCCH DL RRCConnectionSetup 10:36:36 DCCH UL RRCConnectionSetupComplete 10:36:41 DCCH DL DLInformationTransfer 10:36:41 DL TRACKING_AREA_UPDATE_ACCEPT 10:36:41 UL TRACKING_AREA_UPDATE_COMPLETE Since the RRC reestablishment procedure is made up of a couple of steps required to successfully reestablish the connection, a rough time estimate per step can be provided. The typical control plane outage time measured in the field during an RRC connection reestablishment request was measured to be approximately 750 ms. Typically, the time budget is list below: RLF detection needs 200 to 600 ms, cell search needs 100 to 800 ms, SIB reading needs 80 to 320 ms, and reestablishment procedure needs 60 to 200 ms. The example shown in Figure 6.26 gives an interruption time of 790 ms in this case. RRC connection reestablishment can be based on a common solution for all cases: serving cell, incoming and outgoing handover, and unprepared cell. Its operation can be divided in three main flows: RRC connection reestablishment in unprepared cell, RRC connection reestablishment during incoming handover, and RRC connection reestablishment during outgoing handover. The solution of reestablishment during incoming handover is such that no reestablishment info is stored in the target cell until the handover is completed. If a reestablishment occurs prior to the storing of the reestablishment info no matching context will be found in that cell. It will instead be handled as a reestablishment in an unprepared cell (an unprepared cell is a cell 237 UE T311 Start, UE search for a suitable E-UTRA cell RRC Connection Re-establishment Initiated EUTRAN RRC Connection Reestablishment Request T301 Start, Wait for RRC Connection Reestablishment Suitable E-UTRA cell found, T311 Stop RRC Connection Reestablishment received, T301 Stop RRC Connection Reestablishment RRC Connection Reestablishment Complete T311 / T301 expired Go back to IDLE UE eNodeB MME Serving GW PDN GW PCRF HSS 1. Radio Link Failure 2. T311 started 3. RL Restoration before T311 expiry 4. RRC Connection Reestablishment Request 5. RRC Connection Reestablishment Reject 6. UE Context Release Request Cause 7.Release Access Bearers Request 8. Release Access Bearers Response 9. MME-initiated Connection Release Figure 6.25 RRC connection reestablishment timers. UE sends multiple MRs but can’t decode HO command from eNB due to poor radio. This is the last MR before UE loses sync and perform cell reselection. interruption time UE re-establishes RRC connection. Data radio bearer is setup. Figure 6.26 Example of reestablishment time. Drop Call Optimization that does not have a UE context corresponding to the triplet PCI, C‐RNTI, shortMAC‐I) and UE context inquiry procedure is started. ●● ●● ●● If RLF occurs before MR is received, then only source eNB will have the context to support reestablishment If RLF occurs after MR is received and before handover confirm is received, then both eNBs will have the UE context If RLF occurs after handover confirm is received and source has released resources, then only the target eNB will have the UE context For outgoing handover, not all types of handovers are supported. That is, reestablishment is handled by an eNB both when the UE returns to the serving cell during intra‐eNB handover and X2 handover. S1 handover (intra LTE, IRAT, and SRVCC handover) are not supported. The unprepared cell will try to fetch the UE context from all neighbor eNBs (including own eNB) that have a cell with the PCI received in the RRC connection reestablishment request message from the UE. The context fetch procedure consists of proprietary X2AP messages sent between the unprepared target and the serving cell (Figure 6.27). The total time required for a successful RRC reestablishment procedure might vary between 440 ms and almost 2 seconds. Worst case an outage of 2 seconds might be observed during a handover failure or radio link failure when in a VoLTE call. Finally there is a difference in the RRC reestablishment procedure depending on whether the cell is considered as prepared on un‐prepared (Figure 6.28).2 With (multi‐target) RRC connection reestablishment, the UE‐‐specific connection to MME is kept during the time from connection lost until the RRC connection reestablishment procedure is finished. The benefit is improved drop rate and especially for VoLTE calls the user will only see it as a short voice outage and no need for re‐dialing. The following examples are about the failed RRC reestablishment (reject) after RLF due to eNB mis‐configuration during handover (Figure 6.29). The configuration of N310 and T310 will impact of the success of RRC reestablishment. One example shown in Figure 6.30 presents reestablishment failure after RLF expires due to T301. After UE low layer report N310 times of “DL out‐of sync,” UE high layer will start T310 timer and wait for UE low layer report “DL in‐sync,” if UE doesn’t receive “DL in‐sync” before T310 expired, then UE will assume it is “out of sync” in DL. Setting the value low will result in higher risk of out of service and easier to go to RRC idle (drop) or RRC reestablishment. It is recommended the default value: N310 = “N20,” T310 = “ms2000.” In a live network, timers that impact drop call includes: T300, T301, T302, T303, T304, T305, T310, T311, T320, and T321; the proper settings of above timers is important to the call drop. 6.4 RRC Connection Supervision There are two main UE EMM states: deregistered and registered. For deregistered, UE location is unknown to MME, UE cannot be paged. For registered, EPS default bearer has been activated, UE has IP address. EMM‐registered has two substates: ●● ●● Idle (=ECM‐idle + RRC‐idle) Connected (=ECM‐connected + RRC‐connected) 2 For reestablishment in “unprepared” cell is proprietary, thus not standardized by 3GPP. 239 UE Unprepared Target Cell Serving Cell MME RRCConnectionReestablishmentRequest Tell the serving cell, the reestablish order is got Context fetch request T301 Context fetch response 400 ms Context fetch response accept tRelocOverallValue: 5s RRCConnectionReestablishment RRCConnectionReestablishmentComplete Path switch request Path switch request acknowledge RRCConnectionReconfiguration T304 tRrcConnectionReconfiguration: 6s RRCConnectionReconfigurationComplete 500 ms UE context release Counters stepped: Figure 6.27 RRC connection reestablishment. Drop Call Optimization Target eNB 48 ms MAC Contention resolution T301 UE Target eNB UE RA preamble (Msg1) RA preamble (Msg1) RA preamble (Msg2) RRC conn. reestab. req. (Msg 3) RA preamble (Msg2) RRC conn. reestab. req. (Msg 3) Adm. Control. Msg4 with empty RRC comp RRC conn. reestab./reject (Msg 4) RRC conn. reestab. cmpl. RRC conn. reestab./reject (Msg 4) RRC conn. reestab. cmpl. Source eNB Adm. Control. Context fetch Context fetch reply Figure 6.28 RRC reestablishment in case of prepared/un‐prepared cell. RACH failure and RLF detected UE attempting reestablishment re-establishment rejected by eNB UE moves to Idle and start new RRC Connection eNB configures UE with 2xSRB+ 2xAM DRB+1xUM DRB Figure 6.29 Failed reestablishmentafter RLF due to eNB misconfiguration during handover. For ECM‐idle/RRC‐idle, UE location is known at TA level. When UE transition to connected takes place by: i) paging needed if DL user data arrives; ii) UE triggers RRC connection request. UE makes cell reselection and TAU. For ECM‐connected/RRC‐connected, UE location is known at cell level, UE is in RRC‐connected. UE changes cell by handover (Figure 6.31). The radio connection is supervised by both the UE and the eNB. When the connection is considered as “bad,” the UE may be released from RRC connected mode to idle mode. 241 242 LTE Optimization Engineering Handbook RLF detected due to L1 failure UE attempting re-establishment re-establishment fails because of RACH failure for the RRC Re-estb T301 expires and UE moves to Idle and start new RRC Connection eNB configures UE with 2xSRB+ 2xAM Figure 6.30 Reestablishment failure after RLF expires due to T301. Power On Release due to Inactivity Registration (Attach) • Allocate S-TMSI • Allocate IP address + default bearer • Authentication • Establish security context EMM_Deregistered ECM_Idle • Release RRC + S1 connection • Configure DRX for paging • eNB initiated (inactivity timer) EMM_Registered ECM_Idle EMM_Registered ECM_Connected Deregistration (Detach) Change PLMN • Release S-TMSI • Release IP addresses • UE, MME or HSS initiated EMM-REGISTERED Attach Accept, TAU Accept Detach, Attach Reject, Tau Reject EMM-DEREGISTERED New Traffic ECM-IDLE • Activate RRC + S1 connection Timeout of Periodic TA Update • Release S-TMSI and IP addresses S1 Connection Established S1 Connection Released ECM-CONNECTED Figure 6.31 UE EMM states transition. The connection is consedered as “bad” when BLER exceeds a certain threshold during a certain time duration (T310) or all data radio bearers for a UE are inactive for a certain duration, or if the number of RLC retransmissions exceeds a certain threshold. When T310 expires and the connection is still bad, the UE may send RRC connection reestablishment request to indicate to the eNB that a radio link failure has occurred. When T311 expires and no answer is heard from the eNB, the UE may return to the RRC_idle state. If a data radio bearer (DRB) has been inactive in both UL and DL for a certain period, RLC will report inactivity of DRB to the radio connection supervision. If all DRBs are inactive for the duration specified by the tInactivityTimer, radio connection supervision will trigger a UE release. The tInactivityTimer parameter can be set to 0 to switch off supervision or between 10 and 86400 seconds. In addition, the number of retransmissions to RLC failure is specified by the RLC acknowledged mode (AM) parameters (Figure 6.32). Drop Call Optimization RRC_CONNECTED No Inactivity for ‘tInactivityTimer’ 0 = Off 10 to 86400 seconds Maximum number of RLC retransmissions Yes Yes No RLC Acknowledged Mode parameters RRC_IDLE …N310… BLER>Qout RRC Connected T310 BLER>Qout RL Failure -> RRC Connection Reestablishment Request T311 RRC Connection Failure Return to idle RLC Failure If ALL DRBs are inactive UE Release tInactivityTimer Figure 6.32 Radio connection supervision principles. •• Radio Bearer Failure UE Release dlMaxRetxThreshold 243 244 7 Latency Optimization This section focuses on parameters that can improve latency in an LTE network. Latency is the most important KPI for applications with a bursty traffic profile. Latency is important when considering real‐time and interactive services. Latency is generally considered either as control plane latency or as user plane latency. Control plane latency involves the network attachment operation while user plane latency only considers the latency of packets while UE is in connected state. Low user plane latency is essential for delivering interactive services, like gaming and VoIP. The latency related KPI is shown in Table 7.1. To analyze and optimize the latency profile of the LTE network, ping performance will be used as a baseline. Ping is being used for discovering the user plane latency. The main factors that are going to contribute to the latency are the UE state (RRC state), whether it’s a MTC or MOC and the payload (ping size). The LTE latency overview is shown in Figure 7.1. User plane latency is the average time between the first transmission of a data packet and the reception of a physical layer ACK. One‐way user plane latency is half the round trip user plane latency. LTE requirement is that one‐way latency across the radio access network is as low as 5ms in optimal conditions, but that the value is dependent on the system loading and radio propagation conditions. This value corresponds to a round trip time (RTT) of 10 ms, which is challenging to achieve in practice. Control plane latency (connection set up delay) is the time required for the transitions between different LTE states. LTE is based on two main states: RRC_idle and RRC connected (i.e., active). LTE system needs to support transition from idle into active is less than 100ms excluding paging delay and NAS (non‐access stratum) signaling delay. 7.1 User Plane Latency RTT in DL is the interval between sending a datagram to the UE and receiving the corresponding reply by the IP host (peer entity). RTT in UL is the interval between sending a datagram by the UE and receiving the corresponding reply from an IP peer entity directly connected to the Gi interface of the P‐GW. RTT Average Ping _ end _ time Ping _ start _ time 3GPP and NGMN recommend 10 ms1 air interface, 20ms end to end (RTT). For user plan latency budget, median round trip delay between UE and eNB backhaul network was about 1 Measured round trip time with prescheduled uplink meet with 3GPP targrt of 10 ms. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Latency Optimization Table 7.1 The latency‐related KPI. Control Plane User Plane KPI Name Critical KPI Name Critical 1st Page response time Yes Dedicated bearer activation time No UE activation time (idle to active) Yes VoIP call end time (Mobile to Mobile/PSTN) No Service request set up time Yes VoIP Mobile‐to‐Mobile Mouth‐to‐ Ear delay Yes MO VoIP call set up time (Mobile<>PSTN) Yes VoIP Mobile‐to‐Mobile Packet Latency Yes Mobile‐to‐Mobile VoIP call set up time Yes VoIP Mobile<>Land Mouth‐to‐Ear Delay No Ping RTT between UE and application server (unscheduled and prescheduled) Yes VOIP Mobile<>Land over Internet Packet Latency No Initial attach/detach time No VOIP Mobile<>Land over PSTN Packet No Latency eNB UE RAN latency (UE-eNB RTT) is impacted by downlink PRB utilization/traffic load/CQI, uplink PUSCH SINR/Noise etc. RTT: X ms Core Network RTT: Y ms Latency between eNB and core network Assumption: Y=0 ms eNB UE UE eNB latency EPC latency UE core network latency (Network RTT) Internet SGW/PGW Server core network server latency (Internet RTT) Figure 7.1 LTE latency. 19ms, core network transport delay was about 0.4ms round trip. RTT for data applications is measured with the ping application of the UEs operating system or with a comparable measurement tool. It records the time difference between sending an ICMP echo request to an IP host and the reception of the corresponding ICMP echo reply message (Figure 7.2). Prescheduling means UL transmissions without receiving scheduling request, which means eNB blindly grants PUSCH transmission to UEs. If there is no real data to transmit, the UE has to transmit padding bits. In Figure 7.2, UL grants are scheduled after eNB receives an SR (scheduling request). If SR period is 5ms, such that the latency will include all the time taken for SR transmission and receiving, scheduling delay, and grant transmission and receiving. Target figures for the RTT KPI are also distinguished according to the cell position (RF conditions) and ICMP packet size. Figure 7.3 is valid for unloaded cells and good RF conditions. 245 SR Period: configurable to 5ms, 20ms, 40ms (This period grows as the number of UE increases) 3GPP target (Air interface) Echo request Air interface 5 ms SR (Scheduling Request) End -End 0 3-4 ms 10 Pre-scheduled uplink UL Grant Average ~25 ms (measured in the field) ms 5 4 ms 12.4ms 15 20 25 No pre-scheduling 0.24ms BSR + Data ~8ms processing inside eNB (up+down) Ping Ack Echo reply 5 ms eNB SGW PGW server Data Ack eNB UE 6.8ms 0.20ms Pre-scheduled uplink Figure 7.2 3GPP recommend and field test user plan latency budget. 45 40 40 32Bytes 1000Bytes Latency (ms) 35 30 25 1500Bytes 29 25 22 19 20 22 24 25 14 15 10 5 0 60.0% Max Min Aver Ping RTT - Distributions 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 13 14 15 16 17 18 19 20 21 22 23 24 26 27 29 30 31 32 33 36 39 40 45 58 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Ping RTT - PDF 13 14 15 16 17 18 19 20 21 22 23 24 26 27 29 30 31 32 33 36 39 40 45 Figure 7.3 User plane latency – RTT (no prescheduling). Latency Optimization PC UE PING eNB Scheduling Request Router PC Ping times include UL resource allocation and router delay. For persistently scheduled resources, RTT will be 10 ms. ˜10 ms UL Grant PING PING DL Grant + PING PING PING ˜0.1ms (eNB to PC) Figure 7.4 User plane latency distribution. In a live network, the result of average RTT without prescheduling (32 bytes), 90% of samples are below 20 ms, and 98% of samples are below 23 ms. Configure payload size to maximum 2400 bytes shows little difference between 32 bytes with maximum UL grant size configured. Prescheduled user plane latency will meet 3GPP targets, unscheduled user plane latency consistent to specification. UP latency with prescheduling function is approximately 10ms less than prescheduling off. So RTT optimization mainly focus on scheduling section except RF conditions. It notes that overload usually impacts user plane latency (Figure 7.4). 7.2 Control Plane Latency In the ECM_idle state, UE initiated the ping process, or the server to launch the UE ping command, both can trigger the UE state transitions from idle to active. This state transition delay is the time delay of the control plane. Therefore, the test procedure is UE first complete the attachment process, and then wait for the timer inactivity time out, UE enters into the idle state. UE initiated the ping process, making the UE from ECM_idle to ECM_active state. Control plane latency is calculated from the first RACH preamble to RRC connection reconfiguration complete message, which is shown in Figure 7.5. The maximum re‐transmit of the above messages process is shown in Table 7.2. The service request set up time is between the time when the UE sends service request (service request NAS IE is included in the RRC connection set up complete message) to the MME and the time when initial UE context set up complete is received at the MME. It’s a part of the UE idle to active state transition time. 7.3 Random Access Latency Optimization The random access procedure takes two distinct forms: contention‐free and contention‐based RACH access. Contention‐free is applicable only to handover and DL data arrival. Contention‐ based is is performed for the initial access from RRC_idle, RRC connection reestablishment, and DL data arrival during RRC_connected etc. Note that in both events the UE already has a C‐RNTI (Figure 7.6). Contention‐free RACH duration is defined as from message1 to message2, while average RACH duration is around 20ms. Contention‐based RACH duration is defined as from message1 to message4, while average RACH duration is around 90ms (Figure 7.7). 247 248 LTE Optimization Engineering Handbook UE eNB MME IDLE State D 2.RACH Premble B A: processing delay in UE Total time from idle to activate state 4. TA + Scheduling Grant B: processing delay in eNB A 6.RRC Connection Request B C: processing delay in MME 8.RRC Connection Setup A D: delay for RACH Scheduling period > 40 ms >220 ms 10.RRC Connection Setup Complete +NAS Service Request B 12.Initial UE Message C 14. Initial Context Setup Request B A 16. Security Mode Command + RRC Connection Reconfiguration > 50 ms Active State 18. RRC Connection Reconfiguration Complete 19. Initial Context Setup Response Figure 7.5 Procedure for control plane latency. Table 7.2 The maximum re‐transmit of the listed messages. 1 Rach Preamble UL MAX retx (3) 2 Rach Response DL PDSCH No HARQ 3 RRC Con Req UL PUSCH TM CCCH 5 HARQ 4 RRC Con Setup DL PDSCH TM CCCH 5 HARQ 5 RRC Con Setup Complete UL PUSCH AM DCCH 5 HARQ 6 Security Mode Command DL PDSCH AM DCCH 5 HARQ 7 Security Mode Complete UL PUSCH AM DCCH 5 HARQ 8 RRC Con Reconfiguration DL PDSCH AM DCCH 5 HARQ 9 RRC Con Reconfiguration Complete UL PUSCH AM DCCH 5 HARQ 10 UE Capability Enquiry DL PDSCH AM DCCH 5 HARQ 11 UE Capability Information UL PUSCH AM DCCH 5 HARQ 7.4 Attach Latency Optimization This procedure is done when the UE is coming from power off or detach state. During the attach procedure, the UE registers itself to the MME. The MME stores the UE context. The MME and the UE sets up the default data radio bearer as part of the default EPS bearer. The MME keeps the default EPS bearer context even when the UE goes to idle mode after this procedure. The UE also keeps its IP address when in RRC_idle. UE eNB 1. Dedicated Preamble allocation (Preamble ID, RACH Resource ID) 2. RACH Preamble 3. Random Access Response (RAR) UE eNB Contains a UE contention resolution identity. If message 4 is successfully received and the UE contention resolution identity contained in the message matches the content of message 3, the contention resolution is considered successful. 1. RACH Preamble 2. Random Access Response (RAR) 3. RRC Signaling 4. Contention Resolution Figure 7.6 Contention‐free (left) and contention‐based (right) RACH procedure. pdf 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% >0 to < = 10 >10 to < = 20 >20 to < = 30 cdf >30 to < = 40 >40 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Duration (ms) pdf 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% cdf >30 to >50 to >70 to >90 to >110 >130 >150 >170 < = 50 < = 70 < = 90 < = 110 to to to < = 130 < = 150 < = 170 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Duration (ms) Figure 7.7 Contention‐free (left) and contention‐based (right) RACH access latency. 250 LTE Optimization Engineering Handbook UE NAS eNB LTE Uu S1 - MME MME AS Attach Request RRC Connection Request Random Access Request / Response (MSG1 / MSG2) UE Identification / Contention Resolution (MSG3 / MSG4) RRC Connection Setup / RRC Connection Setup Complete ueCapabilityEnquiry / ueCapabilityInformation NAS Authentication Request / NAS Authentication Response The attach time is the time between the sending of the Attach request NAS message from the UE (inside the RRC connection setup complete message) and the receipt at the MME of the Attach complete message from the UE NAS Security Mode Command / NAS Security Mode Complete securityModeCommand / securityModeComplete RRC Connection Reconfiguration / Reconfiguration Complete Attach Accept Attach Complete Figure 7.8 Attach procedure. After RRC connection, eNB forwards attach request to MME initiating the set up of the S1 control plane. This triggers authentication and key agreement procedure involving the HSS and the AUC. Security keys and vectors are generated and passed on to the MME. When UE performs the initial attach procedure, MME need to request for the UE’s IMSI to reconfirm UE’s identity and authentication. Then, MME initiates the establishment of the default date radio bearer as part of the default EPS bearer to finish attach procedure (Figure 7.8). Field result shows that under either stationary(good RF) or mobility, the average attach duration is around 220 to 240 ms, the minimum can even reach to 160 ms. 7.5 Paging Latency Optimization The time duration for paging was measure from paging message to RRC reconfiguration complete message. The latency is determinated by the paramtere DRX paging cycle and RF conditions. Paging procedure and latency analysis is shown in Figure 7.9. The average latency increase with degraded SNR due to fading conditions, the range of variation is usually several milliseconds. Prescheduled ping latency outperforms non‐prescheduled ones as expected, lab result also shows ping latency versus increased path loss in Figure 7.10. 7.6 Parameters Impacting Latency Latency is just one of the performance index. Relevant KPIs should be considered and balanced. Control plane latency involves the network attachment operation while user plane latency only considers the latency of packets while UE is in connected state. Some mechanisms Latency Optimization UE MME eNB Paging Paging Random Access Procedure NAS: Service Request S1 - AP: INITIAL UE MESSAGE(FFS) + NAS Service Request + eNB UE signalling connection ID S1 - AP: INITIAL CONTEXT SETUP REQUEST RRC: Radio Bearer Setup (NAS Message) RRC: Radio Bearer Setup Complete + (NAS message) + MME UE signalling connection ID + Security Context + UE Capacity Information (FFS) +Bearer Setup (Serving SAE - GWTEID, QoS profile) S1 - AP: INITIAL CONTEXT SETUP COMPLETE + eNB UE signalling connection ID + Bearer Setup Confirm (eNB TEID) UE eNB MME Paging 69 ms Delay for RACH Scheduling Period Processing delay in UE Paging RACH Premble TA+ Scheduling Grant RRC Connection Request RRC Connection Setup 76 ms Processing delay in eNB 168 ms Processing delay in eNB Processing delay in UE 10, RRC Connection Setup Complete Figure 7.9 Paging procedure and latency analysis. including scheduling request periodicity, prescheduling, DRX and PRACH parameters, and so on, will impact network latency. Below setting can shorten the latency on air‐interface: ●● ●● ●● Change SR periodicity from 10ms to 5ms (drawback: increase UE power consumption and network interference) Enable prescheduling (drawback: increase UE power consumption and network interference; waste of channel resources; decrease UE throughput when large amount of UEs in the network) Disable DRX (drawback: increase UE power consumption) PRACH parameters impacting control plane latency (attachment operation) and user plane latency are given in Table 7.3. The inactivity timer will ensure that the UE is active even for busy traffic. If no data is send/ received, the UE goes to sleep and wakes up during the next DRX cycle. Figure 7.11 shows the effect of DRX on latency. The result from ping responses suffers the same effect, because they need to wait for the maximum long DRX cycle. 251 LTE Optimization Engineering Handbook Userplane Latency_32Bytes Non–prescheduled vs. pre–scheduled 35 Latency (ms) 30 32.4 31.4 30 25 20 15 17.2 15.6 15.4 10 5 0 A B A high SINR A B A medium SINR Non Pre-scheduled B B low SINR Pre-scheduled Ping latency vs pathloss, AWGN 120 100 Very poor coverage causing higher BLER 80 Ping latency (ms) 252 60 40 20 0 –50 –45 –40 –35 –30 –25 –20 –15 –10 –5 0 pathloss relative to maximum pathloss (dB) Figure 7.10 Paging latency for different SNR locations. Table 7.3 Parameters impacting control plane latency. Name Recommended Name Recommended preambleInitialReceived TargetPower 94dBm macContentionResolutionTimer Sf64 preambleTransMax n8 maxHARQmsg3Tx 4 preambleTransmitPowerStepSize dB4 aUGtriggerDelayforRACHmsg4 5 maximumNumberOfDLTransmisions 4 RACHMessage4 Latency Optimization eNB VoIP Packets VoIP Packet 1 VoIP Packet 2 VoIP Packet 1 Scheduled Packets Transmission Delay PRBs Matrix UE DRX State Talk-Spurt detected No available PRBs On-Duration DRX cycle On-Duration Three types of DRX parameters in live network Support both long and short - DRX drx-Config DRX-Config : setup onDuration Timer : psf2 drx-Inactivity Timer : psf100 drx-Retransmission Timer : psf2 longDRX-CycleStartOffset : sf40 sf40 : 1 shortDRX-Cycle : sf20 dxShortCycle Timer : 1 Figure 7.11 DL UE scheduling with DRX. Only support long - DRX drx-Config DRX-Config : setup onDuration Timer : psf60 drx-Inactivity Timer : psf80 drx-Retransmission Timer : psf8 longDRX-CycleStartOffset : sf320 sf320 : 281 DRX not supported drx-Config DRX-Config : release 253 254 8 Mobility Optimization Mobility is the key procedure for ensuring that users can move freely within a network. The LTE mobility can be divided into “intra‐LTE mobility” and “inter‐LTE mobility” (inter‐working with 2G/3G and CDMA2000). It can be further divided into RRC_connected and RRC_idle mode mobility. Inter‐radio access technology handover (IRAT‐handover) is one type of “inter‐LTE mobility” in RRC_connected mode while cell reselection is referred to as RRC_idle mode mobility. The idle mode tasks can be divided into four processes: PLMN selection, cell selection and reselection, location registration, and manual CSG ID selection. The UE performs a PLMN selection when switched on, and for example, when a new PLMN is found at return from lack of coverage. Cell selection aims to find one suitable cell to camp on, which refers to UE in idle mode by monitoring the signal quality of neighbor cell and the current serving cell to choose the best cell service signal process. Cell reselection aims to camp on the best cell according to the evaluation criteria, which is performed continuously by the UE in RRC_idle. RRC_connected mobility (LTE handover) consists of four distinct phases: measurement configuration, measurement reporting, handover evaluation, and handover execution. Handover process in LTE is hard handover; it means that it has to break the wireless connection first and re‐establish the connection after it handover to a new cell, thus it will impact the user experience in network. It is generally assumed that downlink measurements, done by the UE, are used for the handover decision. Measurements performed by the source eNB may also assist the decision. The reporting can be done on an event triggered and/or periodic basis. In event‐triggered reporting, the UE measures, evaluates, and reports standardized events, which are used for triggering network actions, for example, handover. The events and parameters are controlled by the network and used by the UE, for example, offsets, hysteresis, averaging time, thresholds, and so on. The UE sends a report only when the network is interested in the information, that is, only when event criteria are fulfilled. As an example to fully understand the scope of each event and how the different handover are triggered, Figure 8.1 is provided. The UE will not perform measurements of neighbor cells when the serving cell RSRP is above Threshold1 (Th1). This saves UE battery life and avoids unnecessary handovers in good radio conditions. The eNB will command the UE to perform measurement gaps when one of the Th2 events is reported in a measurement report. Measurement gaps will be cancelled if the serving cell RSRP rises above Th2a. As a last resort, if no suitable intra‐frequency or inter‐frequency/ IRAT cells are found by the UE and the serving cell RSRP falls below Th4, an RRC connection release with re‐direction can be triggered by the eNB. Note that if Th2 values are set to a high value, this will increase the measurement gaps. As DRX cannot be used in parallel with measurement gaps and measurement gaps have higher LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Mobility Optimization No measurements except serving cell Intra–LTE + Inter Freq measurements Intra–LTE measurements Intra–LTE + Inter Freq + WCDMA + GSM measurements RRC Intra–LTE + Inter Connection Freq + WCDMA Release and measurements re-direction RSRP (dBm) Th1 Th2a Th2_InterFreq Th2_WCDMA Th2_GSM Th4 Figure 8.1 Event thresholds configuration for different handover types. priority than DRX the benefits of DRX (i.e., increasing battery time) wouldn’t be observed when Th2 values are set too high. Mobility for UEs is very important to ensure PS service continuity and reduce the interruption, latency, and PS drops. Tuning the handover thresholds/timers is needed to reduce unnecessary handover attempts toward overshooting cells or due to shadowing. 8.1 Mobility Management Mobility management refers to the process of establishing, maintaining, and releasing physical channel between E‐UTRAN and UE, which is shown in Figure 8.2. In the system of E‐UTRAN, according to the connection state of RRC, the mobility management is divided into two categories: connected state and idle state. The purpose of handover is to ensure that a UE in RRC_connected mode is served continuously when it moves. Mobility including intra‐frequency handover, inter‐frequency handover, S1 based WCDMA/ GSM/ CS CS Fallback • Packet Mobility • Packet Handover LTE SRVCC LTE X2 based WCDMA/GSM Figure 8.2 Mobility management. S2a based • Packet Mobility CDMA 2000 • Packet Handover S101/103 based • Packet Mobility S2a based • Packet Mobility • Packet Handover S3/S4 based Gn based • Packet Mobility • Packet Handover • Packet Handover 255 256 LTE Optimization Engineering Handbook Mobility types From LTE LTE LTE LTE LTE Mandatory SIB Optional SIB SIB1-SIB2-SIB3 SIB1-SIB2-SIB3 SIB5 SIB1-SIB2-SIB3 SIB6 SIB1-SIB2-SIB3 SIB7 SIB1-SIB2-SIB3 SIB8 SIB4 LTE-Intra-Freq SIB4 LTE-Inter-Freq To WCDMA GERAN HRPD Figure 8.3 SIB for handover. inter‐RAT handover and handover between TDD and FDD LTE. From the mobility point of view, the mobility can be classified into intra‐LTE handover, coverage trigger session continuity, inter‐frequency load balancing, service‐triggered mobility, and subscriber‐triggered mobility. The parameters of mobility can be derived from SIB message as shown in Figure 8.3. Figure 8.4 shows the different states of the UE/MS in GSM, 3G, and LTE, also shown the transitions of these states while the UE/MS is in RRC_idle and RRC_connected mode. Actually, IRAT mobility from LTE is only supported through session continuity or release with redirect. When an IRAT mobility event is triggered while in LTE network, the eNB shall intiate a UE release with redirect information to reselect to 3G, 2G, or CDMA network. 8.1.1 RRC Connection Management RRC connection management involves RRC connection establishment, RRC connection reconfiguration, RRC connection re‐establishment, and RRC connection release. ●● ●● ●● ●● RRC connection establishment: This procedure is performed to establish an RRC connection. RRC connection establishment involves signaling radio bearer 1 (SRB1) establishment. The procedure is also used to transmit the initial NAS dedicated information or messages from the UE to the E‐UTRAN. RRC connection reconfiguration: This procedure is performed to modify an RRC connection, for example, to establish, modify, or release radio bearers, to perform handovers, and to configure or modify measurements. As a part of the procedure, NAS dedicated information may be transmitted from the E‐UTRAN to the UE. RRC connection re‐establishment: This procedure is performed to re‐establish an RRC connection after a handover failure or radio link failure. RRC connection re‐establishment involves the restoration of SRB1 operation and the re‐activation of security. A UE in RRC_ connected mode, for which security has been activated, may initiate the procedure in order to continue the RRC connection. The connection re‐establishment will succeed only if the cell has a valid UE context. RRC connection release: This procedure is performed to release an RRC connection. RRC connection release involves the release of the established radio bearers and the release of all radio resources. 8.1.2 Measurement and Handover Events The UE performs radio measurements of its surrounding radio environment. The eNB controls the UE measurement in idle mode with the broadcast system information blocks and in 1xRTTCS Active E-UTRA RRC_CONNECTED Handover Handover HRPD Active GSM_Connected CELL_DCH Handover E-UTRA RRC_CONNECTED Handover GPRS Packet transfer mode CELL_FACH Connection establishment/release CELL_PCH URA_PCH CCO with NACC Reselection Connection establishment/release 1xRTT Dormant 1x Reselection E-UTRA RRC_IDLE Reselection LTE Figure 8.4 LTE Inter RAT mobility procedures. HRPD Idle UTRA_Idle Reselection CCO, Reselection Connection establishment/release E-UTRA RRC_IDLE Connection establishment/release Reselection GSM_Idle/GPRS Packet_Idle CCO, Reselection EVDO 3G LTE GSM 258 LTE Optimization Engineering Handbook Measurement object Object ID Object Measurement Report ID ID ID Report Report config. ID ID LTE carrier frequency 1 1 1 1 1 1 Event A1 LTE carrier frequency 2 2 2 2 2 2 Event A3 UMTS carrier frequency 1 3 3 3 3 3 Event B2 UMTS carrier frequency 2 4 4 4 3 4 Event B2 5 5 5 5 GERAN set of carrier frequencies Figure 8.5 Measurement configuration. connected mode with a RRC measurement control message. Two parameters are used to trigger measurements of the neighboring cells, idle mode uses sIntrasearch, and connected mode uses sMeasure. When UE is in connected mode, a UE measurement consists of two parts: measurement object and report configuration. This pair is referenced by a measurement ID. In connected mode, when the eNB provides the UE with a measurement configuration, it includes the following parameters (Figure 8.5): ●● ●● ●● ●● ●● Measurement objects: The objects on which describes the radio technology and frequency to measure on. Reporting configurations: A list of reporting configurations where each reporting configuration consists of reporting criterion and reporting format. Reporting criteria triggers the UE to send a measurement report (e.g., RSRP). Reporting format quantities that the UE includes in the measurement report (e.g., RSRP and RSRQ). Measurement identities: A list of measurement identities where each measurement identity links one measurement object with one reporting configuration. Quantity configurations: One quantity configuration is configured for intra‐frequency measurements, and one per RAT type. Measurement gaps: Periods that the UE may use to perform measurements, that is, no (UL, DL) transmissions are scheduled. Measurement procedures distinguish the following types of cells: the serving cell, cells listed within the measurement objects, and cells that are not listed within the measurement objects but are detected by the UE on the carrier frequencies indicated by the measurement objects. When the measurements triggering conditions are met, the UE initiates the measurement reporting procedure and sends a measurement report message to the eNB. The message includes the information related to the event previously configured by the eNB. From the measurement report sent by the UE, the eNB may take a handover decision. Up to now, LTE measurement items by eNB and UE are listed below: ●● ●● UE measurements: CQI, RSRP, RSRQ, and inter‐RAT measurements for handover. eNB measurements: DL RS Tx power, received interference power, thermal noise power, TA, average RSSI, average SINR, UL CSI, detected PRACH preambles, transport channel BLER, and so on. Mobility Optimization RSRP RSRP A3 trigger A5 trigger RSRP A2 trigger A5 thold, y‘ A1 threshold A3 offset A2 threshold A1 trigger RSRP serving move direction RSRP neigh RSRP serving A5 thold, x‘ RSRP neigh move direction RSRP serving move direction Figure 8.6 Handover events. 3GPP defines in TS36.331 several handover triggers in LTE. The handover events A1, A2, A3, A4, A5, B1, and B2 currently supported by eNB are listed below (Figure 8.6): ●● ●● ●● ●● ●● ●● ●● ●● Event A0: Periodical reporting Event A1: Serving becomes better than threshold Event A2: Serving becomes worse than threshold Event A3: Neighbor becomes offset better than serving Event A4: Neighbor cell becomes better than a threshold value. Event A4 is mainly used to configure UE to perform measurements on inter‐frequency carriers for offloading purpose Event A5: Serving becomes worse than threshold1 and neighbor becomes better than threshold2 Event B1: Inter‐RAT neighbor cell becomes better than a threshold value. Event B1 is use for CS fallback to UTRAN and GERAN Event B2: Serving becomes worse than threshold1 and inter RAT neighbor becomes better than threshold2 The parameters of hysteresis of intra_LTE mobility are concluded in Table 8.1 and Table 8.2. Measurement command message is included in RRC connection reconfiguration message. Measurement configuration is depicted in Figure 8.8. Table 8.1 Use of hysteresis for various mobility events – intra‐frequency mobility. Mobility event Type Event triggered condition Event cancelled condition Start/stop of measurements of intra‐frequency cells A1/A2 RSRPserv < threshold1 Better cell handover (intra‐frequency) A3 RSRPneigh > RSRPserv + a3Offset + hysA3Offset RSRPneigh < RSRPserv + a3Offset ‐ hysA3Offset Coverage handover (intra‐frequency) A5 RSRPserv + hysThreshold3 < threshold3 AND RSRPneigh – hysThreshold3 > threshold3a RSRPserv – hysThreshold3 > threshold3 OR RSRPneigh + hysThreshold3 < threshold3a 259 260 LTE Optimization Engineering Handbook Table 8.2 Use of hysteresis for various mobility events – inter‐frequency mobility. Mobility event Type Start of measurements A2 of inter‐frequency cells Event triggered condition Event cancelled condition RSRPserv < threshold2InterFreq – hysThreshold2InterFreq Better cell handover (inter‐frequency, RSRP based) A3 RSRPneigh – RSRPserv > a3OffsetRsrpInterFreq + hysA3OffsetRsrpInterFreq RSRPneigh – RSRPserv < a3OffsetRsrpInterFreq – hysA3OffsetRsrpInterFreq Better cell handover (inter‐frequency, RSRQ based) A3 RSRQneigh – RSRQserv > a3OffsetRsrqInterFreq + hysA3OffsetRsrqInterFreq RSRQneigh – RSRQserv < a3OffsetRsrqInterFreq – hysA3OffsetRsrqInterFreq Coverage handover (inter‐frequency A5 RSRPserv + hysThreshold3InterFreq < threshold3InterFreq AND RSRPneigh – hysThreshold3InterFreq > threshold3aInterFreq RSRPserv – hysThreshold3InterFreq > threshold3InterFreq OR RSRPneigh + hysThreshold3InterFreq < threshold3aInterFreq Stop of measurements A1 of inter‐frequency cells RSRPserv > threshold2a + hysThreshold2a Note: Event‐A2—serving cell becomes worse than a threshold—can be triggered by different thresholds. Event‐A2 can be used to trigger, for example, RRC connection release with redirection, for inter‐frequency measurements, or inter‐RAT measurements when UE encountering bad coverage can measure on several inter frequency and/or IRAT target frequencies. Mobility control at cell edge needs to define and prioritize the preferred frequencies for measurements and mobility actions and possibility to perform blind handover or release with redirect to that RAT at critical threshold. In LTE network design, there is a search zone introduced shown in Figure 8.7, at search threshold, start inter‐frequency and IRAT measurements. Search zone Search threshold A2 -> Search activity Search for other Freqs/RATs. If found, do HO or RwR. Based on A3/A5 + B2 or Blind RwR/HO Figure 8.7 Search zone. 8.1.3 Handover Procedure Intra‐LTE handover have two versions: S1‐based and X2‐based. The X2 handover is normally used for inter‐eNB handover and is preferred because of its packet‐forwarding feature to avoid data loss. If the X2 interface is not available or the source eNB is configured to use S1 based handover, then S1 handover is used. From the UE perspective, the two procedures are identical. In fact, the UE doesn’t even know which handover procedure is executed. The successful handover was determined by verifying the following message sequence as shown in Figure 8.9: ●● ●● ●● Measurement report sent from UE to source cell. RRC connection reconfiguration message including MobilityControlInfo sent by a source cell to UE, and RRC connection reconfiguration complete sent from UE received by a target cell. Mobility Optimization 8.1.3.1 X2 Handover Based on UE measurement and RRM information, the source eNB decides that a handover to the target eNB is necessary. The source eNB sends a handover request message over the X2 interface to the target eNB. The message contains necessary information to prepare the handover at the target side. The target eNB allocates resources for the target cell and the UE is allocated a new C‐RNTI for identification in the target cell. The target eNB sends a handover request acknowledge message to the source eNB, which in turn sends handover command over the air interface to the UE, including necessary information (e.g., the new C‐RNTI) so that the UE can perform the handover. From statistics in a live network, about 30% of drop calls was due to X2 is failure (Figure 8.10). During the X2 handover procedure, the eNB performs (Figure 8.11): ●● ●● ●● ●● ●● ●● ●● An X2AP handover preparation procedure An X2AP SN status transfer procedure if the PDCP SN status preservation applies for at least one of the (RLC‐AM) radio bearers handed over An RRC connection reconfiguration procedure for mobility within EUTRAN A X2 user‐plane data forwarding An X2AP UE context release procedure An S1AP path switch request procedure A UE context deletion in the source eNB (with associated resources) The source eNB sends PDCP sequence number (SN) information to the target eNB in an SN status transfer message. This information is necessary to avoid missing or duplicating PDCP packets when the uplink and downlink user data paths are switched from the source eNB to the target eNB. Also, the source eNB now forwards the received downlink user data packets to the target eNB instead of sending them to the UE. The downlink user data packets are buffered in the target eNB until the handover is completed. As soon as the handover command (RRC connection reconfiguration) message is received, the UE buffers the uplink user data until the handover has been completed, detaches from the source cell, and synchronizes with the target cell using the non‐contention based random access procedure. Next, the UE sends a handover confirmation (RRC connection reconfigura tion complete) message to the target eNB to indicate that the handover procedure is completed as far as the UE is concerned. Now the UE can start sending the buffered uplink user data and the target eNB can forward the downlink user data to the UE. The uplink user data is sent via the target eNB directly to the serving gateway, since the uplink tunnel endpoint identifier (TEID1) in the S‐GW was conveyed to the target eNB already. The target eNB sends path switch message to MME to inform that the UE has changed cells, including its downlink tunnel endpoint identifier (TEID). MME forwards it to the S‐GW so that the S‐GW can send the downlink user data directly to the target eNB. Before the S‐GW can release any user plane resources toward the source eNB, it sends one or more “end marker” packets to the source eNB as an indication that the downlink data path has been switched. It should be noted that these packets do not contain any user data, and are transparently forwarded by the source eNB to the target eNB to help it decide when the last forwarded packet was received. After receiving an acknowledgment message, the target eNB informs the source eNB about the success of the handover. As a final step, the source eNB releases all air interface and control plane resources associated with the UE context, and the handover is completed. Figure 8.12 is an example from eNB trace, which filter out the UE related message. 1 Different S1 bearers are identified by their tunnel endpoint identifier (TEID), which is allocated by the endpoints (eNB and S-GW) of the GTP tunnel. 261 Measurement identity Reference number in measurement report Measurement objects Reporting configurations Quantity configuration A carrier frequency A list of neighboring cell offsets IRAT neighboring cells (no neighbor list for intra-LTE mobility) Reporting criteria: periodical or event-triggered Reporting format: quantities (e.g. number of cells to report) e.g. RSRP or RSRQ One quantity for intra freq, one for inter and one for each RAT type Figure 8.8 Measurement configuration. LTE Optimization Engineering Handbook UE eNB 1. RRC Connection Reconfiguration (Measurement Control) Legend L3 signalling Packet data L1/L2 signalling UL allocation User data 2. Measurement Reports THOoverall 4. Admission Control and Resource Allocation DL allocation Handover Preparation 3. HO Decision Handover Execution 5. RRC Connection Reconfiguration Detach from old cell and synchronise to new cell 6. Synchronisation 7. UL allocation + TA for UE 8. RRC Connection Reconfiguration Complete 9. Release Resources Packet data Handover Completion 264 Figure 8.9 Intra‐eNB handover. Handover procedure works as expected, data is forwarded over X2, path is switched to target cell and average user plane interruption time was about 200ms but varied between 100ms and 500ms depending on the loading in the network. The source eNB issues X2 handover request message to the target eNB by passing necessary info to prepare the handover and gets the X2 handover request acknowledge message back. The detail content in the two messages is listed in Figure 8.13. Mobility Optimization Source eNB UE Target eNB Serving GW MME 0. Area Restriction Provided 1. RRC Connection Reconfiguration (Measurement Configuration) Packet data Packet data UL allocation Legend L3 signalling 2. Measurement Reports Handover Preparation L1/L2 signalling 3. HO Decision Tx2ACLOCpre User data 4. Handover Request 5. Admission Control and Resource Allocation Tx2ACLOCoverall Tx2ACLOCarea 6. Handover Request Acknowledge DL allocation 7. RRC Connection Reconfiguration Deliver buffered and in transit packets to target eNB Handover Execution Detach from old cell 8. SN Status Transfer Data Forwarding Buffer Packets from Source eNB 6. Synchronisation 10. UL allocation + TA for UE 11. RRC Connection Reconfiguration Complete T 12a. RRC Connection Reconfiguration (Measurement Configuration) x2ACLOCarea/path/assignment Tx2ACLOCcomp 13. Path Switch Request 12b. RRC Connection Reconfiguration Complete 14. User Plane Update Request Packet data 17. Path Switch Request Acknowledge TDATAtwdD2 TDATAtwdsd 16. User Plane Update Response 18. UE Context Release Release S1, X2 signalling and radio resources. Continue forwarding packets Release X2 Signalling Connection Data Forwarding End Marker 19. Release Remaining Resources Packet data Figure 8.10 X2 handover procedure. Figure 8.11 X2AP handover preparation. Begin sending S1-U data Release Other X2 Resources 15. Switch DL Path Packet data Handover Completion End Marker 265 266 LTE Optimization Engineering Handbook Figure 8.12 X2 handover procedure from test tool. X2 HO Request Message HO Preparation Info Serving cell configuration -MIB/SIBs UE radio configuration and capability -existing radio bearers (SRBs/DRBs) UE security capabilites, Last visited cell lists, ... ERAB ID, ERAB level QoS parameters, Target cell ID Global MME indentity …… X2 Handover Request Acknowledge eNB-UE-X2AP-ID Admitted E-RAB list GTP TEID Target eNB to Source eNB Transparent Container A new C-RNTI Target eNB security algorithm identifiers A dedicated RACH preamble (optional) …… Figure 8.13 X2 handover request message and handover request acknowledge message. It is worth to pay attention to the X2 round‐trip delay, if it is large and the UE is moving at high speed, the UE could lose contact with the source eNB due to increasingly poor radio conditions resulting in failure to send handover commands to the UE and possible call drop. Different propagation conditions cause handover performance to be more or less sensitive to Mobility Optimization this delay. For the most delay‐sensitive applications, a one‐way delay over X2 of up to 50 to 70 ms is probably not noticeable, for example, for VoIP. 8.1.3.2 S1 Handover Source eNB decides to initiate S1‐based handover to target eNB based on UE measurement report when there is no X2 connectivity to the target eNB, or by an error indication from the target eNB after an unsuccessful X2‐based handover. The measurement procedures are performed in the same way as for X2 handover. When S1 handover is executed, the source eNB initiates the handover preparation by sending the handover required massage to the serving MME. The source MME sends a forward relocation request to the target MME over the S10 interface. During the S1 handover procedure, the eNB performs: ●● ●● ●● ●● ●● ●● An S1AP handover preparation procedure An S1AP SN status transfer procedure if the PDCP SN status preservation applies for at least one of the (RLC‐AM) radio bearers handed over An RRC connection reconfiguration procedure for mobility within EUTRAN A X2 user‐plane or S1 user‐plane data forwarding An S1AP handover notify procedure A UE context deletion in the source eNB The eNB general procedure for inter‐eNB S1 handover is shown in Figure 8.14. The target MME sends a handover request to the target eNB. The target eNB may admit or reject this request with its admission control function. If it admits the request, the handover request is acknowledged. Then the forward relocation is acknowledged with a response and now the source MME sends the handover command to the source eNB, which forwards it to the UE. Now the UE can perform the random access procedure in the target cell and the handover confirm is sent to the target eNB. Handover notify is sent to target MME in order to indicate to the target MME that the handover has succeeded. The context in the old eNB is released. From an end‐to‐end view, S1 handover can be used to achieve the following: ●● ●● ●● ●● S1 handover with MME and SGW relocation S1 handover with MME relocation (no SGW relocation) S1 handover with SGW relocation (no MME relocation) S1 handover without any EPC node relocation 8.1.3.3 Key point of X2/S1 Handover Intra LTE handover phases can be divided into handover preparation phase, handover execution phase, and handover completion phase. The handover preparation phase is the phase when the source RAN request the target RAN to prepare to accommodate the UE. The target RAN performs admission control and reserves cell resources required for the UE and prepares the UE access and sends the needed access information in a container sent transparent through the source RAN to the UE. The handover execution phase starts with the source RAN command the UE to make a handover. The source RAN performs packet forwarding to target RAN during the execution phase. The phase ends when the UE access the target RAN and confirm the new connection in the target cell. In the handover completion phase the user data path to the SAE gateway is switched over to the target RAN. The handover complete phase starts with the target RAN notice the MME and SAE gateway that the handover has occurred. The MME/SAE gateway switches the bearer from source RAN to target RAN. Bearer between SAE gateway and the RAN is set up per eNB 267 UE RRC Connected Source eNB Target eNB Source S-GW Target S-GW Source MME Target MME 1. RRC Connection Reconfiguration (Measurement Conf) 2. RRC Measurement Report (Event A3) 3. HO Decision 4. S1 Handover Required (Source to target Transparent Container) 5. S10 Forward Relocation Request 6. S11 Create Session Req/Res 7. S1 Handover Request 8. Admission Control 9. S1 Handover Request Acknowledge 10. S10 Forward Relocation Response 11. S11 Create Bearer Req/Res Up Forwarding 12. S1 Handover Commond 13. RRC Connection Reconfiguration Handover Commond Regenerate Security Keys 14. MAC: Random Access Preamble 15. MAC: Random Access Response (UL Allocation+TA) 16. RRC Connection Reconfiguration Complete (Handover Confirm) 17. S1 Handover Notify 18. Data Transfer in Target 19. S10 Forward Relocation RRC Connected Figure 8.14 Inter eNB S1 handover with MME relocation. 20. S1 UE Context Release (Cause: Successful Handover) Complete/ACK Mobility Optimization and not per cell. The target eNB informs the source eNB that the handover is complete. The source eNB finishes the packet forwarding (if used) and releases the UE resources. If the UE is moving to a cell in the same eNB then performs no switching of the bearers. Handover preparation by source eNB and target eNB are shown in Figure 8.15. Handover user plane interrupt time UL or DL interrupt time can be calculated from the instant that UE receives handover command to disconnect user plane to the old cell to the instant that user plane is re‐established. Interrupt time is different depending on a) DL or UL b) DL data forward option. With data forwarding, DL interrupt time is packet delay time, without data forwarding, DL interrupt time is packet loss time. The example in Figure 8.16 showed that user plane S1 handover interruption time is 48 ms. S1 handover failure analysis The symptoms in Figure 8.17 will usually lead to unsuccessful S1 handover, which will bring in low throughput and higher call drop rate in a live network. In Figure 8.17, MME does not reply to S1 handover request till tS1relocprep expired; in this case, we tried to prolong tS1relocprep, and will expect more core network rejection failures. S1 rejection with “unknown‐target ID,” it is suspected that the interworking between source MME and target was unsuccessful due to core network configuration fault or mapping fault. IF S1 rejection is with “ho‐failure‐in‐target‐EPC‐eNB‐or‐target‐system,” in this case the failure cause is quite clear: core network configuration needs to be checked to indicate the fault. 8.2 Mobility Parameter When the UE is powered on, before RRC, it says that the UE is in idle mode. In order to save battery power, UEs enter in idle mode where they are connected to EPC only (not to eUTRAN), that is, context of UE is deleted from eNB and is maintained at MME and SGW only. eNB releases all dedicated resources when UE enters idle mode. In this mode, UE shall only wake up at fixed intervals of time to check whether it has any incoming notifications for data, also UE can wake up at any time when it has any data to send. This mode is called RRC_idle and state is called EMM_registered state. EMM‐registered has two substates: idle (ECM‐idle + RRC‐ idle), and connected (ECM‐connected + RRC‐connected). The two EPS mobility management (EMM) states including EMM‐deregistered and EMM‐registered state, which describe the mobility management states that result from the mobility management procedures, for example, attach, detach, and tracking area update procedures. The two EPS connection management (ECM) states including ECM‐idle and ECM‐connected state, which describes the signaling connectivity between the UE and the EPC. Table 8.3 depicts RRC idle and RRC connected state. Idle mode mobility management is UE chooses a suitable cell to camp according to the system broadcast message sent by the eNB, in order to improve the success rate of UE access and quality of service. Connected mode mobility management is when UE moves in the connected state; the network offers the continuity of the service by handover for UE. UE moves from idle to connected by sending an initial NAS message. The initial NAS message in 3GPP Rel 8 can be attach request, detach request, tracking area update request, service request, or extended service request. When UE goes from RRC_connected to R RC_ idle, the bearers marked X are taken down that is shown in Figure 8.18. 269 X2 Handover Intra-RBS Handover Intra-RBS Handover X2 Handover Receive X2 Handover Request message from Source RBS S1 Handover Get UE Information Get Cell Information License exceeded? -number of connected users X2 Handover of Intra-RBS? Intra-RBS Handover X2 Handover S1 Handover: Place in Source To-Target Transparent S1AP container with target cell details Yes Handover Allowed? -UE Cap OK? -Resources Avail? No X2 Handover of Intra-RBS? Send S1 Handover Required message No Yes Yes Place in S1 Handover Required Message Send X2 Handover Request message Extract Source-ToTarget S1AP Container No No Place in X2 Handove Handover Request Message Intra-RBS Handover Receive S1 Handover Request message to MME Get Handover Preparation Information -UE/Cell Information Place UE/Cell Information in Handover Preparation Information RRC Container Yes S1 Handover Send S1 Handover Failure to MME Send X2 Handover Preparation Failure to Source RBS Build RRC Reconfiguration Message X2 Handover Place in Handover Command of Intra-RBS? RRC Message Yes No Create Target-To-Source eNB container Send S1 Handover Request Ack to MME Send X2 Handover Request Ack to Source RBS Handover preparation by source eNB Figure 8.15 Handover preparation. Handover preparation by target eNB Figure 8.16 User plane S1 handover interruption time. tS1relocprep expired unknown-target ID Figure 8.17 Example of S1 handover failure. Table 8.3 RRC idle and RRC connected state. UE RRC connected RRC idle Network ›Listens to the PDCCH for its assign cell RNTI ›Knows the UE on a cell level ›Has E‐RABs established ›Has a UE context in core nodes and an eNB. ›May send/receive data on the shared channels ›Controls mobility based on UE measurement reports. ›Listens to the PD‐CCH for a paging RNTI ›Knows the UE to within a tracking area. ›Performs random access and connection establishment procedure when it is paged or needs to send/receive data. ›Has a UE context only in core nodes. ›Controls mobility based on system information ›Needs to page the UE to send/ receive data. Note: An EPS bearer carries traffic between the UE and PGW using an enhanced radio access bearer (E‐RAB) between the UE and SGW. An E‐RAB uniquely identifies the concatenation of an S1 bearer and the corresponding data radio bearer. When an E‐RAB exists, there is a one‐to‐one mapping between this E‐RAB and an EPS bearer of the NAS. The E‐RAB is can be thought of a PDP context of previous 3GPP releases. 272 LTE Optimization Engineering Handbook NAS Signaling Connection RRC Connection S1 Connection S11 Tunnel S5 Tunnel Signaling Radio Bearer (SRB) E-UTRAN Radio Access Bearer (E-RAB) Data Radio Bearer (DRB) S1 Bearer S5 Bearer EPS Bearer Figure 8.18 The bearers changed as RRC_connected to RRC_idle. 8.2.1 Attach and Dettach UE needs to register with the network to receive services by the attach messages when a UE is first connecting to the network after being switched on. The UE enters the EMM‐registered state by a successful registration with an attach procedure. The RRC connection is established, with the NAS message attach request, transferred to the network as part of the procedure. The attach request message is provided to the MME in the “initial UE message.” The UE is identified either by the GUTI or the IMSI (in the attach request message). After registration, a context is established for the UE in the MME, and a default bearer is established between the UE and the PDN GW, thus enabling always‐on IP connectivity to the UE. The HSS provides the subscribed QoS (APN‐AMBR, default EPS bearer QCI) for the default APN to the MME. The MME selects and derives the addresses of the SGW and PGW and creates the default EPS bearer in the SGW by sending the message create session request. The MME provides the IMSI, APN name, subscribed QoS, PDN GW S5/S8 address, and the PDN type to the SGW. The default EPS bearer is established using the initial radio bearer establishment procedure. The TEID and the IP address for the UL of the user plane received from the SGW are used. The NAS message attach accept including activate default EPS bearer context request is transferred to the UE during the default radio bearer establishment. If the UE was identified by the GUTI in the attach request message the MME will re‐allocate the GUTI, that is, the new GUTI will be provided in the attach accept message. Finally, the NAS message “attach complete” (including the session management message activate default EPS bearer context accept) is transferred from the UE (Figure 8.19). The MME updates the SGW with the TEID and IP address for the DL of the user plane, received from the eNB by sending the message modify bearer request. There are four tunnels that are set up when a UE attaches to the network, two S1 tunnels, one S5 tunnel and one S11 tunnel. If a user has dedicated bearers each bearer will be a GTP tunnel. First the S11 and S5 tunnel (S5, S1, and S11 tunnels are GTPv2 tunnels) is set up, next is the S1 tunnel from the SGW to the eNB, and finally the S1 tunnel from the UE to the eNB. There is a tunnel setup between the PGW and the PDN, this tunnel is a GTPv1 tunnel. GTPv1 is unsecure and the data in the tunnel can be seen using a protocol analyzer. Data within a GTPv2 tunnel is secure and cannot be seen. Detach is the process of turning a UE off and disconnecting from the network, which is used to remove bearers and clear states in the network. The network or the UE explicitly requests detach and signal with each other. There are three types of detach: UE‐iInitiated detach procedure, MME‐initiated detach procedure, and HSS‐initiated detach procedure, which is shown in Figure 8.20. Mobility Optimization UE eNB MME RRC: RRC Connection setup complete S1AP: Initial UE message NAS: Attach Request NAS: PDN connectivity request NAS: Attach Request NAS: PDN connectivity request SGW PGW HSS the first message sent to the MME to establish a connection Authentication and NAS security procedure S6a: Update Location request S6A (MME to HSS) diameter signalling S6a: Updated Location answer S11: Create bearer request S11: Create bearer response S5: Create bearer request S5: Create bearer response S1AP: Initial Context setup request RRC: RRCConnectionReconfiguration NAS: Attach accept NAS: Activate Default EPS bearer request NAS: Attach accept NAS: Activate Default EPS bearer request RRC: RRC ConReconfigComplete S1AP: Initial Context setup response RRC: UL information Transfer S1AP: Uplink NAS Transport Default bearer establishment The HSS accesses the database and responds with user information to the MME This message from the MME is a three in one message NAS: Attach complete NAS: Activate Default EPS bearer accept NAS: Attach complete NAS: Activate Default EPS bearer accept Figure 8.19 Attach procedure with initial EPS bearer establishment (3GPP TS23.401). The detach procedure allows the UE to inform the network that it does not want to access the EPS any longer or the network to inform the UE that it does not have access to the EPS any longer. The UE sends the NAS message “detach request” to the MME, using the UL NAS signaling transfer procedure. If the reason for the detach was not “power off,” the MME sends the NAS message “detach accept” to the UE, and releases the EPS bearers in the SGW by sending the message delete session request. The S5/S8 bearer is released using the S5/S8 bearer release procedure. 8.2.2 UE Measurement Criterion in Idle Mode and Cell Selection When the UE is switched on, it performs the PLMN and cell selection procedure, which involves sweeping UE supported frequency bands, and searching the carrier frequencies with highest power for cells in a non‐forbidden PLMN. The selection of PLMN can be automatic or manual, and pre‐stored frequency and RAT information can be used to accelerate the procedure. The cell selection criteria is defined in 3GPP standards 36.3042 UE procedures in idle mode to determine which LTE cell to camp on. It is based on the measured reference signal received power (RSRP) level in the cell. Cell selection parameter is shown in Table 8.4. The cell selection criterion is fulfilled if: Srxlev 0,Srxlev Qrxlevmeas Qrxlevmin Qrxlevminoffset Pcompensation 2 The cell camping criteria in Rel 9 is based on both level and quality while cell reselection to a cell with different absolute priority is based on either level or quality. 273 UE eNodeB MME Serving GW PDN GW UE PCRF 1. Uplink NAS Signalling Transfer NAS Message =“Detach Request” eNodeB MME Serving GW PDN GW PCRF 1. Downlink NAS Signalling Transfer NAS Message =“Detach Request” 2. Delete Session Request 2. Downlink NAS Signalling Transfer NAS Message =“Detach Accept” 3. S5/S8 PDN Connectivity Release 3. Delete Session Request 4. Delete Session Response 5. Uplink NAS Signalling Transfer NAS Message =“Detach Accept” 4. S5/S8 PDN Connectivity Release 5. Delete Session Response 6. MME-initiated Connection Release UE 6. MME-initiated Connection Release eNodeB MME Serving GW PDN GW PCRF HSS 1. Cancel-Location-Request IMSI, Cancellation Type = SUBSCRIPTION_WITHDRAWAL 2. Cancel-Location-Answer 3. MME-initiated Detach from ECM-CONNECTED state Figure 8.20 UE‐initiated, MME‐initiated and HSS‐initiated detach from ECM‐connected state. Mobility Optimization Table 8.4 Cell selection parameter. function Srxlev Cell selection RX level value calculated by the UE (dB) Qrxlevmeas Measured cell RX level value (RSRP) Qrxlevmin Indicates the minimum required receive level used in intra‐frequency E‐UTRA cell reselection, which contains information relevant when evaluating if a UE is allowed to access a cell and defines the scheduling of other system information, default: −128dBm Qrxlevminoffset This parameter defines an offset to be applied in cell selection criteria by the UE when it is engaged in a periodic search for a higher‐priority PLMN. (dB). Increasing the value of this parameter would determine an earlier reselection of the target neighboring cell. Decreasing the value would determine a later reselection of the target neighboring cell. If you do not use inter‐PLMN mobility, this parameter is inhibited. Pcompensation Pcompensation is a compensation factor to penalize the low power mobiles. Max(PMax – UE Maximum Output Power, 0) (dB) PMax Indicates the maximum TX power of the UE in the cell. If this parameter is not included in the SIB3, the UE uses the maximum TX power Inter Freq Target Cell SourceCell CDMA Target Cell CellSelectionReselectionConf:: qrxlevmin in SIB1 CellReselectionConfLte:: qrxlevmin in SIB3 CellReselectionConfLte:: qrxlevmin in SIB5 CellReselectionConfUtraFdd or Tdd:: qrxlevmin in SIB6 CellReselectionConfGERAN:: qrxlevmin in SIB7 Snonservingcell = –FLOOR (–20 x log10Ec/Io) in SIB8 (No qrxlevmin) Intra Freq Target Cell WCDMA Target Cell GERAN Target Cell Figure 8.21 Qrxlevmin configuration. Qrxlevmin configures the cell minimum required RSRP level used by the UE in cell reselection as shown in Figure 8.21. Once the cell selection is accomplished, UE dominated reselection is about to happen, the following rules are used by the UE: ●● ●● ●● Intra‐frequency criterion: If Srxlev ≤ sIntraSearch, the UE performs intra‐frequency measurements, if Srxlev > sIntraSearch, the UE does not perform these measurements (Figure 8.22) Inter‐frequency and/or inter‐RAT criterion: If Srxlev ≤ SNonIntraSearch, the UE performs inter ‐frequency and/or inter‐RAT measurements, if Srxlev > SNonIntraSearch, the UE does not perform these measurements sIntraSearch: Specifies the threshold (RSRP) for intra‐frequency measurements, i.e. how bad must the serving cell before the UE starts to measure on neighboring cells Figure 8.23 gives a 4G measurements example. Measurement starts when Srxlev < sIntraSe arch. sIntraSearch can be set to 62 to allow smooth intra‐LTE cell reselection. When RSRP < −68: start to measure intra‐frequency neighbors. When RSRP < −114: start to measure on inter‐ frequency LTE and 3G. 275 276 LTE Optimization Engineering Handbook Candidate Layer Priority relative to Serving Cell Priority Lower Equal Sservingcell > Sservingcell ≤ Sservingcell > Sservingcell ≤ Snonintrasearch Snonintrasearch Snonintrasearch Snonintrasearch Snonintrasearch Measurements Inter-Frequency is broadcast not mandatory Candidate Snonintrasearch Layer is not broadcast Inter-RAT Candidate Layer Snonintrasearch Measurements is broadcast not mandatory Snonintrasearch is not broadcast Measurements Measurements not mandatory mandatory Measurements mandatory Measurements mandatory Measurements mandatory Higher Measurements mandatory Not Applicable Measurements mandatory Figure 8.22 4G measurements. sIntraSearch = 62 qrxLevMin = –130 dBm sNonIntrSearch = 16 RSRP –68 –114 –130 –140 Figure 8.23 4G measurements example. To minimize the impact on UE battery performance, it is possible to set the threshold above which UE may not measure intra‐frequency neighbor cells. When the UE triggers a cell reselection evaluation process, it ranks cells that fulfill the cell selection criteria. The UE ranks the cells according to the R criteria. 8.2.3 Cell Priority In the LTE system, cell reselection from LTE to other LTE frequencies or to the other RATs is based on absolute priorities provided to UEs in the system information (SI). The priority of a given neighbor relation is set by its cellReselectionPriority. All priority settings range from 0 to 7 with 7 being the highest priority. Different priority for different IF/RATs, which can be found in SIB (SIB8 →CDMA‐eHRPD, SIB7 →GSM, SIB6 →WCDMA, SIB5 → inter frequency, SIB3 → for serving cell). UE will periodically search for higher‐priority layer, search for lower priority if below a threshold (Figure 8.24). UE must perform measurements of layers with a priority higher than the priority of the current serving cell regardless of serving cell signal level. Some priority configurations of idle mode, connected mode and CS fallback are shown in Table 8.5 and Table 8.6. 8.3 Intra‐LTE Cell Reselection When the UE is in idle mode, it uses reselection for its mobility. In idle mode mobility it is largely controlled by the UE (based on broadcast parameters) and so the exact behavior may vary somewhat between UE models. UE takes the measurements of the power of signals both for serving and neighbor cells in order to camp on better cell with high priority or with better power signal. The eNB broadcasts idle mode information that assists and controls the UE to Mobility Optimization Layer 7 with priority = 7 (=highest) S crit.< Thresh. Periodic Search Start’s Search Layer 6 with priority = 6 Layer 5 and lower (0 is lowest) priority 7 6 5 f2 f1 f3 frequency UE always tries to reselect on the highest possible priority layer! Figure 8.24 Cell priority. Table 8.5 Idle mode and Connected mode. Idle mode Connected mode MO Parameter Value EUtranFreqRelation (1825/19825) cellReselectionPriority 7 UtranFreqRelation (UARFCN = 10638) cellReselectionPriority 5 UtranFreqRelation (UARFCN = 10662) cellReselectionPriority 4 GeranFreqGroupRelation (GSM_900) cellReselectionPriority 1 GeranFreqGroupRelation (GSM_1800) cellReselectionPriority 0 MO Parameter Value UtranFreqRelation (UARFCN = 10638) connectedModeMobilityPrio 5 UtranFreqRelation (UARFCN = 10662) connectedModeMobilityPrio 4 GeranFreqGroupRelation (GSM_900) connectedModeMobilityPrio 1 GeranFreqGroupRelation (GSM_1800) connectedModeMobilityPrio 0 277 278 LTE Optimization Engineering Handbook Table 8.6 CS fallback. MO Parameter Value GeranFreqGroupRelation (GSM_900) csFallbackPrio 6 GeranFreqGroupRelation (GSM_1800) csFallbackPrio 4 UtranFreqRelation (UARFCN = 10638) csFallbackPrio 1 UtranFreqRelation (UARFCN = 10662) csFallbackPrio 0 select PLMN and best suitable LTE cell according to LTE inter‐carrier priority information and IRAT priority information. 8.3.1 Cell Reselection Procedure The cell reselection process is run when the UE, in “camped normally” state, has found a better neighboring cell than the cell on which it is camping. During cell reselection procedure the UE is not continuously measuring neighbor cells in search of a better cell to camp on but only performs these measurements when the (S) criteria below is met: Srxlev Pcompensation Sintrasearch,where Srxlev max pmaxServingCell Qrxlevmeas qrxLevMin qrxLevMinOffset Pcompensation P;0 Pcompensation factor, which considers the UE power (23dBm) and the maximum allowed power in the cell (PmaxServingCell). This is done in order to prohibit low power UEs to select large cells. In the example: sintrasearch = 62, qrxLevMin = 130 dBm, The UE starts measuring for a better cell when QrxlevMeas is worse than (62−130) = −68dBm. If the UE moves, a new cell might become better than the currently selected cell. In LTE there is a threshold below which the neighbor cell measurements for cell reselection can be triggered. Once the measurements for neighbor cells have been triggered the UE ranks the measured cells which fulfill the S‐criterion according to the R‐criterion: intra‐LTE cell reselection occurs when neighbor cell becomes better than source cell (Rn > Rs) within a time interval (TreselectionEUTRA), as shown in Figure 8.25. R _ s Qmeas , s Qhyst ; R _ n Qmeas ,n Qoffset where Qoffset is: qOffsetCellEUtran: Cell individual offset in the intra‐frequency and equal priority inter‐frequency cell ranking criteria. qOffsetFreq: Frequency specific offset in the equal priority inter‐frequency cell ranking criteria. Different cell reselection parameters settings will lead to different network performance. If sintrasearch sets to low, this means neighbor reselections are happening very late and can result in paging/RACH/RRC setup failures. This has been proposed that it should to be changed to 62, which means neighbors for reselection are always monitored. A drawback for this can be increased battery consumption for UEs. Figure 8.26 presents three sets of parameters configuration and their performance. The recommended setting can reduce the number of reselections with only moderate decrease in RF performance. In LTE, it is possible to use different settings for the cell reselection parameters tReselectionEutra and qHyst based on the estimated speed of the UE. The UE estimates its speed as the number of cell reselections per evaluation period. The speed is divided into normal, medium, and high Mobility Optimization S (dB) 60 Signal quality on neighbor Cell B gets better than on serving Cell A but ... 50 40 Rs = Qmeas,s + QHyst Rn = Qmeas,n – Qoffset Cell B 30 Sintrasearch 20 Srxlev,s (Cell A) Srxlev,n (Cell B) 10 0 Cell A 1 2 –10 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Parameter values considered: • Qrxlevmin = –110 dBm –20 • QHyst, Qoffset, Pcomp assumed 0 Time (sec) ... measurement on any neighbor will start when Cell B selected after Rn > Rs satisfied for TReselectionEUTRAN SservingCell < Sintresearch • Rs = Qrxlev,s; Rn = Qrxlev,n • TReselectionEUTRAN: 3 seconds Figure 8.25 Intra‐frequency cell reselection. Parameter/Metric Qrxlevmin Simulated Setting 1 –120 dBm Original Simulated Setting 2 –120 dBm Recommended Simulated Setting 3 –120 dBm Sintrasearch 62 dB 62 dB 62 dB QHyst 2 dB 4 dB 8 dB 0 0 0 TreselectionEUTRAN 1 sec 2 sec 4 sec DRX Cycle 1.28 s 1.28 s 1.28 s Qoffsets,n Number of reselections 96 42 24 Percentage of time OOS 0% 0% 0% Average RSRP (Camped Cell) Average RSRQ (Camped Cell) –96.74 dBm –8.8 dB –96.84 dBm –97.82 dBm –8.9 dB –9.7 dBm Figure 8.26 Different cell reselection parameters settings. speed. At higher speeds, the tReselectionEutra timer and the qHyst values can be adjusted to trigger the cell reselection quicker and vice versa. 8.3.2 Inter‐Frequency Cell Reselection LTE cell reselection uses priority‐based levels. Parameter sNonIntraSearch decides when UE starts searching and measuring all priority cells. Priority based cell reselection is configured so that LTE frequencies have either higher or lower priority than 2/3G serving cell priority. 279 280 LTE Optimization Engineering Handbook In normal case, LTE frequency is configured to have higher priority, to help push capable UEs toward LTE. Cell reselection priority is the main criteria that determines UEs will camp on which layer. UE will perform call establishment on the layer it is camped on. Idle mode load distribution between layers helps reduce the need for the connected mode load balancing. Load‐ based adaptation of cell reselection thresholds can push cell‐edge UEs down to lower‐priority layer. Inter‐frequency cell reselection measurement trigger depends on the absolute priorities of the serving and non‐serving layers. For each neighboring frequency/layer, it is necessary to define the corresponding priority. UEs close to cell center will camp on high priority although RSRP is higher on the other frequency, UEs close to cell edge will camp according to priority. The idle mode distribution is possible to be adjusted with the thresholds. Inter‐frequency and inter‐RAT cell reselection are decided by higher‐priority cell reselection threshold ThreshXHigh and lower‐priority cell reselection threshold ThreshXLow. ThreshXLow: Threshold for the Srxlev value of the target cell for cell reselection toward a lower‐priority inter‐frequency or inter‐RAT frequency. UE will perform cell reselection toward a lower‐priority inter‐RAT when the Srxlev value of the serving cell is below threshServingLow and the Srxlev value of the target cell is above threshXLow for period of time. ThreshXHigh: Threshold for the Srxlev value of the target cell for cell reselection toward a higher‐priority inter‐frequency or inter‐RAT frequency. The larger of the two values, it is more difficult to select to the corresponding cell. Table 8.7 is for reference. Priority based inter‐frequency/inter‐RAT cell reselection: 1) Low priority to high priority transition is shown in Figure 8.27: Periodic search for higher‐ priority layer Qrxlevmeasneighbour qRxLevMinInterF interFrqThrH 2) High‐priority to low‐priority transition is shown in Figure 8.28: Search for a lower priority if the UE received level is below a threshold Qrxlevmeass qRxLevMinIntraF thresholdSrvLow qRxLevMinOffsett P _ compensation AND Qrxlevmeasn qRxLevMinInterF interFreqThrL P _ compensation Table 8.7 ThreshXHigh and ThreshXLow. Parameter Recommend GERANNFREQGROUP ThreshXHigh −11dBm ThreshXLow −11dBm CDMA2000HRPDBANDCLASS ThreshXHigh −18dBm ThreshXLow −18dBm UTRANFDDNFREQ ThreshXHigh 0dB ThreshXLow 0dB UTRANTDDNFREQ ThreshXHigh 0dB ThreshXLow 0dB E‐UTRANINTERNFREQ ThreshXHigh 0dB ThreshXLow 0dB Mobility Optimization LTE Freq-A PRIO = 2 (highest) LTE Freq-B Serving cell PRIO = 1 PRIO = 1 Srxlev Low prio Treselection Threshx,high PRIO = 2 Cell Re-selection to Freq-A Search for higher prioritized layers at reguler intervals Figure 8.27 Low‐priority to high‐priority transition. LTE Freq-B Serving cell PRIO = 2 LTE Freq-A PRIO = 1 (lowest) Srxlev Low prio Threshx,low Treselection Threshserving,low Search for lower/same priority layers starts Cell Re-selection to Freq-A tReselectionEutra/Utra Higher Priority Cell Srxlev A D Lower Priority Cell Lower Priority Cell (QrRxlevmin + sNonIntra Search) C Threshx,low B Higher Priority Cell Threshserving,low Srxlev = Measured RSRP + Qrxlevmin Srxlev > 0 is a must for suitable cell Figure 8.28 High‐priority to low‐priority transition. A: Start to measure inter‐frequency (RAT) target cell (Srxlev ≤ SNonIntraSearch) B: Srxlev ≤ Threshservinglow, can now attempt to connect to lower‐priority cell C: The lower‐priority target cell Srxlev > ThreshXLow is met, tReselectionEutra/Utra timer starts D: The lower‐priority target cell has been above threshold for tReselectionEutra/Utra, the UE will reselect to lower cell. 3) Equal priority cell transition is shown in Figure 8.29. 281 282 LTE Optimization Engineering Handbook Start to measure inter-frequency target cell (Srxlev≤ SNonIntraSearch) Srxlev Serving Cell tReslectionEutra A C (QrRxlevmin + sNonIntraSearch) Target Cell B Target cell is stronger than serving cell by QHyst. Target cell is above QHyst for time tReselectionEutra, UE will reselect to target cell. QHyst Figure 8.29 Equal priority transition. 8.3.3 Cell Reselection Parameters The Qhyst is a hysteresis value of the serving cell used by the UE for ranking criteria in cell reselection, preventing too frequent reselection back and forth between cells of nearly equal rank. When a neighboring cell is ranked as better than the serving cell (that is, Rn > Rs) during a time interval tReselectionEutra (default value is 2s), the UE performs a cell reselection to the better‐ranked cell. It is worth to note that Qhyst should theoretically be set to a value at least equal to the handover hysteresis as cell reselection in idle mode is not as important as handover in connected mode. Qoffset is an offset in the cell ranking criterion of neighbor LTE cells, pertains to system information block 4 (SIB4) or neighbor cell in measurement configuration in RRC connected mode. It consists of a cell individual part and a frequency specific part. The frequency specific part applies to equal priority inter‐frequency cells only. If cell re‐selection is between the same priority, then, For intra freq, Rankneighbor For inter freq,Rankneighbor RSRPneighbor RSRPneighbor qOffsetCell qOffCell qOffFrq Increasing the value of Qoffset would determine a later reselection of the target neighboring cell (larger target cell list). Decreasing the value of Qoffset would determine an earlier reselection of the target neighboring cell (smaller target cell list) (Figure 8.30). One example of cell re‐selection to the same priority LTE cell between Micro and Macro (applicable if Micro is deployed with the same priority as Macro cell) in a live network, the re‐selection parameters can be set as: F1/F1: If there is a need to promote Micro, one must tune qHyst (direction Micro ‐ > Macro) or qOffsetCell (Macro ‐ > Micro). F1/F2: If there is a need to promote Micro, one must tune qHyst (direction Micro ‐ > Macro) or qOffFrq (Macro ‐ > Micro). qOffsetCell could be used to further differentiation between cells within Micro layer. The recommended re‐selection parameters are shown in Table 8.8. Mobility Optimization SI: RS :U Em ll r es ele easu res cti on RS pa ram RP ete rs ce s ea : RS m UE Cell reselect ion? tReselectionE utra Rn > Rs? UE performs cell reselection autonomously based on measurements Serving cell P SR sR ure Rs = QmeasS + QHyst Rn = QmeasN – Qoffset Neighboring cell RSRP sIntraSearch sNonIntraSearch qHyst(s) Qmeas(n) R(n) R(s) qoffset(s) Qmeas(s) tReselectionEutra time Figure 8.30 Cell reselection parameters. Table 8.8 Re‐selection parameters. Parameter Description Recommend q‐Hyst This parameter configures the hysteresis value of the serving cell used by the UE for ranking criteria in cell reselection 2dB ~4 dB Treselection This parameter specifies the value of the cell reselection UE timer in the cell on the indicated EUTRAN frequency 1s ~2s qOffsetCell cell‐specific offset for reselection ‐ > to differentiate between particular cells within intra‐freq layer 0 qOffFrq frequency‐specific offset for reselection ‐ > to differentiate between frequency layers 0 8.3.4 Inter‐Frequency Reselection Optimization Idle mode inter‐frequency reselection from lower‐ to higher‐priority cell and from higher‐ to lower‐priority cell procedure are shown in Figure 8.31. Parameter sNonIntraSearch dictates when UE starts searching and measuring all priority cells. Higher‐priority carriers and RATs are searched even if UE is in good RF condition, lower 283 UE camp on serving cell N Exist inter-freq neighbor with higher reselection priority? Y Measurement performed by UE t-reselectionEUTRA expires -> cell reselection to target cell Y StargetCell >threshXHigh? N Y SystemInfo SrxLev > SnonIntraSearch? N Reselection priority of Inter-fre neighor is lower than serving fre? Measurement performed by UE Y SrxLev < threshServingL ow & Stargetcell > threshXLow? N N N Figure 8.31 Idle mode inter‐frequency reselection strategy. Fulfill S criteria & Rn > Rs? Y Y Mobility Optimization thresholds are set in a high‐priority carrier to keep the UEs in the carrier, higher thresholds are set in a lower‐priority LTE carrier to encourage the UE to re‐select higher‐priority carriers or another RAT. For reselection from higher priority toward lower‐priority inter‐frequency/IRAT cell, it should reduce ping‐pong by appropriately setting tReselectionRAT (in SIB5 or 6), threshServ ingLow (in SIB3), and threshXLow (in SIB5 or 6). For reselection from lower priority toward higher‐priority inter‐frequency/IRAT cell, it should reduce ping‐pong by appropriately setting tReselectionRAT (in SIB5 or 6) and threshXHigh (in SIB5). An example of the parameters from SIB 3 and 5 messages are shown in Figure 8.32: intra‐frequency (1900–1900) and inter‐frequency (1900–700), and intra‐frequency (700–700) and inter‐ frequency (700–1900). Key idle mode parameter settings for 1900 and 700 MHz band cells can be referred to Table 8.9. As shown in Table 8.9, a UE reselects from a B2 cell to a B17 cell if the serving B2 cell RSRP is < −100 dBm AND the neighbor B17 cell RSRP is > −120 dBm. A UE reselects from a B17 cell to a B2 cell if the neighbor B2 cell RSRP is > −96 dBm. 8.4 Intra‐LTE Handover Optimization Intra‐LTE handover feature manages the UE in connected mode and allows for seamless mobility from one LTE cell to another. In contrast to idle mode, connected mode mobility is entirely managed by the LTE RAN based on configured and received measurement reports from the UE. Only hard handover is supported in LTE. Good handover performance will ensure the UE throughput and experience. By modifying the handover parameters can avoid or reduce the too early, too late and ping‐pong handovers, so as to improve the system performance. In a live network, either event (A3, A4, or A5) can be used in the LTE system with the intra and inter‐frequency handover decision. 8.4.1 A3 and A5 Handover Event A3 means neighbor becomes offset better than serving cell, event A3 is used for c onnected mode handover. The formular for Event A3 triggered is shown below: RSRP at serving cell a3Offset RSRP at neighbour cell The formular for Event A5 triggered is shown below: RSRP at serving cell a5Threshold 1 AND RSRP at target a5Threshold 2 Offset, hysteresis, and timetotrigger values play major role in Intra handover. So if it considers offseta3 = 3dB, hysteresisa3 = 1dB, and timetotriggera3 = 640msec then it could say that neighboring sector has to be 4dB higher than serving cell for 640msec to make the UE to generate event a3 (Figure 8.33). The formular of event A3 entering condition: Mn Ofn Ocn Hys Ms Ofs Ocs Off The formular of event A5 entering condition: Ms hysteresis thresholdEutraRsrp and Mn offsetFreq hysteresis threshold 2 EutraRsrp ●● ●● ●● Mn = measurement result of the neighboring cell [dBm] Ofn = MeasObjectEUTRA::offsetFreq, Ofn is the frequency specific offset of the frequency of the neighbor cell [dB] Ocn = cellIndividualOffset for neighboring cell [dB], can be used to fine tune the handover hysteresis on a cell‐to‐cell basis, which make the handover from cell1 to cell2 more difficult by decreasing the cellIndividualOffset of the cell2 285 Figure 8.32 SIB 3 and 5 messages. Mobility Optimization Table 8.9 Idle mode parameter settings for high‐priority cell and low‐priority cell. LTE Cell Selection & Reselection Parameters B2(1.9GHz) cellReselectionPriority B17(700MHz) 5 4 −106 −124 sNonintraSearch 6 4 threshServingLow 6 4 qRxLevMin threshXHigh(B2 Neighbor) ‐ threshXLow(B17 Neighbor) 10 4 ‐ threshXLow(UMTS) 30 30 threshXHigh(Femto) 14 14 RSRP RSRP Neighbour Cell RSRP a5threshold2 Serving Cell RSRP a3offset Serving Cell RSRP a5Threshold1 RSRP Neighbour Cell time time a3TimeToTrigger a3ReportInterval Meas. Report a5TimeToTrigger a5Report Interval Meas. Report Meas. Report Meas. Report Equal power borders Figure 8.33 Better cell handover concept for A3 and A5. ●● ●● ●● ●● ●● Hys = reportConfigEUTRA::hysteresis [dB] Ms = measurement result of the serving cell [dBm] Ofs = MeasObjectEUTRA::offsetFreq, Ofs is the frequency specific offset of the serving frequency [dB] Ocs = cellIndividualOffset for serving cell [dB] Off = eventA3Offset, Off is the offset parameter for this event [dB] Ocn, Ofn, Ocs and Ofs brings means for optimization of intra‐LTE handovers and/or realization of particular mobility management strategy. It introduces cell‐ and frequency‐specific offsets, which allow for differentiation of handover trigger criteria toward particular frequency layer and/or particular neighbor cells. These mobility offsets can be applied to intra‐ and inter‐frequency A3 and A5 events. Especially, the cellIndOffNeigh parameter may be of special interest in case macro and micro layers run with the same frequency carrier, while offsetFreqInter and offsetFreqIntra may be useful for micro‐macro inter‐frequency scenario. A3 event entry condition (intra‐frequency) RSRPneigh cellIndOffNeigh RSRPserv cellIndOffServ a3Offset hysA3Offset A3 event entry condition (inter‐frequency) RSRPneigh cellIndOffNeigh offsetFreqInter RSRPserv a3OffsetRsrpInterFreq hysA3OffsetRsrpInterFreq cellIndOffServ offsetFreqIntra 287 LTE Optimization Engineering Handbook For handover procedure, two types of RRC connection reconfiguration message need to be investigated as shown in Figure 8.34. The first RRC connection reconfiguration message is sent by source RBS to UE indicating all information for target cell like RACH parameters, root sequence, reference power, PUSCH/PUCCH nominal values, CQI reporting interval, and handover type. If target cell doesn’t have resources, it won’t see this message and handover preparation failure is observed. The second RRC connection reconfiguration message after EPC 10 no . A to w a ll be ta fte ar rg r er et su s eN cc are B es re sfu le l tr as an ed sit ion 288 MME SGW 9. Pa th 8. Pa th itc Sw h h Re qu es t es tA ck no wl ed Target eNB Source eNB 4. 1. A3 (F RR or C M wa Co ea rd nn su in e re g ct m ta ion en rg R tR et e ep ce co or ll i nfi t nf gu ro ra m tio at n io n to UE ) Re qu itc 3. Resource available confirmation along with target cell specific information that UE may need to Sync Sw B N te e ge RA eNB r et pl Ta B et m to s- C arg o C T nc es n Sy oc ith io at to pr d w r u g ig in CH ce nf try A yn co E h R ts s e U g nR ge 5. rou io th UE ct e n 6. on C C R R 7. Figure 8.34 Two types of RRC connection reconfiguration message. ge Mobility Optimization Table 8.10 Intra LTE ‐ A3,A5 parameters. Parameter Description Range Defaults a3Offset The offset value for eventA3 −30…30dB, step 0.5 dB 6(3dB) hysteresisA3 Parameter for entering/leaving measurement report triggering condition. 1dB a3TimeToTrigger The period of time that must be met for a5TimeToTrigger the UE to trigger a measurement report for Event A3 (A5), it depends much on the speed of the UE and the coverage scenarios. 0ms (0), 40ms(1), 64ms(2), 80ms (3), 100ms (4), 128ms(5), 160ms(6), 256ms (7), 320ms (8), 480ms (9), 512ms (10),.…, 5120ms (15) 320ms a3ReportAmount Number of reports when periodical a5ReportAmount reporting is used. 0 means that reports are sent as long as the event is fulfilled. 1r (0), 2r (1), 4r (2), 8r (3), 16r (4), 32r (5), 64r (6), infinity (7) Infinity (7) a3ReportInterval The interval for event triggered a5ReportInterval periodical reporting 240ms 120ms(0), 240ms(1), 480ms(2), 640ms(3), 1024ms(4), 2048ms(5), 5120ms(6), 1.024s (7), 1min(8), … 60min(12) cellIndividual OffsetEUtran Offset value specific to the neighbor cell relationship. This parameter can be applied individually to each neighbor cell with load management purposes. low value will delay the HO, and the higher the value allocated to a neighbor cell, the “more attractive” it will be. ‐ a5Threshold1 RSRP threshold1 used for triggering the EUTRA measurement report for Event A5 0…97 dB, step 1 dB −140 + X ‐ a5Threshold2 RSRP threshold2 used for triggering the EUTRA measurement report for Event A5 0…97 dB, step 1 dB ‐ Note: increasing TimeToTrigger and hysteresis might reduce ping pong effect and unnecessary handover but at the same time increasing too much might lead to handover drop. handover is sent by target (new serving) cell to UE indicating all information for itself like carrier frequency, A3/A2 thresholds and Smeasure parameter values. Normally, there are three ways of optimizing handovers in LTE, via the modification of the parameters a3offset and hysteresisa3, by changing the parameter timetotriggereventa3 and via the modification of the parameter filtercoefficient for event a3. The configuration of A3 should be different in different areas, considering the factor of overlapping, distance of sites etc. For A3 parameters setting shown in Table 8.10, it is recommended to follow the optimization rules below. ●● ●● a3offset should always be larger than hysteresisa3 if you want UE to handover to cells with an RSRP at least equal to the RSRP value of its serving cell, and ensuring a3offset > hysteresisa3 avoids ping‐pongs. The higher the value of a3offset + hysteresisa3 the more difficult for calls do handover to other cells. This is very useful where it has coverage holes (not a one to one deployment scenario on top of 3G cells). The smaller the value of a3offset + hysteresisa3 the faster release the calls to neighboring cells. This is useful in those scenarios where a large number of LTE cells exist in a given geographical area. 289 290 LTE Optimization Engineering Handbook 8.4.2 Data Forwarding This feature of data forwarding works for DL data only and is used during intra‐LTE handover. User‐plane tunnels can be established between the source eNB and the target eNB (or S‐SGW and T‐SGW) during handover preparation, it helps to decreases the packet loss rate, improve end‐user experience in case of time sensitive applications due to data loss or TCP slow‐start characteristics. As soon as the source eNB receives the handover request acknowledge, or as soon as the transmission of the handover command is initiated in the downlink, data forwarding is initiated if necessary. The packet forwarding function shall forward all buffered and incoming PDCP SDUs from the source to the target RAN, that is, IP packets before ROHC and encryption, so ROHC context is not needed to be forwarded. Packet forwarding is only performed in downlink (Figure 8.35 and 8.36). During handover preparation phase, source eNB attempts to allocate forwarding transport bearers toward the target eNB. If this is successful, a forwarding user plane tunnel is established for each E‐RAB between the source and target eNB. During handover execution phase, all buffered and incoming data is forwarded from the source to target eNB. Once handover is complete, the target eNB sends a path switch message to the MME. The MME then sends a user plane update request message to the SGW, which switches the user‐plane path from source to target eNB. The source eNB continues to forward packets while packets are being received from the SGW and the source eNB buffer contains data. When the END MARKER packet is Figure 8.35 Data forwarding. S1 HO Packet Forwarding S-SGW T-SGW IP S-eNB T-eNB X2 HO Packet Forwarding UE S-eNB MME Handover Required Handover Command Handover Command eNB status transfer T-eNB Handover Request Handover Request ACK MME status transfer Handover confirm Data transfer Handover notify Figure 8.36 S1 data forwarding procedure. Mobility Optimization received in the target eNB or the forwarding ordering timer expires, the target eNB discards any further forwarding data packets. Forwarding tunnels are then released. 8.4.3 Intra‐Frequency Handover Optimization In LTE, the earliest a handover can be set up after security mode completed during an attach or idle to active setup. Handover optimization can be done by parameters adjustment, although each of the key handover parameters (hysteresis, eventA3offset, filtercoefficientRSRP/RSRQ, timeToTrigger) have a unique purpose, adjusting them may have many common effects which are illustrated in Table 8.11. Additional handover‐related parameters includes cellIndividualOffset (allows altering effective handover hysteresis in a certain direction for a given serving cell ‐ > neighboring cell pair), maxReportCells (max number of cells reported by UE on the MR), reportingAmount/report ingInterval and s‐Measure (UE will look for neighboring cells only if the serving cell RSRP falls below this threshold). Handover optimization needs to decrease drop call during and after handover procedure. The usual phenomenon is UE cannot successfully access the PRACH in the target cell or UE drops soon after sending RRC connection reconfig complete in target cell. ●● ●● ●● ●● ●● Symptom: T304 expires. In this case timer T304 expires and UE makes RRC connection re‐ establishment with cause “handoverFailure.” (T304 is started after receiving handover command in source cell and stopped in target cell after successful PRACH procedure) Symptom: UE log indicates RRC connection reconfiguration complete (“PRACH msg3”) is transmitted but no further DL messages from target cell are seen and call drops Solution: increase T304 timer to 1000ms or longer, parameter t304IntraLte Solution: increase the max number of UL HARQ retransmissions, parameter harqMaxTrUl. (Parameter harqMaxMsg3 is for contention‐based RA with non‐dedicated preamble) Solution: increase initial preamble target receive power, parameter ulpcIniPrePwr (Value of −90dBm decreased handover drops considerably in a network) Handover optimization also needs to decrease drop call before handover procedure. The usual phenomenon is handover is triggered too late, UE drops before handover command, sometimes ping‐pong handovers increase if A3 used for bad RF. ●● Possible fix: If using A3 handover reduce a3offset and time to trigger. In some cases also filterCoefficientRSRP may need to be reduced. Table 8.11 Handover (HO) parameters optimization. Smaller ●● ●● ●● ●● ●● Faster HO trigger – can be useful in fast‐rising cell situations Increased HO ping‐ponging More frequent throughput interruptions Increased likelihood for UE served by a stronger cell. Hence, can improve HO success rate / throughput / BLER perf but offset by impact from more frequent HO interruptions / HO to momentarily strong cells More suitable for fast moving UE environments, and higher RF loading on the network Larger ●● ●● ●● ●● ●● Slower HO rate Reduced HO ping‐ponging Fewer throughput interruptions Increased likelihood for UE staying on a weaker cell. Reduced HO rate can impact HO success rate / throughput / BLER, but offset by fewer interruptions / HO to more stable cells. More suitable for predominantly stationary usage / slow moving UE environments usage / lighter RF loading 291 292 LTE Optimization Engineering Handbook –70 1 106 211 316 421 526 631 736 841 946 1051 1156 1261 1366 1471 1576 1681 1786 1891 –80 –90 –100 –110 RSRP_instant –120 RSRP_FC(K = 4) RSRP_FC(K = 11) –130 Figure 8.37 filterCoefficientRSRP simulation analysis. ●● One should consider using A5 triggered coverage‐based handover to quickly trigger handover in bad RF. This way, there is no need to reduce A3 trigger time, and less ping‐pong results. In this handover strategy, A3 is the handover for normal usage and A5 is for the bad‐RF quick reaction handover. If handover is triggered too early, the target cell SINR can be too weak when handover occurs. If handover is triggered too late, the source cell SINR can be too low. This can result in an abnormal release before handover. Besides, optimization of hysteresis and timeToTrigger should be performed in conjunction with optimization of parameter filterCoefficientRSRP. Low value of this parameter might allow ping‐pong handover operation. High value of this parameter might delay the handover and possible lead to lost connection to the serving cell. filterCoefficientRSRP simulation analysis is shown in Figure 8.37. It is recommended filterCoefficientRSRP value = “fc8.” Finding the optimum pair of (filter CoefficientRSRP, hysteresis, and timeToTrigger) should be considered to one of the following {(fc6,2,80), (fc8,3,40), (fc8,4,20), (fc5,1,100)}, in current cell and neighbor cell. It was proven that the time to trigger below 100 ms have the same impact on performance as UE measures RSRP every 100ms. 8.4.4 Inter‐Frequency Handover Optimization The eNB must include consideration of inter‐frequency neighbors in all measurement, handover, and redirection algorithms. There are two inter‐frequency mobility preparation methods: blind‐based handover and measurement‐based handover. This includes support of measurement objects (MeasObjectEUTRA) in the RRC connection reconfiguration message that include eutra‐carrierInfo and measurement bandwidth values differing from those of the current cell. Gaps must be supported for inter‐frequency measurements to allow the UE sufficient time to retune its radio the frequency being measured. Also the eNB must support the Mobility Optimization IdleModeMobilityControlInfo (interFreqPriorityList option) in the RRC connection release message, which is used to direct UE camping behavior (Figure 8.38). For inter‐frequency handover, the system supports bad coverage method based on RSRP and RSRQ with below triggers: based on bad coverage, based on interference, based on load, based on service, and based on subscription (SPID3 information from MME). The example of inter‐ frequency handover in Figure 8.39 shows that when UE in layer F1, mobility control is based on coverage, when UE in layer F2, mobility control is based on interference (load). Usually two types of mobility triggers are used to move UEs between layers for inter‐frequency handover control strategy, coverage‐based mobility, and load‐based mobility, used in when losing serving cell coverage and when serving cell load reaches a configurable threshold. The use case is for a cell on a low frequency with coverage relation to a cell at higher frequency. In such case it is possible to handover UEs to the higher frequency layer when the UE is close enough to eNB. For layer F1(higher frequency), when event A2 (RSRP < −100dbm) is met, UE starts inter‐frequency measurement, when event A5 (RSRPf1 < −102dBm & RSRPf2 > −98dBm) is met, load‐based inter‐frequency handover to F2. For layer F2 (lower frequency), when event A2 (RSRQ < −15db or RSRP < −98dBm) is met, start inter‐frequency measurement, when A4 (RSRPf1 > −100dBm) is met, load‐based inter‐ < frequency handover to F1. If both inter‐frequency and inter‐RAT neighbors are provisioned, it needs to set event A5 and event B2 serving cell thresholds such that eventA5 happens first. Any UE on a frequency band cell moving away from the cluster will perform a measurement gap based eventA5 inter‐frequency handoff to another frequency band cell. UE will not be scheduled during measurement gap, which impact UE DL and UL performance. UL performance will is impacted more. Event A2 means serving becomes worse than threshold, after A2 reported, the inter‐frequency measurement will be implemented on UE. If EventA2 reported too early (A2 threshold is high), there may be 20% throughput drop on UL UE, at good coverage samples. In order to reduce the occurrence inter‐frequency measurement, improve the user throughput, the recommended A2 parameter setting (Event A3‐based handover) can be set up according to 95% RSRP samples of the serving cell; the serving cell’s RSRP samples are received from the samples in which the difference with the neighbor cell is not more than 3dB, which is shown in Figure 8.40. On the other hand, if inter‐frequency handover is based on Event A2 + A5, it will lead the handover delay and impact the UE throughput. When A5−1 is too high or A5−2 is too low, ping‐pong inter‐frequency handover will happen, when A5−1 is too low or A5−2 is too high, handover delay will happen. Sometimes A3 is better, event A3 based inter‐frequency coverage/quality handover evaluation is given by: RSRPneighbour RSRQneighbour RSRPserving a3OffsetRsrpInterFreq RSRQserving a3OffsetRsrqInterFreq In a live network, the proposed setting is for UE not to quickly handoff to lower frequency band cell, and measurement gap will turn off on low‐frequency band cells, that UE will not handoff to a high frequency band cell. As an example, inter‐frequency handover takes place 3 The parameter Subscriber Profile ID for RAT/frequency priority (SPID) is specified in 3GPP rel-8. It is an index referring to user information (e.g., mobility profile, service usage profile, or roaming restrictions). The information is terminal specific. 293 1 Wait for measurement reports for poor coverage. UE is connected in the cell. Determine a set of inter-frequency and IRAT frequency candidates in case of poor coverage. The set of frequency candidates is empty yes 3 Do not start measurements for poor coverage. 4 Stop “Target Good Enough” and “Good Coverage” measurements Blind Handover? no 2 Poor coverage reported yes Initiate Handover no Start measurements for poor coverage. 5 8 › Measurement report criteria: A2 Event › Blind handover (more on this later); Blind release with redirect start A1/A5 measurements Should measurements be started? Good coverage reported 7 10 11 Figure 8.38 Inter‐frequency handover. The “Target Good Enough” measurements time out. Handover? yes Initiate release with redirect the reported frequency. 9 Initiate release with redirect to one of the frequencies. “Target Good Enough” reported no 12 Initiate release with redirect to one of the frequencies. yes Start “Target Good Enough” and “Good Coverage” measurements Measurement report criteria: A1 Event 6 no Initiate Handover › Measurement report criteria: A5 Event › Handover, release with redirect Mobility Optimization UE eNB RRC reconfiguration(eventA2) RRC reconfiguration complete Measurement report(eventA2) RRC reconfiguration(eventA1/A5) RRC reconfiguration complete Measurement report (eventA5) RRC reconfiguration (IFHO) RRC reconfiguration complete Figure 8.39 Inter‐frequency handover trigger. 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 –72 –76 –78 –80 –82 –84 –86 –88 –90 –92 –94 –96 –98 –100 –102 –104 –106 –108 –110 –112 –114 The recommended A2 parameter setting Figure 8.40 A2 parameter setting. after serving cell RSRP < = −107dBm (high frequency band) and target cell RSRP > = −110dBm (low‐frequency band), the related parameters are listed below: ●● ●● ●● thresholdEutraRsrp (Entering‐coverage‐alarm) = −103dBm, hysteresis = 1dB thresholdEutraRsrp (eventA5 – high frequency cell) = −105dBm, hysteresis = 2dB threshold2EutraRsrp (eventA5 – low frequency cell) = −112dBm, hysteresis = 2dB The reference parameter settings for high‐/low‐frequency band cells are shown in Table 8.12 and Table 8.13. From summary in Table 8.14, the inter‐frequency handover triggering point would be: ●● ●● UE handover from low‐band layer to high‐band layer if the serving low‐band cell RSRP is < −117 dBm AND the high‐band layer cell RSRP is > −110 dBm. Note that measurement gap is disabled in low band with the assumption that low‐band coverage is ubiquitous. Therefore, the connected mode mobility toward high band is not possible. When UE reaches A2_below serving floor threshold, it will be released and redirected to UMTS carrier. UE handoffs from high‐band layer to low‐band layer if the serving high‐band cell RSRP is < −107 dBm AND the low‐band under layer cell RSRP is > −101 dBm 295 296 LTE Optimization Engineering Handbook Table 8.12 Connected mode parameter settings for high‐frequency band cells. Measurement Event Measurement Purpose ReportConfig Threshold Hysteresis thresholdEutraRsrp B2(EventB2) Mobility‐InterRAT‐to‐UTRA 16 −115 4 thresholdEutraRscp (EventB2) Mobility‐InterRAT‐to‐UTRA 16 −112 4 thresholdEutraRsrp (EventA2) Below‐Serving‐Floor thresholdEutraRsrp (EventA2) Entering‐Coverage‐Alarm 3 −118 1 13 −103 1 thresholdEutraRsrp (EventA1) Leaving‐Coverage‐Alarm thresholdEutraRsrp (EventA5) Mobility‐InterFreq‐to‐EUTRA 9 −98 1 18 −105 2 threshold2EutraRsrp (EventA5) Mobility‐InterFreq‐to‐EUTRA 18 −103 2 Table 8.13 Connected mode parameter settings for low‐frequency band cells. Measurement Event Measurement Purpose ReportConfig Threshold Hysteresis thresholdEutraRsrpB2 (EventB2) Mobility‐InterRAT‐to‐UTRA 14 −115 4 thresholdEutraRscp (EventB2) Mobility‐InterRAT‐to‐UTRA 14 −112 4 thresholdEutraRsrp (EventA2) Below‐Serving‐Floor thresholdEutraRsrp (EventA2) Entering‐Coverage‐Alarm thresholdEutraRsrp (EventA1) Leaving‐Coverage‐Alarm thresholdEutraRsrp (EventA5) Mobility‐InterFreq‐to‐EUTRA threshold2EutraRsrp (EventA5) Mobility‐InterFreq‐to‐EUTRA 19 2 −118 1 12 −114 1 8 −112 1 19 −115 2 −112 2 Table 8.14 Summary of inter‐frequency handover triggers. isMeasurementGapAllowed Event A5 – hysteresis Low Band High Band FALSE TRUE 2 2 Event A5 ‐ thresholdEutraRsrp −115 −105 Event A5 – threshold2EutraRsrp −112 −103 Event A1 ‐ thresholdEutraRsrp −112 −98 Event A1 ‐ hysteresis Event A2(ECA)‐ thresholdEutraRsrp Event A2 ‐ hysteresis Event A2(BSF)–thresholdEutraRsrp 1 1 −114 −103 1 1 −118 −118 8.4.5 Timers for Handover Failures Several timers/mechanisms exist in the database file that influence handover failure rate/ dropped calls, the recommended value of these parameters are listed below. T304 (Recommendation: ms2000): Successful transmission of handover confirm (RRC connection reconfiguration complete) following receipt of handover command. Mobility Optimization dsrTransMax (Recommendation: n64): Maximum number of scheduling requests (SR) UE will send to request an UL grant before declaring SR_Max failure. UE cannot receive UL grants could be due to weak DL SINR or UE running out of UL coverage. T310 (Recommendation: ms1000): DL physical layer break‐down (out‐of‐sync indication within UE’s lower layers). timeAlignmentTimerCommon/Dedicated (Recommendation: 2560ms): UE must receive a time alignment (TA) update within this interval, else declares TA failure. maxRetxThresholdDownlink/uplink (Recommendation: 32): Max number of times eNB on DL/ UE on UL retransmits a RLC packet before declaring max_RLC_RTx failure dopped call. ulSyncTimer (Recommendation: 3sec): eNB declares loss of UL when unable to detect sufficiently strong SRS. 8.5 Neighbor Cell Optimization Neighbor cell optimization must be performed to ensure that UEs in idle or connected mode can promptly perform reselection to or be handed over to optimal serving cells. The purpose of neighbor optimization is to cut the useless growing neighbors and add neighbors with a lot of handover attempts, also include correcting missing reciprocities and deleting one way neighbor relation. It is needed to identify which neighbors are important and which are not, so the engineers will be able to delete the unused or underused neighbor relations and add missing neighbors. After neighbor optimization, it would make the handover in network more smoothly, especially help in drive test audit. Unlike 2/3G, there is no need for transmitting neighbor lists to UE in LTE. 3GPP requires UE to be self‐sufficient in terms of detecting neighbors without eNB’s assistance; however, a neighbor relationship must exist in the cell for it to process handover request. Usually, a network may contain thousands of cells and each cell may in average have 20 to 25 neighbors per frequency. One of the most important item that it would clean up as part of this optimization is going to be the neighbor list, hence, having a good neighbor list is very critical for improving network performance. Missing neighbors can be characterized by a series of measurement reports sent to the eNB from the UE and UE not receiving any response from the eNB as the neighbor it is requesting to be handed over to isn’t in the neighbor list. The main method of neighbor cell list optimization can be based on cluster drive data or other UE measurement report data. The basic idea is that if the cells of which RSRP are detected together in a large amount of bins should be considered as neighbors. It is recommended to run a drive test that covers all the cells within a cluster and its surrounding cells to collect data to support this activity. For neighbor cell optimization, ANR (automated neighbor relations) is an important feature, which will automatically configure necessary measurement reports to detect neighbor cells of intra‐frequency LTE or inter‐RAT. When activated, neighbor relations are automatically added and removed based on these measurement reports. Without ANR, traditional neighbor tuning must be completed. 8.5.1 Intra‐LTE Neighbor Cell Optimization 8.5.1.1 Neighbor Relations Table Within each eNB there is a neighbor cell list (NCL) and a neighbor relation table (NRT, includes intra‐LTE and inter‐RAT). Both of them are logical objects/concepts, so they do not refer to a particular implementation. The NCL contains information exchanged between neighbor sites via the X2 interface such as the cell global IDs and the physical‐layer cell IDs of the cells belonging 297 298 LTE Optimization Engineering Handbook Figure 8.41 SIB4 Content from 3GPP TS36.331. to the neighbor sites and the IP address of those neighbor sites. The NRT contains the neighbor relation information on cell level, that is, for each cell of site, it contains a list of neighbor relations (NR) identified by the ECGI and PCI and for each NR. It is possible to optimize the neighbor relations by blacklisting the inefficient ones. Blacklisting a neighbor relation cell means preventing mobility (either reselection or handover) to that particular (unsuitable) cell (typically a far away cell with some signal resurgence). If a UE is not allowed to reselect a blacklisted cell, UE will not measure blacklisted cells in connected mode. If target eNB is blacklisted in source eNB, source eNB will not send X2 setup request to target eNB, no cell relation and X2 connection will be set up, handover will happen over S1 interface. If source eNB is blacklisted in target eNB, source eNB will send X2 setup request but get negative answer from target eNB. A blacklisted cell will therefore not be reported in any ANR or handover measurement report, and the UE will not measure blacklisted cells in connected mode so the UE will never be instructed from the eNB to handover to a blacklisted cell. Whitelisting a (suitable) neighbor relation (created manually or by ANR) means preventing the eNB from removing it automatically, even if it was not used for a long period (typically due to low traffic in a young network). Adding the neighbor eCGI to the x2BlackList, which will prevent ANR from creating an X2 association toward the target and hence prevent handover. SIB4 also informs about blacklisted cells, up to 16 groups of cells (PCIs) can be blacklisted as shown in Figure 8.41. Besides, the neighbor reciprocity function can identify the cell that is a neighbor of another cell, but the other cell is not a neighbor of the interested cell. 8.5.1.2 ANR ANR supports for LTE inter‐frequency is the detection of new neighbor LTE cells using threshold based on UE measurements, in intra‐frequency ANR for LTE the measurement is A3 while in inter‐frequency ANR for LTE is A5, the related parameters are shown in Table 8.15. The ANR dedicated threshold for RSRP and RSRQ allows ANR to get the CGI before UE reports the cell as best serving and the handover is triggered. It increases the chance for ANR to finalize the neighbor relationship setup before it is used in the handover. Usually, a series of measurement report without any RRC reconfiguration message is a good indicator of missing neighbor. UE is sending several measurement reports for PCI 46 as shown in Figure 8.42, but handover to this cell is never performed because this site is not defined as a Mobility Optimization Table 8.15 ANR parameters. Parameter Description Value x2SetupPolicy ANR setup X2 or not TRUE cellAddRsrpThresholdEutran RSRP threshold for PCI report −100dBm cellAddRsrqThresholdEutran RSRQ threshold for PCI report N/A a3offsetAnrDelta ANR A3 event offset 0dB hysteresisA3 ANR A3 event hysteresis 1dB timeToTriggerA3 ANR A3 event trigger time 640ms a5Threshold1RsrpAnrDelta ANR A5 event offset (serving cell RSRP) 1dB a5Threshold2RsrpAnrDelta ANR A5 event offset (target cell RSRP) 1dB hysteresisA5 ANR A5 event hysteresis 1dB timeToTriggerA5 ANR A5 event trigger time 640ms neighbor. UE is starting to measure on other PCIs defined as neighbor. Handover is performed right after this measurement report. 8.5.2 Suitable Neighbors for Load Balancing Currently, load‐balancing requires a lot of manual configuration. The operator needs to manually specify if each relation is relevant for load‐balancing. In this part, it is discussed what is the suitable neighbours for load‐balancing. First it is needed to decide the calculated UE relevance (Y(A, B)) for each neighbor cell relation which symbolizes the percentage of UEs that have reported the neighbor cell as strongest. This is calculated by performing background measurements that are exactly the same as other load‐ balancing measurements (A4 event is used to off‐loaded UEs to neighbor cells as well as determine the UE distribution) but with lower intensity. Second, passive load‐balancing cell relations are needed to decide as shown in Figure 8.43. Y(A, B) symbolizes the percentage of selected UEs that have ranked the associated neighbor cell B as the strongest in cell A. 8.6 Measurement Gap Measurement gap enables single‐receiver UE to perform measurement on other frequencies or on other radio access technologies. The purpose of these gap assisted measurement is to support inter‐frequency and inter‐RAT mobility in LTE. 8.6.1 Measurement Gap Pattern Measurement gaps allow the UE to measure neighboring cells for several types of handovers: inter‐RAT handover, inter‐frequency handover, network‐assisted cell change, and so on. The eNB will command the UE when to start the measurement gaps via RRC connection reconfigu ration message. The starting point of these is determined so that the measurements gaps do not coincide with the UE’s periodic reporting. During measurement gaps the UE is measuring inter‐frequency/inter‐RAT candidates and does not transmit nor receive any data from the serving cell. UL and DL scheduler track the measurement gap periodicity and gap pattern in order to avoid allocations whilst the UE is not listening. 299 300 LTE Optimization Engineering Handbook RSRP ANR trigger Point Serving Cell RSRP UE starts Measuring HO trigger Point a3offsetAnr Target Cell RSRP sMeasure a3Offset Event A3 Gained Time A5 ANR A5 SourcLTE RSRP (dBM) Waiting for A5 A1 threshold_a1 A2 Distance from eNB LTE IFHO Target LTE RSRP (dBm) time To TriggerA5 threshold1_ANR_a5 threshold_a2 threshold1_a5 threshold2_a5 threshold2_ANR_a5 a5Threshold1RsrpAnrDelta a5Threshold2RsrpAnrDelta Figure 8.42 ANR parameters and example. Mobility Optimization Initial relation Relation with F(A,B)> x1% is moved to passive Relation configured by ANR or operator. Relation with F(A, B)< x2% is moved to” initial relation”. Passive LB UEs to be off-loaded only due to background measurement and indirectly due to A4 measurement on active relations on same frequency Relation with F(A, B) > x3% is moved to active. If no active relation on a frequency & need for load balancing, if sum of F(A, B) for all relations with low enough load > x3%, moved all to active. Active LB UEs to be off-loaded due to started A4 measurement. Figure 8.43 Load‐balancing strategy. Table 8.16 Gap pattern configurations supported by the UE. Gap pattern Id Measurement gap length (MGL, ms) Measurement gap repetition period (MGRP, ms) Minimum available time for inter‐ frequency and inter‐RAT measurements during 480ms period (Tinter1, ms) 0 6 40 60 1 6 80 30 LTE measurement gaps are specified by 3GPP to have 6ms4 gap (see Table 8.16). The periodicity of these gaps is specified in periods of 10ms (1 frame). Larger periodicity of the gaps leads to better UL/DL throughput performance at the cost of inter‐frequency/IRAT monitoring, which could lead to longer required times to perform the handover. By default, Gap 1 is assigned for ANR while Gap 0 is assigned for mobility as shown in Table 8.17. After measurement gap is configured, the UE searches for cells in the target LTE frequency, which will take several measurement gaps since all 504 PCIs need to be scanned. If there are several LTE frequency layers, the search and measurement time is multiplied accordingly since the UE can only tune its RF receiver to one frequency during a measurement gap. The measurement procedure is similar if the target frequency is UTRA. However, in that case, the number of gaps required for cell search depends on the number of neighbor cells, scrambling codes that are not provided in the neighbor list need not be measured by the UE. For GSM target frequency, the UE is required to decode the BSIC on the 8 strongest measured BCCH frequencies. 4 6ms is enough to be able to search for primary and secondary synchronization channels of inter-frequency cells, even for the worst case frame timing where the gap starts at the beginning of the second or seventh subframe of the measured cell. This is because PSS/SSS is transmitted at 5ms interval at the end of the first and sixth subframes. 301 302 LTE Optimization Engineering Handbook Table 8.17 Gap 1 and Gap 0. Gap pattern Feature Event Type Gap 1 IF ANR Inter‐freq A3/A5 Gap 1 IRAT ANR ReportStrongestForSON Gap 0 InterFrequencySessionContinuity A2/Inter‐freq A3/A5 InterFrequencyLTEHandover Gap 0 WcdmaSessionContinuity B2 Gap 0 SRVCCtoUTRAN B2 Gap 0 InterFrequencyLoadBalancing A4 Gap 0 InterRatOffloadToUtran B1 Table 8.18 Maximum allowed time to identify a detectable cell with measurement gaps, per layer. Max time to identify a detectable E‐UTRA FDD inter‐frequency cell Max time to identify a detectable UTRA FDD cell Max time to decode BSIC of a GSM cell (no other layers measured) Long DRX cycle length 40ms gap 80ms gap 40ms gap 80ms gap 40ms gap 80ms gap no C‐DRX 3.84 sec 7.68 sec 2.4 sec 4.8 sec 2.16 sec 5.28 sec 40ms 64ms 2.56 sec 30 sec 80ms 3.2 sec ‐ 128ms 3.2 sec ‐ 256ms 5.12 sec 320ms 6.4 sec 9.6 sec 3.2 sec 5.12 sec ‐ 6.4 sec 6.4 sec ‐ If a cell is detectable and eligible to trigger a measurement report, the maximum time for UE to send the measurement report is called the cell identification time. The maximum allowed identification time depends on the C‐DRX long cycle and the target system, which are listed in Table 8.18. Actually, the UE vendor is free to implement faster search and measurement than required by 3GPP, but network parameters will in most cases need to be optimized for the worst performing fraction of UE models in the network. For this reason, C‐DRX can be disabled for the duration of urgent inter‐frequency measurements to reduce the probability of call drop due to measurement delay. If RRC connection containing inter‐frequency/IRAT ANR comes first, then GP1 (reconfiguration message default value) will be used. If Inter‐frequency/IRAT mobility measurement for handover comes later, then measurement gap will be updated as GP0 (default value). If inter‐frequency/ IRAT mobility and ANR are in the same RRC connection reconfiguration message, GP0 will be used, which is shown in Figure 8.44. An illustrated subframe scheduling in Figure 8.45 has been to show the measurement gap, all DL transmissions granted after measurement start‐4 will not be ACK/NACKed, due to the DL Mobility Optimization Figure 8.44 Gap parameters from RRC connection reconfiguration message. No DL scheduling Frame Subframe Measurement gap N–1 Not immediately transmit after gap N N+1 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 PDCCH/PDSCH 0 1 2 3 4 5 PUSCH/PUCCH 0 1 2 3 4 5 6 7 8 9 8 9 0 1 2 3 4 5 6 7 8 9 6 7 8 9 0 1 2 3 4 5 6 7 8 9 7 8 9 0 1 2 3 4 5 6 7 8 9 PUSCH/PUCCH is transmitted. Figure 8.45 Scheduling strategy during measurement gap. scheduling will stop at measurement start‐4. PUSCH/PUCCH can still be scheduled in the four subframes before measurement gap to send CQI/PMI/RI report, and so on, in order to keep monitoring the radio. Measurement gaps are initiated by eNB with RRC connection reconfiguration message when Event A2 is triggered. The eNB deactivates measurement gaps once a suitable target cell has been reported by the UE or an event A1 is triggered by the UE. When measurement gaps are activated, the eNB informs the UE of the gap pattern. Note that DRX and measurement gaps are not used in parallel. Measurement gaps have higher priority than DRX. The gap pattern is configured by parameter measurementGapsPattern. Note, however, that when Gap is activated on CSFB trigger, the period is hardcoded to 40 ms regardless of the setting of parameter measurementGapsPattern. The starting position of the measurement gap in a given gap period is defined by the UE‐specific Measurement Gap Offset (MGO). It is determined so that the UE performance is degraded as little as possible by measurement gaps. This is achieved by choosing the MGO so as to minimize the collision of measurement gaps with the other periodic transmissions listed below: CQI/ PMI/RI reporting and SRS transmissions, ACK/NACK transmissions on PHICH and PUCCH (except the last ACK/NACK, that is, when the max number of retransmissions has been reached), QCI1 transmissions and SIBs transmissions and so on. 303 304 LTE Optimization Engineering Handbook CQI Report 0 Measurement gap 40 DRX on duration CQI Report 80 CQI Measurement (8 ms before CQI reporting) CQI Report 120 Figure 8.46 CQI report alligned with DRX cycle and measurement gap. Note that the MGOs of the different UEs are also chosen so that their different measurement gaps are distributed over time so that during the measurement gap of some UE, a sufficient number of other UEs can transmit (receive) and the total cell throughput performance will not degraded. The MGO is in the set {0, 1,…, 39} in gap pattern 0 and in the set {0, 1, …, 79} in gap pattern 1. 8.6.2 Measurement Gap Versus Period of CQI Report and DRX Measurement gap needs to consider periodic CQI report and DRX period to not miss CQI reports from the UE sent while the UE is active (on duration). When a measurement gap is started, it needs to determine where the measurement gap should start by presenting a MGO for the UE. The UEs CQI periodic reporting is also configured with an interval and an offset in relation to subframe starts. Measurement gap has to be aligned (equal to or multiples of CQI and DRX cycle) with CQI report configuration to not lose the CQI reports from the UE. Currently DRX is aligned with CQI so that CQI reports are sent while the UE is active (on duration). Measurement gap will be placed before the UEs CQI measurement, to avoid a collision with CQI measurement and reporting and to allow for extended DRX active right after the DRX on period. Figure 8.46 is an illustration of measurement gap period, CQI measurement, CQI report, and DRX cycle. Here it assumes all the cycles are 40ms. 8.6.3 Impact of Throughput on Measurement Gap During a measurement gap the UE cannot be scheduled in uplink or downlink, and therefore, measurement gaps induce throughput loss for the UE, there is no loss of cell throughput, however. From lab trial results, measurement gap with GP0 (40ms) has 25% throughput decrease compared with baseline. Measurement gap with GP1 (80ms) shows 12.8% increase compared with GP0. So up to 15% and 25% DL throughput reductions have been observed for a UE, which is being scheduled with 80 ms and 40 ms measurement gaps respectively. Reducing the amount of time a connected mode UE is required to make inter‐frequency or inter‐RAT measurements can reduce this throughput impact. From inter‐frequency (2.6G@20M <> 1.8G@10M) handover parameter optimization field test, Method1 is basic, Method2 is aggressive, Method3 is Method2‐based plus quick Event A5 handover. Three sets of parameters tested are listed below: ●● ●● ●● Method1: the gap is triggered at −100 dBm, A3 = 3dB Method2: the gap is triggered at 1.8G to −110 dBm, A3 (2.6G →1.8G) =8dB, A3 (1.8G → 2.6G) = −12 dB Method3: method2 based settings plus A5 (2.6G → 1.8G) = −116dBm Mobility Optimization Empirical CDF 1 0.9 basic 0.8 aggressive 0.7 cdf 0.6 basic aggressive aggressive + quick a5 0.5 0.4 aggressive + quick a5 0.3 0.2 0.1 0 0 10 20 30 40 50 60 MAC DL throughput, Mbps 70 80 90 Figure 8.47 Different handover parameters impact DL throughput. From Figure 8.47 it can be seen method2 settings result in UE staying too long (no 2.6G layer in bad coverage). For method3 settings, A5 trigger used to push UE quickly to 1.8G in bad coverage. In this case, Method2 and Method3 are recommended. 8.7 Indoor and Outdoor Mobility Indoor hotspots are assumed to be covered by indoor distributed system as well as small cells for higher efficiency. Indoor wave propagation losses are frequency dependent. Indoor and outdoor mobility optimization evolves priority‐based cell reselection, A2 + A5, A2 + A4, A2 + A3‐based handover, and load balance and indoor leakage optimization items. In case of indoor deployment on a dedicated carrier, the indoor leakage outside the building will be controlled as much as possible. To ensure service continuity between outdoor macro‐ cell network and indoor network, a good overlap must be guaranteed to execute a handover indoor/outdoor. The indoor to outdoor cell overlap is calculated from the scanner measurement. It consists in calculating the RSRP delta between Top1 RSRP of indoor cell and RSRP of outdoor macro cell when Top1 RSRP > = RSRP target + indoor outdoor overlapping margin. Usually, indoor cell will deploy on a dedicated carrier, indoor cells except those covering the first floor must have no F1 outdoor macro cells neighbors. The F1 outdoor macro cell must have as neighbors the indoor cell covering the first floor. Cells covering the building exits located essentially at the first floor, which is called a transition cell (Figure 8.48). For cell reselection parameters setting, it is similar as outdoor cell (with same or different frequency/priority) reselection (Table 8.19). For handover parameters setting, there are usually three configurations as shown in Table 8.20. ●● ●● ●● configuration 1: A2 + A5 for both direction between indoor and outdoor, which can keep the UE indoors as long as possible, and meanwhile, any signal leaks can be controlled effectively configuration 2: A2 + A5 is used for indoor to outdoor, A2 + A4is used for outdoor to indoor configuration 3: A2 + A3 for both direction between indoor and outdoor 305 306 LTE Optimization Engineering Handbook F2 3rd F2 floor Indoor pico cell F2 F2 F2 F2 Outdoor macro cell 2nd floor F1 1st floor Figure 8.48 Transition cell. Table 8.19 Reselection parameters setting between indoor and outdoor (example). 1 2 3 Outdoor cell Indoor cell Outdoor cell priority = 7, indoor cell = 5 Outdoor cell priority = 5, indoor cell = 7 RSRP_out < −102dBm, RSRP_in > −98dBm RSRP_out > −100 dBm, RSRP_in < −104dBm Outdoor cell priority = 7, indoor cell = 7 Outdoor cell priority = 7, indoor cell = 7 RSRP_in‐RSRP_out > 4dB RSRP_out‐RSRP_in > 4dB Outdoor cell priority = 5, indoor cell = 7 Outdoor cell priority = 5, indoor cell = 7 RSRP_in > −98dBm RSRP_out > −100 dBm, RSRP_in < −104dBm Table 8.20 Handover parameters setting between indoor and outdoor (example). config 1 Outdoor cell Indoor cell RSRP_out < −102dBm, RSRP_in > −98dBm RSRP_out > −102dBm, RSRP_in < −106dBm config 2 RSRP_in > −98dBm RSRP_out > −102dBm, RSRP_in < −106dBm config 3 RSRP_in‐RSRP_out > 3dB RSRP_out‐RSRP_in > 3dB 8.8 Inter‐RAT Mobility The part is about the strategy of mobility in RRC idle mode (i.e., inter‐RAT cell reseletction) and in RRC connected mode (redirection, PS HO, CSFB or CCO) between LTE and other RAT (Utran or Geran). When a LTE cluster or site is driven, it is desired that all the time UE are connected to LTE instead of IRAT to other technology. However, this sometimes is not possible, as LTE coverage becomes weaker in some areas it causes UE to switch to IRAT or to a better‐serving UMTS or GSM network. Cell reselection is used for inter‐RAT (IRAT) mobility in idle mode. The UE receives the information from the eNB for IRAT and inter‐frequency cell reselection through system information messages (SIB3 and SIB6) and dedicated signaling that determine when it is appropriate Mobility Optimization RRC-CONNECTED RRC-CONNECTED Bad coverage detection triggers Release with Redirect: Redirect Information Frequency RRC-IDLE Handover Command Move to reserved resources CellReselection according to redirect information (GSM, WCDMA, LTE IF) “RRC-CONNECTED” “RRC-CONNECTED” Figure 8.49 Release with redirect and handover. to begin measuring other RATs and frequencies along with specifics of how the measurements should be triggered, prioritized, and ranked prior to the UE performing IRAT cell reselection. The S criterion is again used to select the good cells for cell reselection. In RRC connected mode, there are in principle two ways of inter‐working between LTE and other RATs triggered based on event A2, and optionally, also by event B2. The inter‐working can be performed by a prepared handover (network controlled) where the UE does not leave the connected state (handover) or by a cell re‐selection/release and redirect (UE controlled or network assisted) where UE via idle state performs network assisted cell reselection (Figure 8.49). 8.8.1 Inter‐RAT Mobility Architecture and Key Technology In Figure 8.50, an inter‐working network between LTE and GERAN/UTRAN is illustrated. The network shall ensure that the user has ongoing service performance by providing inter‐RAT handover functionalities. The UTRAN/GERAN network must be coverage overlapped with the eUTRAN network, and the neighboring frequency and neighboring cell relation shall be configured in the RAN nodes. LTE IRAT capabilities can be divided into two major areas: CS and PS mobility, PS mobility includes cell reselection, cell redirection, NACC, and PS handover mode, which is shown in Figure 8.51. Table 8.21 presents an example of mobility strategy between LTE and UTRAN PS domain. LTE will be broadcast as the highest‐priority RAT in all three technologies, followed by 3G, and then 2G. Reselection will only be made to a lower‐priority RAT when the destination GERAN MS Um Abis BTS A BSC -cs Iu G UE Uu b NodeB Iub RNC Iu-ps UTRAN MSC/VLR Gs GnGp SGSN Gn UE LTE-Uu E-UTRAN FDD X2 UE LTE-Uu E-UTRAN TDD S1-MME S1 -U ME -M S1 E-UTRAN Figure 8.50 IRAT interworking topology. S1-U MME D SGs Gr HSS/HLR S6a Gn PCRF Gx Rx S11 S-GW S5 P-GW/GGSN Gi/SGi PDN 307 308 LTE Optimization Engineering Handbook IRAT MOBILITY Packet service (PS) mobility (including VOIP) Cell Reselection Redirection Circuit Service (CS) Mobility NACC (Network assisted cell Change) PS Handover Figure 8.51 IRAT strategy. Table 8.21 Mobility Between LTE and UTRAN PS domain. LTE => UMTS UMTS=> LTE UE State Strategy Description idle Cell reselection The sib6 message is about the 3G neighbor cells. The system send the sib6 message to the UE side. And according to the sib6 and the LTE signal, the UE decides whether it is necessary to reselect to the 3G. connected R8 redirection eNB sends RRC connection release message with ARFCNof UMTS. UE will access UMTS with the frequency. R9 redirection eNB sends RRC connection release message with ARFCNof UMTS and ralated SI information of the cell. UE will access UMTS with the frequency. SI is acquired by RIM procedure. PS HO When PS HO is triggered, eNB send mobility from EUTRA command message for UE handover to UTRAN. This feature is an end‐to‐end option that requires new functionalities in the UE, the access and the core network. Idle/cell_PCH/ URA_PCH Cell reselection According to SIB19 or UTRAN mobility information ‐ > Dedicated priority information, E‐UTRA frequency, priority and cell reselection threshold, UE reselection from UTRA to LTE. RRC connected redirection UTRA send UE E‐UTRA frequency by RRC connection release (redirection info) message for UE accesses to LTE. cell_DCH PS HO Inter‐RAT PS HO feature can provide lossless handover between eUTRAN and GERAN/UTRAN, but the procedure is rather complicated. c overage is expected to be better. A non‐zero setting of threshXLow for both 3G and 2G targets will be used to achieve this, ensuring that reselection does not occur when LTE and the destination RAT have poor coverage. A UE will reselect to a lower‐priority RAT when both the following criterion are met: RSRP RSCP / RSSI qRxLevMin EUtran threshServingLow qRxLevMin Utran/GERAN threshXLow Mobility Optimization Redirection is a mechanisms to force a UE to switch from RRC‐connected to RRC‐idle mode and reselect to inter‐frequency/inter‐RAT neighbor cell for PS session continuation and/or CS call setup (CS Fallback). Redirection is triggered by either RSRP based event A2 or RSRQ‐ based event A2. Having both RSRP and RSRQ criterion, UE can be redirected to other LTE frequency/RAT layer to avoid performance degradation not only due to poor coverage but also due to extensive interference. Redirection target layer is selected by eNB according to parameter settings and UE capabilities. Selection of target for redirection is done without triggering any target layer measurements, that is, blind selection. NACC: The serving cell controls the redirection using the UE measurement data. UE receives system info from target cell before release from the serving cell. Note that NACC is only applicable to GERAN and is an alternative for GERAN that does not support PS handover (NACC is not applicable to UTRAN to eUTRAN mobility, but to GERAN to UTRAN) PS Handover: In addition to cell control used for redirection, target core communicates with the source core to establish the UE context. Unsent data in the serving RAN will be forwarded to the target UTRAN. This mobility procedure enables the establishment of network resources in the target UTRAN system prior to commanding the UE to move to the target cell. The overview of IRAT classification and mobility actions is shown in Figure 8.52. 8.8.2 LTE to G/U Strategy The UE shall only perform cell reselection evaluation for E‐UTRAN frequencies and inter‐RAT frequencies that are given in system information and for which the UE has a priority provided. Through the configuration of the frequency priority, networks can easily lead terminal reselect to the high‐priority cell to camp, so as to achieve the balance network loading and improve resource utilization, make the UE signal quality better, and so on (Figure 8.53). In idle mode, priority‐based cell reselection is applied, the order of the priorities is LTE > UMTS > GSM. UE will camp on LTE with priority first, then UMTS, GSM finally. In connected mode, UE will move to UMTS (GSM) by redirection/handover in LTE poor coverage. When the user returns to the LTE overlay area in the UMTS (GSM) network, the redirection/handover to LTE can be applied. Figure 8.54 shows an example of LTE<>3G interworking in both idle/connected mode. To realize LTE<>3G interworking in idle mode, the wireless network needs to software upgrade the eNB like 3G neighbor configuration and reselection parameter, RNC needs to configure 4G neighbor configuration and reselection parameter, the core network needs to upgrade MME, MSC (should support SGs interface), and so on. There are four phases in LTE‐ > UMTS and UMTS‐ > LTE cell reselection: start measurement, measurement, decision, and execution. According to SIB3, SIB6, UE proceeds LTE‐ > UMTS reselection based on the priority. For UE in UMTS idle mode state or Cell_PCH/URA_PCH state, according to UMTS SIB19, which contains priority information (for E‐UTRA frequencies and UTRA cells) and LTE serving cell and neighboring cells (Figure 8.55). For LTE to GSM reselection, all descriptions of UTRAN above apply except UTRAN is replaced by GERAN and SIB 6 is replaced by SIB7. SIB7 contains GERAN frequency and neighboring cells for cell reselection mainly. GSM to LTE cell reselection is managed in “blind search” mode including PS cell reselection in NC0 mode (mobile controlled cell reselection). This means that 2G/LTE reselection algorithm is implemented at the MS side and performed autonomously. Inter‐RAT neighboring information and parameters SI2Quater message is broadcasted, all the parameters for the existing algorithm and the priority algorithm are sent. Then it is up to the MS to decide which algorithm to use in idle mode depending on its capabilities. 309 Release with Redirect to LTE (Blind) Fast return (Blind) Reselect to LTE (higher priority) Reselect to LTE (higher priority) HPRIOTHR sNonlntraSearch Reselect to WCDMA (lower priority) threshServingLow a1a2SearchThresholdRsrp b2ThresholdRsrp utranB1ThresholdRscp PS HO to WCDMA qRxLevMin a2CriticalThresholdRsrp Coverage Triggered Session Continuity (Blind) LTE Acceptable Performance Decreasing signal strength threshHigh qRxLevMin (Eutra) CSFB to GSM (Blind) GSM LTE WCDMA CSFB to WCDMA (Blind) Reselect to GSM (lower priority) QRXLEVMINE Coverage Triggered Session Continuity (Blind) threshXLow(Utran) qRxLevMin b2Threshold2RscpUtra qRxLevMin(Utran) threshXLow(Geran) qRxLevMin(Geran) Figure 8.52 IRAT classification and mobility actions. ACCMIN Mobility Optimization 1 2 3 4 5 6 LTE 5 Cell Reselection from low to high priority Cell Reselection from high to low priority 3 Coverage based HO/Redirection 2 1 UMTS 4 6 Service based HO/Redirection CSFB Fast Return 1 GSM 5 6 2 3 Figure 8.53 Inter‐RAT frequencies priority. 1- Event A2 is sent by the UE after 640ms of poor coverage detection 2- Blind RWR to 3G Cell 1- UE Idle mode on LTE Network Starts Measuring on UtranFreq: 2- UE reselects 3G Cell after 2s –122 –120 1- UE Idle on 3G Network starts Measuring on EUtranFreq: 2- UE reselects LTE cell after 2s –118 RSRP (LTE) –112 CONNECTED MODE LTE to 3G Session Continuity IDLE MODE LTE to 3G Reselection IDLE MODE 3G to LTE Reselection RSCP (3G) Figure 8.54 LTE<>3G interworking. For 3G to LTE cell reselection, UE starts measuring IRAT cells in search of a better cell to camp on when the following criteria is met: Srxlev Snonintrasearch OR cellReSelPrio inter RAT cellReSelPrrio serving When the measured RSRP level will be lower than qrxLevMin + sNonintrSearch (i.e., −130 + 20 = −110 dB) the terminal will start IRAT measurements as shown in Figure 8.56. The 311 UE camps on LTE (idle mode) UE camps on 3G (idle mode, or PCH state) N Y Start measurement measurement Y LTE_priority > UMTS_priority LTE(SqualServingCell)≤ 2(QRxLevMin + SNonIntraSearch) N Y Start measure UMTS decision Y execution Start L to U reselection Start measurement Start measure UMTS N UMTS(SrxlevnonServingCell)> 2*(QRxLevMin + ThreshXHigh) for TReseIUtran N measurement UMTS_priority > LTE_priority UMTS(SqualServingCell)≤ Qqualmin N Y Start measure LTE Start measure LTE N N LTE(SqualServingCell)≤ 2(QRxLevMin + ThrshServLow) and UMTS(SrxlevnonServingCell)> 2*(QRxLevMin+ThreshXLow) for TReseIUtran + Sprioritysearch2 N LTE(SrxlevnonServingCell)> 2*(Eqrxlevmin + ThdToHigh) for Treselection decision LTE(SrxlevnonServingCell)> 2*(Eqrxlevmin + ThdToLow) and UMTS (SqualServingCell)≤ Qqualmin for Treselection Y Y Figure 8.55 LTE‐ > UMTS and UMTS‐ > LTE cell reselection. execution Start U to L reselection Y Mobility Optimization qrxLevMin = –130 dBm IRAT measurements will start at –110 dBm sNonIntrSearch = 20 RSRP decreases –110 –130 Figure 8.56 Example of thresholds to start IRAT measurements. Table 8.22 Example parameter values for IRAT reselection in idle mode. Inter‐frequency and inter‐ sNonIntrsearch RAT measurements threshold 20 dB Defines the threshold (in dB) for inter‐RAT and inter‐frequency measurements. Minimum required RX level in cell qrxLevMin −130 dBm Specifies the Mini required RX RSRP level UTRA minimum required received level qrxlevminUTRA −115 dBm Minimum required Rx level in the cell. Cell reselection priority cellReSelPrio 7 Absolute priority of the LTE carrier frequency UTRA carrier frequency absolute priority uCelResPrio 3 Absolute priority of the UTRA carrier frequency UTRA cell reselection timer tResUtra 2s UTRA cell reselection timer Threshold serving low threshSrvLow 14 dB Threshold for the serving frequency used in reselection evaluation toward lower‐ priority LTE frequency or RAT UTRA inter‐frequency threshold low Threshx,low 8dB Threshold used in reselection toward the frequency X priority from a higher‐priority frequency condition regarding priorities is not fulfilled in this case, that is, LTE network has higher priority than 3G but it is enough one of the conditions is fulfilled to start measurements. 3G network can have higher or lower absolute priority than the LTE. This will depend on the operator’s strategy. If the operator wants to keep the terminals in LTE as long as possible it should give lower priority to 3G. Once the IRAT measurements are started, the cell reselection to the best cell on 3G layer will happen if for a time period equal to tResUtra (2 sec in the example): SServingCell threshSrvLow and SnonServingCell , x Threshx , low Based on example values from Table 8.22, when the RSRP level of the serving cell will be lower than −130 + 14 = −116dBm and the 3G level is greater than −115 + 8 = −107 dBm. 313 314 LTE Optimization Engineering Handbook 8.8.3 Reselection Optimization Cell reselection information is related about the available inter‐RAT frequencies, the cell reselection priorities, and the threshold values for cell reselection is provided to the UE through the system information in the cell. Usually, LTE cell priority is 7 and 3G priority is 4. Snonintrasearch drive cell detection frequency and lower‐priority cell detection for UE. If the Srxlev value of the serving cell falls below the threshServingLow value, the UE attempts to reselect a cell on an inter‐RAT frequency with cell reselection priority lower than the frequency where the UE is camping. Cell reselection occurs if the UE finds a cell with an Srxlev value greater than the threshXLow value for that frequency (Figure 8.57). For 3G, the LTE frequencies and the parameters for priority based cell reselection are sent on the broadcast channel in SIB19. For GSM, each GSM cell broadcasts information via the s ystem information message SI2quater, includes neighboring cells (UMTS and LTE), thresholds for IRAT and priority between GSM, UMTS and LTE cells (Figure 8.58). 3G to 4G and 4G to 3G signaling are shown in Figure 8.59. Signal strength Reselection Low Priority Signal to UMTS Snonintrasearch Threshservinglow Threshxlow LTE High Priority Qrxlevmin Starts measuring IF or IRAT freq Condition1: Srxlev (source) < ThreshServingLow Signal strength Low Priority Signal TreselectionUTRA & Condition2: Srxlev (target) > ThreshXLow Reselection to LTE High Priority Signal ThreshxHigh TreselectionRAT Condition: Srxlev (target) > ThreshXHigh Figure 8.57 IRAT measurement and reselection. Time Time Mobility Optimization IDLE CONNECTED RSRP<–118 & [RxLev>–110] LTE RSRP>–116 RSRP >–116 RSRP<–118 & [RSCP>–112 or EcNo>–14] WCDMA EcNo>–12 Fast Return LTE RSRP<–115 & RSCP>–102 RWR WCDMA [RSCP<–112 or EcNo<–14] & RxLev>–95 GSM RSRP<–115 & [RxLev>–109] [RSCP<–105 or EcNo<–12] & RxLev>–95 GSM Figure 8.58 IRAT cell reselection strategy. 8.8.3.1 LTE to UTRAN The following scenario is used when a UE in idle mode state has to move from a LTE coverage area to a 3G coverage area, the UE applies rules as follows. eNB needs to support SIB6 and SIB 3 broadcasting to the UE as shown in Figure 8.60. Table 8.23 depicts the IRAT cell reselection parameters. The SIB 6 message is about the 3G neighbor cells, and according to the message SIB 6 and the LTE signal, the UE decide whether it is necessary to reselect to 3G. SIB 6 message: IRAT (LTE to 3G) (Figure 8.61). To avoid ping pong between LTE and 3G, after a UE reselected to LTE from 3G, the 3G signal has to be strong enough for the UE to consider LTE to 3G reselection again, so that the UE does not measure 3G again. qRxLevmin LTE _ sib1 sNonIntraSearch LTE _ sib3 eUtraTargetFrequencyQrxlevmin 3G _ sib19 eUtraTargetFrequencyThreshxHigh 3G _ sib19 LTE parameter sNonIntraSearch should be set to a value between UMTS parameter threshHigh (threshold for going from UMTS to LTE: −118dBm) and LTE parameter threshServingLow (threshold for going from LTE to UMTS: −122dBm) to avoid ping‐pong between the two RATs: threshHigh sNonIntraSearch threshServingLow 0 Suggested initial values are: threshHigh = 10 (dB), corresponds to −118dBm, threshold for UMTS → LTE reselection, sNonIntraSearch = 8 (dB) corresponds to −120 dBm, threshold for starting UMTS measurements in LTE, threshServingLow = 6 (dB). corresponds to −122dBm, threshold for LTE → UMTS release. The related parameters will be included in Table 8.24. The eNB needs to configure the reselected parameter as shown in Table 8.25. The need to pay attention to the minimum received at the signal level Qrxlevmin can be in a plurality of broadcast news; different broadcast message parameters are different in meaning. Some parameters play a part in the both stages of measurement start and reselection decision (Figure 8.62). Parameter configurations needed to configure the relevant parameters in the corresponding SIB messages, are specifically named in Table 8.25. 315 UE RNC eNodeB MME S-GW SGSN P-GW PCRF HSS 1. Trigger to start TAU MS eNodeB new SGSN old MME S-GW P-GW HSS 0. UE changes to UTRAN or GERAN 2. TAU Request 3. TAU Request 1. Routeing Area Update Request 2. SGSN Context Request 4a. Context Request 2. SGSN Context Response 4b. Context Response 3. Security Functions 5. Security Functions 4. SGSN Context Acknowledge 6. Context ACK 7. Create Session Request old SGSN 8. Modify Bearer Request 6. Update PDP Context Request 9. PCEF Initiated IP-CAN Session Modification 10. Modify Bearer Response 6. Update PDP Context Response 7. Update Location 11. Create Session Response 8. Cancel Location 8. Cancel Location Ack 12. Update Location Request old MME 15a. Iu Release Command 9. Insert Subscriber Data 13a. Cancel Location Request 9. Insert Subscriber Data Ack 13b. Cancel Location ACK 10. Update Location Ack 14. Cancel Location Request 15b. Iu Release Complete 16. Cancel Location ACK 18a. TAU Accept C2 11. Routeing Area Update Accept 17. Update Location ACK C3 12. Routeing Area Update Complete 13. Delete Session Request 13. Delete Session Response 18b. TAU Complete 13. S1-AP: S1 Release 3G to 4G Figure 8.59 IRAT reselection signaling analysis. 4G to 3G Mobility Optimization YES CRP-irat > CRP-servingCell? SIB3 contains cell reselection information (intra-freq, inter-freq, interRAT) and speed dependant reselection parameters UE performs measurements of the higher priority inter-RAT frequency. NO NO UE may choose not to perform Is S-nonintrasearch sent in SIB3? measurement of inter-RAT frequency YES cells of equal or less priority. YES Is S-servingcell > S-nonintrasearch? UE performs measurements of IRAT frequency cells of equal or lower CRP. NO CRP = Cell reselection priority Figure 8.60 IRAT cell reselection. Table 8.23 IRAT cell reselection parameter. Parameters Value Specification SnonIntraSearch 31 Threshold when iRAT measurement are required for cell reselection. ThreshServingLow 24 Threshold that the signal strength of the serving cell must be below for cell reselection toward iRAT cell ThreshX,low 0 Threshold that the signal strength of the target iRAT cell must be above According to Table 8.25, UE will start searching for a lower‐priority 3G frequency when: RSRP qRxLevMin sNonIntraSearch 128 16 112dBm The UE will reselect to 3G when the below is fulfilled for a time tReselectionUtra: RSRP qRxLevMin EUtranCellFDD RSCP threshServingLow qRxLevMin UtranFreqRelation threshXLow 128 8 115 4 120d dBm and , 111dBm In addition, inter‐RAT cell reselection to a even lower‐priority GSM frequency is performed by the UE in LTE when the LTE and 3G criteria listed below are fulfilled during tReselection seconds: ●● ●● ●● Serving LTE cell RSRP below qRxLevMin [EUtranCellFDD] + threshServingLow Target UMTS cell RSCP is NOT above qRxLevMin [UtranFreqRelation] + threshXLow Target GSM cell RxLev is above qRxLevMin [GeranFreqGroupRelation] + threshXLow Using the current parameter values, the UE will reselect to GSM cell when LTE RSRP is below −120 dBm AND UMTS RSCP is NOT above −111 dBm AND GSM Rxlev is above −98 (−100 + 2) dBm for 2 seconds at least. 317 3G reselection Reselection when cell>–90 dBm, threshXLow = 30 Femto reselection Reselection when cell>–106 dBm, threshXHigh = 14 threshXhigh:11 is IE value, real configure value is 22. TDL RSRP>–98dBm Figure 8.61 SIB6 and SIB19. Mobility Optimization Table 8.24 LTE to UTRAN reselection parameters. Parameter Setting qRxLevMin −124 dBm threshServingLow 4 dB qRxLevMinOffset 2 threshXLow 6 sintrasearch 46 dB sNonIntraSearch 8 dB 2 tReselectionUtra b2Threshold1Rsrp −120 a2ThresholdRsrpPrim −118 Table 8.25 Reselection parameters in SIB messages. Impact SIB1 SIB3 Parameter Measure Decision Recommended Qrxlevmin (LTE) √ √ −128 PEMAX (LTE) √ √ 23 Snonintrasearch (LTE) √ Threshservinglow (LTE) √ √ cellReselectionPriority (LTE) SIB6 5~7 Threshxlow (to 3G decision threshold) Qrxlevmin (3G) P‐MaxUTRA (3G) t‐Reselection‐UTRA (3G) 2 cellReselectionPriority (3G) LTE RSRP tReselectionUtra qRxLevMin + sNonIntraSearch qRxLevMin (EUtranCellFDD) + threshServingLow qRxLevMin 2~3 LTE to WCDMA cell reselection WCDMA RSCP qRxLevMin (UtranFreqRelation) + threshXLow qRxLevMin (UtranFreqRelation) Figure 8.62 LTE to 3G reselection (example). 8.8.3.2 UTRAN to LTE UMTS to LTE cell reselection to a higher‐priority LTE frequency is performed by the UE in UMTS if the measured LTE cell RSRP is greater than qRxLevMin + threshHigh during tReselection seconds. The measurement in UMTS for LTE neighbor cells will be enabled continuous. Idle 319 320 LTE Optimization Engineering Handbook LTE RSRP Reselect from WCDMA to LTE threshHigh sNonIntra Search Hysteresis region Stable on either WCDMA or on LTE threshServingLow qRxLevMin Neither is suitable Reselect from LTE to WCDMA Fall off LTE and camp on WCDMA qRxLevMin threshXLow UMTS RSCP Figure 8.63 IRAT reselection strategy. mode cell reselection evaluation is performed every DRX cycle. The UE will reselect to LTE when the below is fulfilled for a time tReselectionUtra: RSRP threshHigh qRxLevMin EUtranFreqRelation 10 128 11 18dBM The related parameters (threshHigh, qRxLevMin etc.) are included in SIB19 in UMTS. The whole IRAT reselection strategy is shown in Figure 8.63. It illustrates the hysteresis and reselection regions for transition between LTE and UMTS in idle mode. 8.8.4 Redirection Optimization Cell redirection is a cell reselection when UE is in RRC connected state. It enables a multi‐RAT UE to be quickly redirected toward a target LTE (2/3G) cell while a UE is camping in a 2/3G (LTE) cell in RRC connected state (and involved in a data transfer). The redirection will make sure the retainability of the PS service and good end‐user experience. 8.8.4.1 LTE to UTRAN The LTE serving cell controls the redirection using UE RF measurement data. Two options are supported: blind redirection and measured redirection. Blind redirection from LTE to UTRA is used if UE does not support event B2 measurement (measurement purpose = “Mobility‐Inter‐RAT‐ to‐UTRA”), or event B2 measurement report is not available when mobility from LTE to UTRA has to be triggered. Compared with a blind redirection without UMTS radio measurements, the redirection with measurement improves the end‐user QoE (quality of experience) by redirecting the UE from an LTE cell to an UMTS overlay in a timely fashion. Measurement‐based redirection from LTE to UMTS is used if PS handover is not supported by the UE or is not activated in eNB. eNB may broadcast six UMTS frequencies in redirection message. The blind redirection is triggered when no PS handover is implemented or activated nor supported between those two systems and when the UE is not supporting inter‐system measurement gap capability. The blind redirection is based on serving cell measurement (Event A2 measurement) only rather than measured redirection, which is based on serving and target measurement (Event A2 and Event B2 measurements). The measured redirection is the same as the previous one with the enhancement on the measurement gap functionality. This procedure is triggered when no PS handover is implemented or activated nor supported and with UE supporting inter‐system measurement gaps on radio frames. The enhancement is mainly related to radio side, which improves UMTS redirection time duration and efficiency. Measurement gaps periods are set by RRC signaling reconfiguration procedure toward UE by fixing, length, periodicity and offset information in Mobility Optimization Sometimes try changing A2 parameter to trigger IRAT earlier and reduce drops during redirect RSRP Serving Cell RSRP threshold4 threshold4 –110 dBm reporting condition met A2 condition met after Time To Trigger a2TimeToTriggerRedirect time Connected Mode Triggered A2 Other case Decision iRAT measurement Blind Redirection Start A1 and B2 A1 reported 1 3 A5B2 timer expires 2 B2 reported Decision Other case Initiate Redirection Figure 8.64 RRC connection release with redirect procedure. radio subframes. The need for measurement gaps by the UE is specified in the UE capabilities. This scenario is using event A2 and B2 type measurement to indicate LTE serving cell degradation and redirection decision (Figure 8.64). Event B2 based redirection parameters are shown in Table 8.26. The example of redirection procedure and related parameters is shown in Figure 8.65. 321 322 LTE Optimization Engineering Handbook Table 8.26 Redirection parameters. Parameters Values UEMeasurementsActive true a1ThresholdRsrpPrim −90 Comments Decides whether iRAT measurement should be started RSRP threshold value for the primary event A1 measurement a2ThresholdRsrpPrim −102 RSRP threshold value for the primary event A2 measurement b2Threshold1Rsrp −104 RSRP threshold value of the serving cell for the Event B2 measurement b2Threshold2RscpUtra −119 RSCP threshold value for the Event B2 measurement a5B2MobilityTimer 3000 Specifies the time the UE are allowed to perform Event A5 and B2 measurements 8.8.4.2 UTRAN to LTE When UE was in connected mode in UMTS, the UE would receive the event 3A‐ or 3C‐related information. And then the UE would start the measurements. UE reported the 3A or 3C, and the RNC will send the RRC release information, which include the target frequency and with redirection information. There are eight frequency sent by RNC in RRC redirection information. And when UE received the command, the redirection will be executed after the TimeToTriger timer reached. Table 8.27 is the threshold of the 3A and 3C event. Figure 8.66 shows the procedure of UTRAN to LTE redirection. When the measurement threshold was satisfied, UE send measurement report to RNC. The redirection decision made by RNC and if the redirection will be executed, then RRC connection release will be sent to UE with target frequency information, finally UE sends the service request to MME and finishes tracking area update procedure. Table 8.28 gives an example of the threshold for trigger of event 3A. If you want to make the redirection from UMTS to LTE easily, thres1 can be set a little higher (−50 dBm), thres2 can be set a little lower (−110 dBm). These two parameters need to be set consist with reselection threshold, for example, the threshold for UTRAN to LTE reselection trigger EqrxlevMinRsrp can be set to −112dBm in this case shown in Table 8.28. 8.8.5 PS Handover Optimization The redirection is based on the RRC release information and found the target frequency and cell to access again. While PS handover flow will include the measurement and decision and action, and it can based on coverage and capacity and RRM will be discussed between serving and target eNB, and then prepare handover and UE finished handover with commands. In theory inter‐RAT PS handover feature can provide lossless handover between eUTRAN and GERAN/UTRAN, but the procedure is rather complicated. 8.8.5.1 LTE to UTRAN This part introduces the UE measurement–based packet switched (PS) handover procedure to move UE from LTE to UTRAN. eNB will trigger the PS handover when UE is leaving LTE coverage area and moving into UTRAN coverage area, and the UE measurement report indicates that the LTE radio condition becomes worse than a threshold and the UTRA radio condition becomes better than a threshold. Comparing with redirection, PS handover from LTE to UTRAN has the advantage of allocating the resources in UTRAN prior to the execution of PS handover. Besides, PS handover has the capability of data forwarding from source LTE to target UTRAN. It thus reduces the service interruption time and ensures better performance to packet loss–sensitive services. The source eNB will give a command to the UE to reselect a cell in the target access network via the RRC connection release. Figure 8.65 Field test of redirection procedure and related parameters. 324 LTE Optimization Engineering Handbook Table 8.27 3A and 3C. Events Event description In Leave 3A RSCPScell < thres1, and RSRPNcell > thres2 QUsed TUsed H 3 a /2 and MOtherRAT CIOOtherRAT TOtherRAT H 3 a /2 QUsed TUsed H 3 a /2 or MOtherRAT CIOOtherRAT TOtherRAT H 3 a /2 3C RSRPNcell > thres MOtherRAT CIOOtherRAT TOtherRAT H 3c /2 MOtherRAT CIOOtherRAT TOtherRAT H 3c /2 UE Trigger phase Node B RNC eNB RRC_PH_CH_RECFG RRC_PH_CH_RECFG_CMP Measurement phase RRC_MEAS_CTRL RRC_MEAS_RPRT RRC_RRC_CONN_REL Decision phase Execution phase RRC_RRC_CONN_REL_CMP TrackingAreaUpdateRequest Figure 8.66 UTRAN to LTE redirection. If PS handover is to be performed, eNB will select the best UTRA cell as the PS handover target cell. eNB will send a handover required message to the MME and start timer TS1relocprep. If the reservation of resources in the target UTRAN cell is successfully completed, MME will send a handover command message to eNB. eNB will then stops the timer TS1relocprep and enter the handover execution phase. If timer TS1relocprep expired, or eNB receives a S1 handover preparation failure message, handover preparation fails. Once a handover command message is received from MME, eNB will stop timer TS1relocprep and start timer TS1relocoverall, eNB will send a MobilityFromEutraCommand to the UE with purpose set to “handover” and targetRAT‐Type set to “ultra.” If UE context release command is received from MME, PS handover is successful. eNB will send a UE context release complete to MME. eNB will stop timer TS1relocoverall and release UE context and associated resources. If timer TS1relocoverall expires, eNB considers the UE to have lost radio coverage and will trigger the release of all UE associated resources by sending an UE context release request to MME and release all UE associated resources in eNB. 8.8.5.2 UTRAN to LTE For UTRAN to E‐UTRAN handover procedures, the UTRAN has two main tasks: processing the handover decision and initiating the handover procedure by sending RANAP relocation required message with relevant content to source SGSN. Prior to the handover preparation phase, UTRAN has to manage UE LTE capacity, monitor the radio condition change and LTE neighboring cells, and eventually activate CM measurements. In the handover execution phase, UTRAN will process CN message RANAP relocation command, command the UE to handover to the target eNB via the message RRC handover from UTRAN command, which contains the transparent container with radio aspects parameters set up by target eNB during preparation Mobility Optimization Table 8.28 Example of the threshold for trigger of event 3A. NLTE RSRP > thresholdOtherSystem + Hysteresis/2 ‐ CIO and SUMTS RSCP < thresholdOwnSyste ‐ Hysteresis/2 thresholdOtherSystem −110 dBm Hysteresis 4 dB thresholdOwnSyste −50 dBm CIO 0 dB TimeToTrigger 640ms phase and perform data forwarding between UE and CN for DL and UL user plane data till effective relocation completion. 8.8.6 Reselection and Redirection Latency The tests show that the delay of LTE to UMTS idle state reselection is about 2.6 to 4.9s. UMTS to LTE reselection delay is about 0.3 to 2s. The control plane interrupt time delay of LTE redirection to UMTS is about 3.0 to 5.5s, it is far from searching with out of LTE service (12 to 32s). The control plane interrupt time delay of UMTS redirection to LTE is about 0.7 to 1s. They are as same as theoretical analysis and laboratory test results. Table 8.29 shows the typical reselection latency between LTE and UMTS. There are some differences in the test results of different network, because of the different broadcast messages in different UMTS systems, and whether the authentication process of inter‐ system interoperability is on. In the same area, the delay of different terminal chips is slightly different, because of the difference of the synchronization to the target cell performance. In addition, an analysis of difference for the “percentage of Time on LTE” between the idle mode users and connected mode users. It aims to focus on improving the connected mode time on LTE. it is needed to find the best transition area for all technologies to improve LTE coverage and end‐user experience with RF optimization. As an example, one key parameter difference between idle‐mode reselection and connected‐mode redirection is the time to trigger as shown in Table 8.30. Table 8.29 Reselection latency. Attempts First broadcast message > RRC connection request/ channel request(s) RRC connection request— > TAU accept/RAU accep(s) Delay(s) TD‐L‐ > TD‐S 500 1.488 2.203 3.691 TD‐S‐ > TD‐L 500 0.114 1.06 1.209 Table 8.30 Parameters comparison of idle‐mode reselection and connected‐mode redirection. Idle mode Connected mode Description Values Idle/connected Qrxlevmin N/A Absolute minimum RF value −122/−122 dBm Treselection Time To TriggerA2Prim Time to Trigger 2000/640 ms ThreshServingLow a2ThresholdRsrpPrim Minimum RSRP value to trigger −116 dBm/−116 dBm qHyst hysteresisA2P rim Hysteresis 0 dBm/1dBm 325 326 LTE Optimization Engineering Handbook 8.8.7 Optimization Case Study One of the things that needs to be monitored in LTE mobility domain is related with the success rate of the handover procedures on source cell. In order to evaluate this it is advised to check success rate indicators, which represents the percentage of the successfully performed handovers. It has possibilities to monitor the different types of LTE‐LTE handover: intra‐eNB, inter‐ eNB, intra‐frequency (X2), inter‐eNB intra‐frequency (S1), inter‐eNB inter‐frequency (X2), and inter‐eNB inter‐frequency (S1). Case 1: ping‐pong reselection In order to prevent the ping‐pong reselection, the reselection threshold of LTE/UMTS/GSM should be considered. In order to prevent the LTE reselection to GSM network quickly after reselection to the UMTS network, 3G signal threshold of LTE to UMTS network needs to be higher than that of UMTS to GSM. In order to prevent ping‐pong reselection between the LTE and UMTS network, the 4G LTE network signal threshold of UMTS to LTE network should be higher than LTE reselection to UMTS threshold (Figure 8.67). Case 2: UE redirection to LTE failure due to EPLMN misconfigured The function of UMTS to LTE redirection is enabled, it is found that after RNC proceeded redirection to LTE, the UE still stayed on UMTS. As shown in Figure 8.68, after UE reported event 3C, RRC connection release was transmitted. LTE frequency was carried in the release message, but the UE was back to UMTS after LTE SIB reading (Figure 8.68). Considering the particularity of the live network, the PLMN of UMTS, LTE of the two operators were inconsistent. Therefore, 3G/4G are required to configure the equivalent PLMN, the inspection found that CS domain of LTE and UMTS were configured the IRAT system equivalent PLMN, but PS domain of UMTS side was not configured the equivalent PLMN of IRAT system, it can be seen by the results of TAU/LAU/RAU attach procedure. After configuring the PS domain equivalent PLMN, UE was redirected to the LTE to initiate the TAU process, as shown in Figure 8.69. For IRAT operation, when UE redirects the PLMN from one system to another one, if the UE cannot access the target system after reading the SIB message, it needs to inspect whether the PLMN is consistent and equivalent PLMN is misconfigured by the core network. 8.9 Handover Interruption Time Optimization After a handover command is triggered, the UE disconnects from the serving eNB before setting up a connection with the target eNB and stops receiving data. This is the point in time where data interruption starts. For long handover interruption time, interference and missing neighbors need to be analyzed. The period of time where the UE can not exchange data is referred to as handover interruption time. It includes the time to execute radio access network procedures, the time for UL and DL radio resource control signaling, and the time taken to notify and execute the data path switching. Handover preparation time is the time duration the eNB takes to prepare the handover. Handover execution time is the time duration RRC signaling is interrupted during handover. Handover preparation time and handover execution time can be according to: Handover preparation time = T (RRC connection reconfiguration) − T (measurement report) Handover execution time = T (RRC connection reconfiguration complete) − T (RRC connection reconfiguration) Table 8.31 gives the most popular handover lantency/interruption KPIs. UE RAT Type UE RAT Type WCDMA LTE WCDMA RAT type RAT type LTE 03 07 11 15 19 23 27 31 35 39 43 47 51 55 59 03 07 11 15 19 23 27 31 35 39 43 47 51 55 59 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: 1: :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 :2 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 Time Figure 8.67 Ping‐pong reselection optimization before and after. 48 52 56 00 4 8 12 16 20 24 28 32 36 40 44 48 52 56 00 04 08 2 6 20 24 28 32 36 40 44 7: 7: 7: 8: :0 :0 8: 8: 8: 8: 8: 8: 8: 8: 8: 8: 8: 8: 9: 9: 9: :1 :1 9: 9: 9: 9: 9: 9: 9: :4 :4 :4 :4 :48 :48 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :49 :49 :4 :4 :4 :4 :4 :4 :4 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 Time 328 LTE Optimization Engineering Handbook Figure 8.68 UE redirection to LTE failure. Figure 8.69 UE redirection to LTE success after PLMN configured. Table 8.31 KPIs focused on the volume. KPI Name HO interruption delay for VoIP HO interruption delay for best effort service LTE TO UMTS PS HO – interruption time and latency Voice call setup time for CSFB to UTRAN using Redirection/PSHO Data interruption time due to LTE to UTRAN CSFB via Redirection Voice call setup preparation time for CSFB to 1xRTT by idle Dual‐Rx UE LTE to CDMA HO interruption time type HO (data) Enhanced non‐optimized LTE‐to‐HRPD handover preparation latency Data interruption time for LTE to GERAN PS mobility using CCO with NACC LTE‐to‐GERAN handover via Redirection interruption time Voice call setup time for CSFB to GERAN using CCO with NACC/Redirection Voice call interruption time for SRVCC HO to UTRAN Mobility Optimization Figure 8.70 Handover interruption time estimation. Take S1 handover, for example, the handover interruption time is estimated in Figure 8.70. For CBRA, the interruption time is estimated to approx 62 ms in DL and 32 ms in UL. Transport delay in both directions is not included in the figures. For CFRA, eNB knows that it is a handover access from the expected UE when receiving random access preamble. This means that the eNB can start transfer of DL data earlier which will reduce the interruption time by approx 10 ms to approx 52 ms. 8.9.1 Control Plane and User Plane Latency Handover control plane interruption time is defined as the duration between the handover command at source eNB and the RRC reconfiguration complete at target eNB. In average the time from measurement report to RRC connection reconfiguration took 40 to 50ms. From RRC connection reconfiguration message to RRC connection reconfiguration complete, it took the UE 30 to 40ms (Figure 8.71). User plane (UP) interruption time has been noticed to be too high in field test, especially with FTP, handover procedure works as specified by 3GPP and user plane interruption times are varying roughly even between 100ms to 350ms most of the time. After disconnecting from the serving eNB, the UE typically waits for the next random access opportunity to execute a random access procedure to acquire service with the target eNB and eventually be able to resume any data exchange. An example of X2 handover UP latency test that uses the different eNBs to identify last/ first package is given in Figure 8.72. It shows before handover, UE1 (IP address 10.63.0.2) received the last package at T1 = 15.850252s, and in the time UE1 stays at eNB1 (IP address 10.100.100.10) (as two S1 links are mapped to a same mirror port, so there are two messages with same source and destination). At T2 = 15.899244s, the first package is received by UE1 after handover to eNB2 (IP address 10.100.100.2). So handover UP latency = 15.899244−15. 850252 = 49 ms. Handover UP is also can be computed the UP handover latency based on the signaling and RLC PDU, shown in Figure 8.73. 329 330 LTE Optimization Engineering Handbook UE Target eNB Source eNB Measurement Report 40–50 ms RRC Connection Reconfiguration Handover Command 30–40 ms RRC Connection Reconfiguration Complete RRC Connection Reconfiguration RRC Connection Reconfiguration Complete Source eNB UE Target eNB Measurement report Last RLC PDU T304 starts RRC Connection Reconfig Minimize PRACH access latency Random Access Procedure T304 stops UP interruption time RRC Connection Reconfig Complete 1st RLC PDU Figure 8.71 Control plane (left) and user plane (right) latency. Figure 8.72 An example of X2 handover UP latency test. After optimization with properly tuned parameters, less than 50 to 60ms FTP layer interruption time can be achieved in good RF conditions. For more handover interruption reduction improvement that aims to shorten the LTE S1 and X2 handover interruption time by triggering the path switch (for X2 handover) and the handover notify (for S1 handover) messages to MME earlier when the target eNB receives the RA message 3 from the UE, which is needed to be earlier than the reception of RRC connection reconfiguration Mobility Optimization Figure 8.73 UP handover latency based on the signaling and RLC PDU. UE Source eNB Target ANR MME SGW RRC:MEASUREMENT REPORT HO Decision X2AP: HANDOVER REQUEST Resource allocation X2AP: HANDOVER REQUEST ACKNOWLEDGE RRC: RCC CONNECTION RECONFIGURATION Switch to Target RAN X2AP: SN STATUS TRANSFER MAC: CBRA RACH PREAMBLE MAC: CBRA RACH RESPONSE RRC CONNECTION REQ REQ/MSG3 / MSG3 S1: PATH SWITCH REQUEST SETUP/MSG4 RRC CONNECTION SETUP / MSG4 S1: PATH SWITCH REQUEST RRC: RRC RRC: RRCCONNECTION CONNECTION RECONFIGURATION RECONFIGURATION COMPLETE COMPLETE DL data S11: UP update req Path Switch Figure 8.74 Reduction on the S1 and X2 handover Interruption time. complete message from the UE but with the same confidence on the UE presence in the target cell. Figure 8.74 shows for the X2 handover case the point when RA message 3 is received in the target eNB and where the path switch request will be triggered. The procedures (in circle) that follow this point will therefore trigger the DL data path switch ~12 ms earlier. 331 332 LTE Optimization Engineering Handbook The system shall be configured to use RA message 3 instead of RRC connection reconfigura tion request for initiating intra eNB switch. During incoming intra‐eNB handover execution and upon the reception of RA message 3, the RAN use this message as an earlier trigger to confirm internal dispatcher of the UE presence in the target cell and initiate the release of resources at the source cell. During incoming X2 handover execution and upon the reception of RA message 3, the RAN use this as an earlier trigger to send the S1‐AP path switch message to MME. During incoming S1 handover execution and upon the reception of RA message 3, the RAN use this message as an earlier trigger to send the S1‐AP handover notify message to MME. 8.9.2 Inter‐RAT Mobility Latency For inter‐RAT mobility, UL trigger can be possible to set per QCI in the same way as bad coverage DL trigger based on GINR. GINR can be calculated as: GINR psdRX psdTX N I where psdRX (Received signal PSD), psdTX is the estimated UL Tx PSD based on power headroom reports from the UE; N + I is the received noise and interference averaged over the bandwidth. Measurement‐based and non‐measurement based “InterRAT cell reselection” and “Release with redirection” from the LTE network to both UMTS and GSM are supported as shown in Figure 8.75. For inter‐RAT mobility latency, timing is an indication only as it is very terminal dependent. Cell search time may be reduced if inter‐RAT measurement done before release with redirect. The UE will take much time for cell search. The UE first performs PLMN selection and finds its home PLMN where the MCC and MNC of the PLMN identity match the MCC and MNC of the USIM’s IMSI. The UE may optionally use information it has previously stored if it is available in order to reduce the time needed to select a PLMN. 8.10 Handover Failure and Improvement Mobility procedure can be divided into handover preparation and handover execution phases. Handover preparation is the phase in which the target cell assigns the necessary radio resources for taking over the connection and sending back a handover command message containing the new radio parameters to the source cell. Handover execution phase starts when the previously received handover command message is sent to the UE and successfully finished after the UE has arrived at the target cell. In most of the cases, handover failure could be due to poor radio conditions or badly tuned handover parameters. For example, with too little overlap between cells, handover may fail, with too much cell overlap, higher interference occurs and cell‐edge throughput can be reduced. So a balance must be achieved by adjusting overlap margins and cell sizes. This can be achieved with parameters and physical changes. Downlink mobility issues include RLC failure on SRB1 that UE doesn’t receive RRC reconfiguration (handover command) or PDCCH decoding error (UE miss detect the PDCCH order). Uplink mobility issues include random access failure in the target cell or target cell didn’t receive RRC reconfiguration complete message from UE. If statistic analysis indicates the main reason for low handover success rate in the network was due to downlink, it could be either due to the DL RLC transmission for RRC handover Re d ire ct Mobility Optimization UL trigger Handover nd Bli Triggered DL trigger (from UE) Non ct re -bli nd A5, B2, B1 i ed R Handover 12 s 5–6 s Cell search 4–5 s 2–4 s 300 ms PS handover PS handover Cell search Cell search SIB reading SIB reading Bearer setup LAU/RAU Bearer setup LAU/RAU Bearer setup LAU/RAU Release with Redirect with NACC (SI) Release with redirect Radio link failure IRAT to WCDMA 15+ s 9–11 s Cell search Cell search Cell search SIB reading SIB reading Bearer setup LAU/RAU Bearer setup LAU/RAU Bearer setup LAU/RAU 8–9 s 300 ms PS handover IRAT to GSM Figure 8.75 Inter‐RAT mobility latency. command reaches max transmit attempts or due to PDCCH decoding issue. Downlink should be focused to improve the handover success rate by improving RLC robustness, improving PDCCH robustness, and reducing downlink interference by means of RF tuning, and so on. For mobility troubleshooting, handover preparation failures and handover execution failures should be identified when monitoring LTE. In Figure 8.76, the example is about handover preparation failure. The source eNB initiates the procedure by sending the handover request message to the target eNB. When the source 333 334 LTE Optimization Engineering Handbook E-RAB QoS information Input Target eNB Resources are granted by target eNB? HO preparation failure N Y Default value for Trelocprep is: 5s Configure required resources Reserve a C RNTI for UE Reserve a RACH preamble (optional) source eNB target eNB HANDOVER REQUEST HANDOVER PREPARATION FAILURE Figure 8.76 Example of handover preparation failure due to Trelocprep expiry. eNB sends the handover request message, it shall start the timer Trelocprep. Upon reception of the handover request acknowledge message the source eNB shall stop the timer Trelocprep, start the timer TX2relocoverall and terminate the handover preparation procedure. If there is no response from the target eNB to the handover request message before timer Trelocprep expires in the source eNB, the source eNB should cancel the handover preparation procedure toward the target eNB by initiating the handover cancel procedure with the appropriate value for the cause IE. Here is another example of handover preparation failure as shown in Figure 8.77. If the target eNB does not admit at least one non‐GBR E‐RAB, or a failure occurs during the handover preparation, the target eNB shall send the handover preparation failure message to the source eNB with the cause IE. If the target eNB receives a handover request message containing RRC context IE that does not include required information as specified in TS 36.331, the target eNB shall send the handover preparation failure message to the source eNB. Figure 8.78 presents the possible causes of handover failures. Handover preparation may fail because the target eNB cannot provide the necessary resources for the handover during congestion or cannot interpret the contents of the handover request message. If the source eNB does not receive a response to its X2AP handover request message from the target eNB, the source eNB will send the handover cancel message. Handover execution failure that may be caused by incorrect parameter settings in the target cell (e.g., PCI collision in the target cell). From the possible causes for handover degradation, it can be concluded according to the mobility optimization steps as shown in Figure 8.79. To better understand what is happening with the handovers use a map to plot the locations of the target and each source cell and also draw a xx km (nearly triple inter‐site distance) ring around the target to identify the source cells with the highest number of failures and add pointers that show the sector’s azimuth. You must first check all KPIs for the site, for example, handover oscillation level, noise floor, interference, high or low traffic, PCI conflicts exist or not, handover preparation success rate, and execution success rate. Mobility Optimization Figure 8.77 Example of handover preparation failure due to no resource granted. The meaning of the different handover cause value is described in Table 8.32. 8.11 Mobility Robustness Optimization In live network, it can be found that in certain scenarios measurement report messages count is significantly higher than number of handovers which indicates inability to execute handover quickly. In such scenarios, UE receives ACKs for PHICH and RLC for the measurement report but does not receive RRC reconfiguration message. Boxes in Figure 8.80 indicate the range of RSRP/RSRQ values where handovers took place easily. 335 336 LTE Optimization Engineering Handbook HO preparation failures HO execution failures Possible Causes Other reasons for poor mobility Possible Causes 1. Resources not granted by target eNB 2. Incorrect parameter settings in target 3. Congestion in target cell 4. Target cannot interpret the contents in handover request message 5. License issue 6. Target cell down 7. Improper MME configuration 8. Insufficient SR and CQI resources for target 1. Incorrect parameter settings in target (PCI collision in target cell) 2. Hardware faults 3. Unreasonable weight of handover attempt to target cells (If handover is triggered too late, the source cell SINR can be too low. This can result in an abnormal release before handover.) 4. Swapped sectors 5. Overshooting 6. Unreasonable neighbor relation Possible Causes 1. Poor radio conditions (interference) 2. Badly tuned handover parameters -Handover hysteresis and time-totrigger settings are required to prevent excessive ping-pong handovers. -Such behavior increases signaling, risk of failure, and decreases throughput Figure 8.78 Possible causes of handover failures. Low HO Succ HoPre fail HoExe fail Target Cell congest? N X2 HO? N Check X2 definition Scell->Multi-tcell Y Expansion, distributary, control coverage S1 HO? Y Check core configuration, cut over... Single neighbor Multi Scell->Tcell Scell alarm? Tcell alarm? Y Y Alarm processing Alarm processing N Overlapping? Check coverage, power Y Control coverage N Tcell interference? N Checking interference Parameter, earfcn/pci, external defining Figure 8.79 Mobility optimization steps. The next step will further analysis of handovers for triggering event per LTE cluster, cell or user. It needs to find that the number of A3‐, A5‐, B2‐based handovers in a live network. Mobility optimization feature enabled in RAN targets the following intra‐LTE mobility issues: ●● ●● ●● Connection failure is due to too early handover: The UE has a connection failure during a handover procedure or soon (<1s) after a successful handover and if the UE re‐establishes in the source cell Connection failure is due to too late handover: The UE has a connection failure during a handover procedure or after a long stay (>1s) in the same cell and if the UE re‐establishes in a different cell, this usually happened just before handover, it was observed that the UE SINR was poor and this caused the UE to drag the call in poor radio environment rather than handover faster to a neighbour cell with better RSRP Connection failure is due to handover to wrong cell: The UE has a connection failure during a handover procedure or soon (<1s) after a successful handover and if the UE re‐establishes in a cell which was neither the original source or the target cell Mobility Optimization Table 8.32 Root cause analysis (X2). Radio Network cause Description Possible Reasons Cell not available The concerned cell is not available. Target cell down Handover(HO) target not allowed HO to the indicated target cell is not allowed for the UE in question Check blacklisted neighbor list Invalid MME Group ID The target eNB doesn’t belong to the same pool area of the source eNB, that is, S1 HOs should be attempted instead. Check MME group id of source and target. Confirm both are belongs to same MME group No radio resources available in target cell The target cell doesn’t have sufficient radio resources available. Target cell congestion during failure needs to be checked Partial HO Provides a reason for the HO cancellation. The target eNB did not admit all E‐RABs included in the HO REQUEST and the source eNB estimated service continuity for the UE would be better by not proceeding with handover toward this particular target eNB. Target eNB will send HO preparation failure message with this cause after “TRELOCprep expiry” happens in source eNB Reduce load in serving cell Load on serving cell needs to be reduced. Applicable for load‐based HO Resource optimization HO The reason for requesting HO is to improve the load distribution with the neighbor cells. Time critical HO HO is requested for time critical reason i.e. this cause value is reserved to represent all critical cases where the connection is likely to be dropped if HO is not performed. A5 triggered (interfrequency attempt) (TDD‐ FDD) HO. HO failure due to poor radio condition. TRELOCprep Expiry HO preparation procedure is cancelled when timer TRELOCprep expires. Target eNB will send HO preparation failure message with this cause after TRELOCprep expiry happens in source eNB Unknown MME Code The target eNB belongs to the same pool area of the source eNB and recognizes the MME Group ID. However, the MME Code is unknown to the target eNB. Check the MME code of target eNB Unknown new eNB UE X2AP ID The action failed because the New eNB UE X2AP ID is unknown X2 link status needs to be checked Unknown old eNB UE X2AP ID The action failed because the Old eNB UE X2AP ID is unknown Unknown pair of UE X2AP ID The action failed because the pair of UE X2 AP IDs is unknown ExistingMeasurementID The action failed because measurement‐ID is already used Unknown eNB measurement ID The action failed because some eNB Measurement‐ID is unknown. Measurement temporarily not available The eNB can temporarily not provide the requested measurement object. Target cell level parameters (mainly HO‐event configuration) needs to be checked 337 Serving vs. Neighbor RSRP Serving RSRP vs. RSRQ –55 –4 –7 –65 –75 Serving RSRQ Neighbor RSRP –10 –85 –95 –105 –13 –16 –19 –22 –25 –115 –28 –125 –125 –115 –105 –95 –85 –75 –65 –31 –125 –55 –115 –105 Serving RSRP Serving vs. Neighbor RSRQ –75 –65 –75 –65 Serving RSRP vs. Neighbor RSRQ –4 –7 –7 –10 –10 –13 –13 Neighbor RSRQ Neighbor RSRQ –85 Serving RSRP –4 –16 –19 –22 –16 –19 –22 –25 –25 –28 –28 –31 –31 –95 –28 –25 –22 –19 –16 –13 –10 –7 Serving RSRQ Figure 8.80 Handovers happen in the range of RSRP/RSRQ values. –4 –31 –125 –115 –105 –95 –85 Serving RSRP Mobility Optimization Handover margin/Offset Cell-A Too early HO Too late HO Cell-B Wrong cell HO Cell-C Figure 8.81 Abnormal handovers. ●● Ping‐pong handover, as known as unnecessary handover or handover oscillation. Typically in this scenario, UE may experience ping‐pong handover between two badly overlapped cells. The major reason is that two cells have a big overlapped area in which two cells alternate to emanate the stronger signal and multiple handovers occur between cell A and cell B within one second after a completion of the handover, back and forth, these results in dropped calls. An oscillating UE is a UE does repeated handover between two or more cells. The UE performs a handover from source eNB to target eNB and then within a certain time performs a handover back again. Oscillating UE is harmful due to increased risk for handover failure, packet loss for UM (unknowledge mode) connections, causes unnecessary network load, and reduced throughput (Figure 8.81). If handover is triggered too early, the target cell SINR can be too weak when handover occurs. If handover is triggered too late, the source cell SINR can be too low. This can result in an abnormal release before handover. There are six types of handover failure are defined as shown in Figure 8.82, corresponding counters/events will peg when the failure cases happen. For too late handover (case 4), before handover is initiated in source eNB, UE sends RRC con nection reestablishment request to re‐establish radio link connection to a non‐source cell. The controlling eNB of the cell sends RRC connection reestablishment reject because it does not have the UE context and sends X2AP RLF indication to source eNB to report RLF failure. Source eNB increments the “HoTooLate” counter. For too early handover (case 2), the eNB succeeds with handover of a UE to a target cell. The UE subsequently lose connection, and the UE does reestablishment in the source cell. The eNB sends RLF indicator to the target eNB. The target eNB detects a too early handover, and sends a handover report to the eNB. X2AP RLF indication is shown below, which is from 3GPP TS 36.423. RLFIndication-IEs X2AP-PROTOCOL-IES ::= { { ID id-FailureCellPCI CRITICALITY ignore { ID id-Re‐establishmentCellECGI CRITICALITY ignore { ID id-FailureCellCRNTI CRITICALITY ignore TYPE PCI TYPE ECGI TYPE CRNTI 339 340 LTE Optimization Engineering Handbook eNB Target eNB eNB Target eNB Handover failure Handover failure eNB UE Reest. UE Reest. Case 3 Too Late Handover, HO initialized Handover UE Reest. eNB RLF RLF INDICATION Case 5 Handover to Wrong Cell, HO not completed eNB Target eNB eNB Target eNB Target eNB HO REPORT Third ENB Handover RLF RLF UE Reest. UE Reest. RLF INDICATION Third ENB UE Reest. RLF INDICATION Case 1 Too Early Handover, HO not completed Target eNB Handover failure RLF INDICATION RLF INDICATION HO REPORT Case 2 Case 4 Too Early Handover, HO completed Too Late Handover, HO not initialized Case 6 Handover to Wrong cell, HO completed Figure 8.82 Handover failures of the six cases. { ID id-ShortMAC‐I { ID id-UE-RLF-Report-Container { ID id-RRCConnSetupIndicator ... CRITICALITY ignore TYPE ShortMAC-I CRITICALITY ignore TYPE UE-RLF-Report-Container CRITICALITY reject TYPE RRCConnSetupIndicator } HandoverReport-IEs X2AP-PROTOCOL-IES ::= { { ID id-HandoverReportType CRITICALITY ignore TYPE HandoverReportType { ID id-Cause CRITICALITY ignore TYPE Cause { ID id-SourceCellECGI CRITICALITY ignore TYPE ECGI { ID id-FailureCellECGI CRITICALITY ignore TYPE ECGI { ID id-Re‐establishmentCellECGI CRITICALITY ignore TYPE ECGI } HandoverReportType ::= ENUMERATED { hoTooEarly, hoToWrongCell, ... } The mobility optimization feature optimizes the cell borders based on the observed output of the different error cases and builds up statistics about the relation. Parameter changes may only take place if enough statistics is gathered. The statistics that is built up is the number of occurrences of the four different types of failures. These data are weighted depending on the cost of the failure and added up per neighbor relation, and then determine if a change is required to move the cell border, usually adjust the cel lIndividualOffset value per relation and applicable on all UE’s in the cell. Figure 8.83 shows a method of handover parameters optimization based on irritative calculation. After change, evaluate the relative difference between the current situation compared what the last change was made. It is worth to note that maximizing the reduction of handover failure rate is not the main objective of this feature, instead tries to find the best balance between handover failures and unnecessary handover. Mobility Optimization Yes dt+1 = 1 dt+1 = –1 No Increase ICO? Decide on change direction No Change ICO by dt+1 dB Gather statistics No Enough Statistics? Yes Evaluate statistics dt+1 = –dlast change Yes Change required? Did results worsen? Yes Figure 8.83 Handover parameters optimization based on irritative calculation. Parameters to be changed in connected mode and idle mode, for connected mode, the UE uses event A3 to evaluate entry, the relation between (Mn‐Ms) and cellIndividual OffsetEutran is: Mn Ms a3offset hysteresisA3 cellIndividualOffsetEutran cellIndividualOffsetEutran , Mn Ms ;cellIndividualOffsetEutran , Mn Ms ; For idle mode, cell reselection formula is: Rs = Qmeas,s + Qhyst, Rn = Qmeas,n – QOffset, Rn > Rs; Relation between (Qmeas,n – Qmeas,s) and Qoffset is Qmeas,n – Qmeas,s > Qhyst + Q offse,. which leads to Qoffset , Qmeas,n – Qmeas,s ; Qoffset , Qmeas,n – Qmeas,s . 8.12 Carrier Aggregation Mobility Optimization Carrier aggregation (CA) is used to increase the bandwidth by combining up to five carriers intra‐band or inter‐band, and thereby increase the peak bitrates. CA UE throughput is much higher than signal carrier, due to additional carrier, frequency selection and scheduling gain and high category UE gain. The UE have one Primary cell (Pcell), this will also be the cell where the UE is connected to, from a EPC perspective. The cell that is not the Pcell will be the Secondary cell (Scell). Mobility is based on Pcell coverage, there is no changes in cell selection/reselection for carrier aggregation. CA configured eNBs are compatible with non‐CA eNBs (for handover, etc.), and that non‐CA UE can co‐exist on a cell simultaneously with CA configured UE. UE uses Pcell to monitor system information, maintain RRC connection, monitor RLF, random access, and so on, all security input and NAS mobility information is communicated 341 342 LTE Optimization Engineering Handbook PCC SCC SCC PCC SCC SCC PCell SCell SCell PCell SCell SCell PUCCH & PUSCH PUSCH Only PDSCH & PDCCH PDSCH & Optional PDCCH Figure 8.84 PCC and SCC. RRC reconfiguration, add Scell IE. Not CA configured RRC reconfiguration, remove Scell IE. Activation MAC control element CA configured Deactivation MAC control element Active Scell triggered based on buffer status CA activated Deactive Scell triggered based on buffer status or poor Scell CQI Figure 8.85 Dynamic Scell selection. using Pcell. UL/DL carrier corresponding to Pcell is primary component carrier (PCC) and to Scell is secondary component carrier (SCC). The Scell operates on a secondary frequency and can be configured once the RRC connection is established. Separate UL PC is required for different CC due to difference in propagation condition in non‐contiguous CA and for different interference conditions in UL CC in contiguous CA (Figure 8.84). The similar Pcell and Scell coverage is the precondition to provide CA capability for the UEs. Before the second carrier cell goes on air, single‐site verification shall be conducted carefully to make sure there are no defects, and second carrier RF optimization tuning shall be done aligning to the first carrier coverage. Similar antenna model, height, downtilt and azimuth is preferred for multi‐carriers. There are always coverage mismatching due to the separate mounted antennas and different bands, which will degraded CA throughput due to poor Scell performance. The mitigation is to activate the feature— dynamic Scell selection as shown in Figure 8.85. It is worth to note that CA configuration/deconfiguration is performed on RRC level, CA activation/deactivation is performed on MAC layer. At attach, reestablishment and incoming handover, the eNB will check CA license, CA neighbor cell configuration (i.e., SCell candidate) and UE capability. Figure 8.86 shows SCell selection is performed by RRC connection reconfiguration in attach procedure. SCell activation/deactivation is performed by MAC control element, which can be shown through RF conditions, resources scheduling, and DL/UL throughput, and so on, in driving test. In CA activated state, the UE is ready to receive data transfer on the SCell (DL assignments sent on SCell, HARQ ACKs are sent to eNB via PCell) and report the SCell’s CSI to eNB via PCell. SCell activate/deactivation can be triggered based on “need” or “coverage.” It notes that eNB will not deconfigure but deactivate the SCell when UE goes out of SCell coverage. PCell always changes due to handover, in the new PCell, old SCell is removed and new SCell is configured. All of this is done in the same RRC reconfiguration message as the handover itself. For CA mobility, a new measurement event A6 is introduced for CA, an intra‐frequency neighbor becomes offset better than SCell for which neighbor cells on an SCC are compared to SCell of that SCC. A6 works in the same way as A3, it reports the strongest cell that matches the configured frequency. If service triggered mobility is used to change the inter‐frequency handover thresholds, CA will continue to be possible with the new combination of PCell and SCell (Figure 8.87). Mobility Optimization Figure 8.86 SCell’s configuration and activation and average DL throughput (example). 343 LTE Optimization Engineering Handbook Low CQI of Scell trigger Scell deactivation Scell deactivation observed by DL MAC transport block Average DL Throughput (Mbps) 344 12 11 10 9 8 7 6 5 4 3 2 1 0 11.5 6.4 5.3 Pcell (CA UE) 4.3 Scell (CA UE) Pcell and Scell (CA UE - Total) Serving Cell (Non CA UE) Figure 8.86 (Continued) Pkt version = 7 RRC Release Number.Major.minor = 10.7.1 Radio Bearer ID = 1, Physical Cell ID = 68 Freq = 5230 SysFrameNum = N/A, SubFrameNum = 0 PDU Number = DL_DCCH Message, Msg Length = 93 SIB Mask in SI = 0x00 Interpreted PDU: Serving Cell Info: PCI = 68 value DL–DCCH–Message :: = { 2. message C1: rrcConnectionReconfiguration: { rrc–TransactionIdentifier 0, criticalExtensions c1 : rrcConnectionReconfiguration–r8 mobilityControlInfo { targetPhysCellId 79, carrierBandwidth { dl–Bandwidth n50 handovertype intraLTE : { sCellToReleaseList–r10 { 1 }, sCellToAddModList–r10 { { sCellIndex–r10 1, cellIdentification–r10 { physCellId–r10 79, },dl–CarrierFreq–r10 2100 Figure 8.87 PCell and SCell handover. Handover Command Handover type: Intra Frequency PCC: PCI = 79 Pcell always changes due to handover In the new Pcell, old Scell is removed and new Scell is configured. RRC reconfig message becomes bigger. SCC configuration during handover SCC: PCI = 79 Mobility Optimization As CA is performed at MAC layer, PDCP and RLC are not aware whether a certain packet will be transmitted in the PCell or in the SCell. All counters for MAC and physical layer are registered on the cell where the data is transmitted. CA expected to increase PDCP throughput in the network since a PDCP packet may be sent using multiple cells in parallel. A CA user using SCell resources, may therefore increase observed PDPC throughput in the user’s PCell but potentially decrease observed PDPC throughput in the SCell. CA‐related KPIs include accessibility, retainability, DL throughput, the number of configured CA UEs, percentage of CA scheduled, percentage of CA traffic, PDCCH utilization, and so on. 8.13 FDD‐TDD Inter‐mode Mobility Optimization LTE supports mobility of TDD/FDD dual band UE during RRC‐connected mode (intra LTE handover) and in idle mode (cell reselection), while this UE is moving between LTE TDD coverage area is adjacent to or overlay with LTE FDD area. FDD‐TDD carrier aggregation combines excellent FDD coverage with large TDD capacity that the way of primary cell on FDD Scell on TDD to boost downlink capacity is many operators’ choice that own spectrums in both LTE modes. Mobility between LTE FDD and TDD will be of increasing importance for operators that have spectrum for both LTE modes, allowing operators to seamlessly offer mobile broadband services on FDD and TDD spectrum, increasing capacity and improving end‐user experience. FDD < ‐ > TDD mobility is very similar to inter‐frequency mobility, including load management. UE Capability of PS handover between FDD and TDD is indicated by FGI (bit 30) (Figure 8.88). Cell selection is priority (LTE FDD high priority) and threshold controlled, when close to antenna the UEs will select high band cell, at cell edge the UEs will select low band cell. Table 8.33 gives the cell reselection parameters in FDD‐TDD deployment areas. For FDD‐TDD connected mode mobility strategy, the eNB takes handover decision based on radio criterion or for load balancing or offload reason. The way to manage handover preference toward a same or different frame structure, that is, TDD‐ > TDD versus TDD‐ > FDD. X2 handover, which is shown in Figure 8.89 is recommended within MME pool area for TDD to FDD handover or FDD to TDD handover (TDD and FDD eNB may be mixed in the same pool area). FDD‐TDD carrier aggregation introduced in 3GPP R12 will be based on existing LTE CA mechanisms. TDD cells are typically deployed on higher bands with reduced coverage. The short duration of uplink in many TDD deployments (3DL:1UL) decreases further the coverage, eight receive antennas can improve TDD cell coverage but there is still significant gap to coverage of FDD cells. FDD‐TDD CA allows to keep PCell on FDD with good uplink coverage and only use the TDD downlink for boosting downlink peak rate. From the link budget comparison D S U D D D S U D D Secondary cell TDD D D D D D D D D D D Primary cell FDD DL aggregation aggregation U U U U U U Figure 8.88 Example of FDD‐TDD carrier aggregation. aggregation U U U U Primary cell FDD UL 345 346 LTE Optimization Engineering Handbook Table 8.33 FDD‐TDD cell reselection parameters. Idle in FDD Start measurement CRP RSRP To FDD 6 Idle in TDD Cell reselection threshold Start measurement Cell reselection CRP RSRP threshold −76 (sIntraSearch:54) N > S + 4 dB 6 Always FDD > −112 To TDD 5 −114 TDD > −112, FDD < −116 5 −76 N > S + 4 dB To 3G 4 −114 3G > −103, FDD < −116 −114 3G > −103, TDD < −116 UE TDD 4 >carrierFreq FDD MME GW >PUCCHConfiDedicated Measurement Report A5 >> tddAckNackFeedback Mode bundling 1 Handover Request >SIB1 Info >>TDD Config 2 Handover Request Acknowledge 3 RRC Connection Reconfiguration Transparent Container dl-CarrierFreq 4 SN Status Transfer >antennaPortsCou nt an2 E-RABs Subject to transfer 5 PDCP Receive Status RRC Connection Reconfiguration Complete UL Count Value 6 DL Count Value Path Switch Request Old eNB X2AP ID New eNB X2AP ID UE Context Release Modify Bearer Request Path Switch Request Ack Figure 8.89 X2‐based handover procedure between TDD and FDD. as shown in Table 8.34, it can be seen that UL cell‐edge throughput is around 300 kbps, and there are around 4dB difference, which result in much more TDD sites required. 8.14 Load Balance 8.14.1 Inter‐Frequency Load Balance As LTE networks are deployed, operators will have multiple combinations of technologies and carriers to serve their subscribers equipped with multi‐mode devices. Operators want to make the best possible use of their spectrum while ensuring the best possible QoS to their subscribers. One way to achieve this is through ensuring proper balancing of the load across its technologies/ carriers. Table 8.34 Link budget comparison. Scenario 2600MHz 20MHz TDD Uplink link budget Data UE output power 23 Resource blocks (RBs) 18 Feeder loss Power per RB Thermal noise 10.5 3 0.301 Gains (antenna UL + DL) Jumper loss Body loss Penetration loss Margins (LNF) Max. pathloss unloaded 900MHz 10MHz FDD 850MHz 10MHz FDD VoIP Data VoIP Data VoIP Data VoIP Data VoIP Data VoIP 2 7 2 8 2 8 2 7 2 7 2 20 14.4 20 14 20 14.1 20 14.3 20 14.4 20 0.012 0.301 0.012 0.33 0.012 0.325 0.012 0.309 0.012 0.305 0.012 −4 −10.8 −4 −10.8 −4 −10.8 −4 −10.8 −4 −10.8 −4 −10.8 −122.4 −129.2 −122.4 −129.2 −122.4 −129.2 −122.4 −129.2 −122.4 −129.2 −122.4 −129.2 3 0 3 0 3 0 3 0 3 0 3 20 0.2 0 18 6.7 128 Utilization 2% Interference margin 0 Max. pathloss 1800MHz 10MHz FDD −174 User bitrate RBS sensitivity 2100MHz 10MHz FDD 0 RBS noise figure SINR 2600MHz 10MHz FDD 128 141.3 0% 13.3 128 132 141.3 131.6 141.3 131.7 141.3 131.9 141.3 131.9 141.3 2% 0% 2% 0% 2% 0% 2% 0% 2% 0% 0 9.4 0 9.8 0 9.7 0 9.5 0 132 132 131.6 131.6 131.7 131.7 131.9 131.9 131.9 Range 0.32 0.41 0.49 0.57 1.12 1.19 ISD 0.48 0.62 0.74 0.86 1.68 1.79 9.4 131.9 348 LTE Optimization Engineering Handbook Inter‐frequency load balancing aims to move connected UEs to another frequency within LTE established by load relations when or before the serving cell becomes loaded. Inter‐frequency load balancing is used to distribute the load across several carriers, traffic will be managed across the two LTE carriers. Load balancing can occur across cells either within the same eNB or in a neighbor eNB, over X2. It is worth to note that load relations can be automatically and dynamically configured of the optimal inter‐frequency. Cell loading information are provided to L3 from L2 in order to calculate DL GBR, DL non‐GBR, and PDCCH load. There are three steps for load balancing as following. Step1: A4 threshold starting the measurement for offload. Estimated DL/UL PRB usage per cell is updated by new admitted requests and periodic modem PRB usage reporting. In addition to the parameters above, the measurement object and Report ConfigID for the inter‐frequency load balancing using A4 threshold mechanism need to be defined in the eNB. For the load information exchange between eNB is over the X2 interface. Load balancing is triggered when current PRB usage exceeds a configurable percentage of total PRBs available in the cell. Step2: UE selection, which means a number of UEs are selected as candidates for load balancing. UE candidates for offloading must support inter‐frequency handover, UE candidates start with lowest priority for offloading. Step3: Execution of offloading to target carrier. MME selects unloaded target carrier for UE measurements, based on X2 resource status reporting information. For equal bandwidth carriers the handover margin can be decided and tuned based on load balancing, for example, for 800M and 1800M layers with 10MHz bandwidth, a3 offsets need to be set asymmetrically, to prevent skewed load. If the carriers have different bandwidths, the throughput difference between layers should be taken into account as well (Figure 8.90). Load balancing can also be triggered when load in source cell (SC) is above the pre‐configured load threshold, and the load difference is larger than the pre‐configured load difference threshold as shown in Figure 8.91. An load balance magnitude to be determined for each target cell (TC), suitable UEs need to be moved (selection/reselection) in order to meet the determined magnitudes. 8.14.2 Inter‐RAT Load Balance The feature of inter‐RAT load balance is applied for reducing the risk of UE trapped in over‐ utilized LTE cells unable to provide acceptable performance and reducing the risk of depleted 3G cells when the ratio of 3G only UE declines. The load balancing features measure the traffic Establish load relations Exchange load info. Load status Select candidates UEs Step1 Step2 Load balance action (handover) New load status Step3 Figure 8.90 Load balance procedure. If load_different> load_different_thresh, load balancing triggered! load_SC load_TC2 Source cell Target cell2 Figure 8.91 Load balance strategy (example). load_different load_TC1 Target cell1 Mobility Optimization F1 LTE cells F1 Double arrows are IFLB relations; single arrows are inter-RAT relations; F2 Frequency F3 3G cells Distance Figure 8.92 IRAT load balance strategy (example). load in each LTE cell with different frequency and 3G cell. LTE cells with inter‐frequency load balance (IFLB) relations exchange traffic load information and also need to monitor own traffic load versus an 3G offload threshold. Figure 8.92 depicts the IRAT load balance strategy. Inter‐frequency load balance attempts to distribute traffic evenly between overlaid LTE inter‐frequency cells that cover the same area. Inter‐RAT offload attempts to offload LTE traffic above an offload threshold to 3G cells that cover the same area. The trigger for the report of measurement results is a B1 event (neighbor becomes better than threshold) for IRAT load balance. The B1 event represents acceptable coverage for offload to the 3G target cell. Reactive load control is triggered when congestion condition is detected during call/bearer admission. Purpose of reactive load control is to move some UEs to less loaded inter‐freq/ inter‐RAT carriers or release some UE/bearers to relieve the congestion condition. Congestion conditions include number of calls exceed limits, number of modem contexts exceed, number of data bearers exceed limits, and number of PRB consumptions exceed limits. 8.14.3 Load Based Idle Mode Mobility Load balancing aims to move connected UEs to another frequency within LTE when or before the serving cell becomes loaded. Cell load is determined by the total UL/DL PRBs consumption. The idle mode load distribution between layers helps reduce the need for the connected mode load balancing. The idle mode distribution is possible to adjust with the thresholds and by that the load distribution. Basic load management should primarily be done through steering of UEs between frequency layers in idle mode. Good idle mode distribution can be achieved by setting a higher cell reselection priority to the higher frequency band, cell‐center UEs will camp on the higher band and cell‐edge UEs will camp on the lower band, distribution can be controlled through parameter threshServingLow. For equal priority based cell reselection, it needs to increase qOffsetCellEUtran for reselection to non‐congested neighbor cells, or decrease qOffsetCellEUtran for reselection to high capacity cell. Priority‐based cell reselection aims to ecourage easier cell reselection in idle mode. If inter‐ frequency carrier/inter‐RAT frequency is with cell reselection priority higher than the serving frequency, then decrease threshXHigh for that frequency relation. If inter‐frequency carrier/ inter‐RAT frequency is with cell reselection priority lower than the serving frequency, then decrease threshXLow for that frequency relation (Figure 8.93). 349 350 LTE Optimization Engineering Handbook RSRP Higher priority Lower priority Lower priority Higher priority RSRP threshold + UE measurement uncertainty UEs camping on blue cell UEs camping on orange cell Cell radius Figure 8.93 Load‐based idle mode mobility. 1 Srxlev > 0 UE starts Inter freq meas (and iRAT) for equal priority cell reselection RSRP Serving Cell 4 Eutran cell is reselected with other freq RSRP equal priority eUTRAN cell tReselectionEUTRAN or tReselectionRAT 3 Qrxlevmin(SIB1) +Qrxlevminoffset(SIB1) +Pcompensation(SIB1) +sNonIntraSearch(SIB3) Automatically adapt SIB3 cell reselection parameters based on serving cell load. Rn > Rs qOffsetCell LTE equal priority cell - criteria not met 2 Qrxlevmin(SIB5) +Qrxlevminoffset(SIB1) +Pcompensation(SIB5) Traditional “S” criteria UE starts ranking cells qHyst time qrxlevminoffset = 0 if same PLMN Figure 8.94 Load‐based adaptation of cell reselection thresholds. Load‐based adaptation of cell reselection thresholds can push cell‐edge UEs down to lower‐ priority layer, as shown in Figure 8.94. This is able to provide the capability to adjust cell selection settings according to cell load, thus helping to balance traffic and load among cells. Load information exchange over X2 interface, allowing the serving eNB to gather load information from the neighboring cells through the X2 interface. Automatically adapt SIB3 cell reselection parameters based on serving cell load, which used to favor idle UEs cell reselection to less loaded neighbour cells, when serving cell is heavily loaded. This feature can be used to impact intra‐frequency, inter‐frequency and/ or IRAT reselection. Here is an example for F1 (high priority) and F2 (low priority) inter‐frequency reselection. Operator can specify the value to use at different load levels for the following parameters: q‐ Hyst, threshServingLow, threshXLow, threshXHigh, s‐nonIntraSearch, s‐nonIntraSearchP, and s‐nonIntraSearchQ, and so on, which are shown in Table 8.35. According to the above parameters settings, downlink/uplink average and edge throughput, the ratios of UE distribution in F1 and F2 band are shown in Table 8.36. Mobility Optimization Table 8.35 Parameters settings for different load levels. Reselection IFHO F2‐ > F1 F1‐ > F2 F2‐ > F1 F1‐band, high priority Load ratio: (F1:F2 = 6:4) threshXHigh = −92 sNonintraSearch = −90, threshServingLow = −93, threshXLow = −125, a2 = −90, a2 = −83, a5−1 = −80, A5−1 = −93, a5−2 = −92 A5−2 = −125 F1‐band, high priority Load ratio: (F1:F2 = 95:5) threshXHigh = −102 a2 = −100, sNonintraSearch = −100, a2 = −68, threshServingLow = −103, a5−1 = −65, A5−1 = −103, a5−2 = −102 A5−2 = −125 threshXLow = −125, F1‐band, high priority Load ratio: (F2:F1 = 99:1) threshXHigh = −105 a2 = −103, sNonintraSearch = −103, a2 = −65, threshServingLow = −106, a5−1 = −62, A5−1 = −106, a5−2 = −105 A5−2 = −125 threshXLow = −125, F2‐band, high priority Load ratio: (F2:F1 = 4:6) sNonintraSearch = −86, threshXHigh = −87 threshServingLow = −88, threshXLow = −102, a2 = −83, a2 = −85, a5−1 = −88, a5−1 = −86, a5−2 = −102 a5−2 = −87 When F1, F2 with equal priority qHyst = 1, qOffsetFreq = 3, a2 = −85, a2 = −83, a5−1 = −86, a5−1 = −88, a5−2 = −88 a5−2 = −86 qHyst = 1, qOffsetFreq = 3, F1‐ > F2 Table 8.36 Performance of different parameters settings. DL/UL avg THP (Mbps) DL/UL edge THP (Mbps) HO IFHO Ratio in F (idle) Ratio in D (idle) F1‐Hi prio −1 (60%; 40%) 18.9/9.3 3.9/0.05 176 61 42% 58% F1‐Hi prio −2 (95%; 5%) 18.6/13.9 4.6/4.9 118 7 11% 89% F1‐Hi prio −3 (99%; 1%) 18.0/14.6 4.7/7.1 106 1 4% 96% F2‐Hi prio −1 (60%; 40%) 22.2/8.8 4.4/3.5 203 101 57% 43% F1, F2 with equal priority 21.7/6.6 5.2/0.05 107 20 90% 10% 8.15 High‐Speed Mobile Optimization The higher the velocity that the UE experiences, the more severe the effect of fast fading that the system suffers. Therefore, it is more difficult to achieve the same performance in a high‐ speed scenario as in a normal speed one. 3GPP has defined high‐speed and performance limits in case of up to 350km/h UE speed. Fast‐moving UEs cause Doppler shifts (frequency offsets) in the received uplink signal as shown in Figure 8.95. Since the UE synchronizes to a Doppler‐ shifted signal in downlink, and in uplink it will be roughly doubled, the Doppler shift is proportional to the velocity of the UE and to the carrier frequency. In this case the eNB uplink signal receiver suffers from a huge Doppler shift, which causes severe performance degradations. 351 352 LTE Optimization Engineering Handbook It is needed to estimate the Doppler spread of the channel, i.e. rate at which the amplitude and phase of the received signal changes. Doppler spread occurs when there are many signal paths with different Doppler shifts, which are added constructively or destructively at the receiver antennas. UE travelling with speed v Maximum Doppler Shift θ Ds/2 Doppler shift + ∆f Minimum Doppler Shift Doppler shift – ∆f Dmin eNodeB A Maximum Doppler Shift Doppler shift + ∆f Ds/2 eNodeB B Perfect frequency sync Frequency offset Railway track Figure 8.95 Doppler shift. Frequency offsets rotate the received signal constellation. If it is too high, the constellation has rotated more than 360 degress and the receiver can’t resolve which frequency offset it is. Scenarios with Doppler shift are mainly line‐of‐sight scenarios where the mobile has a velocity component into the direction of the eNB. Typical scenarios refers to high speed trains where the eNB is mounted near to the tracks or highways where the cars are driving fast and the eNB is located quite near to the highway. The formular of maximum Doppler shifts (fd) and Doppler shift fs(t) are: fd f cu f cd v c 2.57 109 2.69 109 350 / 3.6 1705 Hz, 3.108 fs t f d cos t where f cu , f cd are UL and DL highest carrier frequencies. v is the UE travelling speed, c is the speed of light, and θ(t) is the signal arrival angle shown in Figure 8.96. The formular is that the UE already sees a shifted signal in DL and does frequency synchronization on that and then transmitting in UL the eNB sees roughly twice the shift. Besides, when a UE approaches and passes by the cell antennas will cause a quick change in the frequency offset, that is when a UE moves first toward the eNB and then moves away from it causing positive and negative Doppler shifts respectively. It is difficult to track and compensate for large and fast varying frequency offsets of the UE in this scenario. The effect of the frequency offset is that the subcarriers are no longer orthogonal since the side lobes of the interfering subcarriers no longer has a zero crossing at the main lobe of the Reference symbols Subframe (1 ms) 1 2∆T ≈ 1000 Hz on PUSCH 1750 Hz on PUCCH PUSCH symbols: ∆T = 0.5 ms Figure 8.96 Frequency offset compensation for PUSCH/PUCCH. Mobility Optimization desired subcarrier. High‐frequency offsets cannot be correctly estimated with the frequency offset estimators on PUSCH and PUCCH in LTE. Time difference between reference symbols, ΔT, gives aliasing at frequency larger than 1/2ΔT, which is shown in Figure 8.96. For PUCCH the frequency offset will affect the orthogonality of the orthogonal cover used for format 1/1a/1b (the orthogonality is used to distinguish different users signal using the same cyclic shift). When the frequency offset is severe, these users will disturb with each other. When the frequency offset is very high, users using different orthogonal cover may be impossible to separate, that is, the orthogonal cover will look the same at the receiver. To be able to maintain a connection at high Doppler, the eNB must be sure that the UE transmits the periodic CQI at every occasion. DRX cycle must be set according to CQI period, to ensure the eNB receives CQI periodically. 8.15.1 High‐Speed Mobile Feature In order to estimate and compensate for high frequency offsets, the high‐speed UE feature is introduced in a live network. To mitigate problems in high‐speed deployments eNB shall support a Doppler shift estimation for each UE based on the received signal in different uplink physical channels (PRACH, PUSCH, and PUCCH) and signals (SRS). Those estimates shall be provided to RRM. Some features deal only with enhancing link level performance for high speed users by intoduction of eNB UL receiver improvments (e.g., Doppler shift estimation). For example, antanna tuning‐tilt for high‐speed scenario is calculated below: The tilt of antenna á atan H /D * 360 / 2 * b/2 e _ tilt where H = antenna height – rail height, D = coverage range, b = beam width. For high‐speed mobile UE, the mobility characterization will suffer the issues such as frequent handover, ping‐pong handover, high handover failure rate and drop rate, group handover, and signaling congestion. It is suitable to use frequency diversity mode rather than frequency‐selective scheduling, or transmit diversity rather than spatial multiplexing for a UE at a high speed. Distinguish high‐speed UE via speed estimation by layer1 or layer 3 method, for L1 method, Doppler frequency estimation is used, that will cost 100 ms for estimation. For L3 method, according to v = d/t, UE determines mobility state based on the number of cell changes which occur within a defined period. Judge UE moving direction and set different priorities to forward cell and backward cell as shown in Figure 8.97, which is benefit in accelerating handover and alleviating ping‐pong, for example, CIO = 3 dB for forward cell, CIO = 0 dB for backward cell as below configuration. ●● ●● If UE is moving from west to east:1‐ > 2, 2‐ > 3, 3‐ > 4, CIO = 3; 2‐ > 1, 3‐ > 2, 4‐ > 3, CIO = −3 If UE is moving from east to west:1‐ > 2, 2‐ > 3, 3‐ > 4, CIO = −3; 2‐ > 1, 3‐ > 2, 4‐ > 3, CIO = 3 Besides, railway network plan is usually adpot the “Z” plan shown in Figure 8.97, that can overcome multipath fading. Combined cell shown in Figure 8.98 is another important feature in high‐speed scenario, which configures multiple sector carriers (RRUs) to belong to same cell. All sector carriers are considered as one logical cell, with same PCI, same CRS, and system information as macro cell to fit to light load. Drive test results showing throughput gain when using combined cell compared to separate cells. It is still worth noting that downlink power allocations in terms of reference signal boosting can cause unnecessary large coverage areas and handover execution problems, so typically RS boosting is avoided for the high‐speed scenarios, and uplink power settings with full pathloss compensation while maximizing the signaling robustness is also suggested. 353 354 LTE Optimization Engineering Handbook Cell 4 Cell 3 Cell 2 Cell 1 East West Cell 3 Cell 2 Cell 1 earlier Cell 4 earlier earlier East West Cell 3 Cell 2 Cell 1 earlier earlier Cell 4 earlier Cell B Cell A Cell D Cell C Figure 8.97 UE movement and eNB site planning. 8.15.2 Speed‐Dependent Cell Reselection Cell reselection performance optimization is as similar as handover. The main idea is that speed up the cell reselection in order to have the UEs always under the best cell, ping‐pong cell re‐selections need to be avoided by, for example, speed‐dependent cell re‐selection feature. The speed‐dependent scaling of cell reselection criteria is used to influence the cell reselection criteria for fast‐moving UE. It helps the UE to respond more quickly to cell changes when moving at high speed. A UE may enter three different mobility states: normal mobility, medium Mobility Optimization Separate Vs Combined cell DL Throughput 70 60 Throughput [Mbps] Combined Cell 50 40 30 20 Separate Cells 10 0 0 5 10 15 Time (s) 20 25 30 Separate Vs Combined cell DL Throughput Combined Cell Throughput [Mbps] 20 Separate Cells 10 0 0 5 10 15 Time (s) 20 25 30 Figure 8.98 Combined cell. mobility, and high mobility as shown in Figure 8.99. The medium and high states are specified by the related parameter nCellChangeMedium and nCellChangeHigh resepectively defining number of cell reselections within sliding time window tEvalution [240sec] that determines the UE shall enter mobility states medium or high. High‐mobility state criteria is detected if number of cell reselections during time period tEvaluation exceeds nCellChangeHigh. Medium mobility state is detected if number of cell reselections (n) during time period tEvaluation exceeds nCellChangeMedium and does not exceed nCellChangeHigh. Consecutive cell reselections between two cells are not taken into account: the UE does not count consecutive reselections between the same two cells into mobility state detection criteria if same cell is reselected just after one other reselection. 355 356 LTE Optimization Engineering Handbook NCRItEvaluation nCellChange High nCellChange Medium tHystNormal Enter Enter “Medium” “High” Enter “Medium” Enter “Normal” normal time tEvaluation medium nCellChgHigh > n > nCellChgMed high n > nCellChgHigh Figure 8.99 Mobility states determination of a UE. 8.15.3 PRACH Issues High‐frequency offset impact on PRACH also causes problems. Random access preambles are generated from same root sequence using cyclic shifts, on the receiver, the detector correlates the received signal with the predefined preambles. Frequency offsets will cause correlation peaks for other cyclic shifts of the same sequence and too high‐frequency offsets causes the original peak to be entirely replaced by another peak, thus may cause false detection as shown in Figure 8.100. Fortunately, high‐speed detection can be realized through this PRACH characteristics. If the weighted power sum caused by Doppler shift and spread is greater than a given threshold, the UE is high speed. Impact of high frequency offset f = 0 Hz f = 625 Hz Main peak Additional peak f = 1250 Hz Figure 8.100 Impact on PRACH. Mobility Optimization UE experiences higher frequency offset due to Doppler shift which causes false and aliased peaks. Main peak cyclic shift y Main peak cyclic shift x Detection interval x the additional peaks in a matched filter are handled by extended detection interval Extended detection interval x Detection interval y Additional peak cyclic shift x Cyclic shift y removed in restricted set Figure 8.101 Unrestricted set and restricted set. Table 8.37 Supported cell ranges depending on restricted set of cyclic shift length (delay spread = 5.2us). Ncs Config index Ncs, cyclic shift length 0 15 length of single cyclic shift, μs 9.1 km 1.4 Ncs Config index 8 Ncs, cyclic shift length 68 1 18 12 1.8 9 82 2 22 15.8 2.4 10 100 length of single cyclic shift, μs 59.6 km 9.0 73 11.0 90.2 13.5 3 26 19.6 2.9 11 128 116.9 17.5 4 32 25.3 3.8 12 158 145.5 21.8 5 38 31 4.7 13 202 187.4 28.1 6 46 38.7 5.8 14 237 220.8 33.1 7 55 47.2 7.1 The high‐speed UE feature applies restricted set of cyclic shifts on PRACH (i.e., the RRC parameter highSpeedFlag is set to true), which restricts the number of cyclic shifts that can be used for each sequence to make room for higher‐frequency offsets and allows for extended detection intervals including additional peaks in matched filter output to ensure a properly working PRACH detector at high Doppler shifts. The restricted set of cyclic shifts choose a set of cyclic shifts that allows for secondary peaks, the probability of false detections of the preambles is reduced. In this case, three detection intervals are needed, one for the main peak and one for each of the secondary peaks, compared to the unrestricted set where only one detection interval is needed. This means that at least three times more RACH root sequences are required for a given number of preambles. This increases the number of root sequences needed for PRACH in a cell, but improves the PRACH detector performance and makes it possible to estimate the frequency offset on PRACH. In general, the restricted set can reduce in main peak of matched filter output that does not cause missed detection and additional peaks in matched filter output of receiver will not cause false detection (Figure 8.101). The restricted set preambles are generated by masking some of the cyclic shift positions in order to retain acceptable false alarm rate while maintaining high detection performance for very‐high‐speed UEs. After frequency offset estimation from preamble detector, the result will be sent to PUSCH receiver for compensation. The supported cell ranges for restricted set are shown in Table 8.37. 357 LTE Optimization Engineering Handbook (RACH root sequence) Range: 0...837 Default: 0 (PRACH configuration index) Range: 3...53 Default: 4 Example for SIB2 extract: prach-Config { rootSequenceIndex 264, prach-ConfigInfo { prach-ConfigIndex 3, highSpeedFlag TRUE, zeroCorrelationZoneConfig 9, prach-FreqOffset 4 } }, (PRACH high speed flag) Range: 0 (false), 1 (true) Default: 0 (PRACH cyclic shift) Range: 0...15 Default: 1 (PRACH frequency offset) Range:: 0...94 Default: – Figure 8.102 Example for SIB2. In a live network, UL and DL Doppler shift might not a big problem if eNB have “High‐speed UE” feature, but Doppler spread will degrade UL and DL performance. Cell merge will outperform non‐cell merge because energy overlay gain over bi‐direction Doppler loss. The parameter highSpeedFlag can be got from SIB2 and RRC connection reconfiguration message. Figure 8.102 gives an example of high‐speed PRACH parameters in SIB2. 8.15.4 Solution for Air to Ground Analyze how is LTE standard affected by Doppler shift in case where UE is moving with speed that airplane can reach (relative to the ground), maximum cell size that can be achieved (if currently supported 100km is not enough) and link budget for that solution, interference with services in neighboring frequency bands, and required output power, and so on. 190 Pathloss according to the empirical propagation model (EPM-73) fc = 1 GHz hRBS = 100 m hAirplane = 10000 m 180 fc = 2 GHz fc = 5.8 GHz 170 Path loss (dB) 358 160 fc = 5.8 GHz 150 fc = 2 GHz fc = 1 GHz 140 130 120 110 100 0 50 100 150 200 Distance (km) Figure 8.103 The cell range of air to ground coverage. 250 300 Mobility Optimization Cell size is around 150 to 250 km, which will bring large round trip times (1 ms for 150 km cell radius), and impact on PRACH detector, scheduling, time advance commands, signaling, and so on. For transmission gap in TDD, there needs to be larger than maximum round trip time. High Doppler shift, Downlink (from base station to airplane): fd f cdl v c 1 kHz for f cdl fd 1 GHz Uplink: fd f cul f cdl v c fd 2 kHz for f cul f cdl 1 GHz Path loss and range High altitude for airplanes is the line‐of‐sight propagation and long reflection region. The cell range depends on carrier frequency as shown in Figure 8.103, for 140 dB path loss: 127 km @ 1 GHz, 59 km @ 2 GHz, 22 km @ 5.8 GHz. 359 360 9 Traffic Model of Smartphone and Optimization 9.1 Traffic Model of Smartphone Network traffic is mainly driven by smartphone applications, today’s smartphone applications mostly generate relatively small data bursts. It is important to understand the traffic characteristics for the network optimization purposes. Some years ago traffic was, if not homogeneous, at least a bit easier to predict. For many years, for example, voice has been an important service with very predictable behavior. Today, with the advent of mobile broadband there is really no such service, that is, a service with very predictable behavior that dominates data transfer. Instead, many use mobile broadband all the time to be able to remain connected to their favorite social network. This connection is often done using apps that may be designed without consideration of cellular network characteristics. The nature of app traffic is unpredictable. Also, no two users probably have the same user patterns (update frequency depends on how many friends you have, what they do, OS, etc.). Smartphone’s trigger huge number of small packet in LTE network, always‐on applications require keep‐alive message, frequent transitions between LTE RRC states and causing signaling increase. Numerous applications require that an always‐on mobile‐broadband experience is seamlessly delivered and presented to the end user. Furthermore, many applications may be designed without specific consideration of the characteristics of cellular networks, and consequently may exhibit traffic profiles not well suited to those connections. When attempting to provide such always‐on connectivity at the RAN level, trade‐offs are often encountered between UE power consumption, user experience, data transfer latency, network efficiency, and control plane signaling overhead. Furthermore, the optimum trade‐off point may vary according to application characteristics, or their activity or status. Current trends indicate that the above issues will only increase in significance over the coming years. It is imperative, therefore, that the ability of LTE to efficiently handle and manage such traffic is continually improved. On the other hand, applications characterization drives understanding of demand fundamentals, for example, web page size, packet size, busy hour attempts, avarage call throughput, signaling, and bearer traffic generated by application (Figure 9.1). This control/user plane activity daily profile shows significant differences for different terminal types. Normally, 30% to 60% of all smartphone and M2M terminals are practically “always on,” having user plane activity during all hours of the days. This is due to periodical reporting activity for most M2M terminals and regular background activity for most smartphone terminals. PC terminals are usually not “always on,” but their usage is mostly bound to actual user interaction, for example, web browsing or social networking. Therefore, user plane activity for PC terminals differs significantly from that of smartphones and M2M. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. Traffic Model of Smartphone and Optimization Session Count Session Count CDF Total Load CDF 100% Session Count 80% 60% 40% 20% 1K 10K 0% 100K 1M 10M Session Payload (bytes) 100 90 Control plane 80 Smartphone-centric network 70 Mixed network 60 50 40 30 20 PC-centric network 10 0 0 10 20 30 50 60 40 User plane 70 80 90 100 Avarage call throughput Video Streaming Interactive Video Web Browsing Audio Streaming Email Gaming MMS SMS Busy hour call attempts Figure 9.1 Model reflects variability of mobile application usage. In such traffic model scenario, the maximum number of active users depends on the capacity licenses purchased by an operator. In the case, the maximum number of users is starting to limit user traffic inactivity timer that has mentioned before can be tuned to release inactive users faster. The parameter dictates the time each UE stays connected without downloading or 361 362 LTE Optimization Engineering Handbook Connected Users 90 70 80 60 70 50 60 50 40 40 30 30 20 20 10 10 0 21-01-14 0:00 0 22-01-14 0:00 23-01-14 0:00 Users_connected 24-01-14 0:00 tInactivityTimer RRC Attempts 7000 70 6000 60 5000 50 4000 40 3000 30 2000 20 1000 10 0 21-01-14 0:00 0 22-01-14 0:00 23-01-14 0:00 LTE_02_RrcSetup_Att 24-01-14 0:00 tInactivityTimer Figure 9.2 When inactivity timer decreased, users reduced, RRC attempts increased. uploading data before it is shifted to idle mode. Reducing the value from 61 to 10 seconds can increase signaling load due to increased number of RRC setup attempts but can also reduce the drop rate as less number of users are connected unnecessarily with a chance of dropping the connection as shown in Figure 9.2. This can also reduce congestion in the system due to limitation of connected user’s license in the network. By decreasing inactivity timer, users can be moved earlier from active state to idle state. However, this will have impact on user experience, as there are more frequent session establishments in the case of new data transmission. By changing inactivity timer, an operator can compromised between fast‐release system resources due to inactivity and impact on user experience due delays caused by frequent session establishments. In addition, quality of service (QoS) mechanism controls the service performance, reliability and usability of a mobile service, because different bearer traffic requires different QoS. The following part is about QoS mechanism. 9.1.1 QoS Mechanism The policy control routing function (PCRF) deploys a set of operator‐created business rules that dictate the dservice QoS strategy of the evolved packet system, which is shown in Figure 9.3. Traffic Model of Smartphone and Optimization External IP Networks IP connectivity PGW Gx IP connectivity S5/S8 PCRF SGW QoS parameters: S11 QCI/ GBR/ MBR IP connectivity S1 MME eNodeB IMS Layer MMTEL App SIP P-CSCF Rx Access Layer PCRF Gx 3GPP Bearer Protocols PGW To Request / Grant Access QoS for handling the session media To Request / Get feedback on Loss of Access Qos except handling for Emergency / priority Calls Figure 9.3 LTE QoS mechanism, Rx and Gx interface. The PCRF communicates with the PGW data and push appropriate policy rules across the Gx interface. These QoS requirements can be passed by the PGW to the SGW over the S5/S8 interface. The SGW communicates these requirements to MME over the S11 interface. Finally the MME communicates the QoS requirements to the eNB over the S1‐MME interface. The QoS requirements of the EPS bearer are implemented in the radio network with the UL/DL scheduler and transport network with IP differentiated services code point (IP DSCP) and Ethernet priority bits (Pbits). In LTE network the service QoS is implemented between UE and P‐GW. The QoS framework for UL/DL basically consists of the following building blocks, a radio network, a transport network, the core network, configuration management system, the scheduler for UL/DL data inside the eNB, and the function to control the whole QoS handling. LTE QoS architecture is defined in TS 23.402 and 23.203 standards, UE negotiates its capabilities and expresses its QoS requirements during a SIP session setup or session modification procedure. QoS generally refers to the following parameters, QoS class identifier (QCI). VoLTE requires to support 363 364 LTE Optimization Engineering Handbook UL Service Data Flows EPS Bearer QoS DL Service Data Flows Service Requests - GBR: QCI, ARP, GBR, MBR Packet Filter - Non-GBR: QCI, ARP PCRF Radio Bearer L1/L2 configuration L1/L2 configuration UE-AMBR UE UE-AMBR eNB APN-AMBR PDN GW Gx Figure 9.4 SDF. dedicated service data flows (SDFs1) assigned with desirable QCI for VoIP SIP signaling and RTP packet. In LTE, services are mapped to SDF as shown in Figure 9.4, several SDFs can be treated as an SDF aggregate, all SDFs in an SDF aggregate must have the same QCI/ARP. SDF aggregates mapped to UL/DL EPS bearers which uniquely defines EPS bearers. Traffic flow template (TFT) defines rules so that UE and network knows which IP packet should be sent on particular dedicated bearer. An UL TFT in the UE binds an SDF (or SDF aggregate) to an EPS bearer in the uplink direction, a DL TFT in the SGW binds an SDF (or SDF aggregate) to an EPS bearer in the downlink direction. Every dedicated EPS bearer is associated with a TFT to classify the packet. It usually has rules on the basis of IP packet destination/source or protocol used. The TFT can allow all packages or filter on specific IP address and/or TCP/UDP port, which is shown in Figure 9.5. UE negotiates its capabilities and expresses its QoS requirements during a SIP session setup or session modification procedure, like media type and bit rate, direction of traffic, packet size, packet transport frequency, usage of RTP payload for media, types, and bandwidth adaptation etc. The QoS concept as used in LTE networks is class‐based, where each bearer type is assigned one QoS class identifier (QCI) by the network to ensure proper QoS for bearer traffic in LTE networks. QCI is used to determine packet forwarding treatment, it can be used to mark packets with DSCP. 3GPP has standardized 9 QCI values and mapping to resource type (GBR, non‐GBR), priority, packet delay budget and packet error loss rate (Table 9.1). Besides QCI, allocation and retention priority (ARP) is used to decide whether bearer establishment or modification request can be accepted in case of resource limitations, ARP can also be used to decide which bearer(s) to drop during resource limitations, and has no impact on packet forwarding treatment. The ARP allocated to both the default bearer and the IMS signaling bearer should have a high value this minimize the risk of deactivation of these two bearers that should be always established until the UE disconnects or detaches from the networks. For video telephony, voice bearer has a higher ARP, and video to another bearer has a lower ARP. In a congestion situation, the eNB can then drop the video bearer without affecting the voice bearer. This would improve service continuity. Linked EPS bearer ID (L‐EBI): each dedicated bearer is always linked to one of default bearers. L‐EBI tells dedicated bearer which default bearer it is attached to. (L‐EBI = EBI) The above description is shown in Figure 9.6. Currently a QoS mapping do exist today in both 3G and LTE but is not aligned between the two different radio access technologies. 3GPP TS23.401 has defined mapping between standardized Rel 8 QCIs and pre‐Rel 8 QoS parameter values. 1 SDF: An aggregate set of packet flows associated with an application service. In case of the VoIP service, the SIP based IMS signaling and the RTP based voice media flows together build the service data flow. UE PGW TFT Bearer #1 TFT TFT Bearer #2 TFT TFT Bearer #3 TFT Figure 9.5 TFT. Table 9.1 Standardized QCIs. Packet delay QCI Resource type Priority budget Packet error loss rate Example Services 1 2 100 ms 10−2 Conversational voice 2 4 150 ms 10 Conversational video (Live streaming) 3 5 300 ms GBR −3 10−6 3 50 ms 1 100 ms 10−6 6 7 100 ms 10 7 6 300 ms 8 8 9 9 4 5 Non GBR 10 −3 −3 10−6 Non‐conversational video (Buffered streaming) Real time gaming IMS signaling Voice, video (live streaming), interactive gaming Video (Buffered streaming) TCP‐based (e.g., www, e‐mail, chat, ftp, p2p file sharing, progressive video, etc.) Notes: 1 Packet delay budget (PDB), one-way between UE and gateway. 2 Packet loss rate (PLR) is only air loss counted. 3 VoIP is a GBR service, which means that VoIP users are prioritized against MBB users in the same network. 4 Default bearer is set up with QCI 9 (for non-privileged users) or QCI 8 (for premium users). 5 Guaranteed bit rate (GBR), the minimum guaranteed bit rate per EPS bearer. Specified independently for UL and DL. 6 Maximum bit rate (MBR), the maximum guaranteed bit rate per EPS bearer. Specified independently for UL and DL. 7 APN-aggregate maximum bit rate (A-AMBR), limits the aggregate bit rate that can be expected to be provided across all non-GBR bearers and across all PDN connections of the APN. Specified independently for UL and DL. 8 UE-AMBR limits the aggregate bit rate that can be expected to be provided across all non-GBR bearers among APNs of a UE. Specified independently for UL and DL 366 LTE Optimization Engineering Handbook QoS of EPS bearers Allocation and retention priority (ARP) Qos class identifier (QCI.0–255) Priority Resource type Packet loss rate Packet delay budget Non-GBR Default Bearer QCI 5– 9 A-AMBR UE-AMBR APN IP address ARP Priority level (0–15) (Per bearer) Pre-emption capability (shall not trigger preemption, may trigger pre-emption) (Per bearer) GBR Dedicated Bearer ARP used for call admission control Dedicated Bearer QCI 5–9 A-AMBR UE-AMBR TFT ARP L-EBI Pre-emption vulnerability (not pre-emptable, pre-emptable) (Per bearer) QCI 1–4 AMBR GBR TFT ARP L-EBI Figure 9.6 QoS profile. MME eNB UE INITIAL CONTEXT SETUP REQUEST (E-RAB Level Qos parameters > QCI, ARP) QoS Mapping RNL / TNL Scheduler: Priority / LCG TNL: DSCP rrc Connection Reconfiguration Includes pdcp-SN-Size, UM/AM, snFieldLength, t-Reordering, logicalChannelConfig > priority, and logicalChannelGroup LCG mapping example LCID:1 LCID:2 LCID:x LCID:y LCID:z LCID:w LCID:q SRB1 SRB2 QC11 QC15 QC17 QC18 QC19 LCG0 LCG1 LCG2 LCG3 rrcConnectionReconfigurationComplete INITIAL CONTEXT SETUP RESPONSE Figure 9.7 Uplink QOS mapping. For uplink, the scheduling of the traffics is important to set correct priority per QCI, and also allocate the bearers into logical channel groups (LCG) in such way that the intended scheduling strategy will be used in the UL. LCG mapping in eNB and associated signaling are depicted in Figure 9.7. 9.1.2 Rate Shaping and Traffic Management Wireless networks bandwidth changes constantly and it becomes important to support traffic shaping or bit rate adaptation to guarantee high QoE. Traffic shaping provides a mean to control the volume of traffic being sent into a network in a specified period (bandwidth throttling), or the maximum rate at which the traffic is sent (rate limiting). This control can be accomplished in many ways and for many reasons; however, traffic shaping is always achieved by delaying packets. Rate shaping in EPS is described in 3GPP TS 23.107 as the so‐called token bucket algorithm. The rate shaping function can be seen as a virtual bottleneck that throttles Traffic Model of Smartphone and Optimization Token rate, r Bucket size, b Token bucket Token, representing the allowed data volume Incoming packets R Buffered before sent to next link Incoming E-RAB Requests Admission Control Max Threshold Min Threshold Incoming Rate (Rin) Active Queue Management Queue length > Max: Always drop Min < Queue length < Max: Some drops Queue length > Min: Never drop AMBR/MBR Rate Shaping Rate Shaping according to a token bucket, based on AMBR or MBR (ΣRx ≤ AMBR or R ≤ MBR) Rate after packet shaping (Rshaped) Traffic rate Scheduling Figure 9.8 Rate shaping. the output rate of that buffer to a certain peak burst rate and to a certain average bit rate. The tokens are consumed by the data packets, if there are at least as many tokens in the bucket, the packet is transmitted. This is the nechanism that shaping the aggregate traffic for a user to the aggregate maximum bit‐rate (AMBR). First is admission control for E‐RAB requests. For non‐GBR type E‐RABs admission control is simplified since packets are forwarded on a best‐effort basis. However, GBR‐type E‐RABs requires more elaborate admission control. Request is rejected if the request exceeds available resources. Admission control may consider many different types of resources (shared channel, control channel, transport network resources, memory, etc). After that, active queue management (AQM) is working, which can be viewed as packet‐ level congestion control. It can, for example, drop packets to indicate a congested state to the source at an early stage. GBR queues handled by AQM are not congested, and the scheduler has time enough to serve the queues. Then is data rate shaping concept which function is delaying packets of the data flow so they conform to MBR or UE‐AMBR, using a token bucket as shown in Figure 9.8. 367 368 LTE Optimization Engineering Handbook Traffic Class Signalling Indicator Traffic Handling Priority Quality of Service Class Identifiers Transfer Delay SDU Error Delay GBR GBR MBR / AMBR MBR Allocation Retencion Priority Allocation Retencion Priority .. ... Pre-Rel8 QoS per bearer AMBR Rel8 QoS per bearer Rel8 QoS per UE Set by PDN-GW per bearer Set by HSS per UE QCI (Quality of service class identifier, per bearer); GBR (Guaranteed bit rate, per bearer); MBR (Maximum bit rate, per bearer); ARP (Allocation and retention policy, per bearer); AMBR (Aggregated maximum bit rate, for all bearers for a UE) Figure 9.9 QoS profile. Rate shaping can be seen as an artificial resource limit, resulting in a longer queue if maximum bit rate is violated. The token rate is set to the DL‐AMBR value for DL and to the sum of UL‐ AMBR and all GBR values for UL. The bucket size is selected so that sufficiently large packets are allowed in order to not reduce the efficiency/capacity of the air interface. Standardized 3GPP QoS parameters are realized by using QCI and ARP in LTE and EPC as well as using DiffServ code points (DSCP) in LTE, EPC, and IMS. ARP decides whether new bearer modification or establishment request should be accepted due to resource limitations considering the current resource situation. For GBR bearers the QoS Profile additionally includes a guaranteed bit rate (GBR) and a maximum bit rate (MBR). Further, all Non‐GBR bearers of a UE are limited by an aggregated maximum bit rate, UE AMBR,2 also part of the QoS profile (Figure 9.9). APN‐AMBR shared by all non‐GBR bearers with the same APN, downlink bandwidth management is done in PDN GW and uplink bandwidth management in UE, UE‐AMBR shared by all non‐GBR bearers of the UE. Bearers and the associated establish/modify signaling procedures are required to reserve resources (processing + transmission capacity) before SDFs mapped to that bearer can be accepted into the network. On the backhaul nodes, the network may encounter traffic drops, and poor voice and video quality due to lack of prioritizing and congestion management in the backhaul nodes. To allow traffic separation in transport network, gateways and LTE RAN translate from bearer‐level QoS (QCI) to transport level QoS (DSCP) (Figure 9.10). From the core network the main important parameter for QoS are received, such as the QCI, DSCP, and ARP values. These values are used for QoS translation to find the proper scheduling attributes such as the scheduling strategy, mapping of QCIs to logical channel groups (LCG) for uplink scheduling based on the data as configured by the OSS configuration management system in a so‐called QCI table. DSCPs are used for layer 3 routing and prioritization of different traffic types but also used for mapping to P bit (Priority Bits, the values of which are used for prioritization by transport equipment handling the Ethernet layer) and queues for layer 2 2 The UE AMBR is used in eNB to limit the bit rate of a UE and it is defined by the subscription. The core network uses APN AMBR, which is the aggregated maximum bit rate allowed across all non-GBR bearers associated with the same APN. APN AMBR is not visible in RAN. Traffic Model of Smartphone and Optimization (PCRF) QCI 3GPP QoS Profile Initialize / Upgrade / Downgrade 3GPP QoS Profile 3GPP QoS Profile DSCP DSCP DSCP QCI UL/DL Ethernet Radio pBits (UL) Scheduler Ethernet pBits Ethernet pBits Ethernet Ethernet pBits (DL) pBits (UL) Ethernet pBits eNodeB Transport Transport Packet Core Transport QoS Scheduling Figure 9.10 PCRF controls End‐2‐End QoS including radio and transport. EPC/LTE QoS Profile QCI Mapping function IP datagram ARP Takes place in RBS and AGW (Transport) IP header DSCP Mapping function Ethernet frame MBR/GBR Ethernet header p-bits Data Takes place in devices on edge between L3 and L2 network DSCP Data Figure 9.11 QCI mapping at DSCP and P/bit level. traffic handling on the access transport network, which enabling the transport network to prioritize between different data flows over the S1 interface. DSCP indicates what forwarding treatment, or per‐hop behavior (PHB), shall be applied at each node or router along the path. PHBs can be implemented by employing a range of queue service and/or queue management disciplines on a node’s or router’s output interface queue. The PDN GW and SGW use the QCI information to set the correct DSCP values on the signaling traffic. The DSCP attribute of the QciProfilePredefined defines the mapping between QCI and DSCP. Operator controls QoS by defining QCI’s and related characteristics, mapping services to QoS profiles, packet filters and bearers, mapping QCIs to DSCP (IP layer) and mapping DSCP’s to priority bits (ethernet layer). This corresponds to mapping from RAN to transport and IP network (Figure 9.11). Packet forwarding treatment in transport network is selected based on the DSCP carried in the tunnel header (IP). The recommended QoS mapping is shown in Table 9.2. DSCP is mapped with QCI to realize QoS in LTE network, an example is shown in Figure 9.12. The eNB scheduler is an essential QoS enabler. DL and UL are treated separately. In DL, the priority is simply determined per QCI, traffic can be separated per QCI and a different treatment 369 LTE Optimization Engineering Handbook Table 9.2 Recommended QoS mapping. Traffic Type DSCP Network Synch LU 7 Routing, network control CS6 6 QCI5–IMS Signaling (Non‐GBR) CS5 6 40 S1AP/X2AP‐Inter‐node Signaling CS3 6 24 QCI1–GBR Conversational Voice EF 5 46 QCI3–GBR Real Time Gaming AF41 5 34 QCI2–GBR Conversational Video (Live Streaming) AF42 5 36 QCI4‐GBR Non‐Conversational Video (Buffered Streaming) AF43 5 38 QCI6‐Non‐GBR TCP Specific Services AF31 4 26 OAM Access and Bulk Data CS2 4 16 P‐bits DSCP code 4 CoS 3 CoS 54 1 1 48 2 QCI7‐Non‐GBR Conversational/Speech & Live Streaming AF11 2 10 QCI8‐Non‐GBR TCP’Premium bearer’ AF12 2 12 QCI9‐Non‐GBR TCP Default Bearer AF13 2 14 Guaranteed Bit Rate QCI 1 Voice 3 2 4 3 transport COS1(EF) 2 Unused 3 4 5 Non Guaranteed Bit Rate 370 6 7 8 9 IMS Signaling Video High Priority Data Regular Priority Data Low Priority Data Control(CS4) IMS Core COS2(CS3) COS3(CS2) Default(BE) Default(BE) Figure 9.12 DSCP mapped with QCI. per QCI can be applied. In UL, it is determined per LCG, and based on the currently established dominant bearer of the LCG to ensure that traffic types are separated in the uplink, traffic separation must be ensured between the different data flows within the UE. The QCI of that dominant bearer can be different from UE to UE. The mapping of a radio bearer (or logical channel) to a logical channel group is done at radio bearer setup time by the eNB based on the corresponding QoS attributes of the radio bearers such as QCI. QCIs can be mapped to different LCGs, then buffer status reporting from the UE Traffic Model of Smartphone and Optimization QoS Aware DRBn DRBn DRB1 DRB1 DRBo SRB1/2 is mapped to LCG 0 Ue Scheduler Strict Priority/WFQ per sector eNB uplink UE uplink QoS Aware QoS Aware eNB downlink DRBo Strict Priority/WFQ per Ue Strict Priority/WFQ per sector Figure 9.13 Schematic overview of downlink and uplink scheduling. is done for traffic types separately (UE reports an aggregate buffer status for the combination of radio bearers in a logical channel group. The eNB knows the radio bearers contained in the group and their priorities). The UE prioritizes internally between logical channels according to logical channel priorities (derived from QCI). The parameter controlling the mapping of LCG to QCIs is logicalChannelGroupRef. In the uplink, the UE has four LCGs, which groups together traffic of different logical channels. So a LCG consists of one or more QCIs. For this reason it is important to correctly configure the LCGs. This limitation is applied by 3GPP to keep the buffer status report (BSR) format small in the uplink MAC layer, the buffer status report is sent per LCG and not per logical channel. It also reduces the number of queues required to be supported in the UE side for uplink traffic separation (Figure 9.13). 9.1.3 Traffic Model As we know, smartphones make it easy for people to surf the web and watch online videos, leading to much higher bandwidth use; tablet and notebook devices will send data even higher. iPad‐like devices will chew even more bandwidth than the smartphone because of its larger screen, which is driving bearer and signaling traffic growth. Video, web browsing, and audio streaming will dominate bearer traffic on wireless web, especially user generated content (video, photos, data backup) will stress the uplink. Detailed traffic model incorporating average user behaviors and key application characteristics. For each application, specific characteristics and session attempt rates were estimated. When UE has attached with the LTE network, there are both idle users and connected users. From traffic model perspective, the UE states are connected (UE that is RRC_connected or ECM_connected) and idle (UE that is RRC_idle, ECM_idle). The model for packet data service session consists of one or more subsessions,3 which are the sequences of user activities throughout a particular application. The field based characteristics of each application are defined in order to estimate user plane model, which is shown in Figure 9.14. Traffic profile is mainly defined by: ●● ●● Duration of simulation – define in seconds length of simulation Number of subscribers – number of users with given traffic profile 3 If the dormancy timer is larger than the off-time, then a call may consist of several subsessions. If the dormancy timer is smaller than the off-time, then one call is composed of only one subsession. 371 372 LTE Optimization Engineering Handbook Web browsing BHCA volume Avg per subscriber metric Gaming BHCA volume BHCA p DL throughput q ... r Outcome Streaming Video BHCA duration Handset xx % Device→Appl mapping PC xx % Smartphone xx % on ti ca pli Ap Mix ... Figure 9.14 Traffic model estimated. ●● ●● User type of creation – constant or poisson Non‐real‐time services parameters: number of packet calls, call volume, reading time, number of datagrams in packet call, size of datagram, time between datagrams, and so on. Take an example of inter‐packet arrival time comparison between Youtube and FTP. The mean inter‐packet time of FTP is 0.96 ms, Youtube is 15.9 ms. The maximum inter‐packet time of FTP is 0.89 seconds, Youtube is 1.29 seconds. Around 96.6% of Youtube ethernet packets are 1414 bytes, this corresponds to raw video packet size of 1356 bytes, while the mean FTP packet size is 1402 bytes. 9.2 Smartphone‐Based Optimization Smartphone experience includes response time, always on friendly, and battery life, and so on, which cause challenges to existing LTE networks. The key to good smartphone experience is that UE should maintain active RRC connection; the number of RRC connected users supported by eNB is an important metric for estimating whether eNB is smartphone‐enabled. When UE maintains active RRC connection for longer times, UE battery life and increased handover signaling load needs further consideration. Always‐on applications need to send and receive small packets frequently to keep IP connectivity open—it is called “heart beat” or “keep alive.” Typical frequency is once per minute, or once every few minutes. The amount of data is very small, <<1 kB. 9.3 High‐Traffic Scenario Optimization It is common that networks are configured for low‐load and single‐user performance; however, high‐traffic situations occur daily, and the network must be configured to support all these events at the same time. The higher traffic, the more denser the network is needed, and the more challenges for controlling the inter‐cell interference and physical layer optimization. Figure 9.15 depicts the performance degradation of high‐traffic scenario. Traffic Model of Smartphone and Optimization The served users are reduced UL interference increases with a higher amount of users. The network is starved Which consumes all DL power and leads to power blocking user license limit causes spill over to user Poor CS/PS Accessibility High amount of RRC requests. Leads to high loaded Figure 9.15 The performance of high traffic. With LTE having more and more users, the traffic in some areas have a sharp increase during peak events and then some KPIs such as access rate, scheduling request, and/or CQI resource, PDCCH CCEs utilization in this area changed sharply. LTE high capacity is focused on: ●● ●● ●● ●● ●● Control channel dimensioning (PUCCH/PDCCH resources, such as PUCCH allocation request for CQI/scheduling request resource could not be granted) UL noise rise PRB utilization/cell throughput capacity CPU processor load, RRC connection setup intensity (due to UE releases by inactivity or UE releases by RLC supervision), incoming handover intensity and paging requests, and so on Connected users license 373 374 LTE Optimization Engineering Handbook Typical high‐traffic scene including college scene and large scale entertainment activities (concert/match…). As compared to slightly higher peak traffic events, parameters changes can help improve performance in high‐traffic scenario. After parameters changed depicted in the following chapters, the accessibility success rate can be more than 80% as compared to 50%, retainability success rate can be more than 95% as compared to 85%. DL throughput can be more than 2 Mbps as compared to 1 Mbps. 9.3.1 Resource Configuration When connected users increase, CPU processor, SRS resources, PUCCH resources, or PDCCH resources reach to the intensity limit that will impact the network performance. When high load happened, there will be RRC connection setup rejection, incoming handover preparation request rejection, and drop calls to cause the KPIs degradation. The load management algorithm dynamically sets the permitted intensity limits for RRC connection setups, incoming handover, and pagings based on continuous CPU measurements. When connected users increase, control signaling increases, when control signaling is high enough, data transmissions decrease, due to increased resource usage for non‐data delivery. When data transmission decreases, users stays longer in RRC connected state causing the number of connected users to increase. The number of served users will decrease and non‐served users will consider the radio link to be broken and do reestablish/reconnect causing control signaling to increase (Figure 9.16). Usually the number of users will limit with CPU loading, PUCCH resources and SRS resources, and so on. CPU loading and the number of users are a strong correlation, when CPU load reached 85% (the number of users more than 250 as an example of TD‐LTE, 20 MHz bandwidth), it will result in a RRC request to refuse. PUCCH resource constrained will lead to the RRC establishment failure and insufficient CQI, and SR resources will cause RRC connection reject issue and incoming handover preparation request rejection. When the number of users is more than 250, SRS resource allocation failure may be happened and that will lead to the RRC establishment failure (Figure 9.17). Accessibility success is degraded because PDCCH/PUCCH resources are at the maximum limit. PDCCH is used for UL/DL scheduling assignments, for example, PUSCH/PDSCH resource indication, transport format indication, HARQ information, and PUCCH/PUSCH power control commands. PUCCH is used for transmitting scheduling request, HARQ ACK/ NACK, and CQI. UE is allocated scheduling request and CQI resources during setup procedure, and the resources are kept as long as UE is UL synchronized. UE is not allowed to connect to a cell if there’s no free scheduling request resources. Retainibility and throughput are degraded because of lack of resources and increased interference during the peak event. Cell throughput Connected users Control signaling Capacity limit hit!! “Served users’ Time/load Figure 9.16 Loading versus performance. Traffic Model of Smartphone and Optimization 150 100% 80% 100 60% 40% 50 20% 0% 0 300 30% 200 20% 100 10% 0 CPU loading RRC congestion No. of users No. of users RRC congestion due to PUCCH resource allocation failure 0% 1 2 3 4 5 150 6 7 8 1 100 0.5 50 0 No. of users RRC congestion due to SRS resource allocation failure 0 Figure 9.17 CPU loading, PUCCH resources, and SRS resources with increased users. For UL control channel resources ●● ●● PUCCH: nRB, PUCCH depends on number of scheduling request, HARQ ACK/NACK and CQI resources. PUCCH is allocated by 2 RB at the band edges (RB‐pair), time domain sharing. Each PUCCH is assigned to a UE with a periodicity deciding which subframe UE can access PUCCH (default periodicity for CQI is 80 ms, scheduling request is 10 ms) PRACH: nRB,PRACH in 1 radio frame is independent of bandwidth and is fixed for different cell range For DL control channel resources ●● ●● ●● ●● PDCCH: nRE,PDCCH depends on CFI PCFICH: nRE,PCFICH in one radio frame is 160, independent of bandwidth and # of antenna ports PHICH: nRE,PHICH in one radio frame depends on bandwidth and is fixed for different bandwidth PBCH: nRE,PBCH in one radio frame is 240, independent of bandwidth and # of antenna ports Table 9.3 gives the recommended resource parameters for high‐traffic sites. 9.3.2 Capacity Monitoring In the E‐UTRAN system, resources requiring capacity monitoring are divided into three parts: wireless resources, equipment resources, and transmission resources. This part will focus on the wireless resources. If the evaluation results show capacity overload, corresponding suggestions can be found in Figure 9.18. 375 376 LTE Optimization Engineering Handbook Table 9.3 Resource parameters for high‐traffic sites. parameters Recommend Default Comments noOfPucchSrUsers 200 noOfPucchcqiUsers CommonSrPeriodicity 32 noOfPucchCqiUsers and noOfPucchSrUsers determine the number of resources for CQI and SR 0 32 Aperiodic CQI will be used instead 20 10 pdcchCfiMode 2 0 inactivetimer 30 61 timer T302 10 5 tPollRetrans ul(dl) SRB/DRB 80 50 Active User license Utilization>80% increase PDCCH PRB reduce RRC release, it is currently set to 10 s, lowering this should allow more users to access the network time between two consecutive RRC attempts in case that first attempt is rejected due to congestion reduce RLC signaling by increasing polling interval in UL/DL Y Y 1. Add Active User License N N start DL PRB Utilization>80% END Y PUCCH,SRS,PDCCH Utilization>75% Y Optimize inactivity timer OptimizeConfiguration (e.g.Optimize RB#, sounding period) N 2. Add cells Figure 9.18 Capacity monitoring method. If the active user license utilization is greater than 80%, it needs to expand the license capacity or add a baseband board. If the active user license utilization is greater than 80%, there are a great number of online users and most have no data transmission. In this case, adjust the user inactivity timer base on the operator’s strategies to adjust the number of users in the RRC_ connected state. If the physical resource block (PRB) utilization is greater than 80% and the CPU usage also reaches 80%, it needs to expand the board capacity. If the PRB utilization is greater than 80% and the CPU usage does not reach 80%, perform flow control on cell resources. If PUCCH or SRS resource utilization is greater than 70%, turn on the PUCCH adaption switch or add more cells. Also it is needed to make sure that the eNB can automatically adjusts the number of PDCCH symbols based on the CCE load. If the total number of preambles exceed 75 to 100 times per second in busy hours, adjust the PRACH backoff algorithm or even start the flow control mechanism on the radio interface to smooth the RACH load peak and decrease the average RACH delay. At the same time, considering capacity expansion is required. Traffic Model of Smartphone and Optimization 9.3.3 Special Features and Parameters for High Traffic In case of extreme high capacity, it needs to take all actions to avoid overflow in the system as degrading performance for the end‐users. This special features for high traffic involves many different aspects such as dimensioning, parameter optimization and software features. If number of admitted users becomes higher than the system can handle, this will lead to that the users in the system gets little throughput. The high cell load causes each user already in the system to generate even more load. When number of users increases, the DL data payload drops after reaching a certain level of users. For high‐traffic sectors, the feature of call admission threshold can be changed to more than 90% on all these sectors to allow for more users to connect to LTE, thereby reducing call failure rate. Overload control is the key to ensure a good flow of users in the system. It is needed to control the arrival rate of new users. By just letting all users to access the network will in the end cause the network to stall with consequences. In overload situations it will always be better to serve the users in the network before allowing new users to enter. At high capacity it needs also to ensure that voice users can always access, packet domain users can afford to wait in such congested situations. The features for high traffic are shown in Figure 9.19. The dimensioning perspective is also important. Keep control of interference and maximize the carrier capacity, keep track of your network bottlenecks, and expand capacity in time. Try to be one step ahead when it comes to expand capacity in the radio network. GBR admission control ‐ To ensure that the QoS is maintained for GBR bearers in the system. A headroom is created to allow for fluctuating radio conditions and mobility, also be used to control the ratio between GBR and non‐GBR traffic. Priority paging ‐ At the eNB, the load control and overload protection mechanisms determine if any paging messages need to be discarded. If discards are necessary due to high paging load or high CPU processing load, the paging messages with highest priority are handled first. Load based access barring ‐ Uses barring information from SIB2 broadcast messages to inform the UE of the congestion level in the system. To release inactive UE at high load handover ‐ Automatically alleviates PUCCH resource exhaust at high load, to improve accessibility in high‐load networks, increase efficiency of › Dimensioning – Cater for traffic increase – Monitor capacity bottle necks › Control Arrival rate – Apply admission control – GBR Admission control – MIMO / CA /UL COMP / 4way RX-div – Priority Paging – Load based uplink time alignment timer adjustment – IRAT Offload from LTE › Reduce overhead › Maximizing carrier Efficiency Overload Control – CA based IFLB – Idle mode settings – Admission triggered IFLB – Service/Priority triggered IF HO › Maximize network capabilities – PDCCH Link adaptation – PDCCH Power Boost › Control interference – ICIC / IRC – Frequency selective scheduling – Parameter optimization › Block new access requests – Dynamic Load Control – Release of inactive UE at handover – Load based Access Class Barring Figure 9.19 Features for high traffic. 377 378 LTE Optimization Engineering Handbook Inactivity Timer (10 s) TAT UE SETUP DL TRAFFIC TAC OUT-OF-SYNC PUCCH RESOURCES USED ALLOC PUCCH UE RELEASE NO UL RESOURCES USED EXCEPT RANDOM ACCESS RELEASE PUCCH Figure 9.20 Two timers setting of high load cell. PUCCH scheduling request resource usage. There are two parameters needed to be considered. One is inactivity timer value to 4 sec for high load eNB’s to reduce load on high capacity sites. The other is the time alignment timer (TAT, a timer used for supervision and control of uplink synchronization, range, 0, 500, 750, 1280, 1920, 2560, 5120, 10240[ms], 0 means UEs always in sync), which is needed to set to an operator controlled value for new UEs, it means that these UEs will go out of sync after TAT has expired. The feature has a larger gain in handover intense areas as reducing inter‐cell interference, as fewer UEs are connected at cell border (Figure 9.20). The time alignment timer that will (re‐)start every time that a time alignment command is received. If the time alignment timer expires, the UE concludes that prior to any uplink transmission the random access procedure must be used to obtain uplink sync and all PUCCH resources and any assigned SRS resources will be released. PDCCH link adaptation (LA) could be used in high traffic scene. If no LA is used, 8 CCEs are always required to reach the cell edge. Actually, many users are also likely to require a less robust channel coding on PDCCH than the maximum 8 CCEs. PDCCH link adaptation can increase efficiency by dynamically assigning CCEs based on link quality. So PDCCH LA margin is introduced in LTE, which is added to PDSCH for PDCCH link adaptation, curent value is fixed 10 dB in all radio conditions. As many of the UEs don’t require 8 CCEs, reduction in the PDCCH link adaptation offset reduces the number of medium and good radio condition UEs, which are allocated 8 CCEs. It is recommended to reduce the fixed 10 dB margin to 3 dB. RLC parameters also impact performance in a high‐load cell. The transmitting side of the RLC connection can request a status message through a poll request. Upon expiry of this timer, the transmitting side will consider any PDUs, which have not been positively acknowledged for retransmission. Poll retransmit timer is triggered when the sender doesn’t get an acknowledgement back from the receiver about an earlier poll request regarding reception of successful PDUs. Default value is 50 ms for RLC (DRB) and 45 ms for SRB, it can be changed to 200 ms will give the receiver more time to acknowledge receipt of PDUs, thus lowering the need for retransmission from the sender. The parameter Max_Retx_Threshold determines the number of times a packet is retransmitted at the RLC layer. The objective here is to reduce the number of DL RLC retransmissions (default value is 32, recommend changing is to 16) for high load cell, that is, congested sites where resources can be freed up by reducing this number. Still, UE timers and eNB timers need to be modified in a high‐load cell to improve call failure rate. For example, increase T300 from default 400 ms to 1000 ms, changing T302 value from 4 s Traffic Model of Smartphone and Optimization to 8 s, increase T301 from default value of 400 ms to 1500 ms, changing T304 value from 1 to 2 s, and changing T311 from 5 s to 3 s, and so on. eNB timers are vendor specific, need to be aligned with UE timers. 9.3.4 UL Noise Rise The UL noise rise is the key limiting resource in a scenario with many users. It is the amount of users active in RRC_connected state that cost resources. Each user consumes part of the UL noise rise even though no or very small amount of data is being transmitted. The consequence is that cell shrinks, uplink coverage decreases, and as a result, calls are dropped, accessibility suffers, throughput is starved. UL noise rise can be observed by: ●● ●● ●● Distribution of the SINR values calculated for PUCCH/PUSCH The measured noise and interference power on PUSCH, according to 36.214 of the SINR values calculated for PUSCH The measured noise and interference power on PUCCH, according to 36.214 For a high‐load cell, α factor (i.e., pathloss compensation factor for PUSCH) should be set to default 0.8 to ensure that cell‐edge UEs won’t transmit at higher power (increasing interference), which in turn forces other UEs to transmit with a higher power. p0NominalPucch and p0NominalPusch can be used to adjust power control target, it will trade off between coverage and capacity. Lowing this should help to improve UL noise rise interference. Current setting of the two parameters are as following (Figure 9.21): ●● ●● p0NominalPucch = −116 dBm, higher value can cause high UL interference in high load condition p0NominalPusch = −96 dBm (should be reduced to default −103dBm to reduce UL interference peaks at high load) 9.3.5 Offload Solution and Parameter Settings This part investigates parameter settings to determine if more optimization can be achieved and if some of the settings can be applied market wide to improve high‐capacity cell performance. High PDCCH CCEs utilization In general, PDCCH CCEs utilization above 75% indicates high PDCCH CCEs utilization. High PDCCH CCEs utilization can be caused by large number of active UEs, and/or poor coverage. RRC connection setup failure/incoming handover preparation failure under a high‐ capacity condition usually has high PDCCH utilization behind it. High PDCCH CCEs utilization indicates the SRB traffic is using the majority of the scheduling opportunities and leaves very little opportunities for DRB traffic under high‐load conditions. The eNB is limited on CCEs during low throughput. Scheduling request and/or CQI resource congestion A UE is allowed to connect to a cell if there are free scheduling request and CQI resources. When a UE locates at high‐traffic cell, scheduling request, and/or CQI resource congestion can be caused by large number of users. Even sometimes it was accompanied with retainability degradation, which indicated a drop caused by MME side. The operator needs to configure the number of PUCCH resources for a scheduling request and CQI to control the trade off between the number supported users and the uplink peak throughput. Solution of offloading traffic 379 LTE Optimization Engineering Handbook RSRP vs UE Tx Power/PUSCH Throughput 30 25000 25 A UE Tx Power(dBm) 15 15000 10 10000 5 C 0 UE Tx Power(kbps) 20000 20 B 5000 –5 D –10 2 –6 –6 4 –6 6 –6 8 –7 0 –7 2 –7 4 –7 6 –7 8 –8 0 –8 2 –8 4 –8 6 –8 8 –9 0 –9 2 –9 4 –9 6 –9 –1 8 0 –1 0 0 –1 2 0 –1 4 0 –1 6 0 –1 8 1 –1 0 1 –1 2 1 –1 4 1 –1 6 1 –1 8 2 –1 0 2 –1 2 2 –1 4 26 0 RSRP (dBm) C Total UE Tx Power/RB (pZeroNominalPusch = –96dBm) D Total UE Tx Power/RB (pZeroNominalPusch = –103dBm) A PUSCH Throughput (pZeroNominalPusch = –96dBm) B PUSCH Throughput (pZeroNominalPusch = –103dBm) –95 P0 = –95 P0 = –99 P0 = –103 P0 = –107 P0 = –111 P0 = –115 P0 = –119 –100 –105 IRB,UL [dBm] 380 –110 –115 –120 –125 –130 110 115 120 125 130 135 140 145 150 Lsa,cellrange [dB] Figure 9.21 p0 optimization in a high‐load cell. Besides adding carrier/site to balance the high load, the solution of offloading traffic is as efficient way that is summarized below: ●● By mobility allignment: It can be done by encouraging easier cell reselection in idle mode, aims to increase qOffsetCellEUtran for reselection to non‐congested neighbor cells for equal cell reselection priority, decreases qOffsetCellEUtran for reselection to high‐capacity cell, for priority‐based cell reselection, adjusts parameter threshXHigh and threshXLow, and, encouraging early handover in connected mode, aims to increase cellIndividualOffsetEUtran for a certain EUtranRelation to push some traffic to neighboring non‐congested cells. Traffic Model of Smartphone and Optimization ●● ●● By decrease cell size: RF adjustment to decrease cell coverage, downtilt to reduce coverage footprint on the high‐capacity sector, while may require to up tilt the neighboring non‐congested sectors. Also it can increase cell size defining parameters: Qrxlevmin or, changed Delta PSD for cell specific reference signal relative to the reference from 3 dB to 0 dB to reduce cell size under low inter‐site distance. For further, it can decreases p0NominalPusch and p0NominalPucch in a high‐load condition. By alleviate PDCCH scheduling congestion: Reduce SRB traffic aims to reduce the number of UL/DL RLC retransmissions for SRB and DRB, or, to extends the time for new poll if no RLC status report is received for SRB and DRB. 381 383 Part 3 Voice Optimization of LTE 385 10 Circuit Switched Fallback Optimization Circuit switched fallback (CSFB) addresses the requirements of the first phase of the evolution of mobile voice services, which began on a commercial scale in 2011. CSFB is the solution to the reality of mixed networks today and throughout the transition to ubiquitous all‐LTE networks in the future phases of LTE voice evolution. On the other hand, CSFB is the first step in enabling mainstream LTE handsets with the cost, size, and battery life advantages of single‐ radio solutions to LTE data in combination with 2/3G voice as well as making the initial LTE investments smaller. When the user’s device is paged via LTE with an incoming call, or when the user initiates an outgoing call, the device switches from LTE to 2/3G. This is a circuit switched (CS) fallback function in EPS, which enables the provisioning of the CS services when the UE is served by E‐UTRAN without the need for IP multimedia subsystem (IMS) support. A CS fallback– enabled terminal connected to LTE can use 2/3G to connect to the CS domain in areas where LTE coverage is overlapped with 2/3G coverage, and the call remains in the CS domain until it is completed. Compared with VoIP and standby phones, CSFB is easier to realize, that means we don’t need to do much to upgrade the network, and CSFB can handle various kinds of phones. CSFB may be used as a generic telephony fallback method securing functionality for incoming roamers as well. 10.1 Voice Evolution This voice communication evolution can be characterized into three major phases. In the first phase, currently underway, all voice traffic is handled by legacy CS networks, while data traffic is handled by LTE packet‐switched (PS) networks, 2/3G network used primarily for voice service of legacy terminals. Usually, the operator may deploy LTE with 3GPP CSFB solution across LTE coverage area. In CSFB, whenever an LTE handset generates or receives a voice call it is automatically transferred to the 2/3G networks. Once the call is finished the device reverts to LTE. It relies on interrupting the LTE connection when the terminal is forced to move to the 2/3G network. This might be a big problem, depending on the application that is being used prior to the voice call. The second phase in LTE voice evolution introduces native VoIP on LTE (VoLTE) along with enhanced IP multimedia services such as video telephony, this solution relies on SRVCC (single radio voice call continuity) or PSHO at edge of coverage. Compare to CSFB, VoLTE avoids 4G data service interruptions and preserves the LTE data experience during speech communications. The third phase converges the enhanced capacity and services of all‐IP networks (voice and video over IP and RCS, such as instant messaging, video share, and enhanced/shared phonebooks) for continuous coverage across the broader range of network LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. 386 LTE Optimization Engineering Handbook 2/3G coverage LTE CSFB LTE LTE LTE SRVCC LTE LTE Legacy CS voice CSFB LTE CS CS LTE E LTE LTE LTE E LTE LTE LTE LTE E LTE LTE LTE LTE CS voice service via CSFB IMS Voice service over LTE and CS access LTE Legacy phone CSFB phone CS attach MSS Legacy architecture VoLTE phone CSFB CS attach MSS CSFB Architecture Mainly continuous LTE coverage Areas with continuous LTE coverage Spotty LTE coverage SRVCC SRVCC AKS IMS CSFB CS attach MSS SRVCC / ICS Architecture LTE E LTE E LTE LTE LTE LTE E LTE LTE E LTE LTE LTE LTE LTE LTE LTE E LTE LTE E LTE 2G - 3G LTE LTE E LTE LTE LTE LTE LTE E LTE LTE LTE LTE LTE LTE LTE LTE LTE E LTE LTE E LTE LTE IMS Voice service over LTE and HSPA VoLTE phone LTE HSPA IMS Mobile telephony evolution target Figure 10.1 Voice evolution and network evolution. access methods, including LTE, 3G, and WiFi, with interoperability across operators and legacy telephony domains (Figure 10.1). Migration from CSFB to VoLTE is typically done in several steps starting from CSFB, then combination of CSFB and VoLTE, and finally, full VoLTE where all calls are fully utilizing VoLTE. SRVCC enables handover from VoLTE to CS speech when the UE is running out of LTE coverage. Also reverse SRVCC enables handover from CS speech back to IMS‐based VoIP. The handling of voice traffic on LTE handsets is evolving, as the mobile industry infrastructure evolves toward higher, and eventually ubiquitous LTE availability. When subscribers are attached to LTE access, the UE informs the network what it supports, either data centric, or voice centric. If the UE is voice centric, it shall report what it supports, CS voice only (CSFB), IMS PS voice only (voLTE), CS voice preferred (IMS PS voice as secondary), or IMS PS voice preferred (CS voice as secondary). CS fallback is triggered to overlapping CS domain (2/3G) whenever voice service is requested, and resumed LTE access for PS services after call completion. 10.2 CSFB Network Architecture and Configuration 10.2.1 CSFB Architecture The key point from CSFB view is the SGs interface between MSC and MME that SGs interface connects the MSC/VLR and the serving GPRS support node and mobility management entity (SGSN‐MME). This interface is used for registration in the MSC/VLR of the UE by performing combined procedures, to page the UE on behalf of the MSC/VLR and to convey CS‐related services. CSFB‐evolved node functions require a software upgrade and are listed in Figure 10.2. Following functions shown in Table 10.1 need to be enabled in MSC‐BSC (RNC) and SGSN‐ MME for CSFB. RIM functionality supports NACC/system information distribution for CSFB from LTE to 2/3G. RIM is recommended to pre‐populate LTE eNB with 2/3G SIB data. RIM procedures setup associations between external UTRAN or GERAN cells and the BSS that hosts that cell and receive NACC information for the cell. This process occurs automatically, with RNC/BSC updating eNB of changes to subscribed cell relations. It is required to enable feature “Release with SI” toward 2/3G, which speeds up call setup time toward 2/3G. Circuit Switched Fallback Optimization 1. RIM (O) 2. DMCR (O) 3. Iu-flex/MSC Pool support (O) 4. PS Handover (O) 1. RIM (O) 2. Fast return to LTE (O) 3. A-flex/MSC Pool support (O) 4. PS Handover (O) 5. Dual Transfer Mode (O) UTRAN 1. RIM (O) 2. PS Handover (O) SGSN IuPS Gs Gb Uu GERAN S3 IuCS A MSC-S Um 1. LTE to CS Fallback (M) 2. SMS over SGs (M) 3. MSC Pool (O) 4. MT Roaming Forward (O) 5. Dual Transfer Mode in GSM (O) SGs UE E-UTRAN LTE-Uu 1. CS Fallback (M) 2. DMCR (O) 3. Dual Transfer Mode (O) MME S1 1. CS Fallback to GSM and WCDMA (M) 2. RIM (O) 1. CS Fallback (M) 2. SMS over SGs (M) 3. RIM (O) 4. MSC Pool support (O) M: mandate, O: optional Figure 10.2 CSFB architecture and functions. Table 10.1 MSC‐BSC (RNC) and SGSN‐MME functions. MSC‐BSC SGSN‐MME Feature provide and LKF installation on APG SGs feature activation Check SGs interface parameters SGs setup IS part SGs interface setup and activation SMSoSGs and CSFB requirements: RAN information management (RIM) feature activation RIM support enables SI transfer between RATs SMS over SGs and CSFB to GSM and WCDMA feature activation IP‐based interface configuration SCTP‐based interface configuration TA‐LA to MSC‐S BC mapping SGsAP configuration The support for RIM procedures show in Figure 10.3 that the feature makes it possible for an eNB to receive information for a specific GERAN/UTRAN cell or to create and maintain relationships with GERAN/UTRAN cells. Such relationships are referred to as RIM associations. Once a RIM association is established in the eNB by the operator, the eNB receives updates from the BSC/RNC when SI for the GERAN/UTRAN cell in that RIM association changes. When moving from CSFB region to non‐CSFB region, “data centric” LTE data modems should stay on LTE, “voice centric” LTE smartphones should re‐select to 2/3G (Figure 10.4). CSFB‐configured region is based on tracking area; this CSFB feature can be enabled/disabled per the tracking area. When moving from CSFB region to non‐CSFB region, the tracking area update is triggered, but tracking area update accept contains a tracking area update but does not provide location area information. It is worth to note that UE ID in 2/3G and LTE is different, as shown in Figure 10.5. 10.2.2 Combined Register A dual‐mode UE can attach both to the LTE network, and to 2/3G networks before voice calls via CSFB can be initiated. The UE with CSFB ability not only can access to EURTAN from EPS, but also can access to CS domain from GERAN/UTRAN. 387 388 LTE Optimization Engineering Handbook eNodeB MME SGSN BSC 1. eNB Direct Information Transfer 2. RAN Information Relay 3. RIM PDU 4. RIM PDU 5. RAN Information Relay 6. MME Direct Information Transfer 1. Establish RIM Association through the CN. SGSN MME 2. Report SI changes for Utran cell belonging to RIM association, to the eNB over the lu. RNC E-NodeB UtranCell = N RIM Association ExternalUtranCell = N eNB has the latest system information for the target Utrancell. Figure 10.3 RIM procedure and association. Cs-domain Registration: FAIL Voice centric LTE smartphones should re-select to 2G/3G in non-CSFB LTE TDD LTE TDD TD- SCDMA TD- SCDMA GSM GSM CS-domain Registration: OK CS-domain Registration: OK Figure 10.4 CSFB region (left) and non‐CSFB region (right). M-TMSI (Subscriber confidentiality) SGSN MME P-TMSI EPC RNTI (Scheduling) MSC eNB TMSI IP Address (Packet Forwarding) Figure 10.5 UE ID in 2/3G and LTE. S-GW P-GW Circuit Switched Fallback Optimization SGsAP SGsAP SCTP IP SCTP IP L2 L2 L1 a new IE, mobile class mark 5. MSC performs location update. BSC/ RNC MSC/ VLR 2G/3G Circuit Core G-MSC L1 MSC Server/VLR 4 and 6, Location Update SGs Request/Response 1. Attach request (Type = IMSI/EPS Attach, Additional Info = ”SMS Only or Both”) MME 7. Attach Complete 2. MME performs authentication & bearer establishment. 3. MME derives VLR number for LA update. MME handles the combined PS and CS Attach. For the CS attach it maps TA -> LA and makes a location update over SGs Figure 10.6 Combined register. Table 10.2 SGs procedures and messages. SGs Procedure Used SGsAP messages Location update Location update accept, Location update reject, TMSI reallocation complete EPS detach EPS detach indication, EPS detach ACK. IMSI detach IMSI detach indication, IMSI detach ACK. Paging Paging request, Service request, Paging reject and UE unreachable Alert Alert request, Alert ACK, Alert reject, Activity indication UE activity indication Reset Reset indication and reset ACK SMS Uplink unitdata, Downlink unitdata and release request The SGs‐interface connects the SAE domain with the classic CS domain with MSCs and 2/3G radio network. The SAE domain consists of the EPS with the MME node and the LTE radio network. The SAE domain is completely packet‐based so the only way to provide CS services to SAE is via the SGs interface. The SGs interface use a SCTP (stream control transmission protocol) connection between the MME and the MSC/VLR. It is needed to note that all new interfaces in the SAE architecture begins with an “S.” The SGs‐interface is an “expansion” of the existing Gs interface between the MSC and SGSN. The SGs‐related procedures and messages are shown in Figure 10.6 and Table 10.2. In CSFB deployment, LTE (TA, tracking area)‐2/3G (LA, location area), and LA‐MSC should be mapped correctly in MME. The TA and LA are mapped by connecting them to the same geographical area in the SGSN‐MME. By this mapping, MME can determine the corresponding MSC, and initiate MSC location–combined update request for HSS/HLR. The combined procedures enable a UE supporting both CS and PS services, to connect to both types of services through the EPS network. The UE requests the MME for combined procedures in an attach request or tracking area update (TAU) request message. The MME establishes initial registration of UE, and maintains UE location information updated with the MSC/VLR. The SGSN‐MME sends information regarding the UE’s location to the MSC/VLR. The MSC/ VLR does not have any information about tracking areas, so the SGSN‐MME has to send location information based on the location area in which the UE is located by TA‐LA mapping. 389 390 LTE Optimization Engineering Handbook Figure 10.7 “Combined EPS/IMSI attach” in Attach request message. When a UE send Attach request message to MME, the MME can get a CSFB indicator. The attach type is “combined EPS/IMSI attach” in Attach request message, which is shown in Figure 10.7. In attach request message, the following information can be got: EPS attach type, CSFB capability, and UE’s usage setting. Then the combined (TA/LA) update will be done; it will register with the CS domain in a proper MSC. A mapping between LTE TAs and 2/3G LAs that enables both SGSN‐MME and MSC/VLR to keep track of UE’s location. Whenever the UE moves and changes TA within the LTE network, its position is kept updated in the MSC/VLR according to the TA/LA mapping configured. Then MME will send Attach accept message to UE with the TMSI and LAI, which is assigned by MSC. MME configures the TA‐LA and LA‐MSC mapping relation, determines the MSC, and then MME will do combined registration with MSC, that is, they will trigger MSC register in HLR. After the MME derives the LAI from the TAI of the current LTE cell, then MME sends the “location update request” to the target MSC/VLR. If EMM causes “network failure,” it is found in the Attach accept message, and it means the combined attach was “successful” and is for EPS services only (Figure 10.8). MSC sends the UE a TMSI reconfiguration command by location update accept message, after the TMSI reconfiguration, UE send TMSI reollocation complete message to the core network, indicates that the TMSI reconfiguration has been completed. Once the location update is accepted by MSC, it will create/update the SGs association and mark the subscriber as IMSI and EPS attached. The MSC will then perform the location update procedures toward the HLR in the same way as it is done for an A‐ or Iu‐interface. In attach accept message, the following information TAC, M‐TMSI and LAC can be got. By receiving this message, the UE shall start T3411 and T3402 (refer to Annex). If either of these two timers expires, UE will re‐trigger combined TA/LA updating with IMSI attach. If the UE is not successful for the non‐EPS part LA updating, UE will try five times each time, since after five times, T3402 will start, wait for another 12 mins, and then another five times will initiate. When UE performs detach procedure, it can either be received as IMSI detach, EPS detach, or combined EPS/IMSI detach. The detach procedure can be initiated by UE, MME, or HSS. Depending on detach type, the UE is marked as EPS detached or IMSI detached or both. This procedure is initiated by EPC (MME) by sending an EPS detach indication or IMSI detach indication to the MSC. The MSC removes the SGs association to the EPC (MME). Figure 10.8 Attach request and attach accept message. 392 LTE Optimization Engineering Handbook 10.2.3 CSFB Call Procedure LTE CSFB call procedure can be divided into five parts: LTE attach request, CSFB process, mobile originating call, CS radio bearer session, alerting, and PS radio bearer data transfer. 10.2.3.1 Fallback Options After CSFB has been deployed, eNB supports one of the following methods to fallback to the 2/3G network to make a CS voice call. Depending on feature activation, output from RRM handover algorithm and UE capabilities, the eNB decides which type of CS fallback will be used—RRC release with redirect, PS handover, or (in case of 2G target RAT) inter‐RAT cell change order, and so on (Figure 10.9). ●● ●● ●● ●● Redirect based CSFB to 2/3G – CSFB via redirect allows for a service‐based redirection from LTE to UTRAN or to GERAN during the call setup. The operator can configure priorities for the target frequency bands per cell. eNB selects the highest priority layer supported by UE. UE does not have any pre‐information about the target cell. It is needed to configure the carrier frequency of the target RAT by RRC connection release message, then, UE access to target 2/3G cell (can be based on field measurements or blind). The baseline method is a blind redirect to target RAT. With blind redirect time, the transition from LTE to 2/3G for a non‐stationary UE is around 2 seconds. Redirect based CSFB to 2/3G with SIB (System information block) – Compare to the upper one, the UE does not have to re‐read the whole system information from air interface. RAN information management (RIM) procedure may be deployed for retrieving system information from 2/3G, and it is possible to provide information about the target cell prior to releasing from LTE. The system information provided together with the RRC connection release message will reduce the overall call setup time. Redirect based CSFB to 3G with deferred SIB11/12 reading (deferred measurement control reading (DMCR)) – With this feature UE defers reading system information blocks type 11, 11bis and 12, which may be distributed to several segments depending on the size of the neighbor list in the target 3G cell before RACH access. The gain in setup time depends on SIB contents and scheduling, but several hundred milliseconds are practical by deferring reading SIB the CSFB call setup delay is reduced. Field tests indicate that DMCR reduces the SIB acquisition time from 2 seconds to about 1 second. The information whether the 3G cell supports DMCR is broadcast in SIB3, which the UE will have to acquire before deciding if SIB11/12 reading can be skipped. PSHO based CSFB to 3G—eNB starts upon the CSFB trigger IRAT measurements to identify suitable target cells. eNB performs a PS handover once a suitable UTRAN target cell has been found based on measurements during the handover preparation phase, using handover instead of release with redirect will give a shorter interruption in the ongoing PS data session and target cell SI provided to UE during the handover procedure. Resources are reserved beforehand in target radio and call setup delay is reduced and reliability of call setup is increased. The feature of a PSHO–based CSFB needs a 3G network to be upgraded to the right software. RIM start a CSFB Call without RIM Figure 10.9 RIM. R9 RRC redirect with 3G frequency PSC,SI R8 RRC redirect with 3G frequency measure 3G cells which will fall back and synchronize measure 3G cells which will fall back and synchronize camping and call set up read SI for target cells camping and call set up Circuit Switched Fallback Optimization 10.2.3.2 RRC Release with Redirection In a live network, the fallback method of RRC release with redirection is widely used so that it can be simply deployed. The available variations for CSFB to 2/3G RRC release with redirection are the following: ●● ●● ●● Release with redirection—Basic Release with redirection—SIB skipping Enhanced with redirection—SI tunneling (3GPP Rel 9) The call flows for all three variants are similar and the distinction lies in how the SIB messages are handled. Reading of SIBs in the target cell once the UE has moved to the fallback RAN is a large contributor to the delay budget. The first variant reads all SIBs and thus causes the longest delay. The second reads only the mandatory SIBs—1, 3, 5, and 7—and the remaining SIBs are provided once the UE is connected to the target cell. The third variant reads no SIBs at all and thus has the shortest delay. All SIB information is tunneled via the core from source to target RAN. In Rel‐8, the CSFB release with redirection, at a voice call origination attempt or when receiving a page for CS voice (via SGs interface), the fallback RAT’s frequencies is transmitted by RRC connection release message. The UE changes RAT and starts accessing the new RAT attaching to the indicated target frequency, then transmits a CM service request to the BSC/RNC. Once the CS service request is accepted, the terminal will start the CS call. Rel‐8 CSFB, the release with the redirection procedure is shown in Figure 10.10. For Rel 8 redirection, eNB only carries the frequencies of the target cell, and the UE needs to select a frequency and read the SIB. For Rel 9 redirection (SI tunneling) shown in Figure 10.11, the system information can be regularly updated via RIM signaling and cached at the source eNB, it avoids the delay of configuring and performing target RAT measurements. The main risk is if the provided information is misaligned or outdated, in which case the UE is forced to make the call setup as if there were no system information available. This feature will decrease the amount of time required to setup a CS call since it will not be necessary for the UE to download all SIB containers. UE initiated a CFSB call UE eNB MME Optionally: DL NAS-PDU Transfer „CS SERVICE NOTIFICATION“ (in case of MTC) CSFB release with redirection RRC: ULInformationTransfer (EXTENDED SERVICE REQUEST) S1AP: UL NAS TRANSPORT S1AP: UE CONTEXT MODIFICATION REQUEST (CS Fallback Indicator) UE measures the target frequency and sync to the new RAT Selection of redirect target S1AP: UE CONTEXT MODIFICATION RESPONSE RRC: RRCConnectionRelease L2ACK (With Redirect Info) UE read target RAT’s SIB S1AP: UE CONTEXT RELEASE REQUEST S1AP: UE CONTEXT RELEASE COMMAND UE leaves cell UE access the new RAT and make a call Figure 10.10 Rel 8 redirection procedure. S1AP: UE CONTEXT RELEASE COMPLETE Delete Bearer CSFB is triggered in the eNB through initial context setup request and UE context modification request The MME sends an S1-AP UE context modification request with a CSFB indicator to the eNB 393 394 LTE Optimization Engineering Handbook 1. eNB registers information (id and signaling strength/quality) about neighboring cells from measurement reports from the UE’s. 2. eNB requests system information (SI) for the overlapping 2/3G cells via the packet core network. 3. SI request/response relayed by the MME MM E S1 eNB LTE S3 6. eNB provides SI to UE at CSFB. BS NodeB 2 / 3G 7. UE do not need to read SI in 2/3G => Faster call setup! A-bis Iub BSC RNC Gb IuPs SGSN - 4. SI request/response Gb IuPS relayed by the SGSN 5. BSC/RNC provides SI based on regular intervals (requests) or when ever the configuration is changed. Figure 10.11 RAN information management (RIM). Optional enhancements for 3G, the method of RRC release with redirection (R8) + DMCR (deferred measurement control reading function), is also supported shown in Figure 10.12. UE awaits reading SIB11 and 12 with instructions from the target cell (SIB3 support deferred measurement control reading configuration) and processes other time‐consuming SIBs (related to measurement control). 1. UE falls back to 3G S1 eNB MME LTE SGs 2. UE reads SIB3 = DMCR on/off BS Iub RNC 3G 3a. DMCR on = UE awaits with reading SIB 11, 11bis and SIB 12 => Faster call setup! SIB 11, 11bis and 12 are read later during the call Figure 10.12 Deferred measurement control. 3b. DMCR off = UE reads SIB 11, 11bis and 12 at their regular broadcast intervals (every 1.4 s). Call setup continues after SIB’s are read. Iu MSC Circuit Switched Fallback Optimization CSFB WITH LOAD SHARING Calls redirected to frequencles In Round Robin method F1 F2 F3 ARE OF SAME PRIORITY LTE F2 F3 Calls F1 WCDMA GSM F1 F2 F3 Figure 10.13 Fallback frequency selection. CS fallback will always direct the UE to whichever frequency relation has the highest priority. If the same priority is configured for multiple frequencies, the eNB will redirect the UE based on a simple round robin algorithm for the load sharing, which is shown in Figure 10.13. 10.2.3.3 CSFB Call Procedure The UE sends an NAS extended service request message to the MME requesting a CSFB service when the UE requests the CS service or as a response to a paging message with a CSFB flag. CSFB is triggered in the eNB through initial context setup request and UE context modification request message sent by the MME. For mobile originating (MO) voice calls, the UE sends Extended service request message indicating mobile originating voice call from a CSFB terminal. The eNB triggers release with redirect, redirecting the UE to the CS domain to continue with the call setup. The eNB makes the decision to what neighbor cell the UE should be released to and sends an RRC_release_ with_redirect message to the UE and a S1‐AP UE context release request to the MME. The UE changes RAT and starts accessing the new RAT attaching to the indicated target frequency. Finally, the UE transmits a CS service request to the BSC/RNC or 2/3G network. Once the CS service request is accepted, the UE will start the CS call. The PS services are moved to target the network if possible. MO CSFB call procedure is shown in Figure 10.14. The call flows for mobile (MT) terminating call setup is presented in Figure 10.15. For terminated CS call in LTE, the MSC pages the UE over LTE via the SGs interface, the UE changes radio mode to 2/3G and receives the call. When MSC/VLR pages users, MSC needs to configure the paging parameters of SGs interface and 2/3G network, including paging times, paging interval, TMSI/IMSI paging, whether to continue A/Iu interface paging or SGs and A/Iu interface simultaneously paging after SGs paging failure, and so on. Before MSC/VLR pages the UE, it will check the UE’s SGs association state. If the SGs association is null, paging UE in 2/3G network in accordance with its storage location information, if SGs association is associated, then paging UE by SGs interface, and starts the SGs interface paging timer. SGs paging message include IMSI, LAI and TMSI etc. When MME received paging message from MSC through SGs, it needs to check whether the UE is attached or detached, the SGs interface is associated or not, and the LA/TA matches or not, and so on. If the UE in the LTE network is detached, MME will response paging reject (IMSI detached). If the UE is in RRC connection state, MME directly sends the CS service notification (IMSI or TMSI) to the UE. If the UE is idle, MME sends paging (S‐TMSI) message to 395 eNB UE BSS/RNS MME MSC SGW/PGW SGSN Extended Service Request S1 AP-Request message with CSFB indicator S1 AP-Response message Optional Measurement Report Solicitation RRC connection release , S1 AP-: S1 UE Context Release Request S1 UE Context Release UE changes RAT then LA Update or Combined RA/LA Update or RA Update or LAU and RAU Suspend Suspend Request / Response 8Update bearer(s) CM Service Request A/Iu-cs message (with CM Service Request) Service Reject Location Area Update CS MO call Routing Area Update or Combined RA/LA Update Figure 10.14 MO CSFB call procedure. If the MSC is changed Circuit Switched Fallback Optimization The UE sends a CM service request to the BSC/RNC. Once the request is accepted, the UE starts the CS session Before call proceeding, only MO UE need to analyze. After that, the core network starts paging MT UE, MO and MT UE need to combind analyze. Figure 10.15 MO and MT CSFB call procedure. UE under its TA/TA list for CS and SMS service. When UE received the paging message, MME will send paging response (service request message) through SGS interface to MSC. Finally, the UE informs the MME that it needs to set up the CS call and it needs the CSFB, via the NAS Extended service request message. The UE will be ordered to release the LTE connection it will be redirected to the CS domain. 10.2.4 Mismatch Between TA and LA For CSFB and session continuity, it needs to have a proper TAC‐LAC mapping where the UE on 4G will be redirected to the underlying 2/3G site as MME needs to know to which MSC the UE should be connected to. Operators who have existing 2G or 3G networks should plan their TA boundaries to coincide with their routing area (RA) or location area (LA) boundaries. LA and RA boundaries used for the 2G and 3G systems are likely to be relatively mature and may have already been optimized in terms of their locations. Existing 2/3G counter data can be used to estimate the quantity of paging, which is likely to be experienced across a specific TA. There might be specific requiremets to coordinate TA code (TAC) planning with LA code (LAC) planning in case of mobility during CSFB or in case of cell reselection to 2/3G. When UE makes a combined EPS/IMSI attach or combined EPS/IMSI, by TA update request message, it receives a LAC from MME inside of the attach/TAU accept message. In case of CSFB call, LAC is needed when MSC is paging a UE over SGs interface as MME maps LAC received from MSC to TAC and sends the 397 398 LTE Optimization Engineering Handbook paging to eNBs over S1AP. Therefore, a mapping table between LA identity and TA identity must be configured at the MME. With CSFB enabled, in order to minimize CS call setup delay, it is necessary for the MME to inform the MSC in which LA the UE is located. Delays will be introduced if MSC is informed with the incorrect LA. There could be two scenarios: ●● ●● If the incorrect and actual LA are controlled by the same MSC ⇒ minor delay If the incorrect and actual LA are controlled by different MSC ⇒ more additional delay Thus, the general rule is TA should be correctly mapping to LA, as shown in Figure 10.16. TAL1 TAL3 TAL2 TAC2 TAC6 TAC4 TAC1 TAC10 TAC7 TAC5 TAC3 TAC12 TAC8 TAC9 TAC11 LAC1 LAC2 LAC3 LA Update TA Update Mapping table MME uses TA when roaming in LTE Coverage! LA1 MSC/VLR#1 Figure 10.16 TA and LA mapping. MSC/VLR MSC/VLR#1 MSC/VLR#2 MSC/VLR#2 Identifying LA and MSC/VLR from TA MME TA1 LA LA1 LA2 LA2 TA TA1 TA2 TA3 TA2 TA3 LA2 MSC/VLR#2 Circuit Switched Fallback Optimization 2G/3G RAN MSC Page Page Response/ Call Setup MME AGW LTE PSTN Location Update PDN Same geographical area Figure 10.17 Matched LA/TA. UE eNB BSC/ RNC MME MSS Location Update Time to transition to target RAT Extended service request Transition to target RAT Location Update Request Authentication, identity request, TMSI reallocation Location Update Accept Call setup CM Service Request CS call setup Figure 10.18 Unmatched LA/TA. The TA and LA mapping relationship of MME are pre‐configured. TA should totally match the UE’s true LA in ideal situation. UE need to be registered in geographically “correct” LA while in LTE so that it can respond to a page to the MSC serving the 2/3G cells in the area where the UE currently is located. One LA could contain several TAs, but one TA could only belong to one single LA. Any matched LA/TA, UE will establish a call quickly after falling back to 2/3G and send CM service request or Paging response message. Otherwise, there will be extra LAU procedure. If the MSC changed, the call may failed without LA update procedure (Figure 10.17 and 10.18). Unmatched LA/TA, the target RAT LAC differs from the one UE obtained in LTE, if the two different LAs are controlled by the same MSC, it will cause the UE execute LAU procedure after fallback to 2/3G network. Usually, it will bring 2 more seconds of delay. Before the CM service request or Paging response message on 2/3G, LAU is necessary, and IMEI checking and TMSI reallocation may be done at this point, and UE should also mark Follow‐On‐Request 399 400 LTE Optimization Engineering Handbook This TA Streches over two areas serviced by two different MSCs (or MSC pools) MSC1 LA TA1 = LA1 LA6 LA23 LA22 A TA list is not completely located within an LA, causing LA is different after UE fall back (MSC pool is not changed) TA2 LA2 LA3 TA3 LA5 TA4 = LA4 TA5 = LA5 2G/3G LA LTE TA Figure 10.19 Unmatched LA/TA areas served by same/different MSC pools. (MO) to make NAS connectivity and successful CM_service request or CSMT1 (MT). If the new LA is served by a different MSS that is not the one registered in combined attachment and caused the MT call to fail, because the update location message is received from a MSC, which is not aware of any previous paging request, the 3GPP feature of MTRR/MTRF is needed to avoid call drop. An example of the areas of unmatched LA/TA is shown in Figure 10.19, it notes that with MSC pool solution each MSC within the pool can control all LAs belonging to the pool; therefore, in this case the critical areas are limited to the border between the pool areas. The example shown in Figure 10.20 gives an explanation of how the UE knows if it belongs to a different LAC. 3G SIB1 broadcasts common information to all UEs in the cell related to cell access parameters and information related to scheduling of other SIBs. The NAS IE inside SIB1 defines the LAC to which the cell belongs: “cn‐CommonGSM‐MAP‐NAS‐SysInfo” = 7922 Hex (which means 31010). At this point the UE already knows it is camping in a cell belonging to a different LAC. Actually, in a live network, a 100% correct mapping of LTE and 2/3G cells (TA‐LA mapping) is nearly impossible to maintain, especially due to cell breathing and varying indoor coverage, and so on. To address the setup delays and setup failure risks in MSC border areas, as well as eliminate the LAU delay time, a few features for overcoming the MSC border issue: ●● MSC pool: MSC pool architecture, also known as Iu/A‐Flex, conforms to the 3GPP Rel 5 specifications for connection of RAN nodes to multiple core network nodes. With MSC pool architecture, all MSC servers within a pooled area serve all cells in the pool, eliminating MSC borders and the time delay of inter‐MSC LAUs within the pool. When the UE falls back across the pool and result the call failure problem, there are three solutions: delete the excess neighbor cell, delete the excess frequency points, and replane the 2/3G MSC pool border. 1 CSMT: A flag in location update request message used in CS fallback for MT call for shortening the call setup time. When the MSC receives a LAU request, it shall check for pending terminating CS calls and, if the “CSMT” flag is set, maintain the CS signaling connection after the LAU procedure for pending terminating CS calls. The UE includes the “CSMT” flag in the location update request message that informs the new MSS to delay releasing the NAS signaling connection in order to wait for the incoming call setup from the gateway MSS. Figure 10.20 How the UE knows if it belongs to different LAC. 402 LTE Optimization Engineering Handbook ●● ●● Mobile terminating roaming retry (MTRR, TS 23.018), allows that the MSC/VLR where the UE is registered routes the call further to the MSC where the UE has fallen back to. The probability is higher that this use case occurs if no MSC pool is deployed (at MSC borders) or when MSC pool is deployed it can occur as MSC pool borders. The feature of MTRR needs upgrade the HLR, GMSC, and MSC. Additional delay has to be added for an inter‐ MSC location update case (2‐10 seconds). Mobile terminating roaming forward (MTRF, TS 23.018), which is subscriber registration in the new MSC and call forwarded from old to new MSC, needs additional call setup delay (~5 s) and additional load on HLR2 and MSC. MTRF need MSC software upgraded. MTRF is a newer version of the MTRR standard and it solves the MSC border issue by forwarding calls directly from the old MSC to the new MSC in case a fallback is done over an MSC border. MTRF has the advantage over MTRR of not needing inter‐operator agreements and not re‐routing calls back to the GMSC for a second HLR interrogation. This makes MTRF more reliable and easier to deploy. Field tests show that MTRF activation will enable the failure rates decreased from around 9% to less than 3% in the critical areas. The additional call setup time due to MTRF for a CSFB call is around 1 sec. 10.3 CSFB Performance Optimization The main CSFB‐specific optimization tasks to be discussed in this section are minimization of call setup time, improved call setup success rate, and accelerating the return to LTE after the call is completed. CSFB mobile originated (MO) calls and CSFB mobile terminated (MT) calls are expected to have different performances. MO calls are quite easy in that mobility issues can be handled by the UE autonomously; apart from the lack of coverage, the UE is normally able to find a suitable target cell and establish the call, network is minimally involved in this process. MT calls are more tricky that CN mobility is critical as the call is already routed to a specific MSC, and the incoming call is pending in the MSC. 10.3.1 CSFB Optimization 10.3.1.1 Main Issues of CSFB When originated a CSFB call, the main issues lead to unsuccessful calls are listed below: During paging procedure: the wrong software version of the network, frequency locking of the terminal, terminal software problems, 4G weak coverage, user’s implicit detach, when the user is during fallback, TAU and other procedure conflicts happened, and so on. During fallback procedure: the CSFB function of eNB is not enable, 2/3G frequency configured is incorrect in eNB, weak coverage of 4G or target network, fallback to different MSC POOL, and Pseudo base station case. Call setup procedure: the CSFB function of 2/3G BSC/RNC is not enable, incorrect setting of core network parameter, authentication and handove issues caused by 2/3G weak coverage, the terminal setting blacklist, 2/3G network congestion, 2/3G channel assignment failure, and so on. CSFB call setup time needs special care of CSFB optimization. The main contributor to the additional call setup time is from reading system information (SI) in the new cell. 2 An MTRF call requires some 380% additional load on the HLR and 65% additional load on the VMSC per call since HLR registrations are normally not done on a call basis. Circuit Switched Fallback Optimization UE can not be paged during TAU procedure Paging LTE coverage issue UE can not be paged during reselection between RAT RF optimization and TACLAC match optimization fallback failure Fallback and access to target cell CS call establish failure-MSC Pool issue CS call establish failure-non optimal frequency of target RAT 4G-2/3G neighbour optimization CS call establish failure--Pseudo base station issue Return back to LTE Reselection, fast return failure Figure 10.21 Main issues of CSFB. Return to LTE procedure: When the call is released, shorter time from 2/3G return to LTE is expected, so the reselection failure, reselection with redirect to LTE failure and fast return failure may occur. In summary, the overview of CSFB‐specific optimization tasks is shown in Figure 10.21. 10.3.1.2 CSFB Optimization Method CSFB solution can provide optimal reliability for originating call setup, that is, the same reliability that the target system would offer in a given location. Terminating call setup reliability is more challenging and can be affected by TAC/LAC mismatch and/or excessive delay. For CSFB deployment, the LTE mandate feature includes eNB CSFB function, LTE to 2/3G session continuity, redirect (RIM) function, and so on. 2/3G mandate feature includes RIM function, fast return function etc. Before RF optimization, 2/3G neighbor and frequency allocation, FreqRelation, and TAC to LAC mapping should be checked. For CSFB optimization, the four steps including combined attach, paging, RRC release procedure, and access to 2/3G should be detail analyzed to spot the CSFB failure problem. The CSFB optimization KPIs include success rate, access, fallback and return latency, and so on, as shown in Figure 10.22. Table 10.3 gives a simple description of the main CSFB failures, issue‐related nodes, and solutions. Abnormal call procedure Low CSFB originated call setup success rate analysis procedure is shown in Figure 10.23. According to low fallback success ratio and call setup reject, there are different analysis steps. The causes of CSFB call connection failure include: ●● ●● ●● CSFB fallback failure: CSFB fallback failure will directly impact CSFB call connection 2G secure and authentication procedure failure The added 2G frequency have TCH or SDCCH congestion 403 404 LTE Optimization Engineering Handbook Reasons Extended service request RRC connection release RRC connection request TAU or handover happened RRC connection request (3G) RRC connection request (3G) Fall back with RIM Call confirmed Multi RRC connection request, interference Latency Call confirmed Setup Setup Access failure Access failure Procedure conflict Service setup failure LAU Alerting MultiRAB assignment, physical channel reconfig Interference, sync, poor coverage, wrong parameter setting or other faults TAU or handover happened during extended service request Call drop Pilot pollution Network structure, pilot power setting, antenna height/tilt etc. Call drop Interference, poor coverage, power setting etc. Interference, congestion, wrong parameter setting or other faults Figure 10.22 CSFB optimization method. Table 10.3 The main CSFB failures and solution. Failure reason Issue location Solution Failure before fallback SGs paging failure RF, user Paging optimization Paging success, but fallback failure RF Optimize the issued cell Failure after fallback Assignment failure RF Check MSC and target cell response assignment Call drop RF Optimize target cell Recovery on timer expiry RF, user Check RF conditions and MSC‐related timers Authentication failure after fall back to 2/3G Low CSFB originated call setup success rate CSFB originated call setup reject low CSFB originated call fall back success ratio · UE authentication issue leads to implicit detach · Analyz through EMM cause UE not received RRC release with redirection message,need to check RF coverage and interference high UE search other RAT cell and try to access The number of CSFB originated fall back + the number of CSFB originated call setup reject < The number of CSFB originated request 2/3G issues The feature of CSFB in eNB is not enabled, there is no 2/3G cell frequency in RRC release message LTE misconfigured 2/3G neighbour cell frequency UE search other RAT cell and try to access eNB did not transmit RRC connection release message, the CSFB feature is not enable in eNB The number of UE context modified failure · UE context modified failure is due to frequently handover · Analyze through failure cause The number of UE context setup failure Analyze through failure cause Figure 10.23 The analysis procedure of abnormal call setup. Circuit Switched Fallback Optimization Abnormal combined attach Un‐optimized TAC‐LAC mapping will lead to undesired location update between MSC pools and delays in call setup, in worst cases it will lead to call failure. Check combined attach planning by MME signaling trace, analyze the number of S1_release and LAU message, optimization ECI to CI mapping, and find if there is misconfigured LAC to MSC mapping record in MME, check the type of TAU is “combined TA/LA updating with IMSI attach” or not, and analyze the EMM cause value. There are four main type of abnormal combined attach in a live network, CSFB not p referred, SMS only, EPS only, and low combined attach success rate. According to these four scenarios, core network and RF issues should be spot as shown in Figure 10.24. Abnormal combined attach optimization procedure is shown in Figure 10.25. Paging failure CS paging procedure in case of CSFB is based on three steps: paging is transmitted by core network (CN), UE is reached by paging message on LTE cell, and redirection to target system, and paging answer from the last cell. The causes of CSFB paging failure include: ●● ●● If there is a paging through S‐Gs interface while UE is returning to LTE: While the call ended and UE return to LTE, if TAU procedures haven’t completed, the S‐Gs interface paging will fail in some MME product vendor. LTE network implicit detach UE: If LTE network is implicit detach UE, the second call would fail. There are many possibilities that LTE implicit detach UE, such as defect in equipment Combined attach access CSFB not preferred SMS only check CSFB feature for CN EPS only low success rate check EMM Cause TA-LA mapping 1. poor RF 2. check EMM and Extend EMM Cause Figure 10.24 Abnormal combined attach. Attach success ratio low low low Attach complete ratio Attach complete message is lost, check RF coverage Attach accomplished ratio high The number of EPS only high The number of attach reject message is big. The reason can be found by EMM cause and Extended EMM cause The number of CSFB not preferred The reason can be found by EMM cause, if EMM cause is MSC temporarily not reachable, check TA/LA mapping in MME Figure 10.25 Abnormal combined attach optimization. The number of SMS only Core network should support CSFB 405 406 LTE Optimization Engineering Handbook Check if paging is transmitted by CN N 1. check terminal UE is on the TAU procedure 2. check terminal UE is on IRAT(2/3/4G) procedure 3. check S1 is abnormal or not Y 1. check TAC is wrong or not a. frequently TAU before b. check if last TAU/LAU is abnormal 2. check if RF is poor or not (UE unreachable) 3. check if the number of PDCCH CCE is enough and check if the MCS of paging message is suitable 3. check TMSI is valid or not, check TMSI allocation is success or not 4. check UE issues (paging no response) 5. Paging reject a. IMSI detached for EPS services b. IMSI unknown c. IMSI detached for non-EPS services d. rejected by user Figure 10.26 Paging failure troubleshooting procedure. ●● and the QoS cannot be modified dynamically (check if the Dynamic QoS Modification feature is enabled). Poor wireless environment: Poor wireless environment or high interference will lead to a UE cannot receive or resolve paging message. If paging is not transmitted by CN, it is needed to check that UE is doing data service request or TAU, or UE is interworking between 2/3/4G, these cases maybe caused by CN issues, or there is S1 interface alarm exist. If paging has transmitted by CN, a major cause leading to paging failure on LTE side is the “not camping on a suitable cell,” and it is usually caused by: too many recurrent IRAT reselections including fast return from 2/3G, poor RF, and so on. The troubleshooting procedure is shown in Figure 10.26. Receive RRC connection release issue The normal RRC connection release message includes 2/3G frequency and target cell information and BCCH. In some cases, when UE sends Extended service request, UE starts T3417ext timer, if no RRC connection release is sent from eNB, at the expiration of this timer (10 seconds) UE reselects to UMTS or GSM. eNB logs can be analyzed to find out why eNB does not send RRC connection release with redirect. The causes of CSFB call fallback failure include: ●● ●● ●● ●● Poor wireless environment: Poor wireless environment or high interference will lead RRC connection failure or cannot receive RRC connection release message. CSFB to 2/3G feature is not enable: When the CSFB feature not be enabled in eNB, eNB will still release RRC connection release message with no frequency. In this situation, the UE will fallback to 3G. CSFBPrio (2/3G) is not set correctly. eNBs with unreasonable 2G frequency: When an eNB added too many 2G frequency (more than 15), there is bad quality, high interference, or across MSC pool 2G frequency will all lead to failure in fallback procedure. For the abnormal case, it is needed to check eNB CSFB feature is enable or not, check the cause of service reject and inital context setup failure. RRC connection release issues troubleshooting procedure is shown in Figure 10.27. Circuit Switched Fallback Optimization Receive RRC connection release Y N 1. check RRC connection release contains CSFB frequency or not 2. check if there is a handover command before RRC connection release 3. check if there is a TAU procedure going on 1. check CSFB feature enable 2. check CSFB 2/3G frequency configured or not 3. check if there is implicit detach issue Figure 10.27 RRC connection release issues troubleshooting procedure. Camp on 2/3G Cell Y End N 1. check 2/3G cell RF (not the best frequency) 2. MSC pool issue 3. check if there is a pesedo BTS Figure 10.28 Camp on 2/3G cell issues troubleshooting procedure. Camp on 2/3G cell After UE received RRC connection release message, the next step is camping on the specific 2/3G cell. During the procedure, some issues mostly found in a live network. It needs to check RF, MSC pool, and pesedo BTS issues, as shown in Figure 10.28. If the UE is successful camping on 2/3G cell, it is still needed to analyze the signaling in 2/3G cell, that is, CM service request/RRC connection setup complete to call proceeding, call proceeding to radio bearer setup, radio bearer setup to alerting, and so on. After setup message is transmitted by the MO UE, the network will page the MT UE, and proceed with the extended service request procedure of MT UE (Figure 10.29). Finally, it needs to focus on the return back to LTE issues after the call is completed from RRC connection release in 2/3G to tracking area update accept in LTE. Tracking area update accept message includes TAU result, T3412, GUTI, TAI, and matched LAC information. Usually long fast return (FR) latency and FR failure are due to movement of the UE or frequency planning, or ECI planning is unreasonable, and so on. The main causes of return to LTE failure are listed below. ●● ●● ●● Poor wireless environment in hook area: Only when the RSRP satisfied Qrexlevmin (120dBm‐ 124dBm) can the UE access a LTE network. Thus, if the 4G wireless environment is poor, UE cannot access LTE network and stay in 2G. LTE network implicit detach UE Wrong setting of cell reselection priority: If the cell reselection priority of LTE network hasn’t been set to 7, UE will fail in returning to LTE. 10.3.2 CSFB Main KPI The CS fallback network capability is realized by using the SGs interface mechanism between the MSC server and the MME. The SGs reference point between the MME and MSC server is used for the mobility management and paging procedures between the EPS and CS domain, and it is based on the Gs interface procedures specified in TS 23.060. The SGs reference point is also used for the delivery of both mobile originating and mobile terminating SMS. 407 408 LTE Optimization Engineering Handbook Secuity key, service type, and UE capability etc. Bearer capability, coding scheme and called ID etc. CM Service Request Routing Area Update Request securityModeCommand securityModeComplete Setup MO Call proceeding securityModeCommand securityModeComplete Identity Request Identity Response radio bearer setup radio bearer setup complete Call Proceeding MT Paging/CS service notification Extended service request rrc connection reconfiguration complete rrc connection release rrc connection setup complete Call confirmed radio bearer setup radio bearer setup complete Alerting If apply early assignment, MO RAB setup will not wait for MT Call Alerting Figure 10.29 Signaling in 2/3G cell. CSFB E2E optimization, it is mainly focused on CSFB paging success rate, CSFB call fallback success rate, CSFB call connection success rate, CSFB return back to LTE success rate, and the related latency KPI. CSFB service KPI optimization needs to be carried out with both the core network and wireless side to promote CSFB service quality. The key interfaces involved in CSFB service such as SGs, S1‐MME, and Mc, and so on, and the key related signaling flows need to be analyzed, to find the key process and the performance of each interface in detail, mining the hidden troubles of the interfaces in the network. The CSFB main KPIs from signaling messages are shown in Figure 10.30. CSFB service processes is complex in that it involves the PS and CS domain. A single indicator of OMC can not reflect the problems. It is necessary that CSFB signaling analysis tools require data collection, correlation, and analysis of signaling processes across different domains of the interface, including the CSFB MO/MT call setup success rate, return to LTE success rate, and the related delay analysis through the 2/3/4G network. The multi‐interface signaling message eNodeB BSC MME Paging Request/CS Service Notification Extended Service Request MSC Paging Request 1 2 UE Context Modification Request 1 CSFB paging success rate 2 CSFB call fall back success rate 3 CSFB call connection success rate 4 CSFB return to LTE success rate Service Request UE Context Modification Response UE Context Release Request UE Context Release Command UE Context Release Complete Fall back to 2G LU Request / Paging Response Alerting 3 4 Voice Clear Command 5 Clear Complete TAU Request 6 TAU Accept Figure 10.30 CSFB main KPIs from signaling message. Circuit Switched Fallback Optimization eNB MME BSC MSC (1) Extended Service Request (2) InitialContext Setup Respose (3) UEContext Release Complete (4) CM Service Request (5) Alerting (6) Release Complete (7) Tracking Area Update Request eNB MME (2) Paging/CS Service notification (3) Extended Service Request (5) InitialContext Setup Respose (6) UEContext Release Complete MSC BSC (1) SGSAP Paging Request (4) SGSAP Service Request (4) Paging (7) Paging Response (8) Alerting (10) Tracking Area Update Request (9) Release Complete Figure 10.31 MO and MT KPIs related signaling. by real‐time correlation processing, analysis can realize the end‐to‐end optimization, quickly locate the CSFB terminal and network equipment issues, and improve CSFB service quality and customer perception. MO (mobile originated) and MT (mobile terminated) KPIs are list below, the related signalings are shown in Figure 10.31. ●● ●● ●● ●● ●● ●● ●● ●● ●● ●● ●● MO CSFB success rate = A_Alerting/S1_Extended_service_request MT CSFB success rate = A_Alerting/SGs_paging request MO fallback success rate = A_CR(RRC connection release)/S1_Extended_service_request MT fallback success rate = A_PAGING_RESP/SGS_Paging request MO call setup delay = A_Alerting‐‐‐‐‐S1_Extend_Service_request MTpaging success rate = SGSAP Service request/SGS_Paging request MT call setup delay = A_Alerting‐‐‐‐‐SGs_Paging MO fallback delay = A_CR‐‐‐‐‐S1_Extended_service_request MT fallback delay = A_Paging_response‐‐‐‐‐SGs_paging request Return time to LTE = Tracking area update complete‐rrc connection release Return to LTE success rate = Tracking area update complete/MO CSFB call release. 10.3.3 Fallback RAT Frequency Configuration Optimization In standard CSFB procedure, there is no UE measurement during the fallback, which means CSFB RRC release with redirection is a blind release. UE is allowed to pick any cell on the indicated frequency, or may even try other frequencies/RATs if no cell can be found on the target frequency. The selected RAT frequency is based on UE capability (band/RAT is supported) and prioirty (csFallbackPrio). CSFB frequency priorities defined with the csFallbackPrio parameter for each available frequency relation. Based on these two criterias, when several possible target frequencies have the same priority, the eNB applies a round robin scheme for the selection. It 409 410 LTE Optimization Engineering Handbook Figure 10.32 Wrong GERAN relation frequency configured. should add all the 2/3G frequency into one group so that it can be found all the 2/3G frequency in RRC release message. If 2/3G neighbor frequencies won’t be defined, CSFB call service request would be rejected. Besides, it is needed to define both WCDMA and GSM as candidates, by defining both UtranFreqRelation and GeranFreqGroupRelation. By setting parameter csFallbackPrio for UtranFreqRelation to a higher value would set higher CSFB priority for corresponding UMTS frequency. Take GSM for example, GERAN starting ARFCN is based on the configuration in eNB that know how many ARFCNs are defined in OMC database. To speed up CSFB to GSM, it will be smart to geographically configure ARFCNs based on the collected 2G ARFCN and neighboring 2G ARFCNs. Usually the operator can find that the problem area is the target GSM cells’ frequencies are not set in LTE cell, which causes this call back to a bad GSM cell, may lead to poor call quality and finally causes the connect failure. An example is shown in Figure 10.32. Example 1 Assume CSFB GERAN relation frequency is cell A, cell A and cell B is co‐BCCH, B is located in mountain, and when UE fallback to cell B, the rx level is −93 to −99dBm, which result to UE lost. In this case, frequency point of the MSC pool boundary needs to be optimized, to avoid the same frequency configured across the MSC pool. Meanwhile it needs to control the 4G and 2G coverage, to avoid cross coverage. eNB can not configure the GERAN f requency point not belonging the pool. Through the pool boundary, the configured 2G frequency points need to be verified, if the 2G frequency points’ signal Rx lever of the other pools were too strong, it is proposed the BTS and antenna parameters need to be adjusted to control coverage. For GSM, SDCCH establishment success rate impacts CSFB call setup success rate. SDCCH assignment success depends on GSM network quality and the degree of interference. In the CSFB process, as long as the SDCCH assignment fails, then CSFB call setup will fail. Therefore, the fallback GSM cell can be priority selected from the high SDCCH setup success rate cells. Circuit Switched Fallback Optimization Figure 10.33 GSM neighbor across the pool. Example 2 Before call setup in GSM, MT UE camped on the 4G cell belongs to TAC 16640, MSC pool2. After fallback to GSM, the GSM cell belongs to LAC 16793, MSC pool1. During the procedure, UE did not send paging response, and the call failed. After deleting the GSM neighbor across the pool, the issue was solved. Example 2 is shown in Figure 10.33. 10.3.4 Call Setup Time Latency Optimization CSFB call setup time latency (MO/MT and idle/connected) is calculated as below. For originating call setup the start trigger is when the call is placed, which triggers the Extended service request from the UE to MME. The end trigger is when the UE has moved to 2/3G RAN and received the CS ringing from MSC on the target cell to setup the call on 2/3G. Thus, the call setup delay for originating CSFB call is: Call Setup LatencyCSFB MO ms TCS Alert TExtended Service Requestt In the terminating call setup case, the start trigger is when the UE is paged by the MSC via the message CS service notification and sends Extended service request to MME. The end trigger is when the UE has moved to 2/3G and received the CS ringing from MSC on the target cell to setup the call on 2/3G. Thus, the call setup delay for terminating call setup is: Call Setup LatencyCSFB MT ms TCS Alert TExtended Service Requestt 411 412 LTE Optimization Engineering Handbook Delay is the most critical aspect as CSFB call setup time might be prolonged due to the additional time required to measure the target RAN frequency. It can negatively impact the user perception. MO/MT CSFB call setup time latency is shown in Figure 10.34. In case of terminating calls excessive delay can bring to failures due to timers in the core network. In the best case, the additional delay introduced by CSFB can vary from 0.7 to 2.5 seconds, but it can reach 6−10 seconds or more in case of failures. If originating and terminating UE are both CSFB calls, the additional delay introduced will be much longer. The most critical issue affecting the delay is the redirection to a wrong 2G or 3G frequency, that is, the target RAT/ frequency is not suitable to allow the radio access (Figure 10.35). Let’s take an example of LTE to UMTS, to explain CSFB delay (one side) main contributor and statistics: ●● ●● ●● LTE release delay, time for signaling on LTE: ~100 ms for idle UEs UMTS cell access delay, time to find and select a UMTS cell: 0.4 sec SIB reading, time to read relevant system information from UMTS cell: 0.2 sec CSFB call setup time latency Start: NAS extended service request; End: NAS alerting. Figure 10.34 CSFB call setup time latency. LTE release delay (fallback to 2/3G) time to find and select a 2/3G cell and read relevant system information from 2/3G LTE release delay (fallback to 2/3G) time to find and select a 2/3G cell and read relevant system information from 2/3G LTE release delay (fallback to 2/3G) time to find and select a 2/3G cell and read relevant system information from 2/3G the additional delay introduced by CSFB Figure 10.35 CSFB typical delay. 2/3G call LAU procedure in MSC pool 2/3G call LAU in different MSC pool +MTRF (terminate UE) 2/3G call Circuit Switched Fallback Optimization ●● ●● Call setup delay, time to establish call from RRC connection request to alerting: 6.7−8.5 sec If originating and terminating UE are both CSFB call, the additional delay introduced will be 1.2−1.8 sec under RIM procedure. As mentioned earlier, CSFB will prolong the call setup time. Longer call setup time can result in bad end‐user experience and would be perceived as outages. If we compare a call in the classic GSM network with a CSBF call it will be seen that there are some extra interfaces and procedures that are used. The main contributor to the additional call setup time is from reading system information (SI) in the new cell. CSFB to 2G (GSM) today takes ~1.0−1.5 sec longer than CSFB to 3G in both MO and MT. The reason is slower reading of SIB information in 2G compared to 3G. As you can see the call flows in Figure 10.36, CSFB causes this delay because of the additional interfaces and procedures involved. While in a normal 2G‐2G call it has two nodes and the UE in the CSFB cases it has four nodes involved. Therefore, in 3G, it is typical to cause delays of 1.5 sec while in 2G it increases to 2.5 seconds. As we have discussed in Section 10.2.3.1, there are two features on top of CSFB to improve the call setup delay—one is RIM, the other is DMCR. As mentioned earlier, there are four basic methods for directing the UE to the target system in CSFB call setup, these methods can be either blind or measurement‐based. Call setup time latency of these methods to fallback to 2/3G network is listed in Table 10.4. For redirect, the time to access 3G is typically around one second; obviously this assumes optimized neighbor list and target RAT with sufficient radio quality. Call setup time with PS handover can be optimized by shortening the time to trigger to first measurement report and by optimization of target RAT neighbor lists. In case of a CSFB without PSHO, the data might take 5 to 10 sec to resume, while for CSFB with PSHO it is done almost directly. MSC MSC 2/3G UE MME LTE Page UE Page Page Page Page Page response 2/3G Extended service request Release Release Cell change Call setup Read SI Page response RIM: –1.2 s improvement for 3G –2.0 s improvement for 2G DMCR: 0.7s improvement for 3G Call setup Figure 10.36 Comparison of classic CS call setup in 2/3G (left) and CSFB call setup (right). Table 10.4 Call setup time latency of the four methods. Mobility method between LTE and 3G 3G/LTE in same CSFB MSS Overlay CSFB MSS 1) Redirect based CSFB to 3G +1.0 − 1.5 sec +2.0 − 3.0 sec 2) PSHO based CSFB to 3G, (delta to 1) −0.4 sec −0.4 sec 3) Redirect based CSFB to 3G with SIB, (delta to 1) −0.3 sec −0.3 sec 4) Redirect based CSFB to 3G, deferred SIB 11/12 reading (delta to 1) −0.4 sec −0.4 sec 413 414 LTE Optimization Engineering Handbook CSFB with measurement‐based redirect takes some 440 ms longer compared to CSFB without measurements, timing limit for the number of 2/3G cells and carriers must be considered during parameters setting and network planning. The example delay between Extended service request to RRC connection release with redirect info to 3G is listed in Table 10.5. For GSM neighbor cell frequency planning, it is recommended to configure the overlapping cells’ BCCH (3*7 = 21) frequencies as the fallback frequencies. The measured call setup time can also be improved (for incoming calls) with DRX paging cycle optimization and other parameters. By adjusting, that is, parameter DRXPagingCycle from 1280 ms to 640 m, call setup time can decrease 400 ms, decrease T3413PagingTimer (interval between two paging in S1 interface) also can decrease call setup time. In a live network, the paging parameters should be analyzed and fine‐tuned according to the real traffic. Figure 10.37 gives the parameters related with CSFB call setup latency. For CSFB and session continuity, it needs to have TAC to LAC mapping where the UE on 4G cell will be redirected to the underlying 2/3G site. Every TAC can be mapped to one and only one LAC. Accordingly, one TAC can be defined in two TALs but shall be fully included within one LAC to have 1 TAC to LAC mapping. If TAC and LAC does not match in co‐site LTE and 2/3G cells, it will lead to even longer call setup times due to an additional LAU procedure. At the border of MSC pool, it will even cause connect failure, or increase a 1‐second time delay. Figure 10.38 gives an example of CSFB call setup latency under TAC and LAC matches and not matches. Table 10.5 Call setup time latency of different number of 3G cells and carriers. Modus average time redirect without 3G measurement 110 ms redirect with 3G measurement 1 neighbor 3G cell @ 1 UTRAN carrier 556 ms redirect with 3G measurement 6 neighbor 3G cells @ 1 UTRAN carrier 707 ms redirect with 3G measurement 12 neighbor 3G cells @ 1 UTRAN carrier 760 ms redirect with 3G measurement 6 + 6 neighbor 3G cells @ 2 UTRAN carriers 853 ms UE eNBs in TAI List MME eNB maxNoOfPaging Records: 16 nB: T (1 PO in 1 PF) DRX : 1.28~0.64s N3413 : 4 S1 Paging RRC Paging RRC Paging MSC SGs-Paging-Request T3413: 3~5s pagingDiscardTimer: 3s S1 Paging S1 Paging 1.28 Sec defaultPagingCycle maxNoOfpagingRecords 16 ¼T nB PagingDiscardTimer 3 Sec tInactivityTimer 5–10 Sec S1_T3413-PagingTimer 3 Sec S1 Paging SGs-Paging-Request Service Request CSFB To 2/3G RNC Paging Response or LU Figure 10.37 CSFB call setup latency–related parameters. Figure 10.38 CSFB call setup latency under TAC and LAC matches (left) and not matches (right). 416 LTE Optimization Engineering Handbook Table 10.6 Typical CSFB to GSM call setup latency distribution (second). Extended service request‐‐ > Alerting Extended service request ‐‐ > RRC connection release RRC connection release ‐‐ > CM service request 11.4 0.066 1.941 9.363 1.223 5.1 0.115 1.488 3.545 0.448 5.3 0.11 1.94 3.235 0.474 6.8 1.004 2.39 3.406 0.465 4.6 0.082 1.261 3.257 0.47 14.6 0.202 1.403 12.996 0.47 5.7 0.204 1.83 3.68 0.537 5.5 0.114 2.093 3.346 0.454 6.6 1.509 1.597 3.469 0.444 6.1 0.956 1.822 3.272 0.484 5.6 0.218 1.929 3.491 0.455 5.2 0.521 1.475 3.209 0.63 CM service request‐‐ > Alerting channel release‐‐> TAU Accept 10.3.4.1 ESR to Redirection Optimization For CSFB, UE initiates the call by sending Extended service request (ESR) to MME. The time of transition to target RAT can be measured from Extended service request to RRC connection release. Once the UE is in target RAT, it initiates location update procedure if the unmatched LAC is met or even MSS is changed; this will increase the location update time, moreover, IMEI checking and TMSI reallocation may be done at this point. The duration of the procedure can be measured from RRC connection release to CM Service request. After that, the UE is able to start the actual CS call setup and the duration of this procedure can be measured from CM Service Request to Alerting. After the CS call is finished, the UE should return back to LTE and the duration of this procedure can be measured from channel release to TAU accept. The duration between CM Service request to alerting, is basically the same for ordinary CS call in the target system. Comparing to other network, the delay in 4G (ESR to RRC redirection) is a little bit longer in Table 10.6. Test result showing the variation of ESR to RRC redirection is still great; from procedures comparison between idle and connect mode, it can be seen that CSFB from idle mode took much longer time than connected mode due to its signaling exchange to MME which occupies the longest time (Figure 10.39). For CSFB latency in 4G (ESR ‐ > RRC connection release), as call from idle mode takes much longer time than connection mode, one solution to reduce the 4G CSFB latency is to increase the inactivetimer setting to keep the UE in RRC connection. 10.3.4.2 Twice Paging MSC server paging parameter is 9 s + 6 s, namely if the first paging failure (MSC server does not receive the paging response message), MSC server will send the second paging after 9 s, this is LTE twice paging strategy. Combining MO and MT call signaling to analysis as shown in Figure 10.40, it can be found left signaling reflect that MO’s setup is sent on 10:41:48.789; however, TC fallback time is 10:41:58.189, that means MT UE call prepare to fallback to GSM after 10s waiting. Obviously, ESM rrcConnectionSetup Procedure 0.066s 0.004s Connected mode 0.017s UEContextModification Procedure 0.000s initialContextSetup Procedure (SecurityMode Procedure) 0.001s rrcConnectionRelease Idle mode 0.168s (0.027s) 0.001s rrcConnectionRelease Figure 10.39 ESR to RRC redirection from connected mode and idle mode. UE MME x x x x MSC 1st SGs paging request S1/RRC page S1/RRC page 3s S1/RRC page 3s S1/RRC page 3s 3s Incoming call Ts5 = 9s 2nd SGs paging request UE Unreachable Voice mail service Figure 10.40 LTE twice paging strategy and an example from field test. 418 LTE Optimization Engineering Handbook MT UE did not receive the first paging message due to poor coverage of LTE, and this cause the total call connection period lasts over 18 seconds. 10.3.5 Data Interruption Time If a user is in an active PS data session (e.g., streaming media) when a voice call is initiated, the inter‐RAT transition and routing area update will interrupt the data transfer. The interruption time will depend on the mobility mechanism, as summarized in Figure 10.41 for an example of UMTS. Using handover‐based CSFB, the data stream interruption time of 0.3 seconds is unlikely to be noticeable. The user will experience much higher (5 seconds) data stream interruption Redirection Handover Rel-8/ Rel 9 Handover Rel 9 Rel-8 S1 Tunnel Skip SIBs Basic 0.3 RRC Release 0.2 0.2 0.2 Acquisition on UTRAN 0.2 0.2 0.2 0.4 2.0 Read MIB & SIBs Camp on Cell 0.1 0.1 0.1 Connection Setup 0.3 0.3 0.3 Optional RAU Procedure 4.0 4.0 4.0 4.8 5.1 6.8 Total Data Interruption Time 0.3 Data rate (kbps) ~5–10s w/o PSHO 3) Resume LTE LTE WCDMA GSM (DTM) Time (s) CSFB call 3) Suspend 2) Resume Figure 10.41 Data interruption time. 4) Resume Circuit Switched Fallback Optimization in the redirection‐based Rel 9 SI tunneling and Rel 8 Skip SIBs methods, which may be mitigated in practice by the fact that user attention will already be diverted to initiating an outgoing call or receiving an incoming call. 10.3.6 Return to LTE After Call Complete After the CSFB call is finished, the UE should return back to LTE if LTE network is a preferred access and coverage exists. This is possible by either UE‐controlled normal inter‐RAT cell reselection or by redirect/handover based on a vendor‐proprietary trigger or by fast return mechanism. Return to LTE after call completion procedure is shown in Figure 10.42. Especially, the 2/3G feature release with redirect to LTE functions as well as fast return, to prevent a UE from being stuck in 2/3G after a CS session. It is not a coverage triggered release with redirect, and is triggered based on 2/3G channel switching state changes, it is a blind mobility mechanism. The latency of return back to LTE measurement can be calculated as: Start: UMTS/GSM RR channel release message; End: LTE NAS Tracking area update accept message. Release with redirect to LTE is an IRAT mobility function that provides basic mobility for an LTE‐capable UE from 2/3G to LTE. Redirection to LTE is controlled by the network and can be initiated at several trigger points. A redirection to LTE can be initiated from the active states CELL_DCH, CELL_FACH, or URA_PCH. When the LTE coverage is good, UE redirected quickly to LTE (~2 to 3 s). With bad or non‐existing LTE coverage, UE will search for 10s for LTE frequencies included in the Release with redirect message; if not successful, UE shall search all LTE frequencies, which most likely will return to UTRAN more than 15 s (Figure 10.43). HLR MSC UE RNC Rrc connection release Rrc connection release complete MME eNB RRC connection establishment Tracking Area Update Authentication Request Authentication Response Security Mode Command Security Mode Complete Location update Tracking Area Accept Tracking Area Complete After the RRC connection release, a new combined attach process is need EMM Info UE Context Release Command UE Context Release Complete RRC Connection Release Figure 10.42 Return to LTE procedure. 419 The info of redirect to LTE in the RRC release message was sent, meaning the feature “Release with redirect to LTE” can be achieved. The LTE-CSFB-UE can return to LTE as soon as CS call ending. Figure 10.43 Release with redirect to LTE. Circuit Switched Fallback Optimization For the feature of fast return (FR), there are two methods. One is network side FR, and when the call is ended, the 2/3G network releases the user channels and distributes LTE frequencies, and without waiting for changing to an idle state, the UE will access LTE cells according to the indicated LTE frequencies. The other one is UE independent FR, and when the call ends, based on the cell information before fallback, the UE return 4G independently (need UE support). Field tests show that the minimum fast return to LTE latency is 0.5 s (Figure 10.44). Without the feature of fast return, the normal procedure is that the user stays in 2/3G after the CS call has been released and perform normal idle mode mobility. For idle mode cell reselection, when the call ended, the user stays in 2/3G cells, it can reselect to a high priority LTE cells by reading the 2/3G system information, which contains LTE neighbor cell information. User may stay in 2/3G for at least 10−15 s, as shown in Figure 10.45, which will have impact to data performance. Fast return to LTE after call release is introduced to decrease the packet data transfer outage time in order to ensure that 4G‐capable UEs will spend as much time as possible under LTE coverage. With that feature, multi‐RAT mobiles can be requested to select an LTE cell directly after the speech call release, hence preventing the mobile to first select a 2/3G cell and then re‐select an LTE cell. As soon as the CS speech call is over, the UE is redirected to the LTE layer by methods list below: Fast return is done by including the LTE EARFCN in the IE “Cell selection indicator after release of all TCH and SDCCH” of the channel release message. UE will skip the LA and RA update procedures in 2/3G, and the user can quickly resume data services in LTE. When the network side enable fast return function, it can be seen carrying a 4G macro base stations and ventricular frequency division sites in Release 2/3G talk after the end of the message, so that the UE can be quickly re‐select back to the 4G network. Fast return to LTE at CS call completion needs about 1 second, which is a faster procedure than 2/3G to LTE cell reselection (15–20s). Longer return time to LTE can affect end‐user experience. Longer or delayed return time to LTE can exploit UE behavior and would delay return to LTE through chain reactions in certain circumstances. TAU rejections or PS session deactivation can also result in delays in return time to LTE. 2G 4G Measure LTE neighbor cells CS call release Sync to target LTE cell and read SIB TAU Figure 10.44 Fast return. 4.62s 2.87s 2G Release Read 2G SIB, Location area update 9.73s 1.75s Routing area update 8.53s 1.2s Measure LTE cell Return to LTE UE out of service Figure 10.45 CSFB to 2/3G, cell reselection back to LTE, and typical latency. UE out of service 421 422 LTE Optimization Engineering Handbook 10.4 Short Message Over CSFB Short message in LTE is the same service as in GSM and WCDMA. The solution re‐uses the SGs interface between MSC and MME defined to CSFB to carry SMS messages encapsulated in the NAS signaling channel between MME and terminals. In the SGs interface, NAS procedures are used to translate circuit based SMS messages from the legacy network into the equivalent packet–based SMS messages for the LTE network and vice versa, which does not require “fallback” to 2/3G network to send or receive SMS. Thus, the “SMS over SGs” method can be deployed without deploying CSFB. The solution is ideal for initial LTE services considered to be too complex for these initial terminals and so the NAS signaling based solution developed for CSFB has been extended to offer a low‐cost solution for SMS to data‐only devices. ”SMS over SGs” method can be used in an LTE data only network to provide SMS coverage for LTE UEs. At least one MSC in the legacy network must be equipped, through a software upgrade, to use the SGs interface. If CSFB is set up between the legacy network and the LTE network, then all the MSCs in the legacy network are already set up with the SGs interface. So it does not need to do any additional work to support SMS, if CSFB is already set up. UE needs SMS service but CSFB does not indicate this specific condition with the “SMS‐ only” indication in the EPS/IMSI attach request and combined TA/LA update procedures. This allows an operator to deploy the SGs for SMS delivery over LTE only (without CSFB support), as shown in Figure 10.46. In addition, this allows the MME to use a dedicated algorithm for the selection of the MSC that supports those UEs. In a live network, it is possible that only certain MSCs in the network (one in minimum) is configured to support SGs when the network only supports SMS for SGs operation. However, such a minimal configuration can cause inter‐MSC location updates to be performed at every movement into/out of LTE coverage. SMS is delivered over signaling channel and no data path needs to be established. Once the UE is attached to both MSC and EUTRAN and if MSC wants to deliver a SMS to UE, it will simply send a downlink unit data to MME with SMS content. MME will dump this message in NAS message and send it to UE and in the same way if UE 1. Subscriber registers in MSC by CS signaling over Uu, S1and SGs-interface (attach/Location update). Uu 3. SMS page over SGs, S1, Uu eNB S1 MME LTE SGs 2. Incoming SMS to subscriber in LTE 4. UE responds and receives CS-SMS while roaming in LTE MSC-S Figure 10.46 SMS over SGs Interface. SMS-SC Circuit Switched Fallback Optimization LTE RAN UE UE LTE RAN EPC CS Core EPC CS Core SMS 1. Service Request 1. Paging 2. Uplink SMS Delivery 3. Downlink SMS Delivery report SMS Delivery report 2. Service Request 3. SMS Delivery / Downlink data 4. SMS Delivery report / Uplink data Delivery report Figure 10.47 Overview of mobile originating and terminating SMS. wants to send a SMS it will dump the message in NAS message and send it to MME. Mobile originating and terminating SMS procedure is shown in Figure 10.47. 10.5 Case Study of CSFB Optimization CSFB optimization e2e view The common reasons of CSFB failures include pseudo GSM base station, GSM TCH/SDCCH channel congestion, not received paging messages by the called terminal, 4G coverage issues like poor RSRP as well as with good RSRP but poor SINR, 4G cell overlapping, and overshooting, and so on, neighbor cell issues, PCI mod 3 issues, handover issues, TAC‐LAC mapping issues, and other issues. 10.5.1 Combined TA/LA Updating Issue It is a combined TA/LA updating when CSFB UE is doing TA updating. It required not only tracking area updating success, but also location area updating success. For CSFB optimization, when MT failure occurred, it should be firstly checked whether LA/TA is mapping correctly. In this case as shown in Figure 10.48, only track area updating success in TAU accept but no TA/LA updated, location area updating is failure and cause is MSC temporarily not reachable. Then UE will send track area update request every 10s until reach tracking area updating attempt counter. Finally, the UE starts to select UTRAN radio access technology. The other example is misconfiguration of TAC/LAC that results CSFB UE can not camp on LTE network. When the phone is powered on, 4G signal flashes quickly, then access to 3G immediately. From the signaling shown in Figure 10.49, it is found that when the UE finishes the attach procedure, detach request follows up. UE attach request message and core network gives the attach result, which are shown in Figure 10.50. As shown in Figure 10.49, it is found that after obtained the subscriber information from HSS, the UE sends session request to SGW without combined TA/LA update. After investigation, it is found the reason is that there is no configuration of TAC‐LAC‐MSC mapping in the MME. 423 Figure 10.48 Combined TA/LA updating issue. Circuit Switched Fallback Optimization UE capture the subscriber information from HSS UE sends session request to SGW Figure 10.49 UE signaling after power on. 10.5.2 MTRF Issue Without MTRF feature, MSC pool boarder issue will cause call setup failure. When the UE attached it is associated with an MSC based on the TA/LA mapping, if at the borders of two cells, a CFFB is triggered and the UE selects a 2/3G cell belonging to an MSC (pool) other than what the subscriber is registered with, and in the this case normally the CS call should fail. In order to avoid call setup failures, the MTRF features was implemented and with MTRF the call and subscriber is transferred from the registered MSC to the MSC, which was selected during the CSFB call. MTRF is a supplementary feature to CSFB for boarder cases where the LTE cells overlap with 2/3G cells belonging to different MSC, but it can be used in 2/3G only networks to decrease the call setup failure rates. Usually, there are two scenarios for the MTRF: one on borders where mapping of TA/LA might not be correct and another for physical movement of UE between two MSCs. In a field test shown in Figure 10.51, it is found the call failure in 3G LA 54057(MSC2). After the terminated UE location area update procedure in 54507(MSC2), UE can not receive the setup message, and stay in connected state, after 10 secs receive the signaling connection release message from the network, the RF conditions is really good during the whole procedure. This is the typical issue related with LA update. After signaling analysis, it can be positioned as MTRF issue between MSC1 and MSC2, resulting in the failure of the called. When the UE fallback to different MSC, it will not be able to acquire the information of the initial attached MSC. The cause of the case is the possibility that TAs and LAs are not 100% overlapped at the border between MSCs coverage. For this kind of failure, it needs to make sure the LTE subscriber is physical located in the identified critical area, and the LTE subscriber to be called is registered in old MSC, and further verified that the LTE subscriber sends a location update request with CSMT flag toward the new MSC. 10.5.3 Track Area Update Reject After CSFB When the CSFB call is completed, tracking area update reject usually happened after release with redirect to LTE. It would delay the reestablishment time for PS service. 425 UE request : Combined EPS/IMSI attach Figure 10.50 Attach request and result. Core network assign: EPS only Figure 10.51 Signaling procedure. 428 LTE Optimization Engineering Handbook 10.5.3.1 No EPS Bearer Context Issue UE needs to initiate track area update request after UE finishes call session and return to LTE network, but network replies track area update reject sometimes, the cause value (40) is “No EPS bearer context,” as shown in Figure 10.52. Compare rejecting “Track area update request” with accept “Track area update request,” as shown in Figure 10.53, it shows if there are not any EBI (EPS bearer identity) active specially EBI5 inactive. But accept “Track area update request” with EBI5, which is the default EPS bearer. The UE shall delete the list of equivalent PLMNs and deactivate all the EPS bearer contexts locally, if any, and shall enter the state EMM‐DEREGISTERED.NORMAL‐SERVICE. The UE shall perform a new attach procedure. For some vendor, it had been found if eNB doesn’t enable dynamic QoS modification feature, UE will still use 2Mbps UL/DL bandwidth or be detached directly when it returns LTE form 2G with data traffic. So it is needed to enable the feature so that the different networks can negotiate the user’s QoS dynamically. 10.5.3.2 Implicitly Detach Issue For implicitly detached, the network detaches the UE, without notifying the UE. MME doesn’t inform any other node about it, so when new TAU is coming, it’ll be rejected, and new attach will be performed by UE. This is typically the case when the network presumes that it is not able to communicate with the UE, for example, due to radio conditions. The UE shall delete the list of equivalent PLMNs and delete any mapped EPS security context or partial native EPS security context, and shall enter the state EMM‐DEREGISTERED.NORMAL‐SERVICE. If the rejected request was not for initiating a PDN connection for emergency bearer services, the UE shall perform a new attach procedure. An example shown in Figure 10.54 that as checked UE previous behavior in UMTS network, it is found that UE deactivated PDP context before UE entered LTE network. Further to check the trace signaling in MME, it is found that before UE sends TAU to come back LTE after CSFB, MME receives the “delete bearer request” from SGW and responses the request, so MME deletes the bearer but cannot inform UE, which was in 2/3G, also, that is why the UE TAU fails due to implicitly detach. When UE re‐enters LTE network, UE has no EPS context, and its state in EPC is EMM‐DEREGISTERED, which will cause UE implicitly detach. EPC will reject any behaviors of UE before UE performed a new attach, so “Track area update request” was rejected by EPC (Figure 10.55). UE fallback to UMTS network, when UE has no PS service, PDP context will be deactivated. If UE has PS service, PDP context will be hold on. “Track area update reject” is a normal phenomenon when PDP is deactivated by UE. 10.5.3.3 MS Identity Issue From field tests, sometimes the reason for TAU rejection is “MS identity cannot be derived by the network” cause #9 as shown in Figure 10.56. These cases usually happened for communication problem between SGSN and MME. 10.5.4 Pseudo Base Station The LAU procedure failed after UE fallback to illegal 2/3G base station and call failed. One example shows that the UE starts the CSFB call in the LTE cell and fallback to GSM cell (ARFCN/ BSIC: 67/52,RxLev: −54dBm, LAC: 14555, Cell ID: 4113) in Figure 10.57. Then the LAU request is rejected by the network and call failure. After verification, the GSM cell (LAC:14555, Cell ID:4113) does not exist in the GSM network, which is an illegal GSM base station cell. The solution of these kinds of issues is trying to get rid of the illegal base station or delete the illegal 2G base station frequency point relation in the LTE cell. Network starts the connection release, and at the same time, UE sends an attach request Figure 10.52 Track area update reject. Figure 10.53 Two‐track area update request message. Figure 10.54 Implicitly detach. Figure 10.55 TAU reject caused by implicitly detached. Figure 10.56 MS identity issue. Figure 10.57 Pseudo base station. 434 11 VoLTE Optimization Voice over LTE (VoLTE1) enriches LTE network with more voice‐related capability with common EUTRAN and EPC infrastructure. Common packet‐based mobility is also applicable for both LTE data and VoLTE services. In this book, “VoLTE” is used as the collective notation for both voice and conversational video over LTE. VoLTE voice call is based on GSMA PRD IR.92 and VoLTE video call is based on GSMA PRD IR.94. Introducing VoLTE on IMS provides the service provider with a true converged network where services are available regardless of the access type network. VoLTE provides a first line telephony service with high voice quality and short call setup. Voice and video are using QoS bearers with guaranteed bit rate to secure the service characteristic. Voice over LTE allows very fast call establishment (~1 sec) versus CSFB toward 3G (~5 sec) and even more in case of CSFB toward the GSM (~8 sec). VoLTE avoids 4G data service interruptions and preserves the LTE data experience during speech communications while the throughput of concurrent data session is typically reduced in case of CSFB to 3G and even suspended in case of CSFB to 2G. Table 11.1 depicts the differences between CSFB and VoLTE UE call procedure. The LTE standardization work has established that voice will predominantly be supported by an all‐IP network centralized on the IP multimedia subsystem (IMS). IMS architecture provides integrated voice, data, and multi‐media services interworking between different access networks. IMS‐based VoLTE puts the IMS in the center of the voice core network, managing the connectivity between subscribers and the implementation of policy control. The voice service is then managed by a specially designed VoIP application server. IMS‐based VoLTE is standardized by 3GPP and it’s considered to be the target network infrastructure from long‐ term perspective. Usually initial LTE coverage is non‐contiguous, when UE is out of LTE coverage, single radio voice call continuity (SRVCC,2 voice call handover to CS in GSM or UTRAN) is used to keep the voice call continuity. VoLTE phone can work in a variety of modes, attached to the different networks (2G, 3G and LTE), even also attached to two networks by dual standby terminal and therefore when UE originate/terminate the call, terminal/network needs to select which the network to be accessed. This process is called domain selection according to the network registration information. Terminating access domain selection (T‐ADS) realizes the function of domain selection, T‐ADS is processed by application server to determine if the call is for PS or CS domains. 1 In this part, VoLTE also called non‐native VoLTE, which is SIP client‐based. Third‐party applications can register to IMS system and establish VoLTE call. 2 OTT call might drop in this case. LTE Optimization Engineering Handbook, First Edition. Xincheng Zhang. © 2018 John Wiley & Sons Singapore Pte. Ltd. Published 2018 by John Wiley & Sons Singapore Pte. Ltd. VoLTE Optimization Table 11.1 Difference between CSFB and VoLTE UE call procedure. CSFB UE ‐ LTE access selected VoLTE UE ‐ LTE access selected Detect available network Attach to the EPC and CS network over LTE Setup Internet APN and do some browsing Paging/call preparation between UE and MSC over LTE Place a call/receive a call Detect available network Attach to the LTE network Setup IMS APN and find P‐CSCF(s) Register in IMS Place a call/receive a call (keeping current LTE access) VoLTE signaling and payload packets are supported by VoIP‐specific protocol stacks. E2E connections are managed using SIP with IMS, it requires specific QoS for voice bearers and SIP signaling to achieve performance for a satisfactory user experience and requires specific feature enhancement to achieve capacity and performance. 11.1 VoLTE Architecture and Protocol Stack 11.1.1 VoLTE Architecture Voice service on EUTRA is available when IMS is installed in the core network. The VoLTE network architecture consists of E‐UTRAN, LTE core, PDN, and IMS. It interworks with 3G, which consists of UTRAN, UMTS core, and a circuit switched (CS) network. The MME provides functions that allow LTE and 3G to interwork. Voice service is based on VoIP session, which is controlled by a companion SIP session. Both devices need to be registered on IMS for VoLTE to VoLTE calls. The solution is characterized by a SIP proxy, VoIP application server (AS), and media gateway (MGW) co‐located with the MSC. It uses pre‐configured policy rules on the PDN‐ GW for binding VoIP sessions to EPS bearer and for QoS provisioning. IMS core network is also needed to be with centralized CSCF, HSS, and VoIP AS, using dynamic policy provisioning with co‐located PCRF (Figure 11.1). The P‐CSCF is the first point of contact in IMS system for the UE for mobile access networks. The P‐CSCF forwards the SIP messages received from the UE to an I‐CSCF, E‐CSCF or S‐CSCF (and vice versa). 11.1.2 VoLTE Protocol Stack All new interfaces of VoLTE and the new network protocols are represented in Table 11.2. Protocol stack for VoLTE audio packet, extracted from official document IR.92 – “GSMA PRD IMS Profile for Voice and SMS.” Session initiation protocol (SIP) is a popular protocol used to create, modify, and terminate multimedia sessions, essentially negotiating a media session between two users. SIP is not a transport protocol and does not actually deliver media, leaving that task to RTP/RTCP. Session description protocol (SDP) negotiates the multimedia characteristics of the session between sender and receiver (codecs, addresses, ports, formats, bandwidth require for the session). IMS multimedia uses real‐time transport protocol (RTP) over UDP, RTP was originally defined in 1996 then redefined in RFC 3550 in 2003. RTP added a sequence number in order to 435 436 LTE Optimization Engineering Handbook MSS MGW GERAN/UTRAN Call control signalling User plane traffic Other control signalling SGs D TAS HSS Sv Cx S6a Rx PCRF MME LTE-Uu S1-MME Sh I/S-CSCF BGCF Ma, ISC Mi MGCF Mw P-CSCF Mn Gx S11 Mb eNB S1-U S-GW P-GW MGW Figure 11.1 Nodes for VoLTE. Table 11.2 Interfaces of VoLTE and protocols. Interfaces Components LTE Protocol Sv MSC server – MME(SGSN) GTP‐Cv2 I2* Mj/Mg/Mx MSS‐Server – IMS‐I‐CSCF SIP Cx HSS(NSN)‐IMS Core Diameter Mw P‐CSCF –Core IMS SIP Gm P‐CSCF –UE SIP Mb ATGW‐ MSC server RTP Rx AF‐PCRF Diameter Gx PCEF –PCRF Diameter identify the lost packets. Together with a new timestamp field it allows the receiver to play the packets in the correct order. Other new fields are SSRC (synchronization source, all the packets have the same SSRC identifier indicating that they are from the same source) and CCRC (contribution source) allow the tracking of one or multiple (in case of a conference) sources for the packet. RTP is used in conjunction with the real time transport control protocol (RTCP). While RTP carries the media streams (audio or video), RTCP monitor transmission statistics and quality of service information. RTCP uses a separate flow from RTP. It is transported over UDP as well, and its purpose is to collect statistics on a given media connection including packet loss, jitter, round trip delay, and monitor the quality of the data transmission. RTCP provides feedback on the transmission and reception quality of data carried by RTP periodically. Each RTCP packet always contains either a sender or receiver report, sender report includes an absolute timestamp, VoLTE Optimization IMS Signaling LTE Signaling SIP*/ SigComp/ IPSec VOIP Traffic RTP RTCP UDP TCP NAS IP RRC ROHC PDCP AM RLC AM UM MAC PHY Gm Uu S-GW S1/u UE S1/S8 PDN-GW SGI P-CSCF eNodeB SDPSIP SDPSIP UDP/IP UDP/IP Relay Relay GTPv1-U GTPv1-U GTPv1-U GTPv1-U UDP/IP UDP/IP UDP/IP L2 UDP/IP MAC UDP/IP L2 L2 UDP/IP L2 L2 L2 L1 L1 L1 L1 L1 L1 L1 PDPC PDPC RLC RLC MAC L1 Mb S-GW S1/u Uu UE S1/S8 PDN-GW SGI BGF eNodeB AMR RTP/RTCP RTP/RTCP UDP/IP UDP/IP Relay Relay GTPv1-U PDPC PDPC RLC RLC UDP/IP UDP/IP UDP/IP UDP/IP MAC MAC L2 L2 L2 L1 L1 L1 L1 L1 L2 L1 GTPv1-U GTPv1-U GTPv1-U UDP/IP UDP/IP L2 L2 L1 L1 Figure 11.2 VoLTE‐related protocol stack. to enable synchronization with different streams, receiver report includes number of received packets, to enable QoS determination. For VoLTE, RTCP is not sent during active media transfer but is sent when the call is placed on hold. IMS signaling is carried by SIP messages, which are carried over UDP or TCP, and compared to TCP, UDP has less overhead and no transport layer retransmission (to avoid increase in end‐to‐end delay) (Figure 11.2). The response codes are consistent with, and extend, HTTP 1.1 response codes. There are six classes SIP messages defined: 1xx provisional, 2xx successful, 3xx redirection, 4xx request failure, 5xx server internal error, and 6xx global failure, which are shown in Figure 11.3 and Table 11.3. 437 438 LTE Optimization Engineering Handbook SIP MESSAGES REQUESTS/METHODS REGISTER INVITE ACK BYE CANCEL OPTIONS RESPONSES PROVISIONAL FINAL RFC 3261 100–199 100 Trying 180 Ringing, and 183 Session progress MESSAGE SUBSCRIBE PUBLISH NOTIFY INFO UPDATE REFER PRACK > 199 2xx Success 3xx Redirect 4xx Client Mistake 5xx Server Failure 6xx Global Failure OWN RFCs A Request B Provisional response Provisional response Final response Figure 11.3 SIP messages. 11.1.3 VoLTE Technical Summary VoLTE delivers voice with high QoS (QCI 1, GBR EPS bearer) in transport and radio scheduling, but admission and congestion control may be efficiently used to preserve the voice quality experience in presence of heavy traffic loads, so non‐GBR EPS bearers (QCI = 5) are used for VoLTE Optimization Table 11.3 SIP messages code. 400 Bad Request 401 Unauthorized 402 Payment Required 403 Forbidden 404 Not Found 405 Method Not Allowed 406 Not Acceptable 407 Proxy Authentication Required 200 OK 408 Request Timeout 202 Delivered 409 Conflict 300 Multiple Choices 410 Gone 411 Length Required 301 Moved 413 Request Entity Permanently Too Large 302 Moved 414 Request‐URI Too Long Temporarily 415 Unsupported 305 Use Proxy Media Type 380 Alternative 416 Unsupported Service URI Scheme 100 Trying 180 Ringing 181 Call Is Being Forwarded 182 Queued 183 Session Progress 420 Bad Extension 421 Extension Required 423 Registration Too Brief 480 Temporarily Unavailable 481 Call/Transaction does not exist 482 Loop Detected 483 Too Many Hops 484 Address Incomplete 485 Ambiguous 486 Busy Here 487 Request Terminated 488 Not Acceptable Here 491 Request Pending 493 Undecipherable 500 Server Internal Error 501 Not Implemented 502 Bad Gateway 503 Service Unavailable 504 Gateway Time‐out 505 Version Not Supported 513 Message Too Large 600 Busy Everywhere 603 Decline 604 Does Not Exist Anywhere 606 Not Acceptable SIP and XCAP. In a live network, voice domain selection is IMS PS voice preferred and CS voice is secondary. VoLTE can provide base telephony service and supplementary services and supplementary service management using Ut with XCAP procedures. The main part of VoLTE need focused is described as the following: VoLTE bearer management, includes PDN connection for IMS APN,3 signaling bearer setup, P‐CSCF discovery, home‐routed PDN connection/APN for Ut, handling of loss of PDN connection, signaling, and GBR bearer. IMS feature part, includes ISIM based authentication (USIM fallback), IPSec protection of signaling, both Tel‐URI and SIP URI, SigComp, GBA (recommended) or http digest authentication for Ut, early dialogues and media, and IMS emergency. IMS media, includes AMR narrow‐band and wide‐band codec and payload format, RTP profile/data transport, RTCP usage, and Jitter buffer management. SMS, includes SMS over IP (IMS) and SMSoSGs. Wireless feature as shown in Table 11.4, includes CDRX, semi‐persist scheduling, TTI‐bundling, forward handover with context fetch, and so on. 11.1.4 VoLTE Capability in UE The UE indicates its E‐UTRA capabilities in the UE E‐UTRA capability information element at connection setup (RRC UE capability information message). The FGI (feature group indicator) indicates the functionalities supported by the UE. FGI 3, 7, and 27 will indicate the VoLTE capability of UE, and more information can be found in 3GPP TS 36.331 V9.16.0 (2013‐09), which is shown in Figure 11.4 and Table 11.5. 3 In the VoLTE work in GSMA, access point name (APN) used for IMS services has been defined in GSMA IR.88. The IMS APN shall be added to the list of APN names that are used in the LTE/EPC network, this means that IMS APN is added to the subscriber EPS profile in HSS and policy controller. 439 Table 11.4 VoLTE main wireless feature. Features Description Benefits CDRX Connected mode short DRX allows UE to go to sleep between frames Better talk time HD vocoder Core network and device support for AMR‐WB HD voice with high quality Semi‐persist scheduling(SPS) Efficient scheduling for VoIP‐type traffic System efficiency and capacity TTI‐bundling Bundle 4 TTI together in UL Improve VoIP link budget E‐911 support w/positioning Emergency call with positioning over UP or CP, using E‐CID with TA,OTDOA with PRS and A‐GPS LBS application Voice call continuity(VCC) Voice call continuity to 3G/2G CS and PS domain Needed if LTE coverage is not ubiquitous Forward handover with context fetch Improve handover reliability Robust mobility Bit Definition Note 3 . 5 bit RLC UM SN . 7 bit PDCP SN VoLTE Bit 7 = 1 4 Short DRX cycle Bit 5 = 1 5 Long DRX cycle, DRX command MAC control element 7 RLC UM VoLTE 9 EUTRA RRC_CONNECTED to GERAN GSM_Dedicated handover SRVCC Bit 23 = 1 11 EUTRA RRC_CONNECTED to CDMA2000 1xRTT CS Active handover SRVCC Bit 24 = 1 20 If bit number 7 is set to ‘0’: -SRB1 and SRB2 for DCCH + 8x AM DRB If bit number 7 is set to ‘1’: -SRB1 and SRB2 for DCCH + 8x AM DRB -SRB1 and SRB2 for DCCH + 5x AM DRB + 3x UM DRB 27 . EUTRA RRC_CONNECTED to UTRA FDD or UTRA TDD CELL_DCH CS handover, if the UE supports either only UTRAN FDD or only UTRAN TDD . EUTRA RRC_CONNECTED to UTRA FDD CELL_DCH CS handover, if the UE supports both UTRAN FDD and UTRAN TDD 28 TTI bundling 29 Semi Persistent Scheduling -Regardless of what bit number 7 and bit number 20 is set to, UE shall support at least SRB1 and SRB2 for DCCH + 4x AM DRB -Regardless of what bit number 20 is set to, if bit number 7 is set to ‘1’, UE shall support at least SRB1 and SRB2 for DCCH + 4x AM DRB + 1x UM DRB SRVCC Bit 13 = 1 RRC featureGroupIndicators= ‘11111110 00001101 11011000 10000000’ The indexing starts from index 1, which is the leftmost bit in the field Figure 11.4 FGI bits – extract of VoLTE‐related bits. Figure 11.4 (Continued) 442 LTE Optimization Engineering Handbook Table 11.5 Feature group indicator (bit number 3, 7, and 27). 3 5bit RLC UM SN; 7bit PDCP SN can only be set to 1 if the UE has set bit number 7 to 1. Yes, if UE supports VoLTE 7 RLC UM can only be set to 0 if the UE does not support VoLTE Yes, if UE supports VoLTE ‐ EUTRA RRC_CONNECTED to UTRA FDD/TDD CELL_DCH CS handover, if the UE supports either only UTRAN FDD or only UTRAN TDD ‐ EUTRA RRC_CONNECTED to UTRA FDD CELL_DCH CS handover, if the UE supports both UTRAN FDD and UTRAN TDD related to SRVCC ‐ can only be set to 1 if the UE has set bit number 8 to 1 and supports SR‐VCC from EUTRA defined in TS 24.008 Yes for FDD, if UE supports VoLTE and UTRA FDD 27 11.2 VoIP/Video QoS and Features 11.2.1 VoIP/Video QoS The introduction of VoIP poses a number of challenges over LTE access network. To make VoIP attractive for commercial deployment, the capacity will also be desired to be either comparable to or more than that of legacy circuit system. To provide the quality of CS voice without excessive degrading the system capacity, end‐to‐end quality of service (QoS) support in the wireless and wireline packet network infrastructure is essential. Voice is sensitive to data loss, but robust coding and well‐functioning error concealment units makes voice more tolerant to data loss than the other media types. Voice communication has very stringent end‐to‐end delay requirements and voice is thus very sensitive to delay and jitter. Video telephony is very sensitive to data loss, video telephony requirements on frame errors are almost a factor of 10 less than for voice. Video telephony has also very stringent end‐to‐end delay requirements and is thus very sensitive to delay and jitter. SIP signaling, XCAP4 signaling and other signaling must be error‐free for the transaction to be successful. TCP and SIP retransmissions are the mechanisms that guarantee that the data transfer becomes error free even though the EPS bearer introduces packet losses. SIP signaling is time critical (to allow fast call setup, etc.) and thus have a rank sensitive for delay/jitter, whereas XCAP signaling has less‐stringent delay/jitter requirements. Policy and charging control (PCC) enables QoS supervision and control for the media parts of a SIP session. Policy and charging rule function (PCRF) supports 3GPP standardized PCC procedures and makes policy and charging decisions based on input from user subscription information, services information, and so on. PCRF creates policy rules based on session data and push appropriate policy rules (bandwidth, QoS, traffic flow) to P‐GW, and P‐GW interprets the rules and takes actions to establish required EPS bearers for VoLTE. PCC rules contain service data flow (SDF) description and charging and QoS properties to be applied for the flow identified by the SDF through the Rx and Gx interface. When a VoLTE call is to be setup, a dedicated bearer for voice with QCI1 will be setup, initiated from the P‐CSCF over Rx to PCRF, and then PCRF check the policy control and request 4 XML Configuration Access Protocol (XCAP) is the protocol that is used by the UE to configure various parameters for supplementary services such as call hold and call wait. VoLTE Optimization 1a) SIP INVITE (on SIP bearer) Applic ation 1b) SIP 18x/200 (on SIP bearer) P-CSCF/ IMS AGw 8*) Optional: Bearer established notification LTE radio unit 6a) Session Management Request incl. TFT 2) Application/ Service Info (start/end session) EPS Bearer QoS Rx 7) RRC Reconf. eNodeB 5) E-RAB Setup Request MME EPS Bearer QoS S&P GW 4) Create Bearer Request EPS Bearer QoS 6b) Session Management Response 3) Policy and Charging Rules Provision PCRF EPS Bearer QoS P-CSCF INVTE sip Content-Type: application/sdp Content-Length: 540 v=0 o = sip:+16309798028 s = session c = IN IP4 135.185.13.175 b = CT:1000 t=0 0 m = audio 51306 RTP/AVP 4 a = rtpmap:4 AMR-WB/8000 m = video 7834 RTP/AVP 34 35 a = rtpmap:35 H264/90000 a = rtpmap:34 GVA/90000 a = sendrcv m = audio RTP/RTCP, AMR-WB,BW* m = video RTP/RTCP, H264/GVA,BW* PCRF Audio RTP/RTCP: QCI 1, ARP, GBR Video RTP/RTCP: QCI 6/7, ARP, GBR IMS PGW QCI 5 - SIP SIGNALLING QCI 1 – AMR-WB QCI 6/7 H.264/GVA PGW Packet Filter(s) 1 Packet Filter(s) 1 Packet Filter(s) 1 TFT1 Control Sig IMS (P-CSCF) TFT2 Audio MGW or UE TFT3 Video UE 8 8 8 Figure 11.5 QoS negotiation principles at session setup. the dedicated bearer to the PGW through the Gx interface, and all the way through the network to the UE (Figure 11.5). In LTE, eNB only knows about the service is the QCI, service can be dependent triggered by QCI, for example, scheduler, coverage thresholds (individual bad coverage settings per QCI, for example, QCI1 = −80dBm, QCI9 = −85dBm), RLF settings per QCI, DRX settings per QCI, redirected carrier based on QCI, and handover trigger based on QCI, and so on. Operator also marks DiffServ DSCP (diffserv code point) properly based in the QCI value of the used bearer, which is related to transport QoS. For VoLTE, SIP control signaling and RTP audio packet is assigned with the higher priority of LTE QCI (5 and 1) and DSCP marking over transport network. Transport network QoS is provided by mapping QCI to DSCP for the uplink over S1 and in the downlink for packet forwarding over X2. eNB marks DSCP priority on S1‐U, S1‐MME and X2 interfaces by mapping from LTE QCI, SGW, and PGW marks DSCP priority on S5/S8 and SGi interface by mapping from LTE QCI. 443 444 LTE Optimization Engineering Handbook Example: QCI 5 and 1 to DSCP mapping SIP Speech 11.2.2 Voice Codec For VoLTE UE to VoLTE UE call, voice codec is negotiated by the UEs during the SIP session opening. They have to select a common type. Possible codecs for voice include EVCR‐A, EVCR‐B, AMR, AMR‐WB5 (AMR‐wideband) and even EVS.6 For instance, AMR 12.2, which RTP payload size is 32 bytes, AMR‐WB 12.65 RTP payload size is 33 bytes, AMR WB 23.85 RTP payload size is 60 bytes, and SID RTP payload size is 7 bytes. The UE must be able to operate with any subset of the 8 modes for AMR and any subset of the 9 modes for AMR‐WB. Tandem free operation (TFO) and transcoder free operation (TrFO) must be supported in the IMS core network for CS interworking. VoLTE performance specifications, including video, is listed below. Video quality in terms of the objective estimate is from PEVQ (ITU‐T J.247). Rule of thumb and need to optimize and fine‐tune per operator basis is relevant as a service KPI requirement (Figure 11.6): ●● ●● ●● Speech latency (end‐to‐end delay) requirement is from standard ITU‐T G.114 and 3GPP TS 22.105, no more than 150 ms is preferred, at maximum 400 ms. Packet loss, E2E conversational voice packet loss rate is defined as 1% in TS23.203 for VoLTE audio packet transported between UE and PGW, at maximum 3% FER), video telephone at maximum 1% FER, and data service requires 0% FER. Radio performance is linked to speech quality (MoS) through: M-to-E 300 ms Users very satisfied ec d Co 225 ms 64 ps .2 12 Users satisfied ps kb Some users dis-satisfied kb FER 1% Many users dis-satisfied 5% Figure 11.6 Radio performance is linked to codec, FER, and mouth‐to‐ear delay. 5 AMR‐WB specification: 3GPP TS 26.171 and GSMA IR.36. 6 EVS, codec for enhanced voice services, the EVS standard is the first 3GPP codec to deliver speech and audio in super‐wideband full‐HD voice quality, bringing mobile audio on par with the audio experienced through today‘s digital media services. The algorithm delivers speech and audio up to 20 kHz audio bandwidth, outperforming the audio quality of today’s mobile phone calls. VoLTE Optimization ●● ●● ●● ●● ●● Codec, AMR 12.2 kbps and above provides good speech quality, Frame erasure rate (FER), less than 1% provides good speech quality, Mouth‐to‐ear delay, less than 225 ms provides good speech quality. The packet delay budget (PDB) requires no more than 80 ms for voice, and no more than 130 ms for video. PDB is measured from entering PDCP in eNB to leaving PDCP in the UE and vice versa, which includes 20 ms budget for interactions between PCRF, PGW, and eNB (20 ms (PCRF‐eNB) and 80 ms (air interface). The Number of HARQ retransmissions should be considered that cannot exceed PDB 80 ms in live network. Call setup time is 3 to 6 sec. Video path delay (camera‐to‐display delay) should less than 400 ms. Video frame loss: as well as video frame rate > = 25 Hz and video frame loss <0.2%7 (less than one visible degradation per 20 sec “watching” sequence), unlike for speech, most common is to split a video frame over to RTP packets, hence when looking at the residual packet loss ration (PLR), the target is to be <0.1% for the video call, assuming evenly distributed random packet loss and two packets per video frame. The maximum allowed video frame loss is 1% FER. Video path delay and audio‐video “lip” sync, the audio shall not be more than 25 ms ahead of the video, and video shall at the most be <60 ms ahead of speech to perceive acceptance lip‐ sync performance. During VoLTE call setup, codec negotiation is needed. End‐to‐end codec negotiation with SIP(‐I)‐based networks and HD voice across network borders is enabled. All 3GPP codecs currently supported by SIP(‐I) and G.729 can be negotiated end‐to‐end. Codecs are negotiated between UEs at session setup and potentially at session modification using SDP offer/answer over SIP. The UEs may also explicitly negotiate max bandwidth to be used for voice and video. For example, here are the user accepts and response with DTMF during a VoLTE call. UE1 initiates a call by sending invite message to a SIP client (DTMF host), SDP contains narrowband and wideband codec support for both speech and DTMF (telephone‐event). UE2 responds with 200 OK and receives ACK; 200 OK contains a SDP answer that selects AMR‐ WB as the codec to be used in this call. UE1 INVITE SDP m=audio 49152 RTP/AVPF 97 98 99 100 101 102 a=rtpmap:97 AMR‐WB/16000/1 a=fmtp:97 mode‐change‐ capability=2; max‐red=22 0 a=rtpmap:98 AMR‐WB/16000/1 a=fmtp:98 mode‐change‐ capabi1ity=2; max‐red=22Q; octet‐ align=l a=rtpmap:99 telephone‐event/16000/1 a=fmtp:99 0‐15 a=rtpmap:100 AMR/8000/1 UE2 200 OK SDP m=audio 4 9152 RTP/AVPF 97 98 99 100 101 102 a=rtpmap:97 AMR‐WB/16000/1 a=fmtp:97 mode‐change‐ capability=2; max‐red=220 a=rtpmap:98 AMR‐WB/16000/1 a=fmtp:98 mode‐change‐ capability=2;max‐red=220;octet‐ align=l a=rtpmap:99 telephone‐event/16000/1 a=fmtp:99 0‐15 a=rtpmap:100 AMR/8000/1 7 The measured time between visible degradations for the end user should be once per 20 seconds for video telephony (maybe longer for “premium” video conferencing services.” Given one packet per frame and a frame rate of 25 Hz, that would correspond to a frame error rate of 1/500 = 0.2% end‐to‐end. 445 446 LTE Optimization Engineering Handbook a=fmtp:100 mode‐change‐ capabi1ity=2; max‐red=220 a=rtpmap:101 AMR/8000/1 a=fmtp:101 mode‐change‐ capabi1ity=2; max‐red=220; octet‐ a1ign=1 a=rtpmap:102 telephone—event/8000/I a=fmtp:102 0‐15 a=ptime:20 a=maxptime:24 0 a=sendrecv a=fmtp:100 mode‐change‐ capability=2; max‐red=220 a=rtpmap:101 AMR/8000/1 a=fmtp:101 mode‐change‐capability=2; max‐red=220; octet‐ align=l a=rtpmap:102 telephone‐event/8000/1 a=fmtp:102 0‐15 a=ptime:20 a=maxptime:240 a=sendrecv 11.2.3 Video Codec A conversational video8 service session is an additional IMS service that can be added or removed by the end user as complement to a voice session in IMS or it can be established with voice at the same time. The 3GPP standards offer a variety of terminal, radio, and core network configuration options when launching IMS voice and video services over an LTE/EPC network. Therefore, GSMA has defined two service profiles. IR.92 defines the conversational voice profile and IP.94 defines the conversational video profile. These two profiles define a minimum set of features to be implemented in the terminal and network (Figure 11.7). IR.92/94 phone MMTEL services Supplementary services – Same as IR.92. – Add/drop of video media also in multi-party conferences IMS Features IMS – Video call setup. – Add/drop of video media to established voice call. – “video” media feature tag in SIP headers. IMS Media IMS media – Voice media same as IR.92. – H.264 level 1.2 as mandatory video codec. – RTCP for lip sync and video codec control. EPC EPS bearer management – Voice media as IR.92. – Video media using QCI = 2 GBR, or a non-GBR bearer. LTE LTE radio capabilities – Support for video. – EPS bearer with QCI = 2 implemented with RLC AM or RLC UM data radio bearers. Figure 11.7 IR.94 profile. 8 Video telephony is a service providing interactive communication with a two‐way full‐duplex audio and associated video image, with the output audio and video temporally synchronized. VoLTE Optimization 3GPP has standardized that QCI1 is used for the transmission of voice RTP packets and QCI2 is used for the transmission of video RTP packets. QCI1 and QCI2 will be triggered from PCRF on demand, while QCI2 will always exist together with QCI1. QCI1 and QCI2 will both carry RTCP packets. Actually, video can either be supported on a GBR bearer (QCI2) or on a non GBR bearer9 (QCI6/8). Video standards are mainly from: ●● ●● ●● ITU‐T (International Telecommunication Union, H series standards, e.g., H.261, H.263, and H.264). MPEG (Motion Picture Experts Group), formed by ISO (International Organization for Standards) and IEC (International Electrotechnical Commission), MPEG standards are MPEG1, MPEG2, and MPEG4. 3GPP, support of ITU‐T recommendation H.264 Constrained Baseline Profile (CBP) Level 1.2 as specified in 3GPP TS 26.114, is mandatory in the UE and in the IMS core network. CBP primarily for low‐cost applications, this profile is most typically used in videoconferencing and mobile applications. It corresponds to the subset of features that are in common between the baseline, main, and high profiles. The recommended video codec for conversational video service in VoLTE is H.264.10 It is recommended to use a low complexity profile (baseline profile level) of that video codec. The frame rate should be 25 Hz. Given a video format of 240p, the widescreen format of QVGA, a video bitrate of more than 350 kbps is the target for medium‐motion video and more than 225 kbps for low‐motion content, when shown in native format. If the video is shown upscaled, or if a higher video format is used, significantly higher video bitrates are required. For VGA video format, H.264 video codec, 480p (VGA) native video format (applicable for a smartphone), it is recommended video bitrate should be higher than500 kbps. Table 11.6 shows the various video resolution and the required bitrate. 3G video telephony (176 x 144 pixel = > 25344 pix) requires 64 kbit stream, 12 frames per second (FPS). Video over LTE, two levels of video quality is described for video telephony in this book: one is the lower quality level, which is corresponding to the minimum requirement for video quality specified in GSMA IR.94 that needs to be mentioned due to its reference in Table 11.6 Video telephony bitrate. Level Resolution FPS Requested BW (kbps) 3.1 720P (1280 × 720) 30 2176 3.0 VGA (640 × 480) 30 1216 2.2 VGA (640 × 480) 15 896 1.3 QVGA (320 × 240) 30 640 1.2 QVGA (320 × 240) 15 424 9 Video will in some deployments be sent over a non‐GBR dedicated EPS bearer with, for example, QCI 6 or 8, which uses the same configuration as for the internet data services. 10 H.264 is perhaps best known as being one of the video encoding standards for Blu‐ray Discs, all Blu‐ray Disc players must be able to decode H.264. It is also widely used by streaming internet sources, such as videos from Vimeo, YouTube, and the iTunes Store, web software such as the Adobe Flash Player and Microsoft Silverlight, and also various HDTV broadcasts over terrestrial (ATSC, ISDB‐T, DVB‐T or DVB‐T2), cable (DVB‐C), and satellite (DVB‐S and DVB‐S2). 11 Prefix Q means a quarter of the original resolution, for example, the QVGA resolution is 25% of the VGA resolution. The Q formats are more suitable for handheld devices due to the lower resolution. 447 448 LTE Optimization Engineering Handbook Table 11.7 The recommended video properties. Property VGA resolution Recommendation QVGA resolution (IR.94 minimum requirement) video format (Picture size) VGA (640 × 480) QVGA (320 × 240) video codec At least H.264 Constrained Baseline Profile (3.0) or Constrained High Profile At least H.264 Constrained Baseline Profile (1.2), low complexity video refresh 25 or 30 fps recommended At least 15 fps for the video telephony service Min bit rate 225 kbps (based on ER quality tests) 125 kbps (based on ER quality tests) Recommended bit rate > = 500 kbps > = 350 kbps Video path delay (VPD) Residual packet loss ratio Session Setup time ROHC Scheduler TTI bundling <400 ms <0.1% <4.0 sec ROHC will not be used on the video flow Proportional fair scheduler with minimum rate, rate adaptation Also be used for the video transmission the IR.94 document. This quality level is optimized for operation using a resolution of QVGA11 (320 × 240 pixels). The other higher‐quality level is the preferred level given today’s smartphones with large high‐resolution screens, it is optimized for operation using a VGA resolution (640 × 480) pixels. The following recommendations of video (as per IR. 94) on bit‐rate are suggested in Table 11.7: ●● ●● Provided QVGA, H.26412 Constrained Baseline Profile (CBP), 12.5 or 15 fps (frames per second) shall be higher than 350 kbps in order to achieve a good video quality. VGA, H.264 CBP Level 2.2 or H.264 Constrained High Profile (CHP), 25 or 30 fps, shall be higher than 500 kbps in order to achieve a good video quality. It is important to understand that bitrate with video does not work as bitrate with speech. For speech, the codec choice commonly implies an exact bitrates. For video, the specifications support video from a few kbps up to several hundred Mbps. Even when a specific implementation has expressed a certain level limit, that limit is a maximum limit and any bitrate lower than the limit can be received and may be sent. Video is inherently variable‐rate, the rate control algorithm that allows setting a target bitrate and may also control the bitrate variability. This rate control is not standardized, so there is simply no guaranteed behavior. An average variation per 10 seconds or more video of 5% from an encoder must be considered good constant bitrate video. The peak rate overshoot per 1 second moving average can be 50% and the instaneous bit rate overshoot can be up to 200% when I‐frames are sent. 12 H.264 also known as advanced video coding (AVC), MPEG‐4 Part‐10 or Joint Video Team (JVT) is the most advanced video codec, was standardized in 2003. H.264 is for video compression and is currently one of the most commonly used formats. It is capable of providing good video quality at substantially lower bitrates than H.263 standards. VoLTE Optimization Table 11.8 Effective bitrates. Codec bitrate [kbps] Effective bitrate [kbps] 110 120.1 220 240.2 384 404.2 When it comes to definition of good video quality bitrate is definitely an important factor. Bitrate requirements for video for a given resolution and frame rate depends a lot of the content. Typical low motion video is of the type “talking heads.” Normal usage more likely represents the “medium motion” class. Extensive subjective experiments hint at a preferred bitrate of >350 kbps for handheld devices to meet quality requirements, large tablets will even require higher bitrates for user satisfaction. Actually, the VoLTE device shall dynamically optimize the video encoding to adapt the network bandwidth conditions. Optimally, the client adapts the codec rate to the changed conditions by increasing robustness and lowering the data rate. The most probable cause for this situation is that the radio cannot sustain the guaranteed bit rate for a particular UE due to too low SINR for that particular UE or congestion in the cell. For video, the effective bitrate is calculated as below: Example: calculate the effective bitrate for a codec bitrate of 384 kbps with a frame rate of 30 fps. The frame size is 384000/(8*30) = 1600 bytes, which exceeds the MTU size. Hence this frame needs to be split into two RTP packets, each one with an overhead of 42 bytes. The effective bitrate is (1600 + 2*42)*30*8/1000 = 404.2 kbps. Table 11.8 shows effective bitrates for a set of codec bitrates used for video. One key service KPI is the video quality, measured in the unit MoS (mean opinion score) derived from subjective viewing tests. Video quality depends on different network and terminal capabilities like screen size and resolution, video codec, frame rate as well on the packet loss, delay and jitter in the EPS, and IP backbone. HQ video does require high bitrates and has a high elasticity, which means a high ariation in bandwidth requirements. As shown in Figure 11.8, video bitrates variations related to MOS and different type of motion, there are high bandwidth requirements in case of slow and medium motion13 in combination with an acceptable MOS 3.5. Actually, extensive subjective experiments hint at a preferred bitrate of more than 350 kbps for handheld devices to meet quality requirements, large tablets will require higher bitrates for user satisfaction. Video bitrate demands for high quality video H.264 VC is shown in Figure 11.9. One LTE‐related issue for video telephony is the fact that there are only four logical channel groups in LTE, and one is reserved for radio layer/NAS signaling. Hence, there are only three groups to use for VoLTE and internet services. The probable setting is to have QCI1 and two in same group and QCI = 5 in a separate group. Video telephony user plane of calling side is shown in Figure 11.10. 11.2.4 Radio Bearer for VoLTE For VoLTE, PDN connection (EPS session) handles flows of the IP packets, that are labeled with UE IP addresses, between a UE and a PDN, which is represented by UE IP address and 13 Typical low motion video is of the type “talking heads.” Normal usage more likely represents the “medium motion” class. 449 MOS 5.0 4.5 4.0 3.5 3.0 H.264 MoS scores for QVGA (handheld device) 2.5 Low Motion 2.0 Medium Motion 1.5 1.0 0 125 250 375 500 [kbps] Figure 11.8 Video bitrates variations related to MoS and different type of motion. 4.5 Typical target to be above MOS-VQS 3.5 4.0 3.5 360p upscale 25Hz 480p 25Hz 480p=864 x 480 pixel 3.0 2.5 Typical target to be above 2.0 1.5 225 350 500 650 800 1000 1500 kbps “S-shape” distribution observation when high and low end device implementations are used to play content with varying encoding difficulty Simple content (white shade) Difficult content (grey shade) 4 suggested MOS target 3 300–500 kbps 1–1.5 Mbps Difficult content (sport) Easy content (News) Video bitrate to be above (prerequisite high-end VC implementation) 2 1 0.5 1.0 1.5 2.0 Figure 11.9 Video bitrate demands for high quality video H.264 VC. 2.5 Mbps eNB UE MME PCRF PGW SGW RTP Audio DL stream RTCP Audio DL stream QCl 2 RTP Video DL stream RTCP Video DL stream S5-U DL TEID, 4 associated TFT, Bearer QoS QCl 2 UE IP address listening on: part1 for RTP audio, part1 + 1 for RTCP audio, part2 for RTP video, part2 + 1 for RTCP video, Sending on: part3 for RTP audio, part3 + 1 for RTCP audio, part4 for RTP video, part4 + 1 for RTCP video, S1-U DL TEID QCl 2 Dedicated Radio Bearers QCl 2 S1-U UL TEID QCl 2 S5-U UL TEID, 4 associated TFT, Bearer QoS RTP Audio UL stream RTCP Audio UL stream QCl 2 RTP Video UL stream RTCP Video UL stream Figure 11.10 Video telephony user plane. IMS 452 LTE Optimization Engineering Handbook APN (access point name, IMS APN is for VoLTE). EPS bearer is a pipe through which IP packets are delivered over the LTE network between UE and P‐GW. UE can have multiple EPS bearers concurrently (up to 11 EPS bearers, EBI range is 5‐15, which is allocated by an MME). E‐RAB is a bearer between UE and S‐GW, and consists of a DRB and an S1 bearer. Different E‐RABs are identified by their E‐RAB ID, which is allocated by an MME, E‐RAB ID = EBI. At least the following radio bearer combination at UE and eNB are required for VoLTE: SRB1 + SRB2 + 2x AM DRB + 1x UM DRB. SRB1 and SRB2 is required for RRC and NAS signaling. One acknowledged mode (AM) DRB is required as the default bearer (QCI‐8/9), one AM DRB is required for SIP signaling (QCI‐5), and one unacknowledged mode (UM) DRB is required for VoIP traffic (QCI‐1). SIP signaling default bearer14 is dedicated bearer that is allocated during attach time. SIP signaling with QCI 515 over the default bearer is always established on IMS PDN/PGW to reduce extra message handshaking and call setup delay. Additional dedicated bearer with QCI 1 for VoIP RTP bearer or QCI 6/7 for video bearer is allocated on the same IMS PDN/PGW. Figure 11.11 is an illustration of default and dedicated bearers. S1 Bearer Radio Bearer S5/S8 Bearer E-RAB EPS Bearer PDN UE eNB LTE-Uu S1 S-GW P-GW S5 SGi EPS Bearer ID (Default Bearer) = 5 E-RAB ID = 5 DRB ID IP packet IP packet DL S5 TEID DL S1 TEID DRB ID UL S5 TEID UL S1 TEID EPS Bearer ID (Dedicarted Bearer) = 10, LBI = 5 (Default EPS Bearer ID) E-RAB ID = 10 IP packet DRB ID DRB ID UL S1 TEID DL S1 TEID UL S5 TEID DL S5 TEID IP packet Figure 11.11 Default and dedicated bearers and mapping among EPS bearer IDs. 14 A default bearer is the first EPS bearer established to the network and maintained until UE is switched off or out of coverage, at least one default bearer remains active. A dedicated bearer is established by network to allow flow of traffic between UE and PGW and is maintained until data is transferred (various QoS). An EPS bearer is composed of three parts: the data radio bearer (over the air), the S1‐U bearer (between eNB and S‐GW), and the S5 bearer (between S‐GW and P‐GW). Dedicated bearer acts as an additional bearer on top of default bearers, which provide the tunnel to one or more traffic (VoLTE, video…), it can be GBR or non‐GBR. 15 GSMA IR.92 defines that QCI5 will be the default bearer for SIP signaling/VoLTE services. The UE sets up the default bearer by attaching the IMS APN. Once the IMS APN default bearer (QCI5) is established, the UE can register in the IMS system using SIP signaling. Upon successful registration the UE will make and receive calls using SIP signaling over QCI5. VoLTE Optimization Table 11.9 RB mapping of the logical channels to the UL logical channel groups. LCG ID 0 IR92 1 Scheduling strategy SRBs with highest priority VoLTE 2 QCI5 QCI1 QCI2 Prio. < QCI1 and QCI5 Prio. > QCI1 QCI1 3 QCI5 Prio. > QCI5 HighCap QCI5 Prio. > QCI1 QCI1 QCI 7/8/9 Prio. < QCI2 QCI2 Prio. > QCI2 QCI5 scheduled with non‐GBR, low delay, very low packet loss rate QoS QCI1 scheduled with GBR/MBR DL/UL = 40 kbps, low delay, moderate packet loss rate, uses RLC‐UM and usually only 2 HARQ transmissions QCI2 scheduled with minimum rate proportional fair 400Kbps QCI 7/8/9 scheduled with minimum rate proportional fair 0 kbps The default EPS bearer keeps the UE connected to the network. When there is no user traffic and thus the UE state changes to idle, E‐RAB is deactivated and only the S5 bearer stays on. However, as soon as new user traffic arrives, E‐RAB is re‐established, allowing the traffic to be delivered between the UE and the P‐GW. For uplink, different RB mapping of the logical channels to the UL logical channel groups, QoS for default bearer is provided by HSS, QoS for dedicated bearer is provided by PCRF, shown in Table 11.9. LTE supports up to two VoIP bearers per UE. For UE with two VoIP bearers established and semi‐persistent scheduling (SPS) method is configured, SPS transport block size (TBS) shall be the sum or the maximum of the payload size of the two VoIP bearers. The downlink semi‐ persistent scheduler shall multiplex MAC SDU of two VoIP bearers into one MAC PDU in each SPS period. In the uplink direction, eNB MAC layer shall be able to demultiplex two MAC SDUs from one MAC PDU and deliver each MAC SDU to corresponding RLC entity for each VoIP bearer. The following voice RB metrics are mainly used to measure the quality of service for VoIP. One is bandwidth, the other is E2E delay. The bandwidth depends on the type of voice codec. Generally, the required bandwidth for a single call, one direction, is 12.2 kbps for AMR‐NB codec or 23.85 kbps for AMR‐WB codec. Typically, these codec delivers one voice packet each 20 ms. For AMR‐NB, each packet is sent in one ethernet frame. With every packet of size 256 bits, headers of additional protocol layers are added. These headers include RTP + UDP + IP headers and PDCP/RLC/MAC headers with preamble of sizes 32 + 8 + 8 + 8, respectively. Therefore, a total of 312 bits needs to be transmitted 50 times per second in one direction. The guaranteed bit rate (GBR) for VoLTE is determined by the formula below: VAD BW 1 VAD VoTE _ IP _ PayloadSize Overhead Voice _ SamplingRate Silence _ IP _ PayloadSize Overhead Silence _ SamplingRate 8bit / byte 453 454 LTE Optimization Engineering Handbook Where: ●● ●● ●● ●● ●● ●● ●● ●● VAD is the voice activity factor from 0% to 100% Voice_SamplingRate is the frame transmission frequency per standards: 20 ms Silence_SamplingRate is the silent indicator: 160 ms VoLTE_IP_PayloadSize (32 bytes) + Overhead (40 bytes) for AMR 12.2 kbps = 72 bytes Silence_IP_PayloadSize (5 Bytes) + Overhead (40 bytes) for AMR 12.2 kbps = 45 bytes Number of packets per second: 1000/20 = 50 The GBR for a VoLTE call with 100% activity factor using only AMR 12.2 kbps codec mode should consume a bandwidth of 28.8 kbps BW = [1 × (72)/20 ms + (1 – 1) × (45)/160 ms] × 8 bit/byte = 28.8 kbit/s From Figure 11.12, we can see that the average rate of QCI 1 is 11.8 kbps, QCI 5 is 10.5 kbps, the defaulkt QCI 9 is only 0.04 kbps in a field test. End‐to‐end latency including bearer delay (RTP packets) and signaling delay (SIP signaling). Table 11.10 shows the typical voice delay budget. According to ITU G.114, less than 200 ms of mouth‐to‐ear audio packet delay is needed to make the user very satisfied and no more than 280 ms is needed to make the user satisfied. 11.2.5 RLC UM An RLC entity can be configured to operate in TM, UM, or AM mode, which are shown in Table 11.11. There are RLC services and RLC functions. The RLC functions that are performed by the RLC entities are concatenation, padding, data transfer, error correction, in‐sequence delivery, duplicate detection, flow control, RLC re‐establishment, protocol error detection, and recovery. The RLC UM (unacknowledged) mode provides a unidirectional data transfer service without sending any feedback to the transmitting entity, therefore, there are no re‐transmissions of packets. UM RLC is mainly utilized by delay‐sensitive and error‐tolerant real‐time applications, especially VoIP, and other delayed sensitive streaming services. RLC UM mode can be configured per bearer type, eNB will choose RLC mode to use based on the bearer setup procedure and QoS requirements. RLC UM is useful for services that tolerate a higher packet loss rate but require lower latency, unlike RLC AM with a OTT ARQ reordering mechanism by RLC ARQ. RLC UM is “in‐ascending‐order delivery,” which unlike RLC AM with “in‐sequence order delivery,” same but gaps allowed (Figure 11.13). The RLC UM machine starts re‐ordering when detecting a gap, same as RLC AM. But when the timer expires, UM does not, like AM, do its own retransmission but rather trusts lower layers to have provided enough time for re‐ordering. RLC UM ignores any gaps that might still occur and continues to deliver in ascending order to higher layers. If any missing RLC SDUs arrive later they will be discarded since they arrive outside the re‐ordering window. The RLC mode is configured for each QCI by eNB. If the source eNB has a bearer with QCI configured with the RLC UM feature and the target eNB has mapped the same QCI to RLC AM, the bearer is rejected. Data forwarding at intra‐LTE (X2) handover for RLC (UM) Data forwarding is executed in user plane (UP) tunnels, established between the source eNB and the target eNB during the handover preparation. Data forwarding must be supported for X2 handover for both UM and AM bearers. Also X2 handovers including data forwarding will work for VoIP only and for the combination of VoIP and data simultaneously. There is one tunnel established for DL data forwarding per each E‐RAB for which data forwarding is applied (Figure 11.14). 35 30 25 20 15 10 5 0 100 200 300 PDCPThrput Qci1_UL(kbps) 400 500 600 PDCPThrput Qci5_UL(kbps) 700 0 1000 PDCPThrput Qci9_UL(kbps) 100 Bit rate (Kbit/s) Number of bits [Kbit] 15 SS RE _ _IN N SIO ES _S 3 _18 OG PR SIP 10 50 0 20 _ SIP ING ING _OK R _ 0 180 P_20 IP_ SI K _O S SIP message Average rate 5 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Time 0 Figure 11.12 Data rate of RB in field test. Table 11.10 Voice delay budget. Delay component Range(ms) Comments UE delay (UL/DL) 31‐44 Air interface one‐way delay(UL/DL) 4 ~ 53/1 ~ 37 For 6 HARQ transmission using dynamic scheduling as the worse case eNB delay 2~4 eNB processing delay for packet L1/L2 processing SGW 0.1 ~ 0.5 Packet forwarding PGW 0.1 ~ 0.5 S1‐U 2 ~ 15 S5 2 ~ 15 IP network 18 ~ 41 Propagation delay is mainly proportional to distance(5us/km) Assuming 2 ms processing and queuing and 2000 miles OC3 456 LTE Optimization Engineering Handbook Table 11.11 Three modes of RLC. Transparent Mode Unacknowledged Mode Acknowledged Mode No segmentation and reassembly of RLC SDUs Segmentation and reassembly of RLC SDUs Segmentation and reassembly of RLC SDUs No RLC headers are added RLC headers are added RLC headers are added No delivery guarantees No delivery guarantees Reliable in sequence delivery service Suitable for carrying voice Suitable for carrying streaming traffic Suitable for carrying TCP traffic Transmitter Transmission buffer New RLC AMD PDU-s Receiver Link MUX DEMUX Insequence RLC AMD PDU-s Re-ordering “3” was lost (MAC) Backpressure 1 2 3 4 1 2 4 Time Figure 11.13 RLC UM insequence order delivery. UE Source eNB Target eNB MME SGW Handover request Handover Request ack RRC conn reconf RRC conn reconf Cmp SN Status Transfer Path switch request Modify bearer request GTP-U Modify bearer response UE context release Path switch request ack Figure 11.14 Signaling flow (data forwarding at X2 handover for RLC_UM). VoLTE Optimization RLC UM will be applied to VoLTE (QCI 1 or QCI 2). The RLC in UM feature provides the means to deliver VoIP without delays due to retransmission and may be configured to reduce the control signaling overhead of RLC and PCPC layers. Another advantage is a faster delivery but with a higher risk of packet loss. The QCI = 2 characteristics defined in 23.203 (PELR ≤ 10−3, PDB ≤ 150 ms) can be realized with RLC UM, but also with RLC AM, as RLC AM minimizes packet loss at the expense of an increased delay. However, using RLC AM may give rise to delay spikes, which are caused by the RLC AM protocol while it waits for a retransmission. RLC AM will give rise to quite a few RLC status reports as the overhead is insignificant since video packets are relatively larger. The RLC status reports will just get a “free ride” with the data packets of the reverse direction of the bi‐ directional video flow. The operator can improve the network performance by optimizing RLC parameters, including the maximum number of ARQ retransmission, the length of poll retransmit timer, and especially SN lengths of RLC PDU and PDCP PDU, which are described below. It is recommended to configure the VOIP media bearer (RLC UM) to reduce the control signaling overhead of RLC and PDCP layers by shortening the default sequence number length from 12 bits to 5 bits for RLC and to 7 for PDCP, as specified in 3GPP TS36.322 and TS 36.323,16 respectively. Further information can be found in the reference RLC in unacknowledged mode. The relevant eNB parameters are: ●● ●● rlcSNLength, 5 bits (with rlcMode = UM) pdcpSNLength, 7 bits (with rlcMode = UM) For RLC UM bearers, both RLC and PDCP SN are reset at handover and eNB starts sending UM packets with PDCP SN starting from 0 in the target. Actually, improper configuration for pdcpSNLength and rlcSNLength would also cause negative effects. Setting them to abnormally small values will lead to wraparound on the sequence numbers and hence increase the drop rate if RLC/PDCP SN length for QCI 1 has a mismatch in source and target cell.17 If they are set differently in serving cell and target cell, it would result in handover failures or drops after handover. 11.2.6 Call Procedure VoLTE services high level steps include detect available network, attach to the LTE network, setup IMS APN and find P‐CSCF, register in IMS, and place/receive a call. From the attach procedure, UE may include capabilities such as UE SRVCC capability in the “Attach request” message. In the “Attach accept” message as well as in the “Tracking area update accept” message, MME provides information to the UE on the network capabilities, for example, emergency service support indicator, location service support indicator, and IMS voice over PS session supported indicator. When the PDN connection for the IMS APN is established in the setup IMS APN and find P‐CSCF procedure, the list of P‐CSCF addresses are provided by the PGW to the UE in the “PDN connectivity accept” message and the IMS APN default bearer QCI 5 is established for the IMS signaling. After register in IMS, MO/MT call triggers UE to do service request to move UE to connected mode and then it can send and receive user plane traffic (SIP invite) (Figure 11.15). 16 3GPP TS36.322: Radio link control (RLC) protocol specification. 3GPP TS36.323: Packet data convergence protocol (PDCP) specification 17 In a live network, In some cases it was observed that PDCP SN length was same in both the source and target cell but the HO failed. 457 458 LTE Optimization Engineering Handbook MME eNB S/PGW HSS PCRF CSCF AS Radio setup Registration EPS Registration, Default bearer setup and VoLTE support discovery IMS registration and user authentication IMS VolP session setup Call establishment EPS dedicated bearer setup Ongoing voice call Call release IMS VolP session and dedicated bearer release Figure 11.15 High level e2e VoLTE call flow. 11.2.6.1 LTE Attach and IMS Register As part of attach procedures, VoLTE‐related capabilities and relevant network information (voice and emergency support, and P‐CSCF address) are provided between UE and network. If the IMS APN is not default APN, the UE will initiate a separate PDN connection after attach to establish the IMS PDN. In the attach request message the UE signals if it supports SRVCC to GERAN/UTRAN, or if the UE is performing a combined attach procedure. If it supports SRVCC, it indicates its supported speech codecs for CS speech calls. In the attach accept message the UE is informed whether the network supports IMS VoIP, emergency calls, and EPS/CS location services (Figure 11.16). The UE need to register (authenticate) toward the IMS before any services can be utilized. The registration may take place any time after the UE has attached to LTE and acquired an IP address to P‐CSCF. Once the LTE attach is performed, the terminal start registration at the IMS (S‐CSCF) to be able to receive IMS services. Prior to registering with IMS the UE must establish another PDN connection to the IMS APN by a SIP registration request. VoLTE mandates the use of IMS AKA (i.e., UICC based authentication based on same authentication mechanism as for the access). As a consequence of the authentication, IPsec is established for SIP signaling security between UE and P‐CSCF (Figure 11.17). UE needs to perform re‐registration procedure before registration timer expires. For VoLTE, the application servers also need to know about registration status through a third‐party register. The registration may take place any time after the UE has attached to LTE and acquired an IP address to P‐CSCF. If the user profile includes a service trigger for register, the S‐CSCF sends a third‐party register message to the application server indicated in the trigger, to inform the application server about the change in registration status. The IMS registration latency is: IMS registration latency [ms] = T200 OK − TREGISTER where TREGISTER is the timestamp for SIP register (step 3) being sent from the UE (start trigger for the registration procedure), T200 OK is the timestamp for 200 OK (Figure 11.18). 11.2.6.2 E2E IMS Flow After register to the IMS domain with a SIP register message the UE can initiate a VoLTE session, using SIP invite. The SIP invite is used to find the called party, and negotiate the media to be used. Figure 11.19 shows that the originating (MO) call triggers UE to do a service request VoLTE Optimization UE Attach MME eNB NAS:Attach Request HSS EPG SAPC P-CSCF Diameter:AIR Diameter:AIA NAS:Authentication Request NAS:Authentication Response NAS:Security Mode Command NAS:Security Mode Complete GTPv2:Create Session Request Diameter:CCR EBI:5, QCI:9(Default Bearer) GTPv2:Create Session Response Diameter:CCA S1AP:Initial Context Setup EBI:5 , QCI:9(Default Bearer) EBI:5, QCI:9(Default Bearer) NAS:Attach Accept, activate default EPS bearer request GTPv2:Create Bearer Request EBI:5, QCI:9(Default Bearer) Linked EBI:5, QCI:5(IMS SIP Signaling) S1AP:UE capability info indication S1AP:Initial Context Setup Response NAS:Attach Complete, activate default EPS bearer accept GTPv2:Modify Bearer Request GTPv2:Modify Bearer Response Default Bearer(EBI:5, QCI:9) is established S1AP:E-RAB Setup Request EBI:6, QCI:5(IMS SIP Signaling) NAS:activate dedicated EPS bearer context request EBI:6, QCI:5(IMS SIP Signaling) S1AP:E-RAB Setup Response NAS:activate dedicated EPS bearer context accept GTPv2:Create Bearer Response Dedicated Bearer(EBI:6, QCI:5) is established EBI:6 IMS SIP REGISTER Procedure via Dedicated Bearer for QCI:5 Figure 11.16 Attach procedure for VoLTE user. LTE Attach HSS IMS registration (SIP) (3) Location registration (Download APN for VoLTE) (1) Power-on LTE/EPC MME/SGW (7) Completion of Attach VoLTE terminal (5) Terminal IP address and P-CSCF allocation IMS (4) Setting up of bearer for VoLTE (2) Attach request eNodeB (9) a. Registration and authentication (9) c. Service information download (P-CSCF address) PGW (6) Bearer response (P-CSCF address) (8) IMS registration request (9)b. Authentication (10) Completion of IMS registration Figure 11.17 LTE attach and IMS register. SIP bearer P-CSCF S-CSCF AS 459 460 LTE Optimization Engineering Handbook UE Step 1 and 2 is the LTE attach procedure and UE acquires IP address from PDN gateway. LTE RAN UE power on EPC IMS/MTAS 1: Attach procedure, P-CSCF discovery 2: UE requested PDN connectivity 401 unauthorized response occurs when UE has not yet been assigned a serving CSCF. 3: REGISTER 401 Unauthorized 4: REGISTER 200 OK UE ready to use 5: SUBSCRIBE Step 5 and 6 are optional IMS services for subscribing to registration state. 200 OK 6: NOTIFY 200 OK Figure 11.18 IMS registration latency. UE eNB Service Request MME SGW PGW HSS Service Request Authentication (if required) Authentication (if required) Initial Context Setup Radio Bearer Establishment Initial Context Setup Complete Modify Bearer Request Modify Bearer Response PGW SIP INVITE FROM IMS PCRF DOWNLINK DATA [SIP INVITE] Modify Bearer Request IP-CAN Session Modification Modify Bearer Response SGW MME eNB UE Downlink data Notification DDN ACK Paging Paging Service Request Procedures Stop Paging DOWNLINK DATA [SIP INVITE] Figure 11.19 MO and MT call service request. VoLTE Optimization UE eNB User initate a call Reestablishment of bearers if UE is in idle mode EPC IMS Service Request Procedure SIP INVITE SIP 100 Trying SDP answer with resource reservation. If needed, establish resources for early media (ring tones or announcements) SIP183 Session Progress Dedicated EPS Bearer Establishment Rx/Gx: MMTel Session Est. SIP PRACK Only DL media enabled (for early media) SIP 200 OK Confirmation of resources available SIP UPDATE SIP 200 OK If early media used, ring tone is from network, otherwise generated locally in UE Call picked up and media flowing in both directions. SIP 180 Ringing SIP 200 OK Update of flow status of affected PCC rules (if required) Rx/Gx: Flow/gate update PCC update to open gates for bidirectional media SIP ACK Figure 11.20 Originating call flow. IMS EPC eNB SIP INVITE Paging Service Request Procedure SIP INVITE SIP 100 Trying SDP answer with resource reservation Wake up UE if in idle mode Reestablishment of bearers if UE is in idle mode SIP 183 Session Progress Rx/Gx: MMTel Session Est. Dedicated EPS Bearer Establishment SIP PRACK SIP 200 OK Confirmation of resources available PCC update to open gates for bi-directional media UE SIP UPDATE SIP 200 OK Indication of incoming call SIP 180 Ringing SIP 200 OK Rx/Gx: Flow update Update of flow status of affected PCC rules SIP ACK User answers the call Figure 11.21 Terminating call flow. and move the UE to connected mode and then it can send and receive SIP invite. For terminated (MT) call, the incoming SIP invite results in that a UE in idle mode is paged and as response the UE initiates service request to establish user plane connectivity and the SIP invite can be forwarded to the UE. SIP session establishment start from UE transmits SIP invite to UE received SIP 200 OK message, including dedicated radio bearer setup and session establishment procedure, and so on. The E2E IMS signaling flow is described below and shown in Figure 11.20 and Figure 11.21: ●● ●● 100 Trying: IMS core sends 100 trying message to originating UE once the message is forwarded to the destination. 183 session progress: this message is basically indication that the call is being processed. It’s typically used either to prevent the call from being timed out, as we await the destination user answering the call. It can also be used to support what’s called early media, that is, the 461 462 LTE Optimization Engineering Handbook ●● ●● ●● ●● ability to send media (in‐band ringing or network announcements) to the calling party before the call is completed. Destination UE indicates codec choice in SIP183 session progress. SIP183 message also triggers network‐initiated dedicated EPS bearer setup at source and destination UE. PRACK (provisional response acknowledgement): PRACK improves network reliability by adding an acknowledgment system to the provisional responses (1xx). PRACK is sent in response to provisional response. Source UE sends a SIP PRACK message indicating codec selection and preconditions extension, “Precondition” increases establishment time by 0.5 s in RRC connected state but reduces probability of call establishment failure in live network. Destination UE responds with a SIP 200 OK acknowledging SIP PRACK. Network‐initiated EPS bearer setup and transfer of SIP PRACK and SIP200 messages occurs concurrently. 180 Ringing: Destination UE locally alerts the user and sends across a ring back tone to the source UE through SIP180 ringing. When the user responds to the alert, destination UE sends SIP200 OK response to original SIP Invite. 200 OK: The request was successful. Source UE responds with an SIP ACK message. Media streams are established. An example of originating call and terminating call flow in alive network is shown in Figure 11.22. Source UE IMS client forms a SIP invite message, which includes QoS preconditions18 and order of preference for audio codecs. SIP invite message transmitted as part of IPSec packet to EUTRAN. For QoS preconditions, resources reserved before users receives ring tone, when user answers, the users are guaranteed to be able to talk. However, the reservation of the resources prior alerting the user can increase the call setup time. Two different set of traffic flow templates (TFTs) for audio and video is provided to UE when both reservations are successful. For without QoS preconditions, UE are alerted and can answer the call prior resource reservation, there is no guarantee that the user will be able to get any resources for the call (Figure 11.23). When UE A and UE B are belong to different home network, the E2E call procedure is shown in Figure 11.24. Figure 11.25 shows the SIP invite message details. It is identified by a call identifier, local tag, and a remote tag. A call‐ID contains a globally unique identifier for this call, host name, or IP address. The combination of the To tag, From tag, and call‐ID completely defines a peer‐to‐ peer SIP relationship between two users and is referred to as a dialog. Figure 11.23 shows a SDP message carried as message body in the SIP message, describing the session parameters. Default EPS bearer is used to send the initial SIP invite that contains the SDP information. SDP carries the requested media, bandwidth, and source transport address (audio). 11.2.6.3 Video Phone Session Handling A VoLTE and video‐calling device performs the same network attach, IMS domain authentication and registration procedures as a VoLTE device, plus adding video capability information. At video call setup the client signals in the SDP included in the SIP invite message that two media streams shall be setup by IMS in the EPS. These two media streams shall be full 18 UE supports SIP preconditions per GSMA IR.92 recommendations. The recommendation is to support preconditions for the reason to avoid “ghost ringing” in case resource allocation fails after 180 rings. QoS precondition also prevents that voice/video traffic by mistake is sent on the IMS signaling bearer (which could cause disturbance on other traffic). MO Call Orig NW/IMS MT RACH + RRC Conn. Setup SIP: Invite (SDP Offer) Paging + RRC Conn. Setup SIP: 100 Trying SIP: Invite (SDP Offer) SIP: 100 Trying SIP: 183 Session Progress RRC Reconfig RRC Reconfig SIP: 183 Session Progress Inform User SIP: PRACK SIP: PRACK SIP: 200 OK (PRACK) SIP: 200 OK (PRACK) SIP: UPDATE SIP: UPDATE SIP: 200 OK (UPDATE) SIP: 200 OK (UPDATE) SIP: 180 Ringing SIP: 180 Ringing SIP: 200 OK (SDP Answer) SIP: 200 OK (SDP Answer) SIP: ACK SIP: ACK E2E RTP Audio Figure 11.22 Call procedure with “precondition.” Alert User User Answer UE A eNB EPC IMS QCI = 5 Call initiation Resource reservation Note: Time between resource reservation until QCI = 1 bearer is up is undefined. Ringing Answer RTP RTP QCI = 1 UE A eNB EPC IMS Call initiation Ringing QCI = 5 Answer (SDP) RTP Resource reservation Note: Time between resource reservation until QCI = 1 bearer is up is undefined. QCI = 1 RTP RTP Figure 11.23 With QoS preconditions and without QoS preconditions. UEA ’s home network UEA P-CSCF INVITE INVITE UEB ’s home network S-CSCF 100 Trying AS I-CSCF HSS S-CSCF AS P-CSCF UEB INVITE 100 Trying INVITE 100 Trying LIR LIA INVITE INVITE 100 Trying INVITE INVITE 100 Trying 183 183 183 183 183 PRACK 183 183 183 PRACK PRACK PRACK PRACK PRACK PRACK 200 (OK) 200 (OK) 200 (OK) 200 (OK) 200 (OK) 200 (OK) Dedicated bearer (QC11) established 200 (OK) Dedicated bearer (QC11) established UPDATE UPDATE UPDATE UPDATE UPDATE UPDATE UPDATE 200 (OK) 200 (OK) 200 (OK) 200 (OK) 200 (OK) 200 (OK) 180 (Ringing) 200 (OK) 180 (Ringing) 180 (Ringing) 180 (Ringing) 200 (OK) ACK 180 (Ringing) 200 (OK) 180 (Ringing) 200 (OK) 180 (Ringing) 200 (OK) 200 (OK) 200 (OK) User Answers 200 (OK) 200 (OK) ACK ACK ACK ACK ACK Figure 11.24 Call procedure when UE A and UE B are belong to different home network. Ringing 180 (Ringing) ACK VoLTE Optimization Media negotiation UE IP address IP address used for media stream Time session created and how long intended to last Payload type: 104 Codec: AMR-WB SR: 16000 Hz Codec mode: 0, 1, 2 Mode-change-cap: capability to restrict the mode change Max-red: elapses between the first transmission of a frame and any redundant transmission Required bandwidth for RTP traffic is 30 kbps No required bandwidth for RTCP traffic (RS=0, RR=0) WB DTMF NB DTMF Precondition: resource reservation maxptime: maximum limit of 12 speech frames (240ms) per RTP packet ptime: one speech frame (20ms) encapsulated in each RTP packet Figure 11.25 Session description protocol. UE PGW PCRF IMS Originating network Terminating network IMS PCRF PGW UE Ongoing active voice call User adds video SIP Re-INVITE (voice/video) SIP Re-INVITE (voice/video) SIP Re-INVITE (voice/video) SIP 200 OK P-CSCF updates with new video SIP 200 OK AAR AAA P-CSCF updates with new video IMS interacts with PCC to create a new video bearer for the ongoing call. Dedicated EPS Bearer establishment (video component) AAR AAA Dedicated EPS Bearer establishment (video component) SIP 200 OK Ongoing active call with voice and video Figure 11.26 Adding video to ongoing voice call. duplex and synchronized for lip sync. The P‐CSCF, based on the negotiated SDP, creates one Rx session toward the EPC per call leg and uses the Rx interface toward the PCRF to request two dedicated EPS bearer one for voice and one for video on top of the already existing default EPS bearer used for SIP. The EPS is responsible to set up the two dedicated EPS bearers using the network‐initiated bearer setup procedures in sequence. Adding video to ongoing voice call procedure is described in Figure 11.26. 465 466 LTE Optimization Engineering Handbook 11.2.7 Multiple Bearers Setup and Release LTE permits UE to establish up to eight simultaneous data radio bearers (DRB) and each radio bearer can have a unique QoS profile. Prior to introducing VoLTE, a LTE user typically set up one default bearer for data on QCI {7,8,9}. A default bearer is bearer able to carry all kinds of traffic (no filter) without QoS and it is first established for a new PDN connection and remains established throughout the lifetime of the PDN connection. It is typically created during the attach procedure. The UE will have one IP address used for all data services. All the traffic for the user is sent across one DRB and there is no QoS‐based traffic separation. VoLTE introduces a second default bearer. A VoLTE user in the eNB will always have at least two default bearers configured—one for IMS signaling, one for internet connection. Voice/video traffic shall be separated from data traffic for the UE therefore dedicated bearers will be set up on demand for the voice and video components to enable this traffic separation. QCI1 and QCI2 are dedicated bearers associated to the IMS APN default bearer (QCI5). Dedicated bearers apply packet filters or traffic flow templates (TFTs) on top of the default bearer. The TFTs are installed in the UE to make sure that media traffic is forwarded on the respective dedicated bearers. TFTs consist of IP address, port and protocols to specify which traffic shall run upon the dedicated bearer. Two separate steps are considered before the UE is successfully registered in both LTE and the IMS domain. The first part will be the combined PS and CS (for SMS over sGs) attach procedure including the activation of the default internet APN. The second part is the setup of the IMS default bearer and the IMS registration. After completing these two parts successfully the UE is registered to the network and IMS domain. If both the UE and the network are VoLTE capable, the user is now able to start a VoLTE service. Once the radio bearers are established for the UE (now in RRC_connected mode) at the eNB for the default IMS APN (QCI 5) and default internet APN (QCI 7–9) the UE will only change back to RRC_Idle mode in case there is no activity more for the duration of tInactivityTimer on both default non‐guaranteed bit rate bearers. For VoLTE, voice media should always be mapped to a separate EPS bearer as it has GBR running the RLC unacknowledged mode protocol. Voice service can tolerate error rates on the order of 1%, while benefiting from reduced delays. SIP signaling should also be protected against congestion by being mapped to a separate EPS bearer running RLC acknowledged mode protocol. So, the biggest challenge in the LTE RAN within VoLTE is to handle the relation between signaling (SIP) and media (RTP), the SIP signaling is carried on the QCI 5 bearer, which has highest priority to accommodate, for example, fast call setup. For example, in a live network, 2.6 billion eRAB setups out of which 16 million are VoLTE, that is, only 0.6% of all eRAB attempts are VoLTE (QCI1). In conclusion, an IMS UE will have two bearers allocated all the time, the default bearer to internet APN is QCI = 9 (non‐GBR RLC‐AM), the default bearer to IMS APN is QCI = 5 (non‐ GBR RLC‐AM). During a VoLTE call a third bearer will be activated (QCI = 1), voice bearer QCI1 would be configured during voice call initiation (SIP invite). QCI1 is a dedicated bearer to the IMS APN (GBR RLC‐UM), dedicated bearers are created for QoS differentiation purposes (Figure 11.27). When the UE changes state from EPS connection management (ECM)‐connected to ECM‐ idle all radio resources are released but information about the PDN connection remains stored in the packet core. All radio resources are restored again when a network or UE‐initiated service request is received. QCI 8(9), default bearer setup is always the first step for any service when UE attaches to EPC and remains active as long as UE is attached to EPC (always on). IMS default bearer (QCI5), the VoLTE Optimization MO NW/IMS MT RACH + RRC Conn. Setup (DRB3 + DRB4) SIP: Invite Paging + RRC Conn. Setup (DRB3 + DRB4) QCI 9 and QCI 5 are established during the attach procedure SIP: 100 Trying SIP: Invite SIP: 100 Trying SIP: 180 Ringing SIP: 180 Ringing SIP: 200 OK RRC Reconfig (DRB5) RRC Reconfig (DRB5) SIP: 200 OK SIP: ACK EPS DRB RLC ID ID 5 3 AM 6 4 AM 7 5 UM SIP: ACK QCI Use 9 5 1 BE SIP RTP E2E RTP Audio Connection Established Default bearer established (QCI9 & QCI5) Call established (SIP) Dedicated bearer established (QCI1) time RRC_IDLE RRC_CONNECTED Figure 11.27 Multiple bearers setup. first dedicated radio bearer, is set up for SIP control signaling right after the default bearer setup during the initial attach. QCI5 will be established with the second PDN connection with IMS APN. This will avoid SIP control signaling setup delay when VoLTE call is invoked. QCI 1, MO UE requests a voice call through SIP invite. MT UE CQI1 bearer is only established after MT UE responds with SIP OK to original SIP invite. The above QCI profiles re shown in Table 11.12. Once a VoLTE has been established, the user might want to end the VoLTE call (video phone). It is expected that the user will have two separate buttons on its terminal in order to end the video only or to directly end both voice and video call according to IR.94 from GSM association. Figure 11.28 gives an illustration of VoLTE call end and bearer release procedure. After VoLTE call end, the UE still has both default bearers active, which might be used for additional SIP signaling or uplink and downlink data transmission. In the case the UE (still in RRC_connected mode) does not send or receive anything any more on both the default bearer’s timer, tInactivityTimer, is started, which is currently set to 60s in more live network. Once the timer, tInactivityTimer, expires the UE will move from RRC_connected mode to RRC_idle mode and the radio bearers are released for the non‐GBR default bearers. 11.2.8 VoLTE Call On‐Hold/Call Waiting For VoLTE call on‐hold/call waiting, it has been decided to modify the bandwidth only in the “call waiting” call case since it will need to double the bandwidth. Only at release of the second call will trigger another bearer modification. For the “call on‐hold” call case, no bearer 467 468 LTE Optimization Engineering Handbook Table 11.12 QCI configuration. QCI 1 QCI 2 QCI 5 QCI 8/9 Priority 1 4 2 8/9 Scheduling strategy Delay based Proportional fair with Min rate Resource fair Resource fair or proportional fair Logical channel group 1 2 1 3 DRX Profile 1 2 0 0 DRX Priority 99 100 1 1 RLC mode UM UM AM AM AQM mode GBR (2) GBR (2) OFF (0) Non‐GBR (1) pdb 80 150 100 300 pdbOffset 50 50 0 0 pdcpSNLength 7 12 12 12 rlcSNLength 5 10 10 7 ROHC TRUE FALSE FALSE FALSE Switch from RRC_Connected to RRC_Idle state QCI = 2, Video QCI = 1, Speech QCI = 1, Speech QCI = 8, internet QCI = 8, internet QCI = 8, internet QCI = 5, IMS signaling QCI = 5, IMS signaling QCI = 5, IMS signaling VoLTE + ViLTE DRBs IMS + internet default bearers ViLTE DRB released VoLTE DRB User inactivity tInactivity Timer expires released on default bearers on default bearers Figure 11.28 VoLTE call end and bearer release. modification is triggered since the bandwidth will not be set to “0” but only the packet filters will be adapted, which will not trigger an E‐RAB modification (Figure 11.29). 11.2.9 Differentiated Paging Priority The problem occurs if an operator would need to use a more aggressive paging for VoLTE calls than for VoLTE SMS. Both services are handled by the same IMS APN. If the IMS APN is used for additional IMS services than VoLTE, such as RCS, then additional non–call‐related signaling can be expected on the IMS APN. The same UE will need different paging profiles depending on the service. If the SGW receives a downlink packet while the UE is IDLE, a data downlink notification is sent to MME requesting paging of the UE. It is possible to configure number of tries for the last visited eNB, TA, and TAI list. Profile 1: MME sends paging messages to all eNB in the TAI list held by MME Profile 2: MME starts by paging eNBs in last visited tracking area, if no success the TAI list is paged Profile 3: MME starts by paging last visited eNB, if no success the MME pages eNBs in last visited TA, if still no success all eNBs in the TAI list are paged. VoLTE Optimization eNodeB UE MME 1) E-RAB Modify Request List of RABs and corresponding QoS profile to be modified as well as an optional NAS message per RAB Admission Control (if features are enabled) 1) Admission GRANTED 2) Admission BLOCKED conditional *) E-RAB Modify Response Inside this message “E-RAB Failed to Modify List” Cause Value: Radio resources not available 2) RRC Reconfiguration Message Includes the optional NAS message “Radio Modify Setup” 3) RRC Reconfiguration Complete Message 4) E-RAB Modify Response Contains a list of all successfully modified RABs and possibly a list of all RABs that failed to be modified Figure 11.29 Call flow for the bearer modification due to “call waiting.” The timer for paging messages is configurable in 2 to 15 s. The default value is 3 seconds. For Profile 1, successful paging will take less than 3 s. The number of messages needed is equal to the number of eNBs in the TAI list. For Profile 3, the time for paging a UE that has not moved is less than 3 s. The number of messages needed is ~1. If the UE has moved since last connected, the time for successful paging of the three profiles is less than 3 s, less than 6 s, and less than 9 s depending on how many paging messages are needed. So differentiated paging using separate bearers or APN for VoLTE calls and other IMS services, this way paging profile 1 could be used for VoLTE calls and paging profile 3 be used for other IMS services. In this case that MME will need additional information for deciding which paging profile to be used. One proposal is to provide a service class IE in the data downlink notification message so that MME can decide which paging profile to use. When SGW receives a DL packet and the UE is idle, the DL packet needs to be marked with the service class to be 469 470 LTE Optimization Engineering Handbook used for paging differentiation. SGW will use this information when sending the data downlink notification. A proposal is to have the P‐CSCF mark service class using the DSCP field (IPv4 and IPv6) or alternatively the IPv6 flow label field (not present in IPv4 packets). 11.2.10 Robust Header Compression 11.2.10.1 RoHC Feature IPv4 and IPv6 protocol headers are large relative to voice payload. These headers are transmitted frequently for every voice packet, which has negative performance impacts on over‐the‐air capacity and cell‐edge performance. Header compression is necessary in order to provide voice services on the packet switched (PS) domain with similar packing efficiency associated with the circuit switched (CS) domain. For VoLTE, the IP/UDP/RTP header added to a voice packet for VoIP services is significant (e.g., AMR 12.2 kbps with 32 bytes payload is encapsulated by 40/60 bytes of overhead for IPv4/IPv6). During talk spurt period, ROHC (IETF RFC3095, robust header compression) is expected to compress IP/UDP/RTP header to about 4/6 bytes in average for IPv4/IPv6. Figure 11.30 shows the VoIP frame size of MAC transport block when RoHC is used. For all these codecs, the RTP header part (IP + UDP + RTP) is 40 bytes in IPv4 and 60 bytes in IPv6, and the compressed packet length is between 1 and 42 for IPv4, 1 and 62 for IPv6, which obtaining a reduction up to 50% and more in the best case that shown in Table 11.13. Dimensioning aspect by examples below are described. Example 1: AMR‐WB 12.65 RTP payload is 33bytes, Step 1: [33(AMR‐WB 12.65) +3(RoHC) + 7(PDCP, RLC, MAC, BSR, PHR header)]*8 = 344 bits. Step 2: The desired transport block size (TBS) of 344 is compared with 3GPP TS 36.213 TBS table. The result is 344. The total number of information bits including CRC = 344 + 24 = 368 bits, the effective Layer 1 bit rate: 368/20 = 18.4 kbps. Example 2: AMR 12.2 RTP payload is 32bytes, the effective layer 1 bits with and without RoHC are shown in Table 11.14. Basically the UE could support up to nine different RoHC profiles.19 If it would support any RoHC profile it will always need to support also the uncompressed mode variant. At minimum, UE and network must support “RTP/UDP/IP” profile (0x0001) to compress RTP packets and “UDP/IP” profile (0x0002) to compress RTCP packets, the format of the header of the RoHC packet is different from profile 0x0001. The UE and network must support these profiles for both IPv4 and IPv6. Below an example of a UE, which reports in the RRC UE EUTRA capability message that RoHC profiles it supports. Multiple “packet streams” will be created per bearer that can use different profiles and different modes of operation, each packet stream is identified by context ID (CID). RoHC profiles can be got from S1AP initial context setup message, which is shown in Figure 11.31. 11.2.10.2 Gain by RoHC IETF RoHC is the only protocol selected by 3GPP to support IP packets header compression. The RoHC algorithm establishes a common context at the compressor and decompressor by transmitting full header and then gradually transition to higher level of compression. In RoHC layer, the compressed IP packet streams flow from compressor to the decompressor inside a RoHC channel, which multiplexes different IP packet streams that are compressed differently using different profiles. In the opposite direction, the RoHC channel carries the compressed IP 19 GSMA PRD IR.92 “IMS Profile for Voice and SMS” 4.0, March 2011. VoLTE Optimization VoIP frame Application Layer RTP H VoIP frame UDP H RTP H VoIP frame UDP H RTP H VoIP frame RTP Layer UDP Layer IP H IP Layer 40 bytes (IPv4) or 60 bytes (IPv6) IP H PDCP Layer UDP H PDCP H RTP H RoHC H VoIP frame VoIP frame 4 bytes RLC & MAC Layers MAC H RLC H PDCP H RoHC H VoIP frame Protocol Payload AMR wide-band 23.85 kbps 61 octets AMR wide-band 12.65 kbps 33 octets AMR wide-band 8.85 kbps 24 octets AMR wide-band 6.6 kbps 18 octets AMR 12.2 kbps 32 octets AMR 7.95 kbps 22 octets AMR 5.9 kbps 16 octets AMR 4.75 kbps 14 octets EVRC 8.8 kbps 22 octets SID1 7 octets Protocol Header PDCP 1 octet RLC UM header 1 octet MAC 1 octet (basic header) BSR 2 octets (header in MAC if present) PHR 2 octets RTP/UDP/IP header 1–60 octets Figure 11.30 VoIP frame size with RoHC. 471 472 LTE Optimization Engineering Handbook Table 11.13 Robust header compression gains. Protocol Total hdr size (bytes) Min. compressed hdr size (bytes) Compress gain (%) IP4/TCP 40 4 90 IP4/UDP 28 1 96.4 IP4/UDP/RTP 40 1 97.5 IP6/TCP 60 4 93.3 IP6/UDP 48 3 93.75 IP6/UDP/RTP 60 3 95 Table 11.14 The effective layer 1 bits of AMR 12.2 payload. AMR NB 12.2 w/RoHC AMR NB 12.2 w/o RoHC bits 256 256 RTP/UDP/IP header bits 32 320 PDCP header bits 8 16 RLC header bits 8 16 MAC header bits 8 8 Total size bits 312 616 PHY TBS bits 328 616 Protocol overhead Source payload packet streams, as well as the possible RoHC feedback packets for the associated RoHC channel. Figure 11.32 shows the RoHC channels and contexts. Each context corresponds to an individual compression profile, which has different compression algorithms. A context is identified with a context ID (CID). At RoHC layer, the compressed IP packet streams flow from compressor to the decompressor inside a RoHC channel, which multiplexes different IP packet streams that are compressed differently using different profiles. In the opposite direction, the RoHC channel carries the compressed IP packet streams, as well as the possible RoHC feedback packets for the associated RoHC channel. ROHC has three modes of operation, unidirectional (U‐mode), bidirectional‐optimistic (O‐mode), and bidirectional‐reliable (R‐mode). U‐Mode packets are only sent in one direction, from compressor to decompressor, but will not be used as the preferred mode of operation for VoLTE as it is mainly for unidirectional traffic. This mode, therefore, makes RoHC usable over links where a return path from decompressor to compressor is unavailable or undesirable. O‐ Mode (bidirectional optimistic) is similar to the U‐mode, except that a feedback channel is used to send error recovery requests and (optionally) acknowledgments of significant context updates from the decompressor to compressor. The O‐mode aims to maximize compression efficiency and sparse usage of the feedback channel. R‐Mode (bidirectional reliable) differs in many ways from the previous two. The most important differences are a more intensive usage of the feedback channel and a stricter logic at both the compressor and the decompressor that • 0x0000 ROHC uncompressed (RFC 4995) • 0x0001 ROHC RTP (RFC 3095, RFC 4815), this profile compresses RTP headers efficiently. Such headers are common in a VoIP call or in a video stream; • 0x0002 ROHC UDP (RFC 3095, RFC 4815), this profile will be used to compress the control signaling of the VoIP call, i.e. the RTCP packets • 0x0003 ROHC ESP (RFC 3095, RFC 4815) • 0x0004 ROHC IP (RFC 3843, RFC 4815) • 0x0006 ROHC TCP (RFC 4996) • 0x0101 ROHCv2 RTP (RFC 5225) • 0x0102 ROHCv2 UDP (RFC 5225) • 0x0103 ROHCv2 ESP (RFC 5225) • 0x0104 ROHCv2 IP (RFC 5225) QCI 1 bearer Pkt stream 1 (RTP/UDP/IP): CID #X, profile 0x0001 Pkt stream 2 (UDP/IP): CID #Y, profile 0x0002 Figure 11.31 Possible RoHC profiles. LTE Optimization Engineering Handbook RoHC channel 1 (DTCH UE< = eNB) CID 0 - IP Stream CID 1 - RTP/UDP/IP Stream Decompressor Compressor Feedback Packets CID 2 - TCP/IP Stream Feedback Packets 474 CID 3 - ESP/IP Stream CID 0 - IP Stream CID 1 - RTP/UDP/IP Stream Compressor Decompressor CID 2 - TCP/IP Stream CID 3 - ESP/IP Stream UE RoHC eNB RoHC RoHC channel 2 (DTCH UE = > eNB) IP UDP RTP Data IP RTP Data In-order delivery and duplicate detection Sequence Numbering Compress UDP ROHC Context ROHC Context Decompress Ciphering Deciphering Add PDCP Header Remove header CH Data Figure 11.32 RoHC compression architecture. prevents loss of context synchronization between compressor and decompressor except for very high residual bit error rates. The optimal RoHC operation mode depends on feedback abilities, error probabilities and distributions, effects of header size variation, and so on. RoHC always starts in U‐mode and transitions to another mode based on the decompressor feedback (Figure 11.33). RoHC can provide coverage and capacity utilisation improvements to the network. Figure 11.34 shows the RoHC header size distribution in a live network, 98% of RoHC headers size is smaller or equal to 40 bits. The most important factor for high VoIP capacity is RoHC functionality, especially in uplink. With RoHC functionality, coverage will be improved and inter‐site distance can be increased. For link budget purposes RoHC header size is assumed to be 40 bits. Compare to without RoHC, 3 dB gain due to less bandwidth needed per user (lower MCS used). In conclude, RoHC VoLTE Optimization Unidirectional Mode FO SO O mode R mode bidirectional bidirectional Error feedback No Some Intensive Efficiency Low High High ) ) FB (U FB (U ) U mode unidirectional Packet direction FB (R FB (O ) IR FB (O) Optimistic Mode Reliable Mode FB: feedback channel FB (R) IR FO RoHC feedback packets carry the header compression control information: ACK: decompression success, NACK: decompression failure, STATIC-NACK: static context, invalid/ not established, D_MODE: indicating the desired, compression mode. SO IR FO SO Figure 11.33 RoHC modes. Figure 11.34 RoHC header size distribution. 100% PDF 80% 60% 40% 20% 0% 504 120 24 32 40 56 40 RoHC header size [bits] 100% CDF 95% 90% 85% 80% 75% 24 32 40 56 120 504 RoHC header size [bits] will be enabled for voice (QCI 1) to improve coverage and system capacity. In case there is less data to transmit, UL interference is reduced and thereby increases the over‐the‐air capacity for best effort users and the cell range specifically for the uplink. 11.2.11 Inter‐eNB Uplink CoMP for VoLTE LTE has frequency reuse of 1. That means a lot of interference on cell edges. In effect, cell‐edge UEs are received with similar power by serving a neighbor cell, but to the neighbor cell, this is interference. Uplink coordinated multi‐point reception (UL CoMP) is a feature that combines antenna signals from multiple cells of the same carrier frequency in order to improve uplink 475 476 LTE Optimization Engineering Handbook 8. RLC Data 7. Decoded VoIP data 3a. UL Comp Req. C-RNTI, PCI, MCS, PRB, ... X2 (3b.) Start/Stop 4. –20ms 2. Start Relaxed UL CoMP? Decoding As usual Cooperating eNB Serving eNB Decoding Only send data if CRC = ok 6. VoIP packet 5. UE Grant 1. Measurement report A3 Entering/or leaving Figure 11.35 Inter‐eNB uplink CoMP. throughput by increasing the received signal power and taking this interference and turn it to the useful signal, while the neighbor cell can do it too. The objective of UL CoMP is to improve UL SINR by combining antenna signals from multiple sector carriers belonging to different cells. The benefit is largest for UE that are in the border between two sectors, either two macro sectors, or between a macro sector and a small cell. UL CoMP for VoLTE can realize soft handover for VoLTE bearer, but X2 delay limits the gains. When UL CoMP is used, the expected gains such as increased coverage for VoIP, improved VoIP satisfaction, and reduced call drop rate could be achieved (Figure 11.35). Since the VoIP capacity is limited by the worst users (cell‐edge users), CoMP may be even more beneficial for VoIP than for full buffer traffic. Uplink CoMP can be used if digital processing is centralized in a common point in the network, connecting with remote radio units via fiber. These solutions are not feasible with uncoordinated small cells connected to the macro network via the S1/X2 interfaces. Both macro diversity and interference suppression can be achieved depending on the type of signal combining. Further, a spatial division multiplexing (SDM) scheme can be applied, where multiple users are scheduled at the same time in different nodes, using the same uplink resources. Up to three cells can be included in so called CoMP set ‐ > serving cells can have up to two neighbor cells in each CoMP set. For each UL transmission in the serving cell, a linear SINR average over 4 RX antennas is calculated: SINR _ S _ lin 0.25 * SINR _ S _ lin 1 SINR _ S _ lin 2 SINR _ S _ lin 3 SINR _ S _ lin 4 VoLTE Optimization For each UL transmission in the neighbor cells from the serving cell CoMP set, a linear SINR average over 4 RX antennas is calculated: SINR _ N 1 _ lin 0.25 * SINR _ N 1 _ lin 1 SINR _ N 1 _ lin 3 SINR _ N 1 _ lin 4 SINR _ N 2 _ lin 0.25 * SINR _ N 2 _ lin 1 SINR _ N 2 _ lin 3 SINR _ N 1 _ lin 2 SINR _ N 2 _ lin 2 SINR _ N 2 _ lin 4 11.3 Semi‐Persistent Scheduling and Other Scheduling Methods PDCCH becomes a bottleneck when VoIP capacity is growing and adoption of semi‐persistent scheduling (SPS) allows a reduction of the PDCCH consumption, which in turns allows an increase in the number of data users accessing the cell and improves the cell throughput for non‐VoIP applications. Without SPS, the 100 VoIP bearers would use the entire PDCCH capacity, thus user perception of the VoIP quality is very bad (delays, frames drop) and also cell throughput is heavily impacted as regular traffic could not be granted at all. It notes that TTI bundling and SPS can’t be enabled simultaneously for TDD as defined by 3GPP. 11.3.1 SPS Scheduling Persistent allocation of time and frequency resources can be used for each initial HARQ to each VoIP flow, retransmissions are dynamically scheduled in frequency is also fixed in time due to synchronous HARQ. At the admission accords to a predefine frequency‐time pattern, resource allocation can be with a fixed MCS and number of PRBs, for each initial Tx and subsequent re‐Tx of a VoIP frame, a scheduler grant is sent, the unused semi‐static resource can also be used by the dynamic scheduler. SPS scheduling mechanism is a combination of persistent scheduling for initial transmissions and dynamic for retransmissions. Compared to a dynamic scheduler, SPS minimizes control overhead and low DL grant usage, since grants are required only for retransmissions, but semi‐ persistently scheduled VoIP users do not have the benefit of channel‐sensitive scheduling. Frequency selective dynamic scheduling may provide some gain over frequency hopping for VoIP users at low dopplers and MCS and PRB allocation are based on control channel feedback, buffer occupancy and UE power headroom. Usually, we use dynamic scheduling for select users at the cell edge at low to medium speeds who need the extra performance boost to meet VoIP quality targets, use semi‐persistent scheduling for the remaining users. SPS gives the ability to support non VoIP traffic along with high VoIP load. Table 11.15 describes the comparation of dynamic scheduler and semi‐persistent scheduler. This part of optimization aims to enable the network achieved the target performance by how to use the two‐way scheduling strategy. SPS parameters are included in the “RadioResourceConfigDedicated” IE of the RRC connection reconfiguration message. An example is described below: Example RRC : rrcConnectionReconfiguration ....... sps-Config { semiPersistSchedC-RNTI ‘00...010’B, 477 478 LTE Optimization Engineering Handbook sps-ConfigUL setup : { semiPersistSchedIntervalUL sf640, implicitReleaseAfter e2 ....... } SPS-Config parameter SPS C-RNTI (UE ID used to schedule an SPS unicast transmission, activation, reactivation and retransmission) SPS DL Config SPS interval DL No. of configured HARQ processes for SPS SPS UL Config SPS interval UL Implicit release after (No. of empty transmissions (padding PDU) on the UL before the UE implicitly releases the UL SPS grant) Value 003D – FFF3 10 – 640 ms 1–8 10 – 640 ms 2,3,4,8 Table 11.15 Dynamic scheduler versus semi‐persistent scheduler. Dynamic scheduler Semi‐persistent scheduler VoIP priority over data Adaptive QoS weighting algorithms Fixed absolute priority of SPS VoIP over data PDCCH grant signaling High: At every initial Tx in DL/UL Also at HARQ Retx on DL Low: At SPS activation on DL/UL; At SPS release and HARQ Retx on OL; At explicit UL SPS release. Link adaptation MCS/PRB determined per packet (every 20 ms) based on user’s RF and BLER MCS/PRB determined per talk spurt (every several seconds) UL Power Control Slow power control (200 ms)based on FPC, fixed target SINR based on pathloss Fast power control (~20 ms) with adaptive SINR target based on BLER performance RLC segmentation Yes No 11.3.2 SPS Link Adaptation Different from the dynamic scheduler is the MCS/PRB/modulation assignment, which will vary depending on the air link condition, SPS link adaptation is enhanced to minimize the padding, thereby improving initial BLER. This is done through a combined selection of the number of PRB and MCS (based upon SINR and TBS input to each), instead of a sequential approach, selecting first the MCS then the number of PRB. In the case of the SPS allocation, the MCS is decided at the start of the talk spurt and used throughout the talk spurt. After SPS activation, the same PRBs are used every 20 ms for the initial transmission of the VoIP frame. For example, uplink MCS/PRB/modulation is fixed at 2PRBs, MCS = 10 when SPS is activated. UL SPS also VoLTE Optimization adds an outer loop power control (since there is no rate control) to avoid packet segmentation issue, which makes it even more efficient with scheduling grants compared to dynamic scheduler. DL and UL SPS activation are enabled by RRC configuration, including SPS‐C‐RNTI, UL transmission pattern interval, implicit release after count, and so on. When a SPS bearer is established, the eNB sends a RRC connection reconfiguration message with the new SPS bearer added. To activate the SPS, the DL/UL scheduler sends an SPS grant in PDCCH, addressed to the SPS‐C‐RNTI assigned to the UE. If the UE correctly decodes the grant and correctly receives the first SPS transport block (first SPS MAC packet), it sends an ACK. When the eNB receives the ACK for the SPS packet transmission, the SPS activation is considered successful, and vice versa, the reserved PDCCH resources are freed (as no longer needed). The minimum radio link quality conditions for SPS activation includes the UE is in speech active state, the normalized power headroom, the wideband SINR, no measurement gap actived for the UE and no SPS activation prohibit timer is running. DL and UL SPS activation is shown in Figure 11.36 and Figure 11.37. To deactivate the SPS, the DL/UL scheduler will send another SPS grant, if the UE correctly decodes the grant, it sends an ACK. When the eNB receives the ACK, the SPS deactivation is considered successful. DL always uses explicit release procedure by sending special form of DCI1A to the UE in case of: ●● ●● ●● ●● ●● ●● ●● VoIP inactivity (more than 20 ms‐periods have elapsed with the SPS buffer empty) Measurement gap activation procedure SPS activation failure Potential collision with PDSCH carrying common channels (e.g., paging, D‐BCH) or positioning reference signals or eMBMS After inactivity timer of SPS release expires BLER is higher than threshold for SPS SPS grant parameter value (i.e., MCS or number of PRBs) changed due to GBR modification Figure 11.36 DL SPS activation. 479 Figure 11.37 UL SPS activation. VoLTE Optimization VoIP SID RLC SDU detected Grants are sent until implicit release 2 MAC PDUs with 0 MAC SDUs detected drxShortCycle = 40 ms 4 ms 4 ms 2 MAC PDUs with 0 MAC SDUs sent on sps (SID) BSR + Data SR PDU with 0 MAC SD U PDU with 0 MAC SD U BSR + Data Grant 4 ms grant spsAct Grant grant spsAct (SID) eNB UE Implicit Release 160 ms x 4 ms ms Implicit Release Figure 11.38 SPS implicit released (example). ●● ●● Last VoIP bearer deletion SINR falls below the minimum SINR For SPS In such above cases, SPS is released by an implicit release. An example of implicit release is shown in Figure 11.38 that the release will occur after two consecutive MAC PDUs without any MAC SDUs. When an SPS release is initiated, all transmissions (intial and retransmissions) are managed by the dynamic scheduler. The SPS release is considered successful if an ACK is received. 11.3.3 Delay Based Scheduling On top of the QoS aware scheduler, the idea is that VoIP packet is not scheduled immediately but that several packets are packed in the same transport block. eNB shall check whether there is a need to allocate resource to the UE at every UL scheduled TTI. A delay‐based scheduler will prioritize users whose packets are getting close to their PDB (packet delay budget). The amount of bundling is determined by the setting of bundling time, which, in turn, is influenced by the operator via the parameter PDB (Figure 11.39). The goal of delay‐based scheduling is to ensure that all VoIP packets of UEs in good and poor channel conditions are received correctly within assumed delay target, that is, delay–based scheduling is a mechanism that allows to match the packet aggregation level to the delays on radio interface. Figure 11.39 shows that under relatively fair channel conditions, aggregation can be increased, while in good channel conditions, more packets can be aggregated. The scheduler evaluates the actual scheduling delay of voice packet and compares it with a delay target to decide UE priority, if measured delay is bigger than delay target, UE priority is increased, otherwise UE priority is decreased. The idea is that VoIP packet is not scheduled immediately but several packets are packed in the same transport block. So depending on the configured threshold, one can control on how many VoIP packets are bundled at one transport block by knowing the VoIP packet size of 328 bits. However, there is also a requirement for delay, so that buffering is not possible in all cases. Delay‐based scheduling provides the ability to observability for real‐time service in terms of the degree to which packets meet their packet delay. The benefit of DBS (delay based scheduler) is improved capacity. It provides more opportunities to schedule best effort traffic. Uplink packet sizes can be predicted so as to allow uplink grants requests to be minimized, thereby reducing control channel usage. VoIP packets for UEs in good radio conditions can be bundled without compromising VoIP quality. When VoIP packets are bundled, resources are freed that can be used for scheduling of MBB services. The main key areas where this scheduler contributes to are voice performance, smartphone efficiency, and network efficiency. 481 482 LTE Optimization Engineering Handbook w P1 Scheduler internal parameters P2 P3 PX PDB1 2 el v i at eg r gg on Le 80 v n tio a eg r gg A 3 el Le 60 A 40 Delay Term No io at g re PDB3 100 n g Ag PDB2 20 80 70 60 50 40 30 20 10 0 –10 remaining delay budget [ms] Figure 11.39 Packet delay budget and its configurable attributes. The combined DBS and SABE (service aware buffer estimation) scheduler is prefered in a live network that will automatically keep track of whether the user is in TALK or SID (silence indicator) state in order to control the packet‐bundling periodicity and grant estimation size. The combinded DBS and SABE scheduler adapts to handle different codec rates, RoHC on/off, IPv4 or IPv6, and so on. The delay added by combined DBS and SABE is expected to be less than PDB + pdbOffset; therefore, these two functions should work together without a problem. It’s worth to note that DBS and SABE are not beneficial or efficient for video. It is not possible to recognize a video packet in the system in the same way as a VoIP packet is recognized as it is not possible to determine the age of a video packet. 11.3.4 Pre‐scheduling For LTE uplink, the scheduler is located in the eNB and the buffers are in the UE. Therefore, the uplink scheduler has no direct knowledge of the amount of data available in the UE. The UE can inform the uplink scheduler that it has data by sending a scheduling request (SR) on PUCCH or PUSCH. The terminal can also send a buffer status report (BSR) together with data on PUSCH indicating the size of every priority queue in the UE. There will always be a delay between the time when the data arrives in the buffer and the time when the uplink scheduler is informed. VoIP is a very delay‐sensitive service so minimizing the delay of the uplink will increase the VoIP capacity. Pre‐scheduling is a method to minimize UL delay by means of blindly giving PUSCH grants to a UE in advance, without receiving buffer status information (SR, BSR). Pre‐scheduling VoLTE Optimization PING UE 5 10 15 20 PING 5 SR 1 Grant PING 1 4 1 PING PING PING UE 0 5 10 PING eNB SR encod. and alignment 3 Figure 11.40 Total round trip delay(23 ms) versus total round trip delay with pre‐scheduling (13 ms). eNB processing 3GPP specified delay 1 3 Core network delay 2 1 1 Time [ms] PING Delay Breakdown Server eNB 0 Delay Breakdown Server Grant PING PING PING PING 4 1 1 1 3GPP specified delay 1 3 Core network delay 2 15 20 Time [ms] means the eNB periodicity grants the UE even if UE reported BSR = 0 (UL buffer empty) or the UE has not sent an SR. Pre‐scheduling improves UE response time and thereby reduce latency, but with the cost of higher UE battery power consumption and also occupy the resources of the network without guarantee that these resources will be used for transmission, since the grants are giving blindly to the UE. The UE will respond to the UL grants even if it has no data to send. In this case it will send padding data (the BSR is called “Padding BSR”) and thus UE will not go DRX and UL out‐of‐sync state. All in all, the UE will less often be in a battery saving state when the UE has no data to send when pre‐scheduling is enabled. Pre‐scheduling can be automatically deactivated (timer controlled) to reduce the impact on UE power consumption. As mentioned in Chapter 2, the most common benchmark for observability of the IP packet latency in the network latency is the ping test. It is used to test the reachability of the host on IP network and to measure the round trip delay in the network. The theoretical analysis, which has been confirmed by the measurements in the network, shows that when prescheduling was used in uplink transmission, the latency performance was about 6 ms better than the case with ordinary scheduling request (see Figure 11.40). One example is shown in Figure 11.41, one advantage is paging time can be reduced by 30% to 50% by pre‐scheduling, and another one is the corresponding time inprovement on a complete web download is 3% to 10%. Pre‐scheduling targets the cases where the traffic load in the network is low or moderate. Actually, prescheduling grants are sent only if there are free resources and only in good radio condition. In other hand, counters and drive tests results show that uplink is heavily impacted by the prescheduling, due to increase of UL interference. In particular, the selected transport format and the uplink BLER is much worse when prescheduling is active, power consumption while transmitting doesn’t change too much. 483 484 LTE Optimization Engineering Handbook No Prescheduling, m = 27 ms, std = 3 ms 1 ms Prescheduling, m = 14 ms, std = 1ms 5ms Prescheduling, m = 16 ms, std = 2 ms 10ms Prescheduling, m = 18 ms, std = 3 ms 0 5 10 20 25 [ms] 30 15 35 No Prescheduling, mean DL time = 6325 ms 1 ms Prescheduling, mean DL time = 5690 ms 5 ms Prescheduling, mean DL time = 6065 ms 10 ms Prescheduling, mean DL time = 6139 ms 0 1 2 3 4 5 6 [s] Figure 11.41 The example of pre‐scheduling. 11.4 PRB and MCS Selection Mechanism Varying transport formats (modulation, coding scheme, number of RBs and TBS) result in different SINR requirements. VoLTE data rate may vary depending on codec rates. Coverage for VoLTE users depends on the cell‐edge SINR in relation to chosen TBS. Enhance the UL coverage of VoLTE in bad radio conditions, the mechanism of PRB and MCS selection that the UL scheduler uses to control the minimum number of PRBs and the minimum MCS value assigned to scheduled PRBs. PRB override and MCS override usually working together only for VoIP calls. The scope is to be able to achieve the VoIP codec rate (e.g., 12.65 kbps AMR‐WB) and thus obtain a good call quality even in bad radio conditions. 11.4.1 Optimized Segmentation For a UE in bad radio conditions, the UL scheduler is going to start segmentation of packets in order to use a more robust MCS for the transmission of each individual packet. Segmentation of the VoIP packet introduces delay to transmission of the packet, thus limiting the maximum useful bit rate achievable on the radio link. To ensure the reasonable reliable VoLTE services at poor UL RF condition, two types of constraints need to be addressed to residual block error rate (rBLER) and the packet transfer delay. When the uplink scheduler struggles to maintain its target initial block error rate (iBLER) performance when the total amount of power used for the transmission cannot be increased. The typical behavior of the uplink scheduler is to start segmenting the packets to allow the use VoLTE Optimization of more robust MCS for the transmission of each individual packet. The number of payload bit per transmission is consequently reduced, thereby increasing the amount of power per transmitted bit, which in turns leads to an improved iBLER performance. The packet segmentation introduces delay to the transmission of the packet and inherently introduces a bottleneck to the maximum useful bit rate that can be achieved on the radio link (since the number of bit per TTI is restricted). Beyond a critical level of segmentation (i.e., below a minimum payload size per TTI) the radio link throughput becomes lower than the VoLTE codec rate and the overall packet delay start to build up. Under various fading channel conditions, optimized segmentation yields 3‐4 dB improvement in terms of maximum acceptable uplink path loss (MAPL) for an AMR 12.2 kbps codec managed with uplink dynamic scheduler, comparing with the case no optimized segmentation is enforced. 11.4.2 PRB and MCS Selection If initial BLER is too low, meaning MCS/PRB selection is too conservative, then PRBs used for initial SPS packet transmission will be high even if the retransmission will be low. If initial BLER is too high, meaning MCS/PRB selection is too aggressive, then the retransmission of the SPS packet may be high and also, when users move to a worse RF condition, SPS may not work. In this part, we try to find the right MCS/PRB table to make the initial BLER to be around 10% for SPS packets. There is a trade‐off between the initial BLER for SPS and the total PRB usage. SPS PRB and MCS selection procedure is shown in Figure 11.42. Example The 12.65 kbps AMR‐WB codec delivers a 32‐byte speech frame every 20 ms, consequently the total payload to be transported over four segments is: 3byte RoHC header + 1byte PDCP header + 6 bytes for 1st segment header size + 3 x (4 bytes for subsequent header size) + 32 bytes useful payload = 176 bit header + 256 bits useful payload = 432 bits → Consequently the minimum transport block size per segment is 432/4 = 108 bits, and the four prefered TBS can be selected in Table 11.16. In a live network, as described in the example above, the following MCS rules should therefore be used for enforcing a maximum of four segments per speech frames as different number of scheduled RBs will affect uplink coverage. In Table 11.15, if the PUSCH grant size is 1 PRB, Figure 11.42 SPS PRB and MCS selection. 485 486 LTE Optimization Engineering Handbook Table 11.16 Transport block size. IMCS TB size (in bits) NPRB 0 1 2 3 4 5 6 7 8 9 10/11 1 16 24 32 40 56 72 328 104 120 136 144 2 32 56 72 104 120 144 176 224 256 296 328 3 56 88 144 176 208 224 256 328 392 456 504 4 88 144 176 208 256 328 392 472 536 616 680 5 120 176 208 256 328 424 504 584 680 776 872 the minimum MCS = 8, if the PUSCH grant size is 2 PRB, the minimum MCS = 4, if the PUSCH grant size is 3 PRB, the minimum MCS = 2, and if the PUSCH grant size is 4 PRB, the minimum MCS = 1. 11.5 VoLTE Capacity As specified by 3GPP, the requirement for VoIP capacity is at least 200 users for each cell in the 5 MHz bandwidth. Basically, three things can limit the VoIP capacity in an LTE system: control channel capacity, the number of resource blocks, or UE power. In other words, VoLTE capacity mainly depends on scheduling capacity and PUSCH capacity. More symbols allocated for the PDCCH per TTI gives better capacity, lower voice codec bit rates permits more users, and higher allowed radio interface delay also permits more users. From network design point of view, smaller cell size permits more users. Some features in VoLTE can also help increasing capacity. RoHC is deployed in VoLTE, which increases the VoIP capacity in cases where PDSCH or PUSCH capacity is the limiting factor. RLC UM increases the VoIP capacity in cases where the number of transmissions per TTI is limited. DBS and SABE scheduling can also increase the capacity for VoIP users. The VoLTE capacity is defined according to the number of QCI 1 bearers managed by the cell. Based on the actual bitrate per VoIP call listed in Table 11.17, we can calculate the number of simultaneous VoIP calls. VoIP system capacity is defined as the number of users in the cell when more than 95% of the users are satisfied. VoIP capacity is defined as the min of the two capacities. Outage criteria: less than 5% of VoIP users in the system are in outage; Delay criteria: less than 5% of VoIP users have its 98th percentile delay greater than 50 ms. User outage criteria includes: ●● ●● ●● Packet error rate (PER): the ratio between num of packets in errors after max num of HARQ Tx plus num of packets discarded at UE and num of total packet transmitted by the UE. VoIP outage: a UE is in VoIP outage if its PER exceeds 2% Call dopping criteria: A VoIP call is dropped if more than a certain percentage (10%) of the VoIP packets is lost in an observation window of 10s (sliding window of talk time). Lost VoIP packet criteria: A packet (or any segment) is discarded at eNB if not sent within the delay budget or a packet (or any segment) is discarded at UE if received beyond the delay budget (it can happen in case of retransmission) or a packet (or any segment) is lost due to HARQ failure. VoLTE Optimization Table 11.17 Source codec bit‐rates for the AMR codec (from 3GPP TS26.071) and AMR‐WB codec (from 3GPP TS26.171). Codec mode Source codec bit‐rate Codec mode Source codec bit‐rate AMR_12.20 12.20 kbit/s (GSM EFR) AMR‐WB_23.85 23.85 kbit/s AMR_10.20 10.20 kbit/s AMR‐WB_23.05 23.05 kbit/s AMR_7.95 7.95 kbit/s AMR‐WB_19.85 19.85 kbit/s AMR_7.40 7.4 kbit/s (IS‐641) AMR‐WB_18.25 18.25 kbit/s AMR_6.70 6.70 kbit/s (PDC‐EFR) AMR‐WB_15.85 15.85 kbit/s AMR_5.90 5.90 kbit/s AMR‐WB_14.25 14.25 kbit/s AMR_5.15 5.15 kbit/s AMR‐WB_12.65 12.65 kbit/s AMR_4.75 4.75 kbit/s AMR‐WB_8.85 8.85 kbit/s AMR_SID 1.80 kbit/s AMR‐WB_6.60 6.6 kbit/s AMR‐WB_SID 1.75 kbit/s 11.5.1 Control Channel for VoLTE With a baseline dynamic VoIP packet scheduling approach, where each packet is transmitted by layer1 control signaling, multi‐UE channel sharing is an important aspect when optimizing air interface capacity for VoLTE traffic. PDCCH being transmitted to each scheduled terminal, its overhead can be significant and limiting factor since VoIP consists of many small packets. This can result in insufficient control channel resources for scheduling all physical resource blocks while wasting part of PDSCH capacity. Highest achievable PDCCH capacity [users per cell] is given by: NVoIP ,max N grants SE per TTI consumption 2 N grants N CCE N CCE ,ave Factor 2 in the formular above appears because required Ngrants is sum of requirements for UL and DL. NCCE is the number of allocated CCEs in the cell which depends on bandwidth. NCCE,ave is the average number of CCEs per PDCCH which depends on cell size and load. In a live network, DL scheduling entity (SE) consumption estimation can be got from: DLSEperTTI ,VoLTE VAF 40 1 VAF * 1 % Retrans. 160 In the live network, segmentation and TTI‐Bundling will occur on the UL due to UE UL power limitation, average UL SE considering mixed radio condition (neither segmentation nor TTI‐Bundling, segmentation, TTI Bundling) is given below. ULSEperTTI ,VoLTE VAF 1 VAF 40 160 VAF 1 VAF %Seg . * 2 * 40 160 VAF 1 VAF * 1 % Retrans. %TTI Bunding * 4 * 40 160 1 %Seg . %TTI Bunding * 487 488 LTE Optimization Engineering Handbook Table 11.18 SINR versus CCE. Aggregation level 1 2 4 8 Max number of CCEs 444 222 111 55 Swith point (dB) 6.3 2.4 0.5 −2 According to the network Ng = 1 configuration, PDCCH resources are aggregated in groups of 1, 2, 4, and 8 CCEs, and different aggregation level needs different RS SINR that is shown in Table 11.18. According to the real user distribution in a live network, we can estimate the number of VoIP users in a cell in a simple way, 444 * 3/10 222 * 2/10 111 * 2/10 55 * 3/10 216 In a live network, VoLTE packets consists of two parts: activity (Talk) and inactivity (Silence). Standard voice is packetized in 20‐ms intervals. With DBS it will be 40 ms, for the calculations 20 ms is used, for the inactivity part, SID frames arrive in 160 ms interval. In this case, the VoIP capacity in a cell can still be higher. For UL, the L1/L2 control signaling is carried by PUCCH, CQI information indicating the channel quality estimated by the terminal, and UL scheduling requests indicating that the terminal needs UL resources for PUSCH transmissions, as the number of served UE increases, the CQI overhead can increase significantly, which may become prohibitively large with VoLTE traffic while a large number of UEs must be supported. To maintain the UL control overhead at a reasonable level, CQI feedback should be reduced both in time and in the frequency domain. A reduction of CQI overhead can be achieved with slightly reduced traffic capacity by means of wideband CQI with loss of frequency scheduling gain. 11.5.2 Performance of Mixed VoIP and Data VoLTE call quality is dependant on the LTE network handling thousands of concurrent SIP sessions per second. Due to the fact that VoIP is a guaranteed bit rate (GBR) service it will be served (scheduled during time domain scheduling phase) before all non‐GBR services, dedicated QoS for voice in LTE will impact data applications running on the same limited bandwidth resources. The average data throughput for a user with a VoIP bearer established is decreased comparing to the case when only the data bearer is established. Non‐GBR UE throughput decreases with an increase of the number of VoIP UEs in the cell, due to PDCCH block and the low priority, as compared with VoIP UE. VoIP UE throughput is stable regardless of the number of VoIP UEs in the cell. UL non‐GBR UE throughput is almost zero when the number of VoIP UEs is larger than 130, because non‐GBR traffic has no chance to be scheduled (Figure 11.43). From Figure 11.44, total cell throughput decreases with higher number of VoIP users in the cell. VoIP traffic increases linearly, has higher priority thus is not blocked. Due to lower priority non‐GBR traffic has much less scheduling occasions; therefore, throughput generated by non‐ GBR UEs decreases when more VoIP UEs are in the network. Using SPS for VoIP alleviates the demand for scheduling grants on PDCCH, allowing higher data throughput. UL SPS adds outer loop power control (since there is no rate control) to avoid packet segmentation issue, which makes it even more efficient with scheduling grants c ompared to DS. VoLTE Optimization Parameter Layout Setting 3GPP Macro Case 1 according to TR 25.814 ISD 500 m Link DL + UL Duplex mode TDD DL/UL:TDD Conifg 1 Special subframe: Config 5 according to 3GPP TR 25.814 3GPP_TR25814_3sector_70deg_14dBi TDD DL/UL Config Antenna type Transmission modes TM1 Operational band (MHz) 2 GHz Bandwidth (MHz) 10M Hz eNB 0.8W per PRB Output power indoor penetration loss UE 23dBm Receive Diversity 2 RX Std. dev = 8 dB; corr. distance = 50 m Slow fading Fast fading 20 dB TU3 UL mean # of scheduled UEs Per TTI 10 8 non-GBR VoIP 6 4 2 Ideal Real Ideal Real Ideal Real Ideal Real Ideal Real Ideal Real Ideal Real 0 10 50 100 130 140 150 175 Number of VoIP UEs Per Cell Figure 11.43 10 MHz TDD cell UL mean # of schedeled UEs per TTI. Overall, there is a degradation in data throughput with DS compared to SPS, and the degradation increases with the number of voip users increases. In Figure 11.45, we can see that 100 voice calls, the degradation is around 34%; 50 voice calls, the degradation is about 13%; 20 voice calls, no degradation is found. Both UL and DL show that SPS scheduler improves the residual data throughput when the number of voip users is large, which causes grant shortage, that is, for 50 VoIP users, the UL and DL gain of SPS versus DS is 13% and 5%, respectively. For 100 VoIP users, the UL and DL gain of SPS versus DS is 34% and 13%, respectively. When the number of VoIP users is small, or the grant is not the limiting factor, SPS scheduler performs similar with DS scheduler. In a live network, voice outage in a cell is sensitive to interference over thermal (IoT). Higher IoT level is due to inreasonable network structure, more users, higher SINR targets, and higher bandwidth occupancy with SPS compared to DS. Figure 11.46 gives the relation between IoT and number of UEs. 489 UE Throughput (kbps) UE Throughput (kbps) LTE Optimization Engineering Handbook Mean UL non-GBR UE TP 250 200 150 100 50 0 10 50 100 120 130 140 150 Number of VoIP UEs Per Cell 175 Mean UL VoIP UE TP 8 7 6 5 4 3 10 50 100 120 130 140 150 Number of VoIP UEs Per Cell DL mean TP Cell Throughput (kbps) 7000 Total TP 6000 no-GBR TP VoIP TP 5000 4000 3000 2000 1000 0 10 50 100 130 140 150 Number of VoIP UEs Per Cell 175 UL mean TP 2500 Total TP Cell Throughput (kbps) 490 Non-GBR TP 2000 1500 1000 500 0 10 50 100 130 140 Number of VoIP UEs Per Cell Figure 11.44 Performance of mixed VoIP and data. 150 175 VoLTE Optimization Residual data throughput (Mbps) 16.00 14.00 DS 12.00 SPS 10.00 8.00 6.00 4.00 2.00 0.00 0 50 100 150 200 Number of voice calls Figure 11.45 Benefits of SPS for VoIP on uplink (10 MHz carrier, TDD, multicell). Figure 11.46 IoT versus number of UEs. 8 IoT dB 6 4 2 0 40 50 60 70 80 90 100 110 UE Number 14 12 IoT (dB) 10 8 6 4 DS, AL = 4 2 SPS 0 0 20 50 100 Number of Voice Calls 11.6 VoLTE Coverage As the industry moves toward deployments of VoLTE, one of the most important performance aspects is to be able to provide similar coverage for VoLTE as end users have come to expect from 2/3G voice. However, VoLTE introduces several coverage challenges that were not present in 3G circuit switched voice. For instance, VoLTE deployments have targeted use of high‐ definition voice codes such as AMR 12.65 kbps for improved voice quality, while many 2/3G voice network designs have been based on lower rate narrow band voice even AMR 5.9 kbps, which can provide larger coverage since fewer bits need to be sent. Further, VoLTE has 491 492 LTE Optimization Engineering Handbook a dditional IP, PDCP, RLC, and MAC header overhead compared to circuit switched based voice. This extra IP overhead for VoLTE compared to 2/3G CS voice again means fewer bits for coding, reducing receiver sensitivity. Fortunately, there are aspects of LTE that help make up for the shortcomings that will be discussed in this part. In order to get good VoIP coverage, segmentation and TTI bundling is used. With segmentation, VoIP packets are split into smaller parts in order to distribute transmission over several TTIs. Since the UE output power is limited, it is important to achieve continuous transmission to maximize the transmitted energy. TTI bundling is used to minimize overhead, by autonomous retransmission over four consecutive TTIs. This reduces the need for segmentation and also control signaling needed to schedule each transmission. Also, minimum bandwidth allocation could be adopted to enhance the UL coverage. The techniques discussed in this part will achieve different compromises in terms of experienced BLER, protocol overhead, required scheduling grants, efficient MCS, bandwidth utilization, and packet delay. 11.6.1 VoIP Payload and RoHC VoIP coverage mainly determined by VoIP throughput requirement. VoIP throughput is defined by the voice codec and overhead introduced by lower layers. The VoLTE voice codec, L2/L1 SDU bits is shown in Table 11.19. Different voice code rate have different SINR requirement, reducing AMR codec rate from 12.2 kbps to 5.9 kbps brings additional ~1 dB gain in DL and ~1.2 dB gain in UL. Table 11.20 gives the SINR requirement of three types voice code rate. Use of RoHC can considerably compress the header size in voice packets and reduce the user plane traffic over the air interface. The high compression ratio obtained not only increases network capacity but also provides coverage improvements for VoLTE compared to other VoIP alternatives. The test result of bandwidth consumption for AMR and AMR‐WB (Figure 11.47). Table 11.21 and Table 11.22 gives the UL and DL VoIP bitrate requirement with and without RoHC. Here is an example of field test, under two UEs VoIP call with mobility slow speed (30 km/h EVA), the inter‐arrival time of originator/terminator VoIP calls, and average VoIP throughput are shown in Figure 11.48, Figure 11.49, and Figure 11.50. 1) Originator is under the condition of inter‐arrival VoIP packets are 20 ms, inter‐arrival SID packets are 160 ms; first DTXed packets inter‐arrival is 60 ms. We can get the average VoIP inter‐arrival time is 30 ms, the maximum inter‐arrival time is 167 ms, the minimum inter‐ arrival time is 0 ms. 2) From terminator, impact of the network jitter can be noticed, and we can get the average VoIP inter‐arrival time is 30 ms, the maximum inter‐arrival time is 180 ms, the minimum inter‐arrival time is 0 ms. 3) From the throughput pattern and statistics, we can observe the ‘DTX on’ effect, SID frames will be sent if there is no speech activity. We can get the average throughput is 24 kbps, the maximum throughput is 38 kbps, the minimum throughput is 4 kbps. 11.6.2 RLC Segmentation When a VoLTE UE moves toward the cell edge, SINR received by eNB starts to decrease due to UE power limitation. To maintain the initial block error rate (iBLER), UL dynamic scheduler automatically starts segmenting the VoIP packets from PDCP into multiple smaller packets. This is called a packet segmentation algorithm, which is used as an extension to link adaptation for uplink cell edge coverage improvement. Table 11.19 VoIP payload and L2/L1 throughput. Voice Codec Codec Source Rate [kbps] Frame Size [bit] Padding [bit] PDCP Header [bit] L2 SDU [bit] RLC header [bit] MAC header [bit] CRC [bit] L1 SDU [[bit] AMR‐NB 1.8 1.8 36 4 40 80 8 8 24 120 AMR‐NB 4.75 4.75 95 1 40 136 8 8 24 176 AMR‐NB 5.15 5.15 103 1 40 144 8 8 24 184 AMR‐NB 5.9 5.9 118 2 40 160 8 8 24 200 AMR‐NB 6.7 6.7 134 2 40 176 8 8 24 216 AMR‐NB 7.4 7.4 148 4 40 192 8 8 24 232 AMR‐NB 7.95 7.95 159 1 40 200 8 8 24 240 AMR‐NB 10.2 10.2 204 4 40 248 8 8 24 288 AMR‐NB 12.2 12.2 244 4 40 288 8 8 24 328 AMR‐WB 1.75 1.75 35 5 40 80 8 8 24 120 AMR‐WB 6.6 6.6 132 4 40 176 8 8 24 216 AMR‐WB 8.85 8.85 177 7 40 224 8 8 24 264 AMR‐WB 12.65 12.65 253 3 40 296 8 8 24 336 AMR‐WB 14.25 14.25 285 3 40 328 8 8 24 368 AMR‐WB 15.85 15.85 317 3 40 360 8 8 24 400 AMR‐WB 18.25 18.25 365 3 40 408 8 8 24 448 AMR‐WB 19.85 19.85 397 3 40 440 8 8 24 480 AMR‐WB 23.85 23.85 477 3 40 520 8 8 24 560 Table 11.20 SINR requirement of different voice code rate. Codec AMR‐NB 12.2 Bits/frame DL UL AMR‐NB 7.95 328 AMR‐NB 5.9 240 #PRBs MCS index 10% BLER (1st Tx) 2% BLER (4th Tx) 1 18 [dB] 11,87 [dB] 1,46 2 11 5,54 3 7 6 3 MCS index 200 10% BLER (1st Tx) 2% BLER (4th Tx) [dB] [dB] MCS index 10% BLER (1st Tx) 2% BLER (4th Tx) [dB] [dB] 15 9,65 0,27 13 7,87 –0,81 –2,03 8 3,73 –3,31 7 3,35 –3,61 2,72 –4,09 6 2,23 –4,17 4 0,76 –5,36 –0,10 –5,63 2 –0,92 –6,27 1 –1,55 –6,72 10% BLER (1st Tx) 2% BLER (4th Tx) MCS index 10% BLER (1st Tx) 2% BLER (4th Tx) 10% BLER (1st Tx) 2% BLER (4th Tx) [dB] [dB] [dB] [dB] [dB] [dB] #PRBs MCS index 1 17 11,43 1,95 2 10 5,36 3 7 6 3 MCS index 15 9,37 1,28 13 7,95 0,41 –1,97 8 3,83 –2,75 7 3,11 –3,07 2,54 –3,62 6 1,35 –4,25 4 0,78 –4,73 –0,33 –5,58 2 –1,37 –6,43 1 –1,89 –6,79 VoLTE Optimization 60.0 150% 40.0 100% 20.0 50% 0.0 Normal Consumption Consumption With Consumption With RoHC RoHC & VAD Kbps 0% % Reduction Figure 11.47 RoHC versus VoIP throughput. Table 11.21 VoLTE bitrate estimation – downlink. AMR12.2 AMR WB12.65 AMR WB23.85 No RoHC RoHC No RoHC RoHC No RoHC RoHC Payload and headers 616 320 624 328 848 552 Number of segments 1 1 1 1 1 1 TBS per segment 616 328 648 328 872 568 CRC added 640 352 672 352 896 592 Bitrate[kbps] 30.8 16.4 32.4 16.4 43.6 28.4 SID packet size (bit) 56 56 56 56 56 56 Number of SID packet/sec 10 10 10 10 10 10 SID bitrate[kbps] 0.56 0.56 0.56 0.56 0.56 0.56 Activity factor 60% 60% 60% 60% 60% 60% Overall bitrate[kbps] 18.7 10.1 19.7 10.1 26.4 17.3 Packet segmentation algorithm is not an event‐triggered mechanism, it is done automatically and only for UEs with poor radio channel. In worsening radio conditions scheduler performs packet segmentation on layer 2 in order to use more robust MCS and transmits the packet over multiple TTIs. Since RLC/MAC overhead is transmitted more than once, more resources are consumed to transmit the same amount of user data, more resources on PDCCH are utilized and also on PHICH due to transmission of ACKs/NACKs for HARQ purposes. Also the basic dynamic scheduler has the problem that it may excessively segment voice packets trying to maintain 10% iBLER as the path loss increases, leading to build‐up of voice frame segments in the queue leading to excess delay and this becomes the factor which limits the link budget (Figure 11.51). RLC segmentation reduces the payload bit per transmission and increases the amount of power per transmitted bit. It allows to use more robust MCS. Consequently, iBLER is reduced. Each segment occupies a separate HARQ process, the RLC layer reassembles the packet when all segments are successfully received. Each segment needs an additional grant, more resources on PDCCH are utilized, which increases the burden on the PDCCH. Link adaptation in dynamic scheduler (DS) reduces MCS level to a point where eventually a VoIP packet is segmented into many separate smaller MAC SDUs, and improves the link budget. 495 Table 11.22 VoLTE bitrate estimation – uplink. AMR12.2 AMR WB12.65 AMR WB23.85 Segmentation TTI Bundling Segmentation TTI Bundling Segmentation TTI Bundling No RoHC RoHC No RoHC RoHC No RoHC RoHC No RoHC RoHC No RoHC RoHC No RoHC RoHC Payload with RTP/UDP/IP/PDCP 584 288 584 288 592 296 592 296 816 520 816 520 Payload and all headers 616 320 616 320 624 328 624 328 848 552 848 552 4 4 4 4 4 4 4 4 4 4 4 4 Size per sub‐packet 162 88 616 320 164 90 624 328 220 146 848 552 TBS per segment 176 104 632 328 176 104 632 328 224 176 904 552 CRC added 200 128 656 352 200 128 656 352 248 200 928 576 180.8 110.4 Number of segments Bitrate[kbps] 35.2 20.8 SID packet size (bit) 56 56 Number of SID packet/s 10 10 SID bitrate[kbps] 0.56 0.56 126.4 65.6 35.2 20.8 56 56 56 56 10 10 10 10 0.56 0.56 0.56 0.56 126.4 65.6 44.8 35.2 56 56 56 56 56 56 10 10 10 10 10 10 0.56 0.56 0.56 0.56 0.56 0.56 Activity factor 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% 60% Overall bitrate[kbps] 21.3 12.7 76.1 39.6 21.3 12.7 76.1 39.6 27.1 21.3 108.7 66.5 VoLTE Optimization 180 Inter-Arrivel [ms] 160 140 120 100 80 60 40 20 :04 :17 :24 :30 :37 :43 :00 :55 :02 5.6 :01 :02 :02 :34 :41 :46 :52 :59 :05 :10 :17 :23 :30 :37 :42 :48 :55 2.1 5.7 :12 :18 :25 :33 :38 :45 :52 :58 :5.4 :09 :16 :23 :29 :34 0 180 160 140 120 100 80 60 40 20 0 14.8 17.8 14.8 10.9 17.7 13.2 19.6 15.9 12.2 16.6 13.3 10.4 17.0 14.3 11.2 16.1 12.7 19.2 15.4 10.5 17.5 13.5 10.4 17.2 12.3 18.9 15.8 12.2 16.7 12.9 18.8 15.9 13.2 18.8 15.5 12.4 18.3 15.5 10.0 16.7 13.2 19.7 14.7 Inter-Arival [ms] Figure 11.48 Inter‐arrival time of originator VoIP calls from field test. Figure 11.49 Inter‐arrival time of terminator VoIP calls from field test. Throughput IP [kbits/sec] 40 35 30 25 20 15 10 5 0 Figure 11.50 VoIP throughput distribution. Each segment is wrapped with RLC/MAC header and CRC checksum and is transmitted in a separate transport block causing additional overhead. VoIP packet size with RLC segmentation and overhead analysis are presented in Table 11.23. One example of segmentation of AMR‐NB 12.2 kbps is shown in Figure 11.52. Due to smaller segment sizes, RLC segmentation can help the link budget that the RLC segmentation allows higher total energy accumulation within the delay budget, and more robust MCS can be used increasing coverage but in cost of capacity degradation. 497 498 LTE Optimization Engineering Handbook To fit 3GPP defined TBS of 328 41 Bytes (328b) MAC RLC PDCP ROHC (1Byte) (1Byte) (1Byte) (4Byte) AMR12.65 (33 Byte) Pad (1Byte) CRC (3Byte) RLC payload is segmented MAC RLC (1Byte) (1Byte) Segment 1 (N1 Bytes) CRC (3Byte) MAC RLC (1Byte) (1Byte) Segment 2 (N1 Bytes) CRC (3Byte) MAC RLC (1Byte) (1Byte) Segment 3 (N1 Bytes) CRC (3Byte) Figure 11.51 RLC segmentation. Segmentation is also one of the methods for reducing the number of retransmissions. RLC SDUs are segmented at the RLC layer and the resulting segments transmitted in subsequent TTIs. When a RLC SDU is divided into many segments, each transport block has its own RLC/ MAC header. The transport blocks are transmitted in consecutive TTIs using different HARQ processes. The size of each RLC/MAC header is at least 24 bits but can be more. In addition, layer 1 adds 24 bits of CRC to each transport block. This means that segmentation increases the number of bits used for MAC and RLC headers as well as for CRC. Sending packet segments in consecutive TTIs allows more energy per voice packet to be aggregated, increasing coverage. Packet segmentation can provide up to ~5 dB coverage gain (by sacrificing capacity). Real gains will depend on used codec, link adaptation settings and number of segmentations. In a live network, the typical VoLTE RLC PDU distribution can be shown in Figure 11.53. Table 11.24 gives the coverage gain of AMR12.2 VoIP codec of regular transmission and segmentation. Since the maximum number of supported VoIP users per cell depends mostly on PDCCH resources, it significantly drops with the increase of segmentation order. Each segment (and HARQ retransmission) will require a new PDCCH allocation and a new PHICH in DL (for transmission of ACKs/NACKs per segment). The higher the achieved coverage gain, the greater is the capacity loss. Table 11.25 gives the PDCCH resources occupation with regular transmission and different segmentations. 11.6.3 TTI Bundling TTI Bundling is intended particularly for addressing LTE uplink link budget issues and balance the uplink coverage with the downlink coverage footprint. The main reason for the imbalance of the UL/DL link budgets is that UE maximum transmit power is 200 mW, as opposed to tens of watts for the downlink. UE cannot use enough energy during one TTI in order to send successfully a VoIP packet. The uplink, however, is believed to be coverage limited, and therefore, TTI bundling can be very beneficial when the UE is close to the cell edge specifically for VoLTE services. The eNB activates TTI bundling if the SINR for the UE drops below a configurable threshold when BLER increases and link adaptation has no more options for MCS/PRB reduction. When TTI bundling is activated,TTI bundling uses four automatic retransmissions in four consecutive uplink TTI with a common ACK/NACK for HARQ and that means the receiver waits over the total transmission period (four TTIs) before sending feedback. The four consecutive uplink TTIs are called TTI bundle, and these four transmissions are non‐adaptive with identical MCS/RB location but different redundancy versions. Different redundancy versions are employed to achieve incremental redundancy soft combining gain. Only one uplink grant and one HARQ feedback channel is transmitted for a bundle. HARQ retransmission Table 11.23 VoIP packet size with RLC segmentation and overhead analysis. no segment Segment Size 2 segments Segment Size L1 SDU Voice Codec [bits] [bits] Segment overhead AMR‐NB 1.8 120 120 0% AMR‐NB 4.75 176 176 0% AMR‐NB 5.15 184 184 0% AMR‐NB 5.9 200 200 AMR‐NB 6.7 216 AMR‐NB 7.4 4 segments Segment Size L1 SDU [bits] Segment overhead 80 160 33% 112 224 23% 112 224 0% 120 216 0% 232 232 AMR‐NB 7.95 240 AMR‐NB 10.2 8 segments L1 SDU Segment Size L1 SDU [bits] [bits] Segment overhead [bits] Segment overhead 64 256 100% 56 448 233% 80 320 68% 64 512 159% 22% 80 320 65% 64 512 152% 240 20% 80 320 60% 64 512 140% 128 256 19% 88 352 56% 64 512 130% 0% 136 272 17% 88 352 52% 64 512 121% 240 0% 144 288 17% 96 384 50% 72 576 117% 288 288 0% 168 336 14% 104 416 42% 72 576 97% AMR‐NB 12.2 328 328 0% 184 368 12% 112 448 37% 80 640 85% AMR‐WB 1.75 120 120 0% 80 160 33% 64 256 100% 56 448 233% AMR‐WB 6.6 216 216 0% 128 256 19% 88 352 56% 64 512 130% AMR‐WB 8.85 264 264 0% 152 304 15% 96 384 45% 72 576 106% AMR‐WB 12.65 336 336 0% 192 384 12% 120 480 36% 80 640 83% AMR‐WB 14.25 368 368 0% 208 416 11% 128 512 33% 88 704 76% AMR‐WB 15.85 400 400 0% 224 448 10% 136 544 30% 88 704 70% AMR‐WB 18.25 448 448 0% 248 496 9% 144 576 27% 96 768 63% AMR‐WB 19.85 480 480 0% 264 528 8% 152 608 25% 96 768 58% AMR‐WB 23.85 560 560 0% 304 608 7% 176 704 21% 112 896 50% [bits] [bits] 500 LTE Optimization Engineering Handbook When a RLC SDU is divided into many segments, each transport block has its own RLC/MAC header. No fragmentation, ROHC, same as TTI-B case (OH 23%) MAC (1B) 2 fragments (OH 32%) 3 fragments (OH 37%) 4 fragments (OH 43%) PDCP ROHC (1B) (4B) RLC (1B) AMR12.65 (34B) CRC (3B) 41Bytes (328b) MAC (1B) RLC (1B) MAC (1B) 22Bytes (176b) RLC (13 B) (1B) (20 B) CRC (3B) CRC (3B) 15Bytes (120b) MAC (1B) RLC (1B) (10 B) CRC (3B) (5 B) CRC (3B) 12Bytes (96b) MAC (1B) 8 fragments (OH 58%) RLC (1B) 7Bytes (56b) Number of Segments Segment Size [bits] Maximum Allowable Path Loss* [dB] 1 (no segmentation) 328 161.13 2 184 162.51 4 112 164.22 8 80 165.16 Figure 11.52 Example: AMR‐NB 12.2 (uplink). 40 35 30 PDCP PDU is segmented into small RLC PDUs due to poor RSRP, size from 6 byte to 40 byte 25 20 15 10 5 0 Figure 11.53 RLC PDU distribution. of a TTI bundle is also transmitted as a bundle, occurs 16 TTIs after the previous transmission in order to be synchronized with normal (non‐bundled) HARQ retransmissions (8 TTIs). TTI bundling effectively utilizes more time resources by bundling a single VoIP packet into four consecutive TTIs. UL coverage is improved by allocating more energy per packet avoiding the excessive overhead from packet segmentation. Usually TTI bundling can provide gains up to 2 to 4 dB in uplink coverage. From Figure 11.54, we can see that TTI bundling needs reduced segmentation with less RLC and MAC overhead compared to normal operation, and also less control signaling, that is, HARQ feedback and PDCCH grants sent to UE. Figure 11.55 shows the measurements of RSRP, PUSCH BLER, and POLQA MoS without and with TTI bundling. The HARQ retransmissions ratio increases quickly after certain point in pathloss has been experienced. Once target BLER of ~10% cannot be maintained, MoS decreases quickly. Without TTI bundling, the VoLTE coverage RSRP threhold is around −114 dBm, with TTI Bundling is −117 dBm. Figure 11.56 shows the measurements of UL_SINR without and with TTI bundling. UL_SINR is more than 3 dB better when UE is TTI‐B on than off. The activation and de‐activation of TTI Bundling in the eNB is dynamic and is based on the channel quality. To prevent continuous (ping‐pong) activation and de‐activation of TTI bundling, hysteresis are available to avoid this behavior and unnecessary processor load increase in the eNB. TTI bundling is configured (is activated/deactivated) per UE via RRC control messaging as shown below. VoLTE Optimization Table 11.24 Coverage gain of AMR12.2 VoIP codec of regular transmission and segmentation. regular(1tx) regular(4tx) 2 segments 3 segments From PDCP bits 304 304 304 304 No of segments # 1 1 2 3 L2 overhead (RLC + MAC) bits 16 16 16 16 Segment size bits 320 320 168 120 No of HARQ transmissions # Modulation and coding scheme 1 4 4 4 MCS3 MCS3 MCS2 MCS0 No of PRBs # 6 6 4 5 Actual transport block size bits 328 328 176 120 Energy gain (frequency domain) dB % of VoIP per TTI 0 0 1.76 0.79 93% 93% 86% 84% TX power per VoIP per TTI dBm 22.67 22.67 22.36 22.27 Total energy per packet dBm 22.67 22.67 25.37 27.04 Total energy per packet (HARQ) dBm 22.67 28.69 31.39 33.06 Energy gain (time domain) dB 0 6.02 8.72 10.39 Required SINR (10%/1Tr) dB −1.3 −1.3 −1.1 −3.6 Total delta compared to 10% BLER dimensioning dB 0 6.0 10.3 13.5 Table 11.25 PDCCH resources occupation with regular transmission and segmentation. regular(1tx) regular(4tx) 2 segments 3 segments PRB used per VoIP packet (cell edge) 6 24 32 60 PDCCH used per VoIP packet (cell edge) 1 4 8 12 PDCCH capacity used per VoIP packet (cell edge) 1.2% 4.8% 9.7% 14.5% PDCCH capacity used per VoIP packet (non cell edge) 0.36% 0.36% 0.36% 0.36% No. of supported VoIP users (5% users in cell edge) 247 171 121 93 No. of supported VoIP users (10% users in cell edge) 224 123 77 56 No. of supported VoIP users (40% users in cell edge) 142 46 24 17 value DL-DCCH-Message ::= { message c1 : rrcConnectionReconfiguration : { rrc-TransactionIdentifier 3, criticalExtensions c1 : rrcConnectionReconfiguration-r8 : { mac-MainConfig explicitValue : {ul-SCH-Config { maxHARQ-Tx n16, periodicBSR-Timer sf5, retxBSR-Timer sf320, ttiBundling TRUE }, 501 502 LTE Optimization Engineering Handbook Segmentation –i.e. one RLC SDU using 4 RLC PDUs -Four PDCCH allocations -Four HARQ feedbacks Overhead RLC SDU RLC PDU TB RLC PDU TB RLC PDU RLC SDU RLC PDU TB RLC PDU TB Over head One RLC SDU transmitted in one RLC PDU -One PDCCH allocation -One HARQ feedback TB time time Figure 11.54 Segmentation and TTI bundling. When TTI bundling mode20 is activated, PUSCH single transport block is transmitted over four consecutive TTIs but with different redundancy version, only one UL grant is given for the transmission of the whole bundle, retransmission of a TTI bundle, which is also transmitted as a TTI bundle, occurs 16 TTIs after previous (re)transmission. The procedure of FDD and TDD UL transmission with TTI bundling are shown in the Figure 11.57 and Figure 11.58. TTI bundling gain is introduced by the energy collected from additional transport blocks received during the assumed service delay budget. Higher air link latency budget of VoIP services (typically 50 ms) allows for up to seven packet retransmissions (initial transmission + six retransmissions). From Figure 11.59, we can see that TTI bundling gain can be calculated according to: TTI bundling gain 10 * log 12 / 7 2.34 dB In conclusion, in those poor radio conditions, UL resources over multiple consecutive TTIs can be assigned with a single grant, which decreases the signaling overhead. While TTI bundling is a very useful feature to increase coverage, it comes at the impact of capacity. VoIP capacity will be reduced when large number of VoIP users have TTI bundling enabled, since four consecutive subframes are used by these users. In a live network, we should limit the use of TTI bundling only when it have reached the limits of MCS/PRB override. TTI bundling activation requires a higher layer RRC reconfiguration to enable, and only a single specific format (MCS = 6, 1 PRB) can be used to take advantage of the 4 ms subframe bundling, thus will limit scheduler’s link adaptation flexibility. It should be known that for FDD LTE, HARQ round trip doubled from 8 ms to 16 ms, with TTI bundling, VoLTE data rate is (328 + 24 bits)/(4 HARQ Tx * 4 ms TTI) = 22 kbps. 11.6.4 TTI Bundling Optimization Since TTI bundling will take four subframes of uplink, which impact the capacity greatly, for TTI bundling optimization, it’s reasonable and necessary to restrict the TTI bundling user number, based on the configured SINR and TTI bundling number. TTI bundling increases the probability of successful transport block decoding by eNB. The UE eligibility for TTI bundling is evaluated at context creation (i.e., call setup, handover, call re‐establishment) and updated during normal call operation (bearer creation or deletion or maybe bearer modification). In a situation when BLER increases and link adaptation has no 20 Feature group indicator processing required to determine if UE supports TTI bundling. VoLTE coverage (above POLQA2.5) increased by 2.5dB if harqMaxTrUITti Bundling is 24. TTI Bundling OFF –100 TTI Bundling ON (retrans 24times) +2.5dB gain RSRP[dBm] Poly. (RSRP[dBm]) –108 RSRP : –114.5dBm –112 RSRP : –117dBm –112 –116 –120 –124 –124 TTIB ON POLQA MOS Poly. (POLQA MOS) 4 3 3 POLQA MOS POLQA MOS 5 4 Poly. (POLQA MOS) POLQA : 2.5 2 2 1 POLQA : 2.5 1 0 0 PUSCH BLER[%] 100 PUSCH BLER[%] 100 PUSCH BLER[%] Poly. (PUSCH BLER[%]) 80 PUSCH BLER[%] 80 60 60 40 40 Poly. (PUSCH BLER[%]) 20 20 Figure 11.55 With TTI Bundling the RSRP level goes down to −117 dm, 2.5 dB gain achieved. 59:02.4 57:36.0 56:09.6 54:43.2 53:16.8 51:50.4 50:24.0 48:57.6 47:31.2 51:21.6 49:55.2 48:28.8 47:02.4 45:36.0 44:09.6 42:43.2 0 41:16.8 0 Poly. (RSRP[dBm]) –108 –116 POLQA MOS RSRP[dBm] –104 –120 5 RSRP[dBm] –100 RSRP[dBm] –104 LTE Optimization Engineering Handbook Tti-Feature-ON TTI-Feature-OFF Poly. (Tti-Feature-ON) Poly. (TTI-Feature-OFF) UL HARQ NACK Rate 1 2 3 4 5 RSRP 6 7 8 9 10 11 12 13 14 15 UL_SINR PUSCH BLER POLQA MOS –96 –98 –100 –102 –104 –106 –108 –110 –112 –114 –116 –118 –120 –122 –124 80 70 60 50 40 30 20 10 0 PUSCH BLER [%], POLQA MOS * 10 –3 –2 –1 0 RSRP [dBm] 504 Figure 11.56 UL_SINR is more than 3 dB better when UE is TTI‐B on. TTI# 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 UL grant on 4 redundancy versions of the same packet PDCCH Tx on PUSCH RV0 RV2 RV3 RV1 Decoding ACK/NACK on PHICH RV0 RV2 RV3 RV1 N N Retransmission of bundle 1st bundle transmission HARQ RTT: 16 TTIs Figure 11.57 FDD UL transmission with TTI bundling. more options for MCS/PRB reduction while radio conditions for handover are not fulfilled, TTI bundling can be triggered to keep the voice call quality (in terms of delay and packet loss rate) before UE will either change the cell or RF conditions becomes better (Figure 11.60). The eNB periodically monitors every 100 ms the eligible UEs for TTI bundling activation/deactivation. Once the criteria for entering TTI bundling mode are fulfilled, eNB triggers intra‐cell handover procedure by sending RRC connection reconfiguration message toward the UE (Figure 11.61). VoLTE Optimization TDD Configuration 1 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D D S U U D R R V V 0 2 6 6 R R V V 3 1 8 R R V V 0 2 NACK R R V V 3 1 NACK UL grant R R V V 0 2 R R V V 3 1 8 6 7 ms Figure 11.58 TDD UL/DL configuration 1 with TTI bundling. N PHICH Classical Decoding transmission 8ms RTT PUSCH 328 TTI 1 2 (7 TBSs) 3 4 5 N 6 7 N N PHICH TTI Bundling Decoding 16ms RTT PUSCH 328 328 328 328 RV0 RV1 RV2 RV3 (12 TBSs) TTI 1 2 3 4 5 6 7 N N N 328 328 328 328 328 328 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 8 N N 328 328 328 328 328 328 328 328 RV0 RV1 RV2 RV3 RV0 RV1 RV2 RV3 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Figure 11.59 TTI bundling gain. Channel Quality TTI-B activation LTE frequency2 layer HO conditions fulfilled Decision to switch UE to TTI Bundling LTE frequency1 layer Time TTI Bundling area Figure 11.60 TTI bundling triggered. UE eNB eNB decides to configure UE with TTI bundling mode based on UL channel quality RRC: RRC Connection Reconfiguration (includes MobilityControllnfo) DCCH/UL-SCH/PUSCH (SRB1) PHY: CBRA Random Access Preamble √RACH/PRACH MAC: CBRA Random Access Response √DL-SCH/PDSCH MAC: Random Access Msg3 DCCH/UL-SCH/PUSCH (SRB1) UL Grant PDCCH RRC: RRC Connection Reconfiguration Complete (sent in TTI Bundling mode) DCCH/UL-SCH/PUSCH (SRB1) Figure 11.61 Dynamic switching to/from TTI bundling. 505 LTE Optimization Engineering Handbook TTI bundling activation ccriterias include load criteria and poor RF performance conditions. For load criteria, current total GBR PRB utilization in the cell should be less than TTI bundling activation load threshold, and number of users in the cell already in TTI bundling configuration should be less than the maximum number of TTI bundling users. The consideration to disable the feature might be in heavily loaded eNBs due to high VoIP traffic load at cell edges. If the majority of this VoIP traffic experiences poor channel conditions it may trigger an excessive number of UEs using TTI bundling. This might have impact on the end‐user experience and performance. For poor RF performance criteria, the estimated achievable UL_SINR with 1PRB grant should be less than uplink link budget threshold for a period time, and the measured BLER at TTI bundling measurement points (HARQ Tx point) should bigger than TTI bundling activation BLER threshold within the IIR filter window. With TTI bundling ON, the PRB allocation size for the UE is restricted to NPRB < = 3 and the modulation order is limited to QPSK of the VoLTE service. TTI bundling can improve uplink quality in weak signal (‐120 dBm). BLER is reduced from 40%‐50% to below 1%, which improves MoS from below two (bad quality) to more than three (good quality). The field test result is shown in Table 11.26. TTI bundling is a UE capability that can be used to improve uplink coverage for VoLTE services at lower delay and to reduce the amount of PDCCH control channel usage in cases where normal transmission would have required RLC segmentation. TTI bundling is only beneficial for users close at the cell border that suffer from power shortage. On the other hand, increased latency and jitter after TTI bundling is activated, and sometimes with increased packet drops (PDCP timer discards). The example with 18 UE loading is shown in Figure 11.62 from a field test. Table 11.26 TTI bundling gain. TTIB UE# MoS UL PRB MCS OFF 1 2.4 5.0 0.1 2 1.8 4.9 ON 1 3.2 3.0 2 3.0 3.0 12.0 Tx Power PUSCH PUSCH BLER[%] 24.0 54.4 0.1 23.9 39.6 38 15 −120.6 10.1 23.5 0.6 125 14 −122.1 23.3 0.1 128 15 −115.6 250 200 Latency [ms] 506 150 100 50 0 Figure 11.62 E2E packet latency versus TTI bundling. Phy Thrp [kbps] 47 TTI-B activation PDCP Thrp [kbps] DL RSRP [dBm] 13 −118.7 VoLTE Optimization Codec SINR [dB] AMR wide-band 23.85 kbps EPAS –3.1 EVA70 –4.5 ETU300 –5.6 AMR wide-band 12.65 kbps –4 –7.8 –8.5 AMR 12.2 kbps –4 –7.8 –8.5 AMR 5.9 kbps –3.7 –7.9 –8.9 Figure 11.63 Downlink RS SINR and uplink SINR values for VoIP failure. In field test, actual cause of call drop can be attributed to the UL/DL control channels failing due to poor SINR. It is worth to know that only increasing the TTI bundling threshold (to preserve the UL failure) would not help when the DL control channels fail at the poor SINR (around −5 dB) (Figure 11.63). 11.6.5 Coverage Gain with RLC Segmentation and TTI Bundling Packet segmentation algorithm is used as an extension to link adaptation for uplink coverage improvement, it is done automatically and only for UEs with poor radio channel. Uplink packet are segmented by providing according grants in order have smaller packets with more robust coding but at the cost of additional signaling load and additional header load. As the number of transmitted packets increases, more resources on PDCCH are utilized which also happens on PHICH in DL due to the transmission of ACKs/NACKs for HARQ purposes. The alternative solution is TTI bundling at low/medium loaded networks. The energy per transmitted bit in uplink is increased by allocating one transport block to four ins

LTE optimization engineering handbook (Zhang, Xincheng) (z-lib.org)

Related documents

Study collections

Products

Support

LTE optimization engineering handbook (Zhang, Xincheng) (z-lib.org)

Related documents

Study collections

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib