Radio Interface Engineering Rules Document number: Document issue: Document status: Date: PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 External document Copyright© 2006 Nortel Networks, All Rights Reserved Printed in France NORTEL CONFIDENTIAL The information contained in this document is the property of Nortel Networks. Except as specifically authorized in writing by Nortel Networks, the holder of this document shall keep the information contained herein confidential and shall protect same in whole or in part from disclosure and dissemination to third parties and use same for evaluation, operation and maintenance purposes only. The content of this document is provided for information purposes only and is subject to modification. It does not constitute any representation or warranty from Nortel Networks as to the content or accuracy of the information contained herein, including but not limited to the suitability and performances of the product or its intended application. This is the Way. This is Nortel, Nortel, the Nortel logo, and the Globemark are trademarks of Nortel Networks. All other trademarks are the property of their owners. Radio Interface Engineering Rules PUBLICATION HISTORY Issue 01.01/EN, Preliminary 11/Mar/2005 Creation of the document for the V15.1 release 28/Jun/2005 Issue 01.02/EN, Standard Update of the document for the V15.1 CHR Issue 01.03/EN Preliminary 21/Oct/2005 Update of the document for the V15.1.1 release 14/April/2006 Issue 02.01/EN Preliminary Update of the document for the V16.0 release 28/Sept/2006 Issue 02.02/EN Standard Update of the document for the V16.0 ChR Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 2/84 Radio Interface Engineering Rules CONTENTS 1 INTRODUCTION............................................................................................................................6 1.1. OBJECT .................................................................................................................................6 1.2. SCOPE OF DOCUMENT .......................................................................................................6 1.3. AUDIENCE OF THIS DOCUMENT ........................................................................................6 1.4. DELTA BETWEEN RELEASES .............................................................................................6 1.4.1 1.4.1.1 1.4.1.2 1.4.2 1.4.2.1 1.4.3 1.4.3.1 2. 3. Delta between V15.0 and V15.1 .....................................................................................6 General information .................................................................................................................... 6 Evolutions between releases........................................................................................................ 6 Delta between V15.1 and V15.1.1 ..................................................................................7 Evolutions between releases........................................................................................................ 7 Delta between V15.1.1 and V16.0 ..................................................................................7 Evolutions between releases........................................................................................................ 7 RELATED DOCUMENTS ..............................................................................................................8 2.1. APPLICABLE DOCUMENTS .................................................................................................8 2.2. REFERENCE DOCUMENTS .................................................................................................8 RADIO INTERFACE ENGINEERING RULES ..............................................................................9 3.1. GSM RADIO PROPAGATION AND FREQUENCY ASPECTS ...............................................................9 3.1.1 Frequency spacing..........................................................................................................9 3.1.1.1 Intra_cell ..................................................................................................................................... 9 3.1.1.2 Intra_site.................................................................................................................................... 10 3.1.1.3 Inter_site.................................................................................................................................... 10 3.1.2 Types of frequency hopping..........................................................................................11 3.1.2.1 MA, HSN, MAIO...................................................................................................................... 12 3.1.2.2 Case of 1:1 fractional re-use pattern.......................................................................................... 12 3.1.2.3 Case of 1:3 fractional re-use pattern.......................................................................................... 14 3.1.2.4 1:1 versus 1:3 ............................................................................................................................ 15 3.1.2.5 AD HOC solution...................................................................................................................... 15 3.1.3 Frequency Load ............................................................................................................17 3.1.4 Radio link budget ..........................................................................................................17 3.1.5 Interference measurement ............................................................................................17 3.1.6 Antenna constraints ......................................................................................................17 3.1.6.1 900/1800 cositing ...................................................................................................................... 17 3.1.6.2 Blocking .................................................................................................................................... 18 3.2. RADIO INTERFACE DIMENSIONING .............................................................................................18 3.2.1 Traffic models................................................................................................................18 3.2.2 Dimensioning principles ................................................................................................18 3.2.2.1 Tch dimensioning...................................................................................................................... 18 3.2.2.2 SDCCH dimensioning............................................................................................................... 19 3.2.2.3 BCCH dimensioning ................................................................................................................. 21 3.2.2.4 CCCH dimensioning ................................................................................................................. 22 3.2.2.5 Adaptation of CCCH dimensioning to the paging flow ............................................................ 22 3.3. TDMA CONFIGURATIONS AND PRIORITIES ..................................................................................26 3.3.1 3.3.2 3.3.3 3.3.4 Standard cell .................................................................................................................26 Extended ccch...............................................................................................................27 EXtended cell ................................................................................................................27 TDMA priorities .............................................................................................................28 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 3/84 Radio Interface Engineering Rules 3.3.4.1 Standard cell.............................................................................................................................. 28 3.3.4.2 Extended cell............................................................................................................................. 29 3.4. MONOZONE AND CONCENTRIC CONFIGURATION ........................................................................30 4. GPRS RADIO INTERFACE ENGINEERING RULES .................................................................31 4.1. RADIO RESOURCES .................................................................................................................31 4.1.1 TDMA rules ...................................................................................................................31 4.1.2 PDCH rules ...................................................................................................................32 4.2. CODING SCHEMES CS-1 AND CS-2 ............................................................................................35 4.2.1 Coding schemes description: ........................................................................................35 4.2.2 Rules .............................................................................................................................36 4.3. GPRS CHANNELS’ DIMENSIONING ..............................................................................................37 4.3.1 4.3.2 4.3.2.1 4.3.2.2 5. GPRS impact in the ccch load ......................................................................................37 GPRS channels’ rules ...................................................................................................39 PDCH dimensioning rules......................................................................................................... 39 Dynamic sharing rules............................................................................................................... 40 EDGE RADIO INTERFACE ENGINEERING RULES .................................................................42 5.1. OBJECTIF ................................................................................................................................42 5.2. EDGE, A FORWARD STEP........................................................................................................42 5.3. EDGE DEPLOYMENT STRATEGY ..............................................................................................43 5.4. MAINS ENGINEERING RULES .....................................................................................................44 5.4.1 EDGE TDMA rules ........................................................................................................44 5.5. SPECIFIC DATA FEATURES .......................................................................................................44 5.5.1 Network assisted cell change impact on gprs/edge networks ......................................44 5.5.2 Packet Flow COntext impact on GPRS/EDGE networks..............................................45 5.6. DATA THROUGHPUT ESTIMATION .............................................................................................45 5.6.1 Radio conditions (C/I and Eb/No) distributions on the cell............................................48 5.6.1.1 C/N distribution......................................................................................................................... 48 5.6.1.2 C/N at cell edge ......................................................................................................................... 49 5.6.1.2.1 C/N distribution estimation ................................................................................................ 50 5.6.1.3 C/I distribution .......................................................................................................................... 50 5.6.1.3.1 C/I at cell edge.................................................................................................................... 50 5.6.1.3.2 C/I distribution ................................................................................................................... 51 5.6.2 Mean Throughput per TS calculation............................................................................52 5.6.2.1 BLER distributions on the cell .................................................................................................. 52 5.6.2.2 Throughput distributions on the cell ......................................................................................... 54 5.6.2.3 Link adaptation.......................................................................................................................... 55 5.6.2.4 Mean throughput / TS ............................................................................................................... 56 5.6.3 Comments on data throughput calculations..................................................................58 5.6.3.1 C/(I+N) versus BLER estimation .............................................................................................. 58 5.6.3.2 Limitations ................................................................................................................................ 58 5.7. EDGE PERFORMANCES ESTIMATIONS .....................................................................................58 5.7.1 Incremental Redundancy ..............................................................................................60 5.7.1.1 Incremental Redundancy Functionality..................................................................................... 60 5.7.1.1.1 Retransmission table in UL ................................................................................................ 60 5.7.1.1.2 Retransmission table in DL ................................................................................................ 61 5.7.1.2 Incremental redundancy performance improvement ................................................................. 61 5.7.1.3 MCS usage distribution............................................................................................................. 63 5.7.2 High Power Amplifier impact on Data performances ....................................................66 5.7.3 Frequency Reuse..........................................................................................................68 5.7.4 Important hypothesis and assumptions supposed........................................................73 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 4/84 Radio Interface Engineering Rules 5.7.5 BLER study ...................................................................................................................74 5.8. UL PERFORMANCE ANALYSIS ..................................................................................................76 5.8.1 Different mobile type performance impact ....................................................................77 5.9. LAB TEST RESULTS .................................................................................................................78 5.9.1 UL Lab Results..............................................................................................................78 5.9.2 DL Lab Results..............................................................................................................79 5.9.3 Real Link Adaptation impact on mean data throughput................................................79 5.9.4 Applicative versus RLC/MAC throughput relationship ..................................................80 5.10. CONCLUSION ..........................................................................................................................81 6. ABBREVIATIONS AND DEFINITIONS.......................................................................................82 6.1. ABBREVIATIONS ......................................................................................................................82 6.2. DEFINITIONS ...........................................................................................................................84 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 5/84 Radio Interface Engineering Rules 1 INTRODUCTION 1.1. OBJECT This document is an inventory of all the engineering rules concerning the GSM/ GPRS/EDGE radio Interface. An engineering methodology to estimate EDGE introduction and performances in case of EDGE deployment on existing GSM/ GPRS network is also detailed in this document. . 1.2. SCOPE OF DOCUMENT This document is applicable from V16.0 release. 1.3. AUDIENCE OF THIS DOCUMENT This document is intended primarily for customer network designers and applications engineers involved in GSM networks engineering with Nortel GSM networks. 1.4. DELTA BETWEEN RELEASES 1.4.1 DELTA BETWEEN V15.0 AND V15.1 1.4.1.1 GENERAL INFORMATION This document is created in V15.1 release since the BSS Engineering Rules document has been split in different elementary rules. Nevertheless the majority of the rules and recommendations of this document where already applicable to V15.0. The main evolutions are presented below. 1.4.1.2 EVOLUTIONS BETWEEN RELEASES New rules introduced and rules modified due to V15.1 evolutions • Air_TDMA_005 • Air_FH_006 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 6/84 Radio Interface Engineering Rules Rules modified or added due to lack of precisions or return from experience (but not linked to V15.1 content) Air_CS_002 Rules suppressed • None Other main changes in the document • This document allows to group together three documents which are the Radio interface, GPRS Interface and the new EDGE radio interface. The last one for the moment is not detailed with all engineering rules associated. Only EDGE TDMA rules are added now. The next step may be for V15.1 CHR will be to provide thanks the EDGE engineering study the mains rules deduced. 1.4.2 DELTA BETWEEN V15.1 AND V15.1.1 1.4.2.1 EVOLUTIONS BETWEEN RELEASES New rules introduced and rules modified due to V15.1.1 evolutions • None Rules modified or added due to lack of precisions or return from experience (but not linked to V15.1 content) • Air_PDCH_004 • Air_TDMA_003 Rules suppressed • None Other main changes in the document • Addition of a new chapter providing information about Air capacity for paging. 1.4.3 DELTA BETWEEN V15.1.1 AND V16.0 1.4.3.1 EVOLUTIONS BETWEEN RELEASES New rules introduced and rules modified due to V16.0 evolutions • Air_NACC_001 • Air_NACC_002 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 7/84 Radio Interface Engineering Rules Rules modified or added due to lack of precisions or return from experience (but not linked to V16.0 content) • Air_PDCH_004 • Air_freq_001 • Air_freq_002 • Air_freq_003 • Air_FH_001 • Air_FH_002 • Air_FH_003 • Air_HSN_001 • Air_HSN_002 Rules suppressed • None Other main changes in the document • 2. 2.1. 2.2. None RELATED DOCUMENTS APPLICABLE DOCUMENTS [A1] PE/DCL/DD/014281 PCUSN Engineering Rules [A2] PE/DCL/DD/014284 Agprs Interface Engineering Rules REFERENCE DOCUMENTS [R1] PE/DCL/DD/014280 BSC/TCU3000 Engineering Rules [R2] PE/DCL/DD/014286 A and Ater Interfaces Engineering Rules [R3] PE/DCL/DD/014285 Abis Engineering Rules [R4] PE/IRC/APP/008966 EDGE Engineering Guideline [R5] PR/BTS/DD/4091 EGPRS Handbook Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 8/84 Radio Interface Engineering Rules 3. RADIO INTERFACE ENGINEERING RULES 3.1. GSM RADIO PROPAGATION AND FREQUENCY ASPECTS 3.1.1 FREQUENCY SPACING 3.1.1.1 INTRA_CELL Rule Air_Freq_001 (O) Nortel BTS is using hybrid couplers technology. Considering the UL power control activated, Nortel recommends 400khz frequency spacing between TRX on a same cell with or without frequency hopping, in order to respect the following minimum ratios recommended in 05.05 to guarantee voice quality C/I>=-9dB for first adjacent channels( 200kHz), C/I>=-41dB for second adjacent channels( 400KHz), C/I>=-49dB for third adjacent channels ( 600KHz). Justification: The most constraining case is the following one. A UE at 30m from the BTS, and another one at the cell edge. Based on many measurement campaingn analysis the signal received at 30m from the BTS is -40dBm, the minimum signal received at the cell edge is -102dBm. The pathloss difference is 62dB, in that case the UE near the BTS can create high UL interferences from the UE far from the BTS. The UL power control allows at least a 24dB dynamic in the reduction of the UE output power. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 9/84 Radio Interface Engineering Rules The pathloss difference with the UL poWer control is 38dB. 400KHz frequency spacing allow sufficient isolation. 3.1.1.2 INTRA_SITE Rule Air_Freq_002 (O) The recommendations are the same than the ones of the Rule Air_Freq_001. The most constraining case is shown in the picture below, when two mobiles are in the adjacent sectors overlap area, one mobile is far from the BTS in communication with sector 1 and another mobile near the BTS in communication with sector2. In that case, the highest pathloss difference between the UE is in the worst case the same than the one calculated previously, the recommendation is 400KHz frequency spacing. 3.1.1.3 INTER_SITE Rule Air_Freq_003 (O) For the inter site scenario, the most constraining case, is when 2 mobiles are in the overlap area between 2 cells (see the picture below). Each mobile UL transmission is an interference for the other cell and each cell DL signal is an interference for the mobile not in communication with it. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 10/84 Radio Interface Engineering Rules As the pathloss difference between a mobile and each cell is less than the margin taken for the HO (less than 6dB), the C/I can be equal to -6dB. 200KHz frequency spacing are necessary to ensure this, as they guarantee 18dB isolation 3.1.2 TYPES OF FREQUENCY HOPPING According to the type of combiners used in the BTS, two main types of frequency hopping mechanism can be used: - Synthesized mode It consists in changing the RF channel of each TDMA frame except the BCCH, within each cell of the network. Each hopping TDMA frame uses a RF channel picked in a given frequency group which size can be higher than the number of TRXs in the cell. Rule Air_FH_001 (O) When the BTS is equipped with hybrid combiners (H2D, H4D) or duplexers, only the synthesized frequency hopping mode can be used. Rule Air_FH_002 (O) It is not recommended to include the BCCH frequency in the list of hopping frequencies whatever the choice of frequency hopping type. Rule Air_FH_003 (O) In a cell where GPRS/EDGE service is activated and frequency hopping is used for GPRS/EDGE TDMA, the maximum number of hopping frequencies in this cell should be as shown in the following table: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 11/84 Radio Interface Engineering Rules BSS Release Max. Number of Frequencies (CA) Max. Number of Hopping Frequencies (MA) V15.0 / V15.0.1 55 55-n V15.1/V15.1.1 52 52-n From V16.0 50 50-n where n is the number of non-hopping frequencies in cell. Justification: In case of using a higher number of hopping frequencies, there is no sufficient place to carry the complete GPRS mobile allocation bitmap in the corresponding field of the SI13. This truncates some mandatory fields of the SI13 and has as consequence a GPRS/EDGE service dysfunction. 3.1.2.1 MA, HSN, MAIO The following rules apply in the case where only Nortel BTS are used. (means that they are synchronized). 3.1.2.2 CASE OF 1:1 FRACTIONAL RE-USE PATTERN Rule Air_HSN_001 (M) In case of 1:1 fractional re-use pattern, The HSN must be the same for all the cells within a site. To get full benefit of frequency hopping, a minimum of 6 different frequencies shall be used in each cell. Justification: Not following this rule will lead to frequency collision. From Nortel Experience, 6 frequencies lead to a good protection against fading effects. Rule Air_MAIO_001 (M) The choice of MAIO depends on the frequency load. Nortel’s experience recommends: If frequency loads <=16.6%, the MAIO are given according to step 2 rule and considering the TRX per order (TRX1 of cell1, TRX1 of cell2, TRX2 of cell1...) Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 12/84 Radio Interface Engineering Rules Example: • TRX1 of cell 1: MAIO 0 • TRX1 of cell 2: MAIO 2 • TRX1 of cell 3: MAIO 4 • TRX2 of cell 1: MAIO 6 • TRX2 of cell 2: MAIO 8 If frequency load >16.6%, the MAIO are given according to step 2 rule as long as the frequency is available (i.e. it has not already been chosen for a previous MAIO). Then, the MAIO are given according to step 1 rule which leads to adjacent frequencies. Step 1 MAIO are given to the cells which have the smallest overlap with neighbors or which have the smallest traffic. Example: There is one site, 3 cells and 10 frequencies in the group. Frequency load = 20% (2 hopping TRXs per cell). MAIO for hopping TRX1 and TRX2 of each cell: • cell 1: 0 & 6 • cell 2: 2 & 8 • cell 3: 4 & 9 If HSN make us start by f3, the frequencies in grey cases will be chosen. MAIO 8 9 0 1 2 3 4 5 6 7 Cell1 f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 Cell2 f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 Cell3 f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 There are adjacent interference (f1,f2,f3). Justification: In case of freq. load <= 16.6%, there are no intra site collision (adjacent or co channel). All frequencies inside the group are adjacent so the step 2 rule allows maximizing the space between the frequencies and ensuring the non-adjacency of the frequencies. However in case of freq. load > 16.6%, there are intra site collision (adjacent). The step 1 rule will create adjacent interference. It is not recommended to use 1:1 frequency reuse pattern if the frequency load is >20%. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 13/84 Radio Interface Engineering Rules 3.1.2.3 CASE OF 1:3 FRACTIONAL RE-USE PATTERN Rule Air_HSN_002 (O) In case of 1:3 fractional re-use pattern, The HSN must be the same for all the cells within a site. To get full benefit of frequency hopping, a minimum of 6 different frequencies shall be used in each cell. Justification: Not following this rule will lead to frequency collision. From Nortel Experience, 6 frequencies lead to a good protection against fading effects. Rule Air_MAIO_002 (O) The choice of MAIO depends on the frequency load In case of frequency load <= 50%, The frequencies must be non continuous per group. The MAIO are given by considering the cells by order, then the rule is step 2 within the cell, shift of 1 for cell2 and cell 3 same as cell1. Example: MAIO 0 1 2 3 4 5 Cell1 f1 F4 f7 F10 f13 f16 Cell2 f2 f5 f8 f11 f14 f17 Cell3 f3 f6 f9 f12 f15 f18 cell1_TRX1: MAIO 0 cell1_TRX2: MAIO 2 cell1_TRX3: MAIO 4 cell2_TRX1: MAIO 1 cell2_TRX2: MAIO 3 cell2_TRX3: MAIO 5 cell3_TRX1: MAIO 0 cell3_TRX2: MAIO 2 cell3_TRX3: MAIO 4 If case of frequency load > 50% The MAIO are given according to step 2 rule as long as the frequency is available (i.e. it has not already been chosen for a previous MAIO). Then, Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 14/84 Radio Interface Engineering Rules the MAIO are given according to step 1 rule. Step 1 MAIO are given to the cells which have the smallest overlap with neighbors or which have the smallest traffic. Justification: This recommendation comes from Nortel’s experience. 3.1.2.4 1:1 VERSUS 1:3 Rule Air_FH_005 (O) If available hopping frequencies < 27 and the grid of cells is irregular then the 1:1 fractional reuse pattern is more suitable. If available hopping frequencies >= 27, then 1:1 and 1:3 fractional reuse pattern are quite similar. But if the grid of cells is regular and the height of antenna are regular then the 1:3 fractional reuse pattern is more suitable. Justification: This recommendation comes from Nortel’s experience (radio simulation and field experience). 1:1 fractional reuse pattern will be better if available hopping frequencies <= 27 and irregular grid as it will be better in terms of FER and SQI. If the grid is regular, the 1:3 frequency reuse pattern will not bring about co channel interference as the 1:1 will do in same situation. 3.1.2.5 AD HOC SOLUTION 1:1 and 1:3 are not the only solution for frequency planning Due to the network densification, the frequency planning becomes more and more difficult. 1:1 or 1:3 solutions need too many frequencies to be efficient in high dense urban environments like New York In that case the best way is to use an Automatic Frequency Planning tool (AFP). This tool find the best frequency plan based on the simulated interference matrix, and some QOS and C/I targets values. With a great propagation model used in the RF planning tool the AFP can provides a very precise frequency plan. The principle is to calculate a cost based on the input targets, and the solution proposes at the end of the computation is the one with the lower cost Below an example of AFP settings Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 15/84 Radio Interface Engineering Rules The user of this kind of tool should have an experiment on frequency planning in order to provide the best inputs. AFP is best way to obtain a good frequency plan; many operators use it in order to be more efficient. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 16/84 Radio Interface Engineering Rules 3.1.3 FREQUENCY LOAD The fractional reuse pattern that can be used in a network depends on the frequency load. Rule Air_FL_001 (O) For choosing the fractional re-use pattern, Nortel recommends to follow this table. Fractional re-use pattern FrequencyLoad max FreqLoadSite max 1:1 20% 20% 1:3 50% 16% Figure 3-2 Frequency load versus fractional reuse pattern Justification: The results in this table come from Nortel simulation and field experience. 3.1.4 RADIO LINK BUDGET For the radio link budget, there are no specific Nortel Networks engineering rules. Please refer to the BPUG to get the value of parameters necessary for the radio link budget. 3.1.5 INTERFERENCE MEASUREMENT Please refer to BPUG document. 3.1.6 ANTENNA CONSTRAINTS 3.1.6.1 900/1800 COSITING For conservative electromagnetic compatibility purpose, some separations between antennas are required for isolation. Rule Air_Ant_001 (O) In case of a co-siting situation, Nortel experimentation is: In terms of electromagnetic compatibility, a separation of 50 cm (horizontal and vertical) is necessary between one GSM 900 (emission or reception) antenna and any GSM 1800 antenna. Justification: Any separation of 50 cm (vertical or horizontal) between the GSM 900 and the GSM 1800 antenna makes sure that no interference between GSM 900 system and GSM 1800 system will be encountered. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 17/84 Radio Interface Engineering Rules Rule Air_Ant_002 (O) In terms of space diversity, an horizontal separation by 1.6 to 3.2 m (3.20 to 6.4m) is required between the two GSM 1800 (900) antennas for efficient diversity gain. 2.4 to 4 m vertical separation can also be used. 3.1.6.2 BLOCKING Rule Air_Ant_003 (HC) For Nortel BTS products, Nortel guarantees: In case of GSM 1800 transceiver in a GSM 900 receiver, the level of blocking is -13 dBm for out band and in band emission In case of GSM 900 transceiver in a GSM 1800 receiver, the level of blocking is -35 dBm for out band and in band emission. 3.2. RADIO INTERFACE DIMENSIONING 3.2.1 TRAFFIC MODELS Nortel has some traffic models to help operators to dimension their networks. Please refer to the BSC overview document. 3.2.2 DIMENSIONING PRINCIPLES 3.2.2.1 TCH DIMENSIONING Rule Air_TCH_001 (O) For TCH blocking rate, Nortel recommends TCH blocking rate = 2% in normal loaded network to 5% in very loaded network for traffic and data on the radio interface, depending of the traffic load of the network. TCH number is computed with the Erlang Law. Hence we can deduce the number of DRXs. For Nortel BTS, one cabinet can have at most 12 DRX for S12000, 8 DRX for S8000, 4 DRX for S4000 BTS, and S2000 H/L and e-cell BTS have 2 DRX per cabinet. The maximum configurations are DLU dependant. All the configurations allowed are given in the BTS Engineering Rules documents. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 18/84 Radio Interface Engineering Rules 3.2.2.2 SDCCH DIMENSIONING Rule Air_SDCCH_001 (O) For SDCCH blocking rate, Nortel recommends SDCCH Blocking rate Middle LAC LAC border Normal load 0.1% 0.2% Very loaded 0.1% 0.2% Figure 3-3 SDCCH blocking rate Justification: This recommendation comes from Nortel field experience - Rules related to DCU2 Rule Air_SDCCH_002 (M) In case of BTS with DCU2: The TRX cannot support more than 1 SDCCH TS (SDCCH/8, SDCCH/8 + CBCH or BCCH combined) on the same TDMA. Justification: For reasons of DCU2 and MNU load. Rule Air_SDCCH_003 (M) In case of BTS with DCU2 On the TDMA carrying the BCCH (combined or not), TS7 = TCH Both consecutive logical channels (TS1, TS2) or (TS3, TS4) or (TS5,TS6) or (TS7, TS0) must respect the following constraints: • (TSa,TSb) = SDCCH/8 and TCH with (a,b)=(1,2) or (2,1) or (3,4) • (TSa,TSb) = SDCCH/8 + CBCH and TCH with (a,b)=(0,7) or (1,2) or (2,1) or (3,4) • (TSa,TSb) = 2 TCH for all (a, b) Justification: In order to balance the load, 1 BCCH and 1 SDCCH are not managed by the same processor so are not on two consecutive TS. Rules related to all types of TRX (DCU4, DRX, eDRX). The OMC-R check is introduced in V14 release. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 19/84 Radio Interface Engineering Rules Rule Air_SDCCH_004 (HC) A SDCCH TS is either a BCCH COMBINED (4 sub channels), a SDCCH/8 (8 sub channels), a CBCH SDCCH_4 (3 subchannels + 1 CBCH) or a CBCH SDCCH_8 (7 subchannels + 1 CBCH) For a regular cell: The TRX cannot support more than 2 SDCCH TS on the same TDMA. The 2 SDCCH TS must be carried respectively by an odd and an even TS. For an extended cell: The TRX cannot support more than 1 SDCCH TS per TDMA (This unique SDCCH TS must be on an even TS.) Justification: For load reason no more than 2 SDCCH TS are supported more on one TDMA. In order to balance the load on the different processors, the 2 SDCCH are carried on an even and an odd TS so there are not supported by the same processor. - Rules related to DCU2 combined with DCU4 Rule Air_SDCCH_005 (M) In the case of DCU2/DCU4 cell configuration and for defense reasons, the TRX must not support more than one SDCCH/8 on the same TDMA. - General rules Rule Air_SDCCH_006 (HC) The SDCCH/8 + CBCH position must be <= TS3 Justification: REC 05.02 Multiplexing and multiple accesses on the radio path Rule Air_SDCCH_007 (O) SDCCH channels are preferably placed on the hopping TDMA. Justification: It allows To better appreciate in advance the power the MS should use once on TCH To free space on the BCCH frame for cell tiering or packet data communications Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 20/84 Radio Interface Engineering Rules Rule Air_SDCCH_008 (O) It is recommended to spread evenly the SDCCH onto the various TDMA available in a cell (in particular avoid mixing SDCCH and BCCH on the same TDMA if enough TRX are available in the cell) Justification: This recommendation allows splitting the load on the processors. In case of cell tiering or GPRS, this rule allows to reserve channels for traffic. Rule Air_SDCCH_009 (O) SDCCH channels are preferably placed on the same TS numbers to allow gathering of Immediate Assignment messages 3.2.2.3 BCCH DIMENSIONING Rule Air_BCCH_001 (M) One BCCH is required per cell. The BCCH is supported by the TS0 of the beacon frequency. In the case of one TRX per cell, BCCH can be combined with one SDCCH/4 (4 SDCCH channels) in order to have 7 TCH instead of 6. This configuration can be applied under certain conditions such as the LA size. Actually, if the size of the LA is too large, a great amount of paging will be generated and the PCH (which is limited in this configuration) will not be able to flow all the paging messages. Note: A lot of parameters can be configured to optimize the BCCH dimensioning, among them numberOfslotsSpreadTrans, numberOfBlockforAccessGrant..., please refer to BPUG for more information. Rule Air_BCCH_002 (O) It is mandatory that the radio TDMA carrying BCCH channel should have highest priority. It is strongly recommended that the BCCH priority is set to "0". This TDMA must be the one and only one TDMA with the highest priority. Refer to the table “TDMA Priorities". Note: This new rule is introduced to secure the BCCH. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 21/84 Radio Interface Engineering Rules 3.2.2.4 CCCH DIMENSIONING The number of CCCH radio time slots is 1 per cell In some cases (microcell, dualband) we may need more than 1 CCCH. From V12, thanks to extended CCCH, it is possible to support multiple CCCH per TDMA supporting the BCCH in order to support a higher signaling traffic. Hence up to 4 TS (TS0, TS2, TS4, TS6) per TDMA on BCCH frequency may be use to support a high cell traffic. The addition of CCCH TS will depend on the traffic model, the LAC repartition and the environment. Rule Agprs_CCCH_001 (O) 2 CCCH TS may be necessary in a single layer cell if the number of TRX per cell is > 6 and the offered traffic per LAC is > 1200 Erl with 1 CCCH TS. 2 CCCH TS may be necessary in a multi layer cell if the number of TRX per cell is > 5 with 1 CCCH TS. 3.2.2.5 ADAPTATION OF CCCH DIMENSIONING TO THE PAGING FLOW The paging capacity of the Air interface is linked to different radio parameters. Paging messages are sent on Air interface in CCCH blocks on the PCH radio channel (in Nortel implementation cs-paging and ps-paging are both sent on these radio channels). The capacity of the CCCH blocks depends on the type of pagings (IMSI or TMSI): one CCCH blocks supports either: • paging on IMSI • 1 paging on IMSI and 2 paging on TMSI • 4 paging on TMSI. In a 51 TS multi-frame, the number of CCCH blocks depends on the BCCH TDMA configuration: TDMA BCCH type Number of CCCH blocks in a multi-frame = nCCCH BCCH combined 3 BCCCH not combined 9 Extended CCCH 18 Extended CCCH 2 27 Extended CCCH 3 36 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 22/84 Radio Interface Engineering Rules In a given multi-frame some CCCH blocks are used for paging and some are used for immediate assignment messages. The parameter noOfBlocksForAccessGrant allows to reserve some blocks for immediate assignment messages. In that case the number of CCCH blocks for paging in a multiframe is nCCCH - noOfBlocksForAccessGrant. (Note that when the value noOfBlocksForAccessGrant is set to 0, blocks are preempted when needed for the immediate assignment messages). A given mobile will listen to CCCH blocks corresponding to its "paging group".The paging group of a mobile is determined by its IMSI number and by the "number of multi frames between paging". This number is given by the parameter noOfMultiframeBetweenPaging. It corresponds to the number of multi-frames used to recover the same paging group for a given mobile. The number of different paging groups possible is the number of possible CCCH blocks available for paging during these noOfMultiframeBetweenPaging multi-frames. The paging are distributed on these possible paging groups according to the IMSI number of the called mobile (with the use of a "modulo" on the IMSI value). The different pagings are thus distributed in: (nCCCH -noOfBlocksForAccessGrant)*noOfMultiframeBetweenPaging Thus when a paging arrives for a mobile, this one will be sent in the appropriate paging group: the BTS will have to wait for this paging group before sending this paging. Moreover if the BTS has already some paging to send in this paging group; it will wait for a new corresponding paging group before sending this paging message. With Nortel BSS, the BTS is able to perform by itself some paging repetitions. Thus in the different paging groups, the BTS will sent either new paging or some paging repetitions. Nevertheless, new paging messages are always handled in priority compare to paging repetitions. (By this way, if a number of n repetitions is wanted at BH, it is possible to specify a value m > n for this number of requested BTS repetitions: this will allow during non busy hours to try to perform the m retries while may be at BH the dimensioning is only able to support n repetitions). In summary, the Air interface paging capacity will depend on: - The ratio of IMSI paging vs TMSI paging - The TDMA BCCH type - The value of the parameter noOfBlocksForAccessGrant - The latency before being able to send a given paging in the appropriate paging group - The number of wanted repetitions at BTS. The following tables provide some indications of the Air interface capacity for paging depending on some typical parameters (paging are supposed to arrive here uniformly shared between the different paging groups and the latency is modeled through a Pollaczek-Khintchine formula in order to limit the delay to a maximum corresponding to one noOfMultiframeBetweenPaging duration). Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 23/84 Radio Interface Engineering Rules A ratio of 10% vs 90% between IMSI paging and TMSI paging is considered here. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 24/84 Radio Interface Engineering Rules Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 25/84 Radio Interface Engineering Rules 3.3. TDMA CONFIGURATIONS AND PRIORITIES 3.3.1 STANDARD CELL Recommended TDMA are of different types. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 26/84 Radio Interface Engineering Rules 3.3.2 EXTENDED CCCH 3.3.3 EXTENDED CELL Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 27/84 Radio Interface Engineering Rules 3.3.4 TDMA PRIORITIES 3.3.4.1 STANDARD CELL The following values are adapted to Nortel standard call profile with about 50% AMR HR penetration. As it corresponds to an average call profile over the network, these values could be adapted regarding the specificities of particular parts of the network. Moreover for more aggressive call profile the number of SDCCH resources need to be adapted Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 28/84 Radio Interface Engineering Rules 3.3.4.2 EXTENDED CELL Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 29/84 Radio Interface Engineering Rules 3.4. MONOZONE AND CONCENTRIC CONFIGURATION In different cases, some transceivers remain disabled/dependency due to a bad configuration at the OMC-R. To avoid such situations the following rules must be followed: Rule Air_zone_001 (HC) If the concentric cell attribute is set to "monozone", the OMC-R has to check that there is no transceiver associated to a transceiverZone, before setting the BTS unlocked. Rule Air_zone_002 (HC) If the concentric cell attribute is set to "monozone", the OMC-R has to check that the transceiver equipment class of all transceiverEquipments is set to 1, before setting the BTS unlocked. Rule Air_zone_003 (HC) If the concentric cell attribute is set to "concentric" and transceiver equipment class is the same for all transceiverEquipments, the OMC-R has to check that the number of transceiverEquipments of both zones is equal or greater than the number of transceivers of the both zones, before setting the BTS unlocked. Rule Air_zone_004 (HC) If the concentric cell attribute is not set to "monozone" and there are two classes of transceiverEquipments, for each transceiverZone, the OMC-R has to check that the number of transceiverEquipments is equal or greater than the number of transceivers, before setting the BTS unlocked. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 30/84 Radio Interface Engineering Rules 4. GPRS RADIO INTERFACE ENGINEERING RULES 4.1. RADIO RESOURCES A cell supporting GPRS uses physical channels to handle GPRS traffic. Those channels (PDCH) shared by the GPRS subscribers can be taken from the available resources in the GSM cell (see 12.3.2). The GPRS timeslot can be configured over any TDMA of the cell, it can be a new TDMA or an already existing TDMA used by GSM. High data rates cannot be reached when no new TRXs are added. The BSC dynamically allocates the resources between GSM and GPRS. The GPRS multi-frame structure is different from GSM. It consists of 52 TDMA frames divided into 12 blocks (each of 4 frames), with 2 idle frames and 2 frames used for timing advance: Figure 4-1 GPRS MULTI-FRAME STRUCTURE 4.1.1 TDMA RULES Rule Air_TDMA_001 (HC) Several GPRS TDMA per cell can be supported starting from V14. GPRS service can be supported on any TDMA on the cell. This TDMA can be in the Outer Zone and the Inner Zone of a Dual Band Cell. We can have a TDMA in the Inner and second one in the Outer Zone. If the inner zone coverage is lower to the outer zone coverage, we recommend configuring PDTCH only in the outer zone. Note: This feature is very helpful in case of a high load on a particular TDMA and a lack of GPRS resources. The operator can use additional TDMAs in order to decrease the load on existing GPRS resources and increase data bandwidth. Nortel simulations have shown that for a cell of 8 GPRS dedicated PDTCHs, with 10 GPRS subscribers (4+1 Multislot class) the mean throughput will be 38Kbps/user for FTP and 23Kbps/user for HTML and email applications. To keep this quality of service with an increasing number of GPRS users, we must add more PDTCHs on a new TDMA. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 31/84 Radio Interface Engineering Rules Rule Air_TDMA_003 (HC) The Frequency Hopping for the PDTCH channels can be applied only in Outer Zone for a Dual Zone Cell. No frequency hopping on a PDTCH channels is possible in the Inner zone. In outer zone, only one hopping frequency law can be declared for the TDMA with PDTCH configured. Rule Air_TDMA_004 (HC) Up to 32 GPRS subs MS can be supported by TDMA frame with a maximum of 32 TFI values. Justification: This limitation is a consequence of the Temporary Block Flow Identifier (TFI) coding. One TFI differentiates 2 Temporary Block Flows (TBFs) which have a common PDCH allocated, one in UL and one in DL. Since it is coded on 5 bits, there are only 32 different possible TFI values, which will allow (1 UL + 1 DL) × 32 TBFs. We have 32 TFI and we always allocate one in UL and one in DL for one MS (due to TBF Keep Alive). 4.1.2 PDCH RULES Rule Air_PDCH_001 (HC) Up to 8 PDCH can be allocated to one user with the “TS aggregation” according to its multi-slot capability. Thus it has to be taken into account for radio resource dimensioning when several users are in the same cell. A percentage of MS activity has to be used (in call profile). Note: The MS multi-slot capability provides a more efficient use of the air interface resources. The TS aggregation can be done dynamically. A higher bandwidth can be offered to the end user according to its multi-slot capability (all multi-slot classes 1 to 29 are supported). This can be useful for the operator or the end user since it can provide lower costs with higher bit rates. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 32/84 Radio Interface Engineering Rules Rule Air_PDCH_002 (HC) Each radio PDTCH can allocated to up to 7 subscribers (not 8, due to USF utilization). One PDTCH can handle up to 16 MS at most. Due to the system limitation (USF limitation in UL) a maximum of 7 simultaneous users are allowed per PDCH on the UL. Justification: The Timing Advance Index is coded on 4 bits, which gives a total of 16 different TAI per PDTCH. On the other hand, the Uplink Status Flag is coded on 3 bits (system limitation), which allows a maximum of 8 values in the UL. As USF=0 is reserved; only a maximum of 7 subscribers can share one TS on the UL. When USF=0 is used, none of the MSs is allowed to transmit in the following UL Radio Block, so one of the MSs that only has a DL TBF can acknowledge the DL data coming from the PCUSN when the PCUSN polls it. This rule is related to rule Air_TDMA_004 (HC): If all the MSs on TS1 takes this TS1 as control TS, we can have only 16 MS on that TS1. If a 2+1 MS takes its control information (TAI) on TS2 and is sharing TS1 and TS2 with other MSs, it won't be counted among the maximum of 16 MSs allowed on TS1 since a MS takes only 1 TAI on one of the TSs it is using. So the number of MSs allowed per TDMA varies between 16 and 32 (depending on the number of PDTCH per TDMA). Taking into account the USF (per PDTCH), only 7 MSs can have UL+DL transfers and 32-7=25 MSs can have only DL transfers. Rule Air_PDCH_003 (HC) With TS Partitioning, operators will be able to handle more users per PDCH, which leads to more users per cell at more variable data rates. This feature is therefore best suited when many users share more than one PDCH in one cell to avoid very low throughputs. Justification: TS partitioning feature allows finer granularity which gives more efficiency by multiplexing one PDCH between several users (up to 7) sharing the same TS. When several users share oneTS, data throughput available per user diminishes. A new MS can access to the resource as soon as it receives the UL assignment from the PCUSN (see figure 12.2). Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 33/84 Radio Interface Engineering Rules Figure 4-2 TS PARTITIONING BENEFITS Rule Air_PDCH_004 (HC) If NMO1 feature is activated, it’s recommended to set the parameter minNbrGprsTs different from 0. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 34/84 Radio Interface Engineering Rules 4.2. CODING SCHEMES CS-1 AND CS-2 4.2.1 CODING SCHEMES DESCRIPTION: In good radio conditions CS-2 provides a higher throughput than CS-1. CS-1 is used to transfer the signaling while CS-2 allows a more efficient air-interface throughput for data transfer. Figure 4-3 SIMULATION RESULTS GIVING THE RLC/MAC THROUGHPUT (PER CS) VS BLER CS-1 has the same coding as SDCCH, giving a payload rate of 9.05 Kbps and a maximum data throughput of 8Kbps at the RLC/MAC layer. CS-2 is a punctured version of CS-1; the payload rate is 13.4Kbps and the maximum data throughput (at the RLC/MAC layer) allowed by this coding scheme is 12 Kbps per TS. (For more details about coding scheme see "Access Network Parameters’ User Guide"). Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 35/84 Radio Interface Engineering Rules 4.2.2 RULES Rule Air_CS_001 (O) CS-1 and CS-2 are selected on a GPRS TDMA basis depending on the operator’s radio engineering. It is recommended to use CS-2 rather than CS-1 in order to offer better results for data transfer. Signaling must be conveyed using CS-1. Justification: Simulation results achieved by Nortel show that it would be better to configure CS-2 in most cases. Measurement results confirm that in a typical GSM cell, with a BLER below 10%, CS-2 has often better throughput compared to CS-1 (see 12.2.1). Rule Air_CS_002 (O) No impact will occur on the Abis interface with the first GPRS introduction. The same mapping for GPRS is adopted as for GSM TSs on the Abis interface. Thus no changes will occur regarding the Abis dimensioning and architecture. For more details on the Abis mapping, see the document “Abis Interface Engineering Rules”. Justification: The first release of GPRS will only use CS-1 and CS-2, GPRS radio resources will then use 16K TS like GSM TSs since CS-1 bit rate is 9.05 Kbps and CS-2 bit rate is 13.4Kbps. The separation between GSM and GPRS TSs will be done at the BSC level. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 36/84 Radio Interface Engineering Rules 4.3. GPRS CHANNELS’ DIMENSIONING 4.3.1 GPRS IMPACT IN THE CCCH LOAD Rule Air_DimCCCH_001 (O) GPRS introduction in a GSM network has a low impact on CCCH. GSM traffic has the most important impact on CCCH capacity. Existing CCCH can cope with a traffic level compatible with what is foreseen at the starting of a GPRS network. Therefore it doesn’t generate changes in CCCH (RACH, AGCH, and PCH) dimensioning. The introduction of GPRS in the first steps (low penetration rate) doesn’t show major impact on the CCCH capacity. Justification: Considering the following hypothesis: -One Routing Area = one Location Area (all PS and CS paging messages sent to the same cells of the LAC), -Using the Nortel standard GPRS call profile, -Location Area of 80000 users and 11 paging message/cell/s, -With 5%, 10 and 25% of GPRS penetration rate, -In order to offer a minimum GPRS throughput (for 25% Class B GPRS MS penetration), only fixed PDCH are considered. If the GPRS class B MS penetration rate is less than 25% the data throughput will be higher. With a total UL_TBF_CCCH = 16.10 /subscriber/BH (16.1 channel request and 16.1 Immediate_Assignment) And a total DL_TBF_CCCH = 13.15 /subscriber/BH (DL immediate assignment messages) NB: The results focus only on the GPRS impact over CCCH capacity. With these hypotheses, 1 CCCH per cell is enough with a standard GSM call profile. For the same conditions, with a heavier GSM traffic (short call GSM call profile) and 21.3 paging message/cell/s, the signaling traffic can exceed the CCCH capacity for 15 and 16 TRX configurations. This means that a future traffic increase will need more CCCH capacity. So it would be safer to add a new CCCH channel in this case. This shows the direct impact of GSM traffic on the signaling traffic. The number of GPRS paging messages is negligible comparing to GSM. Additionally, “TBF pre-establishment” and “TBF keep alive” decrease the number of TBF establishments, thus their activation will induce a significant decrease in CCCH traffic load for GPRS. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 37/84 Radio Interface Engineering Rules The following table shows a comparison between standard GSM call profile and Short call GSM call profile with GPRS class B penetration rate from 5 to 25% with respect to the CCCH capacity. These recommendations are drawn with Nortel's GPRS call profile. Since the results can vary a lot with the call profile and network inputs (LAC size, Number of users per LAC…), specific studies must be done to give conclusions for operators. Additionally, the number of users per LAC has an impact on the CCCH capacity. If the number of users increases in the LAC, we will reach faster the CCCH capacity limit. Rule Air_DimCCCH_002 (O) For the evolution of the GPRS services, an increase will be experienced in the data traffic (call profile change). This will induce more signaling. In these cases the use of extended CCCH is recommended to cope with the additional signaling load. The CCCH load depends on the following parameters: • GPRS call profile • GSM call profile Extended CCCH is available starting from V12.4c. The CCCH can be spread over up to four time slots per BTS (TS0, 2, 4 and 6) instead of using only one (TS0). Extended CCCH will allow the BTS to handle larger volumes of traffic and thus to improve the call set-up time. Drawbacks The use of new CCCH TS when needed (extended CCCH) has to be done carefully regarding the overall dimensioning process since adding new CCCH TS decreases the number of available traffic TCHs in the cell. Thus more resources need to be added and this will impact the cell configuration. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 38/84 Radio Interface Engineering Rules 4.3.2 GPRS CHANNELS’ RULES The traditional Erlang-B law is commonly used for circuit switched networks for radio resource dimensioning. This law is no longer valid for GPRS since it’s a packet queuing system with resource sharing. The major challenge is to find a traffic model representing a subscriber and its behavior. Both voice traffic and data traffic have to be considered to find the necessary resources for GPRS service. 4.3.2.1 PDCH DIMENSIONING RULES Rule Air_DimPDCH_001 (O) Dimensioning has to use throughput at the application layer. The throughput is a vital part of the service offered to the end user. Dimensioning the air interface is an underlying issue since the radio interface is the weakest link of the chain. The effective parameters: • throughput of a subscriber depends on several Radio conditions (BLER) • Coding scheme • MS Multi-slot capability • Number of GPRS subscribers per cell • Number of active subscribers • GSM/GPRS TS dynamic sharing • GPRS busy hour • GSM busy hour • Call profile We look forward to finding the number of PDCHs needed per cell to offer the requested service. This result must be in keeping with the rules listed above (see 12..1 “Radio Resources”). Dimensioning methodology First we derive the bit rate per PDCH using BLER and CS-2 as inputs for the simulator: Then we derive the needed bit rate service to offer on the Busy Hour (BH). We use the data volume generated (using the call profile) by the active users on the BH: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 39/84 Radio Interface Engineering Rules The mean bit rate to offer can be fixed by the operator, in this case there is no need to compute it. • At this stage we compute the number of needed PDCHs: Number of needed PDCH= [(Number of active users)*(mean bit rate at BH)]/ (%TS_Oc × bit rate/PDCH) Where %TS_Oc is the TS occupancy rate. The features’ improvements (TBF establishment improvements, TS partitioning, USF) help increasing this coefficient towards higher efficiency of the radio resources. This will diminish the needed number of PDCHs. Although the radio dimensioning is a very important step in the GPRS deployment, the operator might dimension its network with a 2 (or 3) year traffic forecast. This will diminish the “re-dimensioning” effects since the first GPRS introduction period is only an intermediary step towards a more stable network. 4.3.2.2 DYNAMIC SHARING RULES Rules Rule Air_DynSh_001 (O) GPRS service capacity is directly related to the number of radio resources (PDTCH) that GPRS is allowed to use. The “TS dynamic sharing” feature is necessary to lower the GPRS impact on GSM, particularly when the operator chooses to use shared TS. The operator has the choice in a pool of 3 types of resources: (GSM dedicated TCHs; Shared TCH-PDTCHs; dedicated PDTCHs). Dynamic sharing between GSM and GPRS increases the radio interface efficiency by sharing radio resources between circuit and packet services. This efficiency has to be taken into account for GPRS radio resource dimensioning to find the needed number of PDTCH to handle the packet data traffic. Justification: The air interface efficiency in GSM goes up to 70% with a 2% blocking rate. Nortel’s simulations of this feature show an efficiency increase up to 20% due to the introduction of this feature. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 40/84 Radio Interface Engineering Rules As a consequence of this rule, we can state that this feature allows the GPRS introduction with a low impact on the cell configuration and network architecture. If the operator doesn’t configure any fixed GPRS radio TS, no blocking rate increase is experienced due to this feature. Drawbacks Fixed GPRS PDTCHs will decrease the cell capacity unless additional TDMA is configured, which may require new network planning and configuration during GPRS deployment. Additionally, TS dynamic sharing will have a direct impact on the BSC’s CPU load. At each allocation or release of a GPRS radio TS; the BSC has to perform 2 switching connections, which will increase the load. Fixed and Shared TS separation The number of fixed and shared GPRS TS in a GSM network has to be specified per cell using: • GSM blocking rate allowed (Degree of GSM priority over GPRS), • Number of existing TRX, • Measured cell load (Erlang, circuit GSM) • GPRS traffic estimation, • MS and network capabilities. Once the number of needed PDTCHs is derived (see Rule Air_DimPDCH_001 (O)), we must find an optimised set of TSs (GSM dedicated TCHs; Shared TCHPDCHs; dedicated PDCHs) on the TDMA supporting GPRS. Shared TCH-PDCHs number has to be chosen regarding the GSM TCHs blocking rate at the voice BH. The measured cell load will give us the usage efficiency rate of the radio resources. Thus blocking and efficiency rates have to be constantly monitored. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 41/84 Radio Interface Engineering Rules 5. EDGE RADIO INTERFACE ENGINEERING RULES 5.1. OBJECTIF The aim of this part is to detail Nortel Networks’ engineering methodology to estimate EDGE performances if EDGE would be deployed on existing GSM network. Incremental redundancy impacts on capacity estimations as well as frequency reuse are also analyzed. Advantages of new powerful PA impact on DL performances are explained. Even if DL performances are mainly studied, different type of UL mobiles available on the market are also commented. Test carried on platform and live network experience feedbacks from Engineering Test Plan are also detailed. Finally, comparison between GPRS and EDGE in term of user capacity is studied for an hypothetic FTP connection and specific allocator. There is an evident inaccuracy in the estimation capacity performances but the tendencies obtained give a good overview of expected performances. 5.2. EDGE, A FORWARD STEP GSM is a constantly evolving technology, and the next major milestone is Enhanced Data Rates for GSM Evolution (EDGE). Data application requirement implies the evolution of radio access techniques to satisfy user’s demand. 3G CDMA systems offers data application possibility (main difference in term of application between 2.5G and 3G is video conference capability) but the high cost to deploy them and early deployment makes us consider the possibility to upgrade existing infrastructure to adapt GSM networks and compete against new 3G generation system or complement its coverage. GPRS was the first step; EDGE the step forward. EDGE improves GPRS throughput performances by 3 times and allows dealing with 5 times more users. Engineering methodology has changed between voice and data. For speech, end user quality is equivalent to air interference quality. For data, excellent user quality is assumed but it is less related to air interference quality, which could be bad. Although air interface quality fixes throughput, one user can use several radio channels as time slots are aggregated, allowing achieving higher throughputs. Therefore, data throughput is the QoS parameter for data services since perfect reception is assumed. No engineering change between GPRS and EDGE is applied. Higher data throughput than GPRS is achieved thanks to new modulation usage, which allows reaching higher data rates. Whereas GSM and GPRS use a Gaussian Minimum Shift Keying (GMSK) modulation scheme, EDGE uses both GMSK and a modulation called Eight Phase Shift Keying (8-PSK). 8-PSK can communicate three bits of information in each radio symbol whereas GMSK can offer only one bit. The higher the MCS, the higher the user data rate (these data rates are at RLC/MAC level), since there are fewer bits of protection and less puncturing. It also means that the highest MCSs will only be used in good radio conditions whereas in bad radio conditions, the lowest MCSs will be used. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 42/84 Radio Interface Engineering Rules Complementary to radio modulation, EDGE takes benefit from other improvements such as: more performing link adaptation, incremental redundancy or advanced RLC/MAC protocol. By adapting the coding schemes to the radio channel conditions dynamically, it is possible to optimise communication performances and throughput. This is done by link adaptation: through radio measurements, the network chooses the best MCS and adapts it. Estimated best MCS is used in each position of the cell. All values presented in this document assumes ideal link adaptation algorithm. Moreover, another feature is only available for EDGE: this is incremental redundancy, IR, which is the possibility to retransmit a data block using a different puncturing method and also recombining with received erroneous packets. By this way, probability to receive a correct block is increased. The RLC/MAC layer has been significantly improved in EDGE development. For handsets supporting multiple TS, performance limitations in GPRS due to the limited size of the acknowledge window is not reproduced in EDGE, i.e. in GPRS RLC window size is 64, i.e. the transmitter cannot transmit block N+64 if block N has not been correctly acknowledged by the receiver. In EDGE, windows size has been extended to 1024 blocks, avoiding loss of incorrect blocks because of too bad radio conditions. Thus, thanks to all these improvements, EDGE offers better performances than GPRS. 5.3. EDGE DEPLOYMENT STRATEGY There are two methodologies to design a data network: • First one is based on the deployment on current / existing network characterized by a C/(I+N) distribution. User throughput will be distributed in the cell according to the existent network air-interface quality. • Second one consists in implementing a new network for data or changing existing cell planning, frequency planning or network densification, in order to guarantee a given throughput or a given QoS to subscribers. Several performance parameters are significant for an EDGE network design. Main three parameters are mentioned in this document: mean data throughput per TS overall cell, guaranteed data throughput, i.e. maximum data throughput at cell edge with cell edge reliability and MCS usage distribution. First two throughput metrics depend on the number of erroneous blocks that need to be retransmitted on air interface. BLER, i.e. Block Error Rate, distribution, which is different according to the MCS, depends on several parameters: BTS type (macro, micro cell); MS performances, frequency band; environment and mobiles’ speed (TU3, TU50…), features of signal processing (diversity, frequency hopping, incremental redundancy…) and also on the radio condition of the cell, which is related to cell planning and the frequency plan. MCS usage distribution determines which and where each MCS is used and it has a big impact on backhaul dimensioning. This document is focused on the performances of an EDGE networks deployed over an existing GSM network. GSM networks typical values of coverage and frequency plan are considered. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 43/84 Radio Interface Engineering Rules 5.4. MAINS ENGINEERING RULES 5.4.1 EDGE TDMA RULES Rule Air_TDMA_005 (O) Several EDGE TDMAs per cell can be supported starting from V15.1. GPRS/EDGE service can be supported on any TDMA of a given cell, with multiple GPRS/EDGE TDMAs in a same cell. Those TDMAs can be both in the Outer Zone and the Inner Zone of a Dual Band Cell. In case the inner zone coverage is smaller than that of the outer zone coverage, we recommend configuring PDTCH only in the outer zone. Besides, in case where the operator wants to support EDGE MS which do not support the frequency band of the Inner zone (ex: Roamers), then Nortel recommends not to configure any PDTCH in the Inner Zone. 5.5. SPECIFIC DATA FEATURES 5.5.1 NETWORK ASSISTED CELL CHANGE IMPACT ON GPRS/EDGE NETWORKS This feature allows for the rel'4 MS an assistance during the cell re-selection during the on-going TBF. The NACC feature is limited only to the intra-BSC cell re-selection for the moment. For MS without NACC, the MS abruptly aborts the on-going TBF without indicating it to the network. With MS supporting NACC, the MS dialogues with the PCU in order to keep the initial TBF established for a maximum of 960ms during the cell reselection. This avoids having a loss of service during re-selection by performing a smooth transition between the TBF. For the NACC usage it's necessary to respect the following rules: Rule Air_NACC_001 (M) In a given BSC area, the (ARFCN, BSIC) shall be unique Note: A notification is raised by PCU when this situation occurs in a cell Rule Air_NACC_002 (M) The neighbor cells of a BSC shall not have the same (ARFCN, BSIC) than any of the cell of that BSC Area. Note: This configuration shall be ensured by the customer and cannot be detected by the BSS (for example, with 2 BSCs configured on 2 different OMC-R, or in the boundary with another BSS vendor, or in boundary with another country). Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 44/84 Radio Interface Engineering Rules 5.5.2 PACKET FLOW CONTEXT IMPACT ON GPRS/EDGE NETWORKS This feature allows offering a support to Rel99 QoS management inside GPRS/EDGE network. The PFC feature helps in supporting applications requiring a big bandwidth with a good QoE to end users. In order to ensure a significant success of admission control for RealTime MS, BSS resources have to be allocated in coherence with the operator objectives. According to theses PFC targets, the following recommendations present the principle of resources dimensioning for the PDTCH and Jokers in the concerned cells for Air interface. The Agprs interface document present the dimensioning principles for dynamic Agprs resources and constraints. PDTCH and Jokers dimensioning in the cell for Abis and Air interfaces: N _ rtpfc _ pdtch ≤ N * GBR pfcTsGuaranteedBitRateDl * ( 1 − pfcMinNrtBandwidthPerTs ) Npdtch >= N_rtpfc_pdtch Njokers >= N_rtpfc_pdtch * pfcMinNumberOfJokersforTsGuaranteedBitRate NDS0jokers = Njokers/4 If a significant bandwidth for the non PFC mobiles is reserved by the operator, then the PDTCH and Jokers may of course be set to values higher than the one presented here. 5.6. DATA THROUGHPUT ESTIMATION Mean data throughput offered by EDGE on a given cell depends on BLER distribution on this cell. BLER distribution depends on radio conditions. Estimations between BLER and radio conditions have been obtained through software signal processing simulator at air-interference layer, where transmitter, receiver, interference and multipath channel are modelled [R1]. Following global methodology is implemented to calculate EDGE mean throughput offered per TS on a given scenario: • Implemented frequency plan on network determines C/I at cell edge. C/I distribution on cell is obtained applying typical propagation model. • Cell size, once the environment is characterized, determines Eb/No at cell edge. Eb/No distribution on cell is obtained applying typical propagation model. • Thermal noise and interference are independent but with combined effects, thus C/(N+I) is considered. BLER distributions on cell is deduced for each MCS, from BLER=f[C/(I+N)]. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 45/84 Radio Interface Engineering Rules • Throughput distribution on cell is deduced from these BLER distributions for each MCS. User throughput at RLC/MAC layer is roughly estimated by the following simplified formula: Effective Throughput=MaxThroughput_MCS*(1-BLER) (1) • Finally, throughput is estimated integrating data throughput over coverage area. Calculations at cell edge determine guaranteed data throughput with certain reliability since worst radio conditions are found. BLER distribution over cell defines MCS distribution usage if perfect link adaptation is considered, i.e. best data throughput is considered at each position over all MCS available. Figure 1 gives an overview of the methodology: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 46/84 Radio Interface Engineering Rules Voice cell range Frequency plan Eb/No distribution on the cell C/I distribution on the cell C/(N+I) distribution on the cell R&D simulations: BLER = f (C/(N+I)) BLER distributions on the cell for each MCS Throughput distributions on the cell for each MCS Mean throughput / TS on the cell Figure 5-1: Throughput calculation methodology scheme Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 47/84 Radio Interface Engineering Rules Note that this methodology is applicable to either DL or UL. Following steps are detailed in the next sections: - Calculations of C/I and Eb/No distributions on cell - Calculations of throughput distributions on cell - Calculations of mean throughput per TS on cell 5.6.1 RADIO CONDITIONS (C/I AND EB/NO) DISTRIBUTIONS ON THE CELL Let be a cell of range R. Cell is divided in concentric cells of radius ri, so that 0 ≤ ri ≤ R, in order to determine Eb/No and C/I distributions on this cell (figure 2). r6 r7 r8 r9 r10=R r5 r4 r3 r2 ro r1 R Figure 5-2: Cell partitioning (schematic representation) 5.6.1.1 C/N DISTRIBUTION The C/N distribution on the cell depends on voice pathloss. From this pathloss, C/N at cell edge can be deduced and then Eb/No distribution. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 48/84 Radio Interface Engineering Rules 5.6.1.2 C/N AT CELL EDGE The cell radius R corresponds to a given voice pathloss. MS and BS sensitivities are deduced from this pathloss using the following formulas: MS sensitivity calculation: PathLoss _Voice = BS _TxPower − DL _ Losses − MS _ Sensitivity (2) BS sensitivity calculation: PathLoss _Voice = MS _TxPower −UL _ Losses − BS _ Sensitivit y (3) Note: DL_Losses and UL_Losses correspond to all losses introduced in the transmission – reception chain, in particular cable losses, BS antenna gain, slant losses and penetration factor. Note that no body losses are included in these factors since for data, handset is assumed to be separated enough from the body. Then, MS and BS Eb/No at cell edge can be calculated using formula below: Where: - NF stands for noise factor. NF is 3 dB for BS and 8 dB for MS. Being: M = number of bits per symbol: M = 1 for GMSK modulated signal (MCS-1 to MCS-4) M = 3 for 8-PSK modulated signal (MCS-5 to MCS-9) As M is different according to the modulation, Eb/No at cell edge is also different. 3 bits per symbol are transmitted in 8-PSK while 1 bit per symbol is being transmitted in GMSK modulation. In other words, there are two Eb/No distributions: One for GMSK MCS and another for 8-PSK MCS. The difference between the two distributions is 10*log(3) factor. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 49/84 Radio Interface Engineering Rules 5.6.1.2.1 C/N DISTRIBUTION ESTIMATION A simplified propagation model is used to calculate C/N distribution over the cell depending on each ri, i.e. UE cell position: where: • 5.6.1.3 α is propagation coefficient. Following values are used: o 3.522 in dense urban, urban and suburban environments o 3.441 in rural environment • R is cell radius • r is expressed in km C/I DISTRIBUTION C/I distribution on the cell depends on implemented frequency plan and radio conditions. From this frequency plan, C/I at cell edge can be deduced and then C/I distribution. It is very difficult to simulate C/I distribution as radio conditions vary quickly. 5.6.1.3.1 C/I AT CELL EDGE C/I at cell edge depend on frequency plan implemented on the network. Nortel Networks experience and simulations have shown that 4*12 for non-hopping frequency plan allow ensuring a minimum C/I of 12dB with 95% of reliability. So, the C/I at cell edge is assumed equal to 12 dB. Note: of course, if the frequency reuse pattern is higher, the C/I at cell edge is greater. Following figure shows the C/I cdf depending on frequency plan. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 50/84 Radio Interface Engineering Rules Comparatif entre les différent motifs 100 90 80 60 50 40 % P(C/I >= X) 70 30 20 10 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 X : Seuil de C/I (dB) Figure 5-3 C/I Cdf depending on frequency reuse pattern 5.6.1.3.2 C/I DISTRIBUTION Following formula is used to calculate C/I distribution over the cell depending on UE position, i.e. where: - α is the propagation coefficient. Following values are used: o 3.522 in dense urban, urban and suburban environments o 3.441 in rural environment - R is cell radius - r is expressed in km Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 51/84 Plan opérationnel Motif 1-1 Motif 1-3 Motif 2-6 Motif 3-9 Motif 4-12 Motif 7-21 Radio Interface Engineering Rules 5.6.2 MEAN THROUGHPUT PER TS CALCULATION 5.6.2.1 BLER DISTRIBUTIONS ON THE CELL BLER distributions are deduced using the R&D curves [R5]. These curves are actually the results of simulations, giving performances for each MCS in term of BLER uplink versus C/I or Eb/No according to: o Environment o Frequency o Activated features, redundancy… e.g. diversity, frequency reuse, incremental For instance, hereafter are two simulations of BLER versus Eb/No (figure 4) and C/I (figure 5) in following scenarios: TU50 propagation profile; 1800 MHz; no FH, no diversity and no incremental redundancy: Figure 5-4: BLER = f (Eb/No) in TU50, no FH, no IR, no Diversity and at 1800 MHz Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 52/84 Radio Interface Engineering Rules Figure 5-5: BLER = f (C/I) in TU50, no FH, no IR, no Diversity and at 1800 MHz These curves show clearly that BLER performances for a given radio condition (C/I or C/N) are not the same according to MCS considered. It means that BLER distribution needs to be calculated for each MCS. Thermal noise and interference noise effects are combined, thus, both are added to calculate C/(N+I). R&D simulations are done for C/I, Eb/No and/or C/N. If both curves are superposed for the same environment, C/(I+N) is expected to be found between them. Thus for BLER distribution, mean values between BLER found with Eb/No curves and BLER found with C/I curves is considered. Methodology to calculate BLER distribution on the cell for one MCS is the following: o Consider each couple {C/I; C/N}i calculated previously for each ri o C/(I+N) is estimated following expression: o Deduce from the R&D simulations (BLER versus C/(I+N) in C/I curves and BLER versus C/(I+N) in Eb/No curves for each considered MCS) TWO BLER values {BLERC/I; BLEREb/No}i associated to C/(N+I) o In order to estimate BLER depending on C(I+N) and to take into account noise characterization (either coverage or interference limited environment) {BLERC/I; BLEREb/No}I are weighted following a baricenter weighting Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 53/84 Radio Interface Engineering Rules Note: Nortel Networks simulations are based on eDRX Nortel Networks performances. Standard GSM 45.005 specifies MS performances in term of sensitivity and C/I only at one BLER value (10% or 30% according to the MCS), if these figures are compared with R&D Nortel Networks simulations, 4 dB degradation are found, either for sensitivity or interferences figures. For throughput calculation in DL, Nortel Networks prefers to use its own eDRX performance figures without diversity; but considering a margin so that the performances given by simulations are coherent with those given by the standard at the BLER of standards (i.e. 10% or 30%). UE signal processing algorithms implemented in the UE equipment should not be less performing in the UE than in simulations considered. Thus, 0 dB C/I margins are taken in the calculations in downlink. Sensibility performances are different between UE and BS because of hardware equipment performances. As it has been said, 4 dB degradation is found between standard 05.05 and R&D signal processing simulations. Thus, 4 dB is the worst conditions. On the other hand, it can be expected, UE performances will be better than standard, UE suppliers will leave a margins to guarantee the standard. Therefore, 3 dB margin is proposed in C/N simulations in downlink. Implementation margins could be also considered independently from those margins to adapt eDRX performances to UE. In Uplink, 1 dB margin is proposed by default in uplink. 5.6.2.2 THROUGHPUT DISTRIBUTIONS ON THE CELL Having BLER distributions on the cell for each MCS, i.e. BLER value for each ri between 0 and R, throughput distributions can be easily determined. Following formula is used to calculate effective throughput per TS according to the BLER. where MaxThroughput are according to MCS: h e r Table1: Maximum Throughput per TS according to the MCS Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 54/84 Radio Interface Engineering Rules Note: these throughput values are at RLC/MAC level. 5.6.2.3 LINK ADAPTATION Link Adaptation (LA) allows changing MCS used in the transmission of RLC data blocks according to radio conditions. To maximize data throughput, EDGE network should choose in each point of the cell MCS, which maximizes throughput: this algorithm is called Link Adaptation. LA algorithm inside PCU selects the MCS for RLC data blocks based on the LQM received from MS via PDAN for DL TBF, or received from BTS in-band for UL TBF. The Link Quality Measurements (LQM) Report is made of 4 parameters: o MEAN-BEP [8-PSK / GMSK] o CV-BEP [8-PSK / GMSK] stands for “Mean Bit Error Probability” stands for “Coefficient of Variation of BEP” LA algorithm relies on the use of MEAN_BEP and CV_BEP to derive the MCS number that provides the maximum throughput. Only the measurements (MEAN-BEP / CVBEP) corresponding to the modulation really used for transmission are meaningful. These MEAN_BEP and CV_BEP are estimated form a training sequence, thus, the higher the MEAN_BEP the higher MCS is used MEAN_BEP and CV_BEP are computed by the mobile and the BTS respectively for DL and UL direction. They are filtered before being sent to the PCU as part of the LQM report. The reported values are not the actual values, but integers, which can take values from 0 to 31 for MEAN_BEP and from 0 to 7 for CV_BEP. Performances which are presented in this document assume an ideal link adaptation, only one throughput value (throughputi) is kept for each ri of the cell (0 ≤ ri ≤ R): this throughput value is the best one among 9 throughputs offered by the different MCS. This means that only one throughput distribution results from link adaptation. MCS usage on the cell is deduced from this throughput distribution (see figure 5-6) Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 55/84 Radio Interface Engineering Rules Figure 5-6: MCS distribution / usage on the cell (schematic representation) Mean Throughput / TS 5.6.2.4 MEAN THROUGHPUT / TS Having throughput value per TS for each ri of the cell, mean throughput per TS is estimated by integration on coverage area (see figure 6). Following formula is used: where: • SurfaceTotale is coverage area, i.e. • Si is surface of the crown of range δ, centred on ri (actually δ= ri) Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 56/84 Radio Interface Engineering Rules Note that So and SR have different expressions since their range is /2: Figure 5-7.- Mean data throughput integration Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 57/84 Radio Interface Engineering Rules 5.6.3 COMMENTS ON DATA THROUGHPUT CALCULATIONS 5.6.3.1 C/(I+N) VERSUS BLER ESTIMATION Since no R&D simulations have been done in C/(I+N), values are estimated form C/I and C/N curves as explained in chapter 5.5.2.1. 5.6.3.2 LIMITATIONS Several hypotheses are considered to simplify calculations: regular distributed omnicell; uniform traffic distribution; MS performances are deduced from eDRX performances; perfect propagation model without fading effects; no antenna diagram is considered; instantaneous perfect link adaptation… 5.7. EDGE PERFORMANCES ESTIMATIONS In this chapter, EDGE performance capacities are estimated. Impact on capacity of incremental redundancy and frequency reuse are studied. EDGE performance estimation depends critically on hypothesis and input parameters. Thus, no estimation performance value has any sense if hypothesis associated are not clearly specified. In this chapter, EDGE performance values are given in order to show an idea on capacity performances but case-by-case capacity study should be done to for each network separately. Maximum cell ranges are considered, thus performances improve critically if inter-site distance is reduced. For performance estimations calculated in this document, following default network characteristics are considered: • 4x12 frequency plan is considered by default estimating 12 dB C/I at cell edge. • A cell size of 2.29 km at 900 MHz in TU3 and 6.95 km in TU50 is considered using a default link budget considering a S222 BS configuration available. Tables 2-6 detail link budget default parameters: User Equipment Performances Maximum UE Tx Power (dBm) / MEAN UE Tx antenna gain (dBi) UE Rx noise figure (dB) UE Rx sensitivity (dBm) Table 2: User Equipment performances BS Performances Maximum BS Tx power (dBm) / MEAN BS Tx/Rx antenna gain (dBi) BS Rx noise figure (dB) BS Rx sensitivity (dBm) 33 0 8 -104 44.8 18 3 -114 Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 58/84 Radio Interface Engineering Rules Slant losses (dB) 1.5 BS Rx cable & connector losses (dB) 3 Site antenna heigh (m) 30 Table 3: Base Station performances and characteristics Configuration Coupler Loss (dB) Table 4: Coupler Losses S222 1.4 S444 4.7 Margins Area Reliability desired Shadowing Margin Building penetration factor (dB) Body loss (dB) Table 5: Link Budget Margins TU3 TU50 Selected environment correction factor 0 -12 (dB) Table 6: Selected environment correction S888 7.8 90% 3.4 15 3 RU130/RA250 -22 Thus, taking into account all these input values, 5.7 dB C/N is found in UL at cell edge for GMSK MCS and 11.1 dB C/N in DL. C/I at cell DL C/No at UL C/No at Environment edge cell edge cell edge Urban TU3 12dB 11.1 dB 5.7 dB Suburban TU50 12dB 11.1 dB 5.7 dB Table 7: C/I and C/N hypothesis considered for EDGE performances estimations in different environments Polarisation diversity gain is considered in UL calculations and no diversity is considered in downlink. ePa output power is considered the same for GMSK and 8-PSK modulation. 11% mean downlink capacity per TS is lost if 3 dB output power loss is found for 8-PSK modulation. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 59/84 Radio Interface Engineering Rules Mean data throughput has been maximized independently from transmission impact. There is a trade-off to be analysed between data throughput and backhaul dimensioning. Backhaul dimensioning has an important impact on average throughput because operator may want to sacrifice a little bit of radio throughput to get significant gains on backhaul and equipment. As a typical example, if MCS 3, 6 and 8 are not used in a typical urban environment, 33% on backhaul equipment is saved (which translates into additional PCUSN and BSC savings because of connectivity needs) at the expense of only 3.5 % mean throughput degradation. 5.7.1 INCREMENTAL REDUNDANCY 5.7.1.1 INCREMENTAL REDUNDANCY FUNCTIONALITY In data systems, either packets are found correctly received or packets are requested to be retransmitted. While in GPRS, wrong received packets were discarded, EDGE allows to reuse old wrong received packet to combine them with fresh new packets to increase probability of receive correctly a packet. Incremental redundancy is highly related to retransmission. In order to allow this recombination gain between packet, certain constraints should be respected in term of MCS retransmission. The MCSs are divided into different families A, B and C. Each family has a different basic unit of payload. Different code rates within a family are achieved by transmitting a different number of payload units within one Radio Block. The following table lists the different coding schemes: When a retransmission is needed, a MCS belonging to the same family as the original block must be selected. If the LA-CommandedMCS does not belong to that family, another MCS has to be chosen. The bloc may be segmented in two half-blocks if needed, in order to be sent with MCS 2 or 3 from any 8-PSK MCS. (Re-segmentation activated in DL and UL in Nortel implementation) 5.7.1.1.1 RETRANSMISSION TABLE IN UL The following table synthesizes the situation for UL retransmission. This table is normative (see [3GPP TS 04.60]). Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 60/84 Radio Interface Engineering Rules Figure 5-8: UL retransmission table Note: In one particular case, the MCS selected for retransmission may be higher than the MCS selected by the link adaptation algorithm. If the original block was transmitted with family A and MCS 2 is commanded according to LA, MCS3 will be used. 5.7.1.1.2 RETRANSMISSION TABLE IN DL The following table synthesizes the situation for DL retransmission. This table is not normative. Nortel optimized the table to take advantage of Incremental Redundancy (IR) on mobile side: the retransmission may be sent in higher MCS than the LA commanded MCS in order to maximize incremental redundancy feature. (see bold MCS where higher MCS are proposed) 5.7.1.2 INCREMENTAL REDUNDANCY PERFORMANCE IMPROVEMENT Two different environments are analysed and compared: TU3 and TU50. Impact of IR is underlined. Frequency reuse is considered, UL IR is available from V16.0. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 61/84 Radio Interface Engineering Rules Tables 8 show mean data throughput per timeslot estimations and maximum data throughput per timeslot at cell edge. Incremental redundancy gain is also deduced. Mean Data Throughput Max Data Throughput at cell edge TU3 UL 900MHz IFH UL no IR 22 kbps 9.2 kbps UL IR 23.8 kbps 10.5 kbps Gain IR 8.20% 14.10% Table 8.1: Mean data throughput and maximum data throughput at cell edge in UL TU3 in 900MHz without and with IR Mean Data Throughput Max Data Throughput at cell edge TU3 DL 900MHz IFH UL no IR 21.5 kbps 8.2 kbps UL IR 22.1 kbps 9.7 kbps Gain IR 2.80% 18.30% Table 8.2: Mean data throughput and maximum data throughput at cell edge in DL TU3 in 900MHz without and with IR Mean Data Throughput Max Data Throughput at cell edge TU50 UL 900MHz IFH UL no IR 22.6 kbps 10.3 kbps UL IR 23.4 kbps 11 kbps Gain IR 3.40% 6.40% Table 8.3: Mean data throughput and maximum data throughput at cell edge in UL TU50 in 900MHz without and with IR Mean Data Throughput Max Data Throughput at cell edge TU50 DL 900MHz IFH UL no IR 21.5 kbps 8.4 kbps UL IR 21.5 kbps 9 kbps Gain IR 2.90% 7.10% Table 8.4: Mean data throughput and maximum data throughput at cell edge in DL TU50 in 900MHz without and with IR Note: For calculations of the throughput in DL, Nortel Networks eDRX performances without diversity have been considered. No engineering margin in eDRX C/I performances without diversity is taken since no difference should be found between eDRX and UE signal processing implementation. 3dB engineering margin is taken in Eb/No eDRX performances since eDRX performances are expected to be better than UE, although UE performances are expected to be better than standards. For UL data throughput calculations, 1dB implementation margin is considered. Main conclusions on mean data throughput per TS for an ideal frequency hopping network are: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 62/84 Radio Interface Engineering Rules 5.7.1.3 • Around 22-23 kbps /TS are achieved in UL while 20-22 kbps / TS are expected in DL. • As shown in tables 8, IR brings a gain around 8% in the mean data throughput. Anyway, where IR becomes really interesting is in worst conditions, i.e. at cell edge and in downlink where more retransmissions and recombination are needed. Up to 60% increase of guaranteed capacity at cell edge is obtained thanks to incremental redundancy. • Users’ speed degrades throughput performances. Higher MCS are affected the most, thus performances of mean data throughput are decreased. Around 5% loss is found between a user moving at 3km/h and a 50 km/h under same radio conditions. MCS USAGE DISTRIBUTION Table 12 and 14 show MCS used depending on UE-BS distance and maximum data throughput at this cell position. Moreover, mean data throughput and guaranteed data throughput are also shown for UL TU3 1800 no FH with IR and DL TU50 1800 no FH without IR respectively Table 13 and 15 shows MCS usage distribution for the same environment. Distance (km) 0.00 0.11 0.23 0.34 0.46 0.57 0.69 0.80 0.92 1.03 1.14 1.26 1.37 1.49 1.60 1.72 1.83 1.95 2.06 2.17 2.29 Throughput / TS MCS usage (kbps) (%) 59.2 MCS-9 59.2 59.2 59.1 58.4 56.2 52.0 46.4 40.4 35.3 MCS-8 30.4 26.3 MCS-7 23.2 20.5 18.2 16.1 14.3 12.8 11.9 MCS-4 11.1 10.5 MCS-3 Mean Throughput/TS (kbps): Mean Throughput / TS calculation 0.0 0.3 0.6 0.9 1.2 1.4 1.6 1.6 1.6 1.6 1.5 1.4 1.4 1.3 1.3 1.2 1.1 1.1 1.1 1.1 0.5 23.8 Table 12: MCS usage distribution and data throughput depending on cell position in UL TU3 900MHz IFH with IR Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 63/84 Radio Interface Engineering Rules MCS-1 MCS-2 MCS-3 MCS-4 MCS-5 MCS-6 MCS-7 MCS-8 MCS-9 % Users 0.0 0.0 4.9 18.5 0.0 0.0 49.0 9.5 18.1 % Users Cumulative 100.0 100.0 100.0 95.1 76.6 76.6 76.6 27.6 18.1 Mean BLER Mean BLER when used 0.0 0.0 29.4 34.8 0.0 0.0 59.5 39.9 14.1 43.4 BLER at cell edge 16.1 21.0 29.4 41.3 62.9 72.3 79.1 90.3 94.3 Table 13: MSC usage distribution in UL TU3 900MHz IFH with IR Figure: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 64/84 Radio Interface Engineering Rules Distance (km) 0.00 0.35 0.70 1.04 1.39 1.74 2.09 2.43 2.78 3.13 3.48 3.82 4.17 4.52 4.87 5.22 5.56 5.91 6.26 6.61 6.95 Throughput / TS MCS usage (kbps) (%) 59.2 MCS-9 59.2 59.2 58.7 56.6 52.2 46.1 MCS-8 41.1 MCS-7 36.3 30.4 25.7 MCS-6 23.1 20.4 17.4 15.5 MCS-5 13.5 11.9 10.2 9.6 MCS-2 9.1 8.4 Mean Throughput/TS (kbps): Mean Throughput / TS calculation 0.0 0.3 0.6 0.9 1.1 1.3 1.4 1.4 1.5 1.4 1.3 1.3 1.2 1.1 1.1 1.0 1.0 0.9 0.9 0.9 0.4 20.9 Table 14: MCS usage distribution and data throughput depending on cell position in DL TU50 900MHz IFH without IR MCS-1 MCS-2 MCS-3 MCS-4 MCS-5 MCS-6 MCS-7 MCS-8 MCS-9 % Users 0.0 23.4 0.0 0.0 31.0 23.0 12.0 3.0 7.6 % Users Cumulative 100.0 100.0 76.6 76.6 76.6 45.6 22.6 10.6 7.6 Mean BLER Mean BLER when used 0.0 18.4 0.0 0.0 43.6 27.9 20.8 15.3 5.2 27.6 BLER at cell edge 11.3 25.1 59.4 87.3 69.8 81.1 96.9 99.3 100.0 Table 15: MSC usage distribution in DL TU50 900MHz IFH without IR Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 65/84 Radio Interface Engineering Rules Figure: Main conclusions of MCS usage distribution in non-hopping frequency reuse patterns can be stated as follows: • MCS 9 is used in a small cell coverage area near the BS. • MCS 6, MCS 7 and MCS 8 are mainly used over the cell. • MCS usage distribution shows 8-PSK MCS are mainly used all over the cell but GMSK MCS are necessary to prevent worst punctual cases at cell edge, mainly in downlink. Frequency reuse degrades MCS data performances and it modifies critically MCS usage distribution as it is detailed in the next chapter (same simulations without frequency hopping have shown that MSC9 is much more used than with frequency hopping) 5.7.2 HIGH POWER AMPLIFIER IMPACT ON DATA PERFORMANCES Nortel Networks proposes to increase power amplifier capabilities. ePA which allows to transmit at 30W. For GMSK and 8-PSK can be replaced by a HePA (High ePA) which is able to transmit at 60W. For GMSK and 45 W. for 8-PSK modulation. Different advantages justify HePA deployment: • Increase on coverage or area reliability. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 66/84 Radio Interface Engineering Rules • Compensate coupling losses when higher configurations should be deployed. • Increase on data performances in coverage limited environments. For the sake of quantifying HePA gain and MCPA, following environment is proposed: TU3 dense urban environment with 510 meters cell range (maximum coverage once link budget is defined) at 1800 MHz. 12 dB C/I are supposed at cell edge. H2D coupling loss. Same example is calculated considering H4D and 420 meters cell radius. Table xxx presents comparison between ePA, HePA and MCPA impact on data throughput: MCS usage distribution is also modified since radio coverage conditions are improved. As shown in table xxx, higher MCS codecs are further used: It should be underlined, that conditions proposed are significantly coverage limited. If intercell distance would be reduced for any network design reason, e.g. capacity cell splitting, HePA and MCPA gain will be reduced significantly. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 67/84 Radio Interface Engineering Rules Besides, same coverage conditions have been considered, i.e. 510 m and 420m. Therefore, all power gain has been translated in a performance gain instead of a coverage improvement. Deploying a HePA can change link balance limitations, e.g. typically for H2D coupling loss. For an ePA, downlink link budget is found to be the most constraining link while when HePA is deployed, UL limitations are found (HePA does not impact uplink link budget). This raises the question if it is worth to deploy a TMA to improve UL constrains. Customers’ requirements should be taken into account in order to answer that question. Deployment of a TMA is useful if higher coverage gain is required or if UL performance becomes limiting factor. Moreover, TMA deployment would degrade slightly DL performances since small DL loss in output power should be considered. Therefore, no TMA addition is recommended when HePA is deployed if it is not justified for UL performance improvement requirements. 5.7.3 FREQUENCY REUSE Frequency reuse brings frequency and interferer diversity in the process of reception; since changing the RF carrier while hopping allows to average the effects of strong fading and allows spreading interferences across the network. Frequency reuse is particularly efficient for slow moving mobiles and for a sufficient number of hopping frequencies. It allows reducing the number of frequencies per TRX and, thus, the capacity is increased since more TRX can be deployed in the same frequency band. Frequency load is crucial parameter in a network with frequency reuse. It is simply the ratio between number of hopping TRXs in a cell and number of hopping frequencies. Thus, it represents the time fraction for a given frequency being used in the network. In order to determine impact of frequency reuse in EDGE MCS, frequency reuse gain of each MCS is compared with speech’s gain. Figure 5-9 shows sensitivity versus BLER performances for TU3, 900 MHz, no diversity no frequency hopping. In order to guarantee 10% BLER, -107.3 dB sensitivity is found. 4 dB gains on sensitivity is found due to frequency hopping as it is shown in figure 5-10, where –111.3 dB is found to achieve 10% BLER. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 68/84 Radio Interface Engineering Rules Figure 5-9- BLER versus sensitivity for speech in TU3 without frequency hopping Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 69/84 Radio Interface Engineering Rules Figure 5-10- BLER versus sensitivity for speech in TU3 with frequency hopping Frequency reuse does not provide same benefits for data than for voice. If BLER gains (for data) versus FER gains (for speech) are considered: spreading the errors does not necessarily bring any improvement, as packets have to be received errorfree (used error correction algorithms must be able to correct all errors). In fact, FH introduces errors in the bits flow which are difficult to correct if the signal correction process is very poor. The more demanding on radio quality the protocol is, the less benefits are given by frequency reuse. Therefore, impact of frequency reuse for each MCS has to be calculated. Table 16 shows impact for TU3, 900 MHz, and no diversity for different MCS. Frequency reuse impact on EDGE MCS is estimated in [R1]: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 70/84 Radio Interface Engineering Rules As it has been test in the field, 16% frequency load brings same quality for speech, considering DTX and power control inactivated, than a non-hopping frequency plan network [R4]. Therefore, depending on frequency load deployed in the network, different performances are found. Assuming that 4 dB gain brings same quality than 16% frequency load for speech, and knowing the impact of frequency hopping on each MCS, frequency load required to achieve same performances as 4*12 frequency plan can be calculated for each MCS using a linear approximation following equation 15. Different frequency load for each MCS is required to achieve same performances as non-hopping frequency plan network are shown in table 17: Table 17.- FL for each MCS to obtain same performances as non-hopping frequency plan Even if frequency reuse gain depends on considered BLER, in Annex 2, it is shown how using 10% BLER value is a good approximation for all MCS. If frequency load in a network correspond to table 17 MCS FL, MCS performances of this MCS are the same that no frequency hopping network. If FL is lower, frequency hopping brings a gain in the performances of the MCS but if it is bigger degradation has to be considered. Equation (16) estimates the gain or degradation depending on frequency load implemented and MCS studied. Degradation values are applied to C/(I+N). Same values are applied independently from the environment. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 71/84 Radio Interface Engineering Rules Table 18-20 show capacity increase thanks to frequency reuse. Dense Urban, TU3, 1800 MHz, no IR, downlink with diversity, 510 m. cell size study has been analysed. Study compares capacity available for different frequency band, e.g. 5MHz, 7.5 MHz and 10 MHz, considering same QoS in each scenario for users, i.e. same data throughput. Capacity is then calculated as mean data throughput times the number of TS available divided by the number of TS required to obtain the same QoS as nonhopping frequency reuse pattern network. Downlink cases are mainly analysed in the document since asymmetrical traffic expectations make downlink limited system. As can be expected, frequency hopping degrades mean data throughput per timeslot. But, frequency hopping associated with fractional reuse pattern allows increasing the number of timeslots per site. Using a typical 16% frequency load allows increasing 37% capacity with regard to non-frequency hopping dimensioning if 10 MHz is available. Capacity gain is critically dependant on frequency band available and frequency load deployed in the network as it is shown in tables 18-20. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 72/84 Radio Interface Engineering Rules 5.7.4 IMPORTANT HYPOTHESIS AND ASSUMPTIONS SUPPOSED • dB UE noise figure and 8 dB BS noise figure are considered • Engineering margins are left to adapt BS eDRX performances to UE performance estimations. • For the simplicity of the analysis, no call profile is considered and number of either EDGE or speech TS are fixed. • Moreover, voice and data busy hours are considered the same (worst case) • Perfect link adaptation is considered independently from BLER estimated, thus no stability problem is considered in the backhaul even if BLER grow to more than 50% BLER. • Results are purely radio capacity, thus no delay on protocols, neither connection times nor release times are considered. • For mean data throughput calculations, users are supposed to request maximum available data throughput. • Real distribution of BCCH and SDCCH are considered depending on base station configuration. • Uniform user distribution is considered. • Maximum inter-site distance is considered following Nortel Networks default link budgets. • Frequency reuse plans are considered always in 1:1 with 16% frequency load independently from EDGE TS load. • 70% channel occupancy is considered for both, speech and EDGE data TS. Reference value of FL of 16% is estimated. • 4*12 BCCH frequency plan is always considered, which corresponds to 12 dB C/I with 95% reliability. 12 dB C/I are found considering full power TX by the BS. Thus, worst case it is considered. There will be more interference because of data but C/I should not be lower than 12 dB. • ePA usage with same output power for GMSK and 8-PSK modulation is considered. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 73/84 Radio Interface Engineering Rules 5.7.5 BLER STUDY Another study, which could be done with EDGE Radio Performance Excel tool is to estimate mean BLER over the cell and expected BLER at cell edge. For all presented exercise in this document, no limitations, in term of BLER, have been taken into account. To calculate different MCS data throughput simple formula was considered: EffectiveThroughput=MaxThroughput_MCS*(1-BLER) (21) Link adaptation was supposed to choose MCS, which optimizes maximum data throughput independently from BLER estimated for each MCS. Thanks to RLC/MAC layer improvement from GPRS, maximum EDGE BLER could be higher than 10% BLER which is found in GPRS as limiting point. 100% maximum BLER is proposed (even if it is unthinkable for good network stability) to calculate maximum radio data throughput. Further studies will help to determine this value. High BLER values are considered for 8-PSK modulations. 8-PSK data throughput performances compared with GMSK modulations makes 8-PSK MCS suitable to be chosen for ideal link adaptation algorithm than GMSK. For the sake of giving a simple example, MCS5 working at 50 % BLER effective data throughput equals 11.2 kbps is suitable for ideal link adaptation than MCS2 working at 2% BLER which represents 10.976 kbps effective data throughput ! Therefore, all mean BLERs given in this chapter could be limited with a low loss in mean data throughput. Following table shows mean BLER for an environment TU3; UL; 1800 MHz; IR where 8-PSK modulations are recommended to maximize data throughput. A MCS usage distribution is shown as well as mean BLER per MCS and overall means BLER over the cell: For example, MCS 7 is used through 27% of the cell area with a mean BLER of 52.4%. Over the whole cell, meaning BLER of chosen MCS to maximize data throughput is 30.7% BLER. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 74/84 Radio Interface Engineering Rules In an environment where GMSK is used, mean BLER is critically decreased. Following table stands for TU50 UL 1800 MHz, no IR, 12dB C/I and 1.56 km cell range for a S444 maximum BS configuration: MCS-1 MCS-2 MCS-3 MCS-4 MCS-5 MCS-6 MCS-7 MCS-8 MCS-9 % Users 0.0 0.0 4.9 18.5 0.0 0.0 49.0 9.5 18.1 % Users Cumulative 100.0 100.0 100.0 95.1 76.6 76.6 76.6 27.6 18.1 Mean BLER Mean BLER when used 0.0 0.0 29.4 34.8 0.0 0.0 59.5 39.9 14.1 43.4 BLER at cell edge 16.1 21.0 29.4 41.3 62.9 72.3 79.1 90.3 94.3 Table 13: MSC usage distribution in UL TU3 900MHz IFH with IR MCS-1 MCS-2 MCS-3 MCS-4 MCS-5 MCS-6 MCS-7 MCS-8 MCS-9 % Users 0.0 23.4 0.0 0.0 31.0 23.0 12.0 3.0 7.6 % Users Cumulative 100.0 100.0 76.6 76.6 76.6 45.6 22.6 10.6 7.6 Mean BLER Mean BLER when used 0.0 18.4 0.0 0.0 43.6 27.9 20.8 15.3 5.2 27.6 BLER at cell edge 11.3 25.1 59.4 87.3 69.8 81.1 96.9 99.3 100.0 Table 15: MSC usage distribution in DL TU50 900MHz IFH without IR Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 75/84 Radio Interface Engineering Rules Main conclusions in BLER study: 5.8. • Ideal link adaptation chooses MCS independently from associated BLER to optimise data throughput. A further simulation study should be done to estimate, which is the maximum BLER allowed in EDGE system to keep stability. • Incremental redundancy reduces BLER at cell edge. BLER from MCS 2 does not overcome 20% BLER without frequency hopping in any environment. • Frequency hopping degrades critically BLER performances. The validity of all results presented in this document depends on stability because of high BLERs. UL PERFORMANCE ANALYSIS EDGE UL performances are affected by the fact that it presents the advantage to have diversity available but on the other hand, IR will not be available until V16 release. Different types of mobiles are found in the market. 5 different GMSK mobiles are possible. Power Class 4 mobiles for 900 MHz (33 dBm output GMSK power) are widely expected while Power Class 1 mobile for 1800 MHz (30 dBm output GMSK power) are supposed. In term of 8-PSK modulation mobile performance, table xxx shows difference of power performances depending on type of mobile and power class. E2 mobiles are widely Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 76/84 Radio Interface Engineering Rules expected to be found on the market. E2 Power Class mobiles have a different output power between GSMK and 8-PSK modulation. 6 dB less 8-PSK modulation power are found with regard to common 33dBm for GMSK modulation and 4 dB less power are expected. 5.8.1 DIFFERENT MOBILE TYPE PERFORMANCE IMPACT As explained in last chapter, different types of mobiles are found in the market, presenting different radio performances in terms of 8-PSK modulation output power. The less 8-PSK output power is available; less performing in terms of data throughput is expected. In order to quantify radio throughput degradation following environment is proposed: 1800 MHz, 510 meters cell range, H2D configuration, no FH available, no IR, 12 dB C/I at cell edge (some coverage and interference propagation model as DL is deduced) Degradation depends critically on environment defined. If interference UL limited environment is studied, mean data throughput degradation could be neglected. E2 class mobiles usage affects also MCS usage distribution. The better power capabilities, higher MCS could be used. Usage of E2 and E3 mobiles provoke massive usage of GMSK modulation in UL. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 77/84 Radio Interface Engineering Rules 5.9. LAB TEST RESULTS Several tests simulating different radio condition environments have been carried in platform as well as different mobile performances have been tested. CAUTION: Measured performances in lab are applicative throughput while radio RLC/MAC is shown in the rest of the document. This makes that comparison between both figures should be done carefully. Chapter 9.4 shows a comparison between measured and simulated performances with all hypotheses detailed supposed in order to compare both performances. 5.9.1 UL LAB RESULTS The goal of the test was to get for each EDGE coding schemes (MCS) and for the default LA table, the graphic representation of the averaged end-to-end throughput that can be obtained for different C/I values in uplink direction (FTP-PUT). TU3 radio environment is simulated. During the whole test, the “C” value is fixed to -65 dBm, while the “I” value while GMSK co-channel interference generated by an interferer tool transmitting are produced at a constant level of -70 dBm, is modified through an attenuator. 10 transfer iterations are executed for C/I value. Each throughput point of the curve is the averaged result over the 10 iterations. Motorola T725 (3+1) mobile handset was used. Diversity is available. UL Applicative data throughput test results are shown in figure below: Throughput in Degraded Radio Conditions C/I - FTP Put 50 s8k 1900MHz with Duplexer Diversity enabled. No Hopping Cochannel GMSK interference TU3 Environment MS Class 4 : 3+1 For LA, transfers with dropped IP frames due to CRC are kept. For MCS curves, they are removed. 45 40 35 Throughput (kbps) MCS9 MCS8 30 MCS7 MCS6 25 MCS5 MCS3 20 MCS2 LA Curve 15 10 5 0 0 10 20 30 40 50 C/I at BTS (dB) Figure- UL applicative throughput test results Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 78/84 Radio Interface Engineering Rules 5.9.2 DL LAB RESULTS The goal of the test was to get for each EDGE coding schemes (MCS) and for the default LA table, the graphic representation of the averaged end-to-end throughput that can be obtained for different C/I values in downlink direction (FTP-GET). This is done in a TU3 radio environment. During the whole test, the “C” value is fixed to -65 dBm, while the “I” value while GMSK co-channel interference generated by an interferer tool transmitting are produced at a constant level of -70 dBm, is modified through an attenuator. BROADCOM GC82 (4+1) card was used. 10 transfer iterations are executed for C/I value. Each throughput point of the curve is the averaged result over the 10 iterations. DL Applicative data throughput test results are shown in figure below: Throughput in Degraded Radio Conditions C/I - FTP Get s8k 1900MHz with Duplexer (bsTxPwrMax=41) No Hopping Cochannel GMSK interference TU3 Environment MS Class 8 : 4+1 For LA, transfers with dropped IP frames due to CRC are kept. For MCS curves, they are removed. 200 Throughput (kbps) 150 MCS9 MCS8 MCS7 MCS6 MCS5 100 MCS3 MCS2 LA Curve 50 0 0 10 20 30 40 50 C/I at Ms (dB) Figure .- DL applicative throughput test results 5.9.3 REAL LINK ADAPTATION IMPACT ON MEAN DATA THROUGHPUT Real UL and DL link adaptation measured curves using default adaptation table set (see figure xxx and figure xxx) show impact of real link adaptation with regard to the optimum LA curves, which would correspond to the choice of the MCS which would maximize data throughput. Applying engineering methodology detailed in chapter 6, and using UL and DL curve performance obtained through lab measurements, (only C/I versus throughput is considered), following mean data throughput performances are obtained: Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 79/84 Radio Interface Engineering Rules From 3% to 6% degradation is found due to real link adaptation algorithms with regard to ideal link adaptation. DL link adaptation looks like to be more performant than UL. This better functionality could be attributed to incremental redundancy link adaptation aggressiveness. MCS used are normally higher than link adaptation requests when are retransmitted in order to take profit of IR functionality. This allows recovering some throughput compared to ideal MCS par MCS optimized link adaptation. UL incremental redundancy is still not implemented in the product and it is expected in V16. 5.9.4 APPLICATIVE VERSUS RLC/MAC THROUGHPUT RELATIONSHIP Not deterministic link between applicative and RLC/MAC throughput can be found. Several reasons are listed below: - Different MCS are used during a communications, having each MCS a different header payload. - Throughput is critically related to application: • TCP throughput is related to network TCP/IP parameters • Depends on packet size (e.g. impact of slow starting affects applicative throughput) - UL and DL radio conditions are difficult to link and their relationship affects other link performances: • Acknowledgements should be correctly received - BLER impact on IP level cannot be evaluated linearly: • Low BLER: retransmissions at RLC/MAC level, which reduce applicative throughput. No IP packet loss occurs; thus, lineal relationship could be used. • High BLER: retransmissions at RLC/MAC level, which reduce applicative throughput and also IP packet loss occurs which no lineal relationship can be done. Some test on perfect radio conditions were carried on platform, i.e. BLER=0% for all codecs. Different tests using a single codec were done. Big files of 8Mb were transferred in order to avoid slow starting effect. Table xxx shows radio throughput compared with TCP applicative layer results per codec. Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 80/84 Radio Interface Engineering Rules DL MCS (kbps) MCS-2 MCS-3 MCS-5 MCS-6 MCS-7 MCS-8 MCS-9 Average 4 TS TCP Applicative Radio versus MCS Max MCS Max layer to Radio Applicative Throughput Throughput Performances Throughput 41,43 11,2 44,8 8,13% 54,57 14,8 59,2 8,48% 82,19 22,4 89,6 9,02% 108,58 29,6 118,4 9,04% 162,39 44,8 179,2 10,35% 195,17 54,4 217,6 11,49% 210,61 59,2 236,8 12,44% Table xxx.- Radio RLC/MAC versus TCP Applicative throughput comparison As it is shown in table xxx, difference between radio and TCP applicative throughput is dependent on MCS and it can be quantified between 8-12% in perfect radio conditions. 5.10. CONCLUSION Nortel Networks EDGE engineering data throughput performances estimation methodology has been detailed. Different feature impacts on performances have been analysed. Following conclusions are found for an only EDGE traffic networks scenario: • Incremental redundancy brings 15% mean data throughput/TS increase. Its main advantage is reached in worst conditions, i.e. on cell edge where guaranteed data throughput is increased up to more than 50% data throughput. • Frequency hopping has shown to increase capacity more than 30% depending on frequency band available and frequency load deployed. Performances are critically dependent on scenario, frequency band available, EDGE penetration, EDGE TS allocation strategy, reference taken for comparison... Thus, all estimated performances are for the sake of giving examples and correspond to a typical environment but case-by-case study should be reconsidered using each customer network characteristics. Performance calculation inaccuracy concerns most of results presented but not the tendencies obtained. It has been detailed how EDGE TS allocation strategy depends on EDGE penetration and frequency band available. • When best data throughputs and low data capacity are expected, PDTCH on BCCH is best approach • When data capacity is required, it is worth to sacrifice a little bit of peak data throughput to have a gain on data capacity by setting the PDTCH on a hopping layer. Finally, GPRS versus EDGE offered traffic load is compared to underline the advantage on traffic load when bigger data throughput par TS is available. Data Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 81/84 Radio Interface Engineering Rules throughput does not only improve the transmission time but increases the number of user which could be processed in the network if comparison is done at the same QoS with GPRS. 6. ABBREVIATIONS AND DEFINITIONS 6.1. ABBREVIATIONS A ETSI generic name for BSS-NSS i/f Abis ETSI generic name for BTS-BSC i/f Agprs NORTEL specific name for BSC-PCU i/f BCCH Broadcast control channel BH Busy hour BLER Block error rate BSC BSS Base station subsystem BTS BSSGP BSS GPRS protocol BVC CCCH Common control channel CS CS-paging Circuit Switched-paging CCU Channel codec unit DL Downlink FN Frame number Gb ETSI generic name for PCU-SGSN i/f GGSN Gateway GPRS support node Gi ETSI generic name for GGSN-PDN i/f GMM GPRS mobility management Gn ETSI generic name for SGSN-GGSN i/f GPRS General packet radio service GSL GPRS signaling link Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 82/84 Radio Interface Engineering Rules I/F Interface IP Internet protocol LA Location area LAC Location area code LAI Location area identity LAPD Link access protocol on D channel LLC Logical link control MAC Medium access control MCS Modulation and Coding Scheme MO Mobile originated MS Mobile station MT Mobile terminated NS Network service NSAPI Network service access point identifier NSS Network and switching subsystem NTS Number of TS assigned to the UL TBF O&M Operation and maintenance OAM Operation administration maintenance OML OAM link PACCH Packet associated control channel PAREJ Packet access reject PBCCH Packet broadcast control channel PCA Packet control acknowledgement PCCCH Packet common control channel PCM Pulse coded modulation PCU Packet control unit PCUSN Packet control unit support node PCUSA Packet control unit system application PDAN Packet DL Ack/Nack PDAS Packet DL assignment PDCH Packet data channel PDN Packet data network PDP Packet data protocol PDTCH Packet data traffic channel Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 83/84 Radio Interface Engineering Rules 6.2. PDU Packet Data Unit PRR Packet resource request PS-paging Packet Switched-paging PSTN Public switched telephone network QoS Quality of service RA Rural Area (type of radio environment) RA Routing area RAC Routing area code RAI Routing area identifier RACH Random access channel RLC Radio link control RRM Radio Resource Management RSL Radio signaling link SAPI Service access point identifier SGSN Serving GPRS support node SSN Starting sequence number TA Timing advance TBF Temporary block flow TCP Transmission control protocol TDMA Time division multiple access TFI Temporary flow identity TLLI Temporary logical link identifier TRX BTS transceiver entity TS Timeslot UL Uplink USF UL state flag DEFINITIONS For the rules presented in the document, (O) means the rule is optional, (M) means the rule is mandatory, (HC) means the rule is hard-coded. Z END OF DOCUMENT Y Nortel confidential PE/DCL/DD/014283 02.02 / EN Standard 28/Sept/2006 Page 84/84