Survey of Next-Generation Optical Access System Concepts

advertisement
Consortium groups the following Organizations:
Partner Name
Short name
JCP-CONSULT SAS
JCP
DEUTSCHE TELEKOM AG
DTAG
INTERDISCIPLINARY INSTITUTE FOR BROADBAND IBBT
TECHNOLOGY
TECHNISCHE UNIVERSITAET MUENCHEN
TUM
KUNGLIGA TEKNISKA HOEGSKOLAN
KTH
ADVA AG OPTICAL NETWORKING
ADVA
ERICSSON AB
EA
ERICSSON TELECOMUNICAZIONI
TEI
ACREO AB.
ACREO
MAGYAR
TELEKOM
TAVKOZLESI
NYILVANOSAN MAGYAR
MUKODO RESZVENYTARSASAG
TELEKOM
RT
SLOVAK TELEKOM AS
SLOVAK
TELEKOM
UNIVERSITY OF ESSEX
UESSEX
Country
FR
D
B
D
S
D
S
I
S
H
Survey of Next-Generation Optical Access System
Concepts
SK
UK
Deliverable 4.1
Abstract:
Document to be considerate as Word Template for all deliverables
OASE_D4.1_WP4_EAB_210612_v3.0.doc
Version: 3.0
Last Update: 21/06/2012
“The research leading to these results hasDistribution
received funding
from the European Community's Seventh
Level:
Framework Programme (FP7/2007-2013) under
agreement n° 249025”
PU grant
(Public)

Distribution level
PU = Public,
RE = Restricted to a group of the specified Consortium,
PP = Restricted to other program participants (including Commission Services),
CO= Confidential, only for members of the OASE Consortium (including the Commission Services)
Partner Name
DEUTSCHE TELEKOM AG
INTERDISCIPLINARY INSTITUTE FOR BROADBAND
TECHNOLOGY
KUNGLIGA TEKNISKA HOEGSKOLAN
ADVA AG OPTICAL NETWORKING
ERICSSON AB
ERICSSON TELECOMUNICAZIONI S.P.A
ACREO AB.
MAGYAR TELEKOM TAVKOZLESI NYILVANOSAN
MUKODO RESZVENYTARSASAG
UNIVERSITY of ESSEX
Short name Country
DTAG
D
IBBT
B
KTH
ADVA
EA
TEI
ACREO
MT
S
D
S
I
S
H
UESSEX
UK
Abstract:
This deliverable presents a survey of system concepts for next-generation optical access based
on OASE requirements. For the considered systems, key evolving components/subsystems are
identified and a survey for each component/subsystem presented. A broad range of system
concepts and key technologies are covered, including key aspects such as maturity, cost,
capacity and reach.
“The research leading to these results has received funding from the European Union's Seventh
Framework Programme (FP7/2007-2013) under grant agreement n° 249025”
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 2 of 168
Document Identity
Title:
Subject:
Number:
File name:
Registration Date:
Last Update:
Survey of Next-Generation Optical Access System Concepts
WP4 “System concepts for Next Generation optical access networks”
D4.1
OASE_D4.1_WP4_EAB_210612_v3.0.doc
th
March, 12 2010
21/06/2012
Revision History
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Version
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
0
Comments:
1
Comments:
2
Comments:
3
Comments:
Edition
1
Author(s)
A. Bianchi (TEI)
Date
03/12/2010
2
B. Skubic (EAB)
04/10/2010
Sub-chapters brought out of document and edited individually
3
B.Skubic (EAB), A. Bianchi (TEI)
10/8/2010
Compilation of contributions from ADVA, ACREO, DTAG, EAB, IBBT, KTH, TEI, MT, UESSEX
4
B. Skubic (EAB), A. Bianchi (TEI)
10/12/2010
Version released for internal review
5
M. Parker (UESSEX)
10/21/2010
6
B. Skubic (EAB)
10/25/2010
7
M. Roppelt (ADVA)
10/25/2010
8
J. Chen (KTH)
10/25/2010
9
B. Lannoo (IBBT)
10/27/2010
10
B. Skubic (EAB)
Included update from KTH and DTAG
11
A. Bianchi (TEI)
10/27/2010
12
J. Chen (KTH)
10/27/2010
13
M. Roppelt (ADVA)
10/27/2010
14
B. Lannoo (IBBT)
10/28/2010
15
B. Skubic (EAB)
Final editing
0
E. Malésys (JCP)
Formatting/ released version
0
B. Skubic (EAB)
Incorporation of reviewers’ comments
0
B. Skubic (EAB)
Incorporation of reviewers’ comments
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
10/27/2010
10/29/2010
10/29/2010
06/06/2011
21/06/2012
Page: 3 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table of Contents
EXECUTIVE SUMMARY ................................................................................................................................... 6
REFERRED DOCUMENTS ................................................................................................................................ 8
FIGURES AND TABLES .................................................................................................................................. 16
ABBREVIATIONS ............................................................................................................................................. 21
1.
INTRODUCTION ..................................................................................................................................... 29
2.
TASK DESCRIPTION AND SCOPE ...................................................................................................... 29
2.1
3.
ASSUMPTIONS AND METHODOLOGY ............................................................................................. 31
3.1
3.2
3.3
4.
NGOA SYSTEM CONCEPTS.................................................................................................................. 31
KEY ASPECTS FOR THE ASSESSMENT ................................................................................................... 33
COST AND POWER ASSUMPTIONS ........................................................................................................ 35
SYSTEM CONCEPTS FOR NGOA ........................................................................................................ 36
4.1
4.1.1
4.1.2
4.1.3
4.1.4
4.2
4.2.1
4.2.2
4.2.3
4.3
4.3.1
4.3.2
4.4
4.5
4.5.1
4.5.2
4.5.3
4.5.4
4.5.5
4.6
4.6.1
4.6.2
4.6.3
4.7
4.7.1
4.7.2
4.8
4.8.1
4.8.2
4.8.3
4.8.4
5.
RELATIONSHIP TO OTHER WORKPACKAGES ........................................................................................ 30
TDM-PON ......................................................................................................................................... 36
Serial 40G NRZ ............................................................................................................................. 40
Serial 40G QPSK .......................................................................................................................... 41
Serial 40G DQPSK ....................................................................................................................... 41
Serial 4x10G NRZ ......................................................................................................................... 42
WDM-PON ........................................................................................................................................ 43
WDM-PON with (Tunable) Lasers and Laser-Arrays ................................................................... 43
WDM-PON with seeded Reflective Transmitters .......................................................................... 50
WDM-PON with wavelength reuse ............................................................................................... 55
OFDM-PON ....................................................................................................................................... 57
Serial 40G CO-OFDM .................................................................................................................. 58
Serial 40G DDO-OFDM ............................................................................................................... 59
CDM .................................................................................................................................................. 61
WDM+XXM HYBRID ........................................................................................................................ 66
Hybrid WDM/TDM-PON .............................................................................................................. 67
Hybrid WDM/CDM-PON .............................................................................................................. 72
Hybrid WDM/OFDM-PON ........................................................................................................... 77
Hybrid WDM/SCM-PON............................................................................................................... 79
UDWDM ....................................................................................................................................... 80
HYBRID ACTIVE WDM ....................................................................................................................... 81
Hybrid active WDM with active P2P access ................................................................................. 81
Hybrid active WDM-PON ............................................................................................................. 82
Comparison of active/passive hybrids and WDM-PON ................................................................ 83
NEXT GENERATION AON: RPR, WDM, ETHERNET ........................................................................... 85
GbE access .................................................................................................................................... 85
10 Gb/s Ethernet access and distribution/aggregation ................................................................. 86
RADIO OVER FIBRE BACKHAULING ..................................................................................................... 87
Radio-over-Fibre (RoF) Technologies .......................................................................................... 89
CPRI .............................................................................................................................................. 90
OBSAI............................................................................................................................................ 90
Hybrid Optical-Wireless PON Architectures ................................................................................ 91
SURVEY OF EVOLVING COMPONENTS/SUBSYSTEMS ............................................................... 92
5.1
WDM COMPONENTS/SUBSYSTEMS .................................................................................................... 93
5.1.1
Wavelength selective components for WDM systems .................................................................... 93
5.1.2
Example of FTTH-PON-based on ASE-injected FP-LDs .............................................................. 94
5.2
ONT BASED ON RSOA, REAM, REAT .............................................................................................. 95
5.2.1
ONT based on RSOA ..................................................................................................................... 95
5.2.2
ONT based on REAM .................................................................................................................... 98
5.2.3
ONT based on REAT ................................................................................................................... 100
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 4 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
5.3
TUNABLE LASERS ............................................................................................................................. 101
5.4
WAVELENGTH SELECTIVE RECEIVERS............................................................................................... 115
5.4.1
Tunable Filters ............................................................................................................................ 115
5.4.2
Coherent Receivers ..................................................................................................................... 122
5.5
BURST MODE RECEIVERS .................................................................................................................. 133
5.6
ADC/DAC ....................................................................................................................................... 140
5.7
DISPERSION COMPENSATION ............................................................................................................. 143
5.7.1
Compensation at the transmitter ................................................................................................. 143
5.7.2
Compensation at the receiver ...................................................................................................... 143
5.8
PASSIVE WAVELENGTH SELECTIVE DEVICES ..................................................................................... 144
5.8.1
Thin film filter-based WDM components .................................................................................... 144
5.8.2
Fibre Bragg grating-based WDM components ........................................................................... 145
5.8.3
Etched Diffraction Grating-based WDM components ................................................................ 146
5.8.4
Arrayed Waveguide Grating-based WDM components .............................................................. 147
5.8.5
Technology constrains for EDG and AWG WDM components ................................................... 149
5.9
REACH EXTENDERS .......................................................................................................................... 149
5.10
SWITCHING COMPONENTS ................................................................................................................. 153
5.10.1
Power consumption ................................................................................................................ 153
5.10.2
Challenges on high speed switching process: ........................................................................ 154
5.10.3
Network stability ..................................................................................................................... 154
6.
SUMMARY .............................................................................................................................................. 155
7.
APPENDIX .............................................................................................................................................. 160
7.1
7.2
7.3
COST AND ENERGY CONSUMPTION DATA ........................................................................................ 160
EXAMPLES DEVICES ON THE MARKET BASED ON TFF ....................................................................... 161
EXAMPLES DEVICES ON THE MARKET BASED ON AWG..................................................................... 165
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 5 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Executive summary
This document provides a survey of system concepts/technologies with potential of meeting
the requirements of next-generation optical access (NGOA) as defined in OASE Deliverable
2.1. Key requirements include residential peak data rates of ≥1 Gb/s, support for 256 to 1024
customers per feeder fibre, support for 20 to 40 km passive reach, and support for 60 to 90 km
extended reach for the protection path.
For each next-generation system candidate, a detailed technical description is made as well as
a coarse investigation with respect to key aspects such as bandwidth, number of customers per
feeder fibre, reach, cost and power consumption. Furthermore, based on the presented survey
of next-generation systems, key evolving components/subsystems are identified and
investigated as a basis for an assessment for system maturity and time to market. A detailed
technical description of each evolving component/subsystem is made as well as a coarse
assessment based on key aspects. Key components include WDM components/subsystems,
tunable lasers, reflective transmitters, wavelength selective receivers (such as tunable filters
and coherent receivers), burst mode receivers, analogue-to-digital converters (ADC), digitalto-analogue converters (DAC), dispersion compensation and passive wavelength selective
devices. Reach extension technology as well as switching technologies are also discussed.
The main results for each considered system concept are summarized below:
-
Time division multiplexing (TDM) – passive optical networks (PON): Options based
on higher rate on-off keying as well as advanced modulation and coherent detection
are investigated. TDM-PON approaches present significant challenges with respect to
simultaneously increasing bandwidth, reach and splitting ratio, whilst maintaining low
cost and power consumption.
-
Active optical networks (AON): AON imply either large port count at the central
office in a centralized architecture or low degree of node consolidation in a
decentralized architecture. Both these factors have implications on cost and power
consumption.
-
WDM-PON: The WDM-PON concepts are here categorized based on optical network
unit (ONU) design, i.e. based on tunable lasers, reflective lasers or reflective lasers
employing wavelength reuse. Ultra-dense WDM-PON based on coherent detection is
also considered. Advantages of WDM-PON include long reach and large per customer
sustainable bandwidth. A drawback is the limited fan-out of traditional WDM-PON
approaches, limiting the number of customers per feeder fibre.
-
Orthogonal frequency division multiplexing (OFDM)-PON: OFDM is based on data
transmission over several densily spaced subcarriers. The OFDM format, with its
tolerance to chromatic dispersion, offers long reach and large resource flexibility.
Several variants of OFDM-PON are described. OFDM-PON presents similar
challenges as TDM-PON concerning limited total capacity and limited splitting ratio.
System complexity and processing requirements present challenges with respect to
power consumption and cost.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 6 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
-
Code Division Multiplexing (CDM): Different CDM implementation options and fanout possibilities are discussed with focus on coherent optical CDM using onedimensional code words. CDM presents significant challenges with respect to fan-out
requirements.
-
Radio-over-fibre technologies and digital interfaces for radio access network (RAN)
transport are also discussed considering their potential role for future RAN
architecture over the NGOA.
-
Hybrid concepts: Hybrid concepts are motivated by the fact that each of the
aforementioned pure system concepts individually may have difficulties in fulfilling
the complete set of NGOA requirements. Hence, hybrid concepts have been proposed
that combine advantages of different concepts. Typically the advantages that are
exploited are the increased overall capacity of WDM and the efficient resource sharing
of TDM, OFDM or CDM. Concepts that involve different types of active remote
nodes have also been considered. A preliminary cost and power consumption analysis
of different variants shows that the most promising configurations are hybrid
WDM/TDM-PON as well as various active hybrid variants (WDM-PON with AON
access and two stage WDM-PON).
As an outcome of the survey it is seen that the main candidates for further consideration
within the OASE project based on the posed requirements are different variants of pure
WDM-PON, hybrid WDM/TDM-PON, AON as well as various active hybrid variants based
on WDM-PON and AON.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 7 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Referred documents
Project Contract - Annex 1 “Description of the Work” (DoW)
OASE deliverable D2.1 – “Requirements”
Gigabit-capable passive optical networks (GPON), ITU-T G.984.x series of
recommendations [Online]. Available: http://www.itu.int/rec/T-REC-G/e
[4] IEEE Standard for Information technology-Specific requirements - Part 3 [Online],
IEEE Standard 802.3, 2008. Available: http://standards.ieee.org/getieee802/802.3.html
[5] 10-Gigabit-capable passive optical networks (XG-PON), ITU-T G.987.x series of
recommendations [Online]. Available: http://www.itu.int/rec/T-REC-G/e
[6] J. Kani and K. Suzuki, "Standardization Trends of Next-generation 10 Gigabit-class
Passive Optical Network Systems," NTT Technical Review, Vol. 7, No. 11, November
2009.
[7] Optical system design and engineering considerations, ITU-T Rec. G.Sup39 (12-2008).
[8] J. Yu, et al., “Applications of 40-Gb/s chirp-managed laser in access and metro
networks,” IEEE Journal of Lightwave. Technology, Vol. 27, No. 3, pp. 253–255,
February 2009.
[9] B. Mason, et al., “40Gb/s photonic integrated receiver with –. 17dBm sensitivity,”
OFC2002, March 2002, FB10.
[10] OFDM Introduction: Lecture Online, http://bwrc.eecs.berkeley.edu/classes/ee225c/
Lectures/Lec16_ofdm.pdf
[11] Characteristics of a single-mode optical fibre cable, ITU-T G.652 series of
recommendations (10-2000).
[12] I. Van de Voorde, C. Martin, J. Vandewege, and X.Z. Qiu, “The SuperPON
Demonstrator: an Exploration of Possible Evolution Paths for Optical Access
Networks”, IEEE Communications magazine, pp. 74-82, February 2000.
[1]
[2]
[3]
[13] C. Antony, et al., “Demonstration of a Carrier Distributed, 8192‐Split Hybrid DWDM
‐TDMA Pon Over 124km Field‐Installed Fibers” OFC 2010, San Diego, March
2010, PDPD8.
[14] F. Ponzini, et al., “Evolution scenario towards WDM-PON [Invited],” J. Opt. Commun.
Network, Vol. 1, No. 4, September 2009.
[15] K. Kanonakis, et al., “ACCORDANCE: A Novel OFDMA-PON Paradigm for UltraHigh Capacity Converged Wireline-Wireless Access Networks,” ICTON 2010, Munich
2010.
[16] N. Cvijetic, D. Qian, J. Hu, and T. Wang, "44-Gb/s/λ Upstream OFDMA-PON
Transmission with Polarization-Insensitive Source-Free ONUs," OFC2010, San Diego,
March 2010, OTuO2.
[17] W. Shieh, H. Bao, and Y. Tang, “Coherent optical OFDM: theory and design,” Optical
Express, Vol. 16, No. 2, p. 841, January 2008.
[18] B. J. C. Schmidt, A. J. Lowery, and J. Armstrong, “Experimental Demonstrations of
Electronic Dispersion Compensation for Long-Haul Transmission Using DirectDetection Optical OFDM,” IEEE J. Lightwave Technology, Vol. 26, No. 1, January
2008.
[19] D. Qian, N. Cvijetic, J. Hu, and T. Wang, “108 Gb/s OFDMA-PON with Polarization
Multiplexing and Direct Detection,” OFC 2009, San Diego, March 2009, PDPD5.
[20] T. N. Duong, et al., "Very High Bit Rate Transmission for NGPON using AMOOFDM
Direct Modulation of Linear Laser," OFC 2010, San Diego, March 2010, OTuO3.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 8 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[21] X. Wang, et al., “Experimental demonstration of 511-chip 640 Gchip/s superstructured
FBG for high performance optical code processing,” ECOC2004, Stockholm,
September 2004, Tu1.3.7.
[22] S. Etemad et al., “Spectrally efficient optical CDMA using coherent phase-frequency
coding,” IEEE Photonics Technology Letters, Vol. 17, No. 4, pp. 929–931, April 2005.
[23] Z. Jiang, et al., “Four-user 10-Gb/s spectrally phase-coded O-CDMA system operating
at ~30 fJ/bit,” IEEE Photonics Technology Letters, Vol. 17, No. 3, pp. 705–707, March
2005.
[24] B. Chen, et al., “Add/drop multiplexing and TDM signal transmission in an optical
CDMA ring network,” OSA J. Optical Networking, Vol. 6, pp. 969-974, 2007.
[25] S.P. Wan and Y. Hu, “Two-dimensional optical CDMA differential system with
Prime/OOC codes,” IEEE Photonics Technology Letters, Vol. 13, pp. 1373–1375, 2001.
[26] S.M.R.M. Nezam, et al., “Experimental Demonstration of OCDMA Transmission Using
a Three-Dimensional (Time-Wavelength-Polarization) Codeset”, IEEE J. Lightwave
Technology, Vol. 23, pp. 3282-3289, 2005.
[27] K. I. Kitayama, X. Wang, and N. Wada, “OCDMA Over WDM PON – Solution Path to
Gigabit-Symmetric FTTH”, IEEE/OSA J. Lightwave Technology, Vol. 24, pp. 16541662, 2006.
[28] K. Fouli, and M. Maier, “OCDMA and optical coding: principles, applications, and
challenges,” IEEE Commun. Mag., Vol. 45, pp. 27-34, 2007.
[29] X. Wang, et al., “Demonstration of Over 128Gb/s Capacity (12-User x 10.71 Gb/s/User)
Asynchronous OCDMA Using FEC and AWG-Based Multipoint Optical
Ecoder/Decoder,” IEEE Photonics Technology Letters, Vol. 18, pp. 1603-1605, 2006.
[30] N. Kataoka, N. Wada, G. Cincotti, K. Kitayama, ” Dispersion-Compensation-Free,
Long-Reach OCDMA-PON System with Passive Remote Node at Arbitrary Position
using Single Multi-port Encoder/Decoder”, CLEO, 2010.
[31] C. S. Brès, I. Glesk, R. J. Runser, and P. R. Prucnal, “All-optical OCDMA code-drop
unit for transparent ring networks,” IEEE Photonics Technology Letters Vol. 17, pp.
1088–1090, 2005.
[32] D. Wake. Radio over Fiber Systems for Mobile Applications. In: Radio over Fiber
Technologies for Mobile Communications Networks. Ed.: H. Al-Raweshidy, S.
Komaki. Artech, Boston, pp. 217-241, 2002.
[33] D. G. Moodie, et al., “Applications of electroabsorption modulators in high bit-rate
extended reach transmission systems,” OFC 2003, Invited Paper TuP1, pp. 267-268,
2003.
[34] H. Al-Raweshidy and S. Komaki (ed.), Radio over Fiber Technologies for Mobile
Communications Networks, Artech House 2002.
[35] Ken-Ichi Kitayama, et al., “An Approach to single component antenna base stations for
broadband millimeter-wave fiber-radio access systems,” IEEE Transactions on
Microwave Theory and Techniques, Vol. 48, No. 12, pp. 2588-2595, 2000.
[36] CPRI Specification V4.1 (2009-02-18): “Common Public Radio Interface (CPRI);
Interface Specification”, Online, http://www.cpri.info/
[37] IEEE Std 802.3-2005: "Part 3: Carrier sense multiple access with collision detection
(CSMA/CD) access method and physical layer specifications", December 2005.
[38] ISO/IEC 14165-115 – Information Technology – Fibre Channel Part 115 Physical
Interface (FC-PI), February 2006.
[39] INCITS (ANSI) Revision 8, T11/08-138v1– Fibre channel Physical Interface-4 (FC-PI4), May 2008.
[40] Infiniband Trade Associationm, “Infiniband Architecture”, Rel. 1.1, Vol. 2, November
2002.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 9 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[41] Open Base Station Architecture Initiative BTS System Reference Document (Version
2.0).
[42] L.G. Kazovsky, et al., “Next-Generation Optical Access Networks,” IEEE Journal of
Lightwave Technology, Vol. 25, No. 11, November 2007.
[43] C. Hoymann, “Analysis and performance evaluation of the OFDM-based metropolitan
area network IEEE 802.16,” Computer Networks (Elsevier), Vol. 49, Issue 3, October
2005.
[44] IEEE 802.11 WG, Part 11: Wireless LAN Medium Access Control (MAC) and Physical
Layer (PHY) Specification, 1999.
[45] Y. Xiao, “IEEE 802.11n: Enhancements for Higher Throughput in Wireless LANs,”
IEEE Wireless Communications, Vol. 12, Issue 6, pp. 82-91, December 2005.
[46] W.-T. Shaw, et al., “Hybrid Architecture and Integrated Routing in a Scalable OpticalWireless Access Network,” IEEE Journal of Lightwave Technology, Vol. 25, No. 11,
November 2007.
[47] S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid Wireless-Optical Broadband Access
Network (WOBAN): A Review of Relevant Challenges,” IEEE Journal of Lightwave.
Technology, Vol. 25, No. 11, November 2007.
[48] G. Shen, R.S. Tucker and C.-J. Chae, “Fixed Mobile Convergence Architectures for
Broadband Access: Integration of EPON and WiMAX,” IEEE Commun. Mag., Vol. 45,
Issue 8, pp. 44-55, August 2007.
[49] S.-J. Park, et al., “Fiber-to-the-Home Services Based on Wavelength-DivisionMultiplexing Passive Optical Network,” IEEE Journal of Lightwave Technology, Vol.
22, Issue 11, pp. 2582-2591, November 2004.
[50] J.-M. Kang, S.-K. Han, “A Novel Hybrid WDM/SCM-PON Sharing Wavelength for
Up- and Down-Link Using Reflective Semiconductor Optical Amplifier,” IEEE
Photonics Technology Letters, Vol. 18, No. 3, February 2006; E. Wong, K. L. Leem, T.
B. Anderson, “Directly Modulated Self-Seeding Reflective Semiconductor Optical
Amplifiers as Colorless Transmitters in Wavelength Division Multiplexed Passive
Optical Networks,” IEEE Journal of Lightwave Technology, Vol. 25, Issue 1, pp. 67-74,
2007.
[51] P. Healey, et al., “Spectral slicing WDM-PON using wavelength-seeded reflective
SOAs,” Electronics Letters, Vol. 37, No. 19, pp. 1181-1182, September 2001.
[52] Berrettini et al., “Testbed for experimental analysis on seamless evolution architectures
from GPON to high capacity WDM-PON,” Int. J. Communication Networks and
Distributed Systems, Vol. 5, No. 1-2, pp. 193-210, 2010.
[53] M.S. Lee, B-T Lee, B.Y. Yoon and B.T. Kim, “Bidirectional amplification with linear
optical amplifier in WDM-PON,” OFC2006, San Diego, March 2006.
[54] Ultra-wideband Radio over Optical Fibre (UROOF). FP6 European integrated project.
www.ist-uroof.org
[55] Fully-converged quintuple-play integrated optical-wireless access architectures
(FIVER). FP7 STREP project. www.ict-fiver.eu
[56] J. Buus, and E.J. Murphy, “Tunable Lasers in Optical Networks,” IEEE Journal of
Lightwave Technology, Vol. 24, No. 1, January 2006.
[57] L.A. Coldren, “Tunable Semiconductor Lasers – a tutorial (Invited),” OFC2003,
Atlanta,
March
2003,
ThF1,
Online:
http://www.ece.ucsb.edu/Faculty/Coldren/papers/TSL.pdf
[58] L.A. Coldren, et al., “Tunable Semiconductor Lasers: A Tutorial,” IEEE Journal of
Lightwave Technology, Vol. 22, No. 1, January 2004.
[59] S. Dahlfort, et al., “Tunable Transmitters and receivers for WDM PON,” FSAN NGPON2 Workshop, San Francisco, February 2010.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 10 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[60] G. Busico, “Introduction to DS-DBR,” Oclaro technical slides, 2009, Contact:
[email protected]
[61] A.J. Ward, et al., “Widely Tunable DS-DBR Laser With Monolithically Integrated
SOA: Design and Performance,” IEEE J. Selected Topics In Quantum Electronics, Vol.
11, No. 1, January/February 2005.
[62] Y. Liu, et al., “Directly-Modulated DS-DBR Tunable Laser for Uncooled C-band WDM
System,” OFC2006, Anaheim, March 2006.
[63] K. Grobe, and D. Nesset, “Low-cost Tunables for WDM-PON,” FSAN NG-PON2
Workshop, San Francisco, February 2010.
[64] M.J. Wale, et al., “Technology Routes for Cost-Effective, High Performance, NextGeneration Access Components,” EPIC Workshop, FTTH Council Europe,
Copenhagen, February 2009.
[65] S. Zou, et al., “100mW phase-shifted 1550nm BH DFB arrays with a 10-micron pitch,”
OFC2005, Anaheim, March 2005, JWA28.
[66] G.W. Yoffe, et al., “Widely-tunable 30mW laser source with sub-500kHz linewidth
using DFB array,” 21st LEOS Annual Meeting, Newport Beach, November 2008,
ThBB2.
[67] J.P. Wilde, G.W. Yoffe, and J.M. Kahn, “Frequency Noise Characterization of a Widely
Tunable Narrow-Linewidth DFB Laser Array Source,” NFOEC2009, San Diego, March
2009, JWA33
[68] ChemOptics, “Preliminary Technical Specifications Tunable Laser Module”, January
2009, Contact: 104-11 Moonji-Dong, Yusung-Gu, Deajeon 305-380, South Korea,
URL: www.chemoptics.co.kr, E-mail: [email protected]
[69] H. Rohde, “Questions for Tunable/selectable TRx technology - Capability of
Tunable/Selectable TxRx,” FSAN NG-PON2 Workshop, San Francisco, February 2010.
[70] M. Jiang, et al., “Error Free 2.5 Gb/s transmission over 125 km conventional fiber of a
directly modulated widely tunable vertical cavity surface emitting laser,” OFC2001,
Anaheim, March 2001, TuJ3-1.
[71] P. Debernardi et al. “Modal Properties of Long-Wavelength Tunable MEMS-VCSELs
With Curved Mirrors: Comparison of Experiment and Modeling,” IEEE Journal of
Quantum Electronnics, Vol. 44, No. 4, April 2008.
[72] C.C. Renaud, et al. “100 GHz Spaced 10 Gbit/s WDM over 10 °C to 70 °C using an
uncooled DBR laser,” OFC2006, Anaheim, March 2006, OWI76.
[73] D. Sadot, and E. Boimovich, “Tunable Optical Filters for Dense WDM Networks,”
IEEE Commun. Mag., Vol. 36, No. 12, December 1998.
[74] Hyung-Kew Lee, et al., “A Wide-Range Linearly Tunable Optical Filter Using Lorentz
Force,” IEEE Photonics Technology Letters, Vol. 16, No. 9, September 2004.
[75] Hailiang Zhang, et al. “Novel tunable liquid crystal Fabry-Perot filters for Fiber-Optical
system, in: Metro and Access Networks,” Wanyi Gu, Jianhui Zhou, Jin-Yi Pan, Editors,
Proceedings of SPIE, 64 Vol. 4583, 2001.
[76] D. Hohlfeld, et al. “Tunable Thermo-Optic Filter for WDM Applications,” Proc. IEEE,
pp. 564-567, August 2002.
[77] T. Nakazawa, et al. “Ti:LiNbO3 Acousto-Optic Tunable Filter (AOTF),” FUJITSU Sci.
Tech. J., Vol. 35, No. 1, pp. 107-112, July 1999.
[78] M. Ferchichi, et al. “Design of Temperature-Strain Tunable UDWDM, DWDM, WDM
FBG filter for Passive Optical Network Access,” 2nd ICTON - Mediterranean Winter
2008, Marrakech, Morocco, December 2008, Fr1A.5
[79] M. Lequime, “Tunable thin-film filters: review and perspectives,” Proceedings SPIE
Vol. 5250 (Advances in Optical Thin Films), February 2004.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 11 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[80] F. Havermeyer, et al. “Volume holographic grating-based continuously tunable optical
filter,” SPIE Journal Opt. Eng., Vol. 43, No. 9, September 2004.
[81] Fiber Fabry-Perot Tunable Filter | FFP-TF2, Data Sheet, Micron Optics, Inc. 2008,
Online:
http://www.micronoptics.com/uploads/library/documents/Datasheets/Micron%20Optics
%20FILTERS%20-%20FFP-TF2%20%20-%200812.1.pdf
[82] Thin-film Variable Filters, Data Sheet, Lightwaves2020, 2010, Online:
http://www.lightwaves2020.com/publications/lightvision/newsletter-Feb-2010-4p.pdf
[83] R. Murano, et al. “Tunable 2.5Gb/s Receiver for Wavelength-Agile DWDM-PON,”
OFC2008, San Diego, March 2008.
[84] R. Murano, et al. “Tunable GPON Receivers Enable Phased Migration to 1Gb/s per
Subscriber,” OFC2009, San Diego, March 2009.
[85] F. Riemenschneider, et al. “Low-Cost Electrothermally Tunable Optical Microcavities
Based on GaAs,” IEEE Photonics Technology Letters, Vol. 14, No. 11, November
2002.
[86] F. Derr, “Coherent Optical QPSK Intradyne System: Concept and Digital Receiver
Realization,” IEEE Journal of Lightwave Technology, Vol. 10, No. 9, September 1992.
[87] C.R.S. Fludger, et al. “10 x 111 Gbit/s, 50 GHz spaced, POLMUX-RZ-DQPSK
transmission over 2375 km employing coherent equalisation,” OFC2007, Anaheim,
March 2007, PDP22.
[88] C.R.S. Fludger, et al. “Coherent Equalization and POLMUX-RZ-DQPSK for Robust
100-GE Transmission,” IEEE Journal of Lightwave Technology, Vol. 26, No. 1,
January 2008.
[89] F. Gini, G. B. Giannakis, “Frequency Offset and Symbol Timing Recovery in FlatFading Channels: A Cyclostationary Approach,” IEEE Trans. Communications, Vol. 46,
No. 3, March 1998.
[90] F. M. Gardner, “A BPSK/QPSK timing-error detector for sampled receivers, IEEE
Trans. Communications,” Vol. COM-34, pp. 423-429, May 1986.
[91] M. Oerder, and H. Meyr, “Digital Filter and Square Timing Recovery,” IEEE Trans.
Commun., Vol. COM-36, pp. 605-612, May 1988.
[92] S.J. Savory, et al. “Digital Coherent Receivers for Uncompensated 42.8Gbit/s
Transmission over High PMD Fibre,” ECOC2007, Berlin, September 2007.
[93] S.J. Savory, et al. “Electronic compensation of chromatic dispersion using a digital
coherent receiver,” Opt. Express, Vol. 15, Issue 5, pp. 2120-2126, 2007.
[94] C.R. Johnson, et al. “Blind Equalization Using the Constant Modulus Criterion: A
Review,” Invited Paper, Proc. IEEE, Vol. 86, No. 10, October 1998.
[95] A. Vgenis, et al. “Nonsingular Constant Modulus Equalizer for PDM-QPSK Coherent
Optical Receivers,” IEEE Photonics Technology Letters, Vol. 22, No. 1, January 2010.
[96] B. Widrow and M. E. Hoff, “Adaptive Switching Circuits,” IRE WESCON Conv. Rec.,
Part 4, pp. 96–104, 1960.
[97] Guifang Li, “Recent advances in coherent optical communication,” Advances in Optics
and Photonics, Vol. 1, Issue 2, pp. 279–307, February 2009.
[98] Yan Han and Guifang Li, “Coherent optical communication using polarization multipleinput-multiple-output,” OSA Optics Express, Vol. 13, No. 19, pp. 7527-7534,
September 2005.
[99] A.J. Viterbi and A.N. Viterbi, “Nonlinear Estimation of PSK-Modulated Carrier Phase
with Application to Burst Digital Transmission,” IEEE Trans. Information Theory, Vol.
IT-29, No. 4, July 1983.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 12 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[100] M.G. Taylor, “Phase Estimation Methods for Optical Coherent Detection Using Digital
Signal Processing,” IEEE Journal of Lightwave Technologies, Vol. 27, No. 7, April
2009.
[101] M.G. Taylor, “Coherent Detection for Optical Communications using Digital Signal
Processing,” OFC2007, Anaheim, March 2007.
[102] N. Wiener, Extrapolation, Interpolation, and Smoothing of Stationary Time Series,
Wiley, New York 1949.
[103] Shaoliang Zhang, et al. “Novel Ultra Wide-Range Frequency Offset Estimation for
Digital Coherent Optical Receiver,” OFC2010 (EDC Session), San Diego, March 2010.
[104] T. Pfau, et al. “Coherent optical communication: Towards realtime systems at 40 Gbit/s
and beyond,” OSA Optics Express, Vol. 16, No. 2, pp. 866, January 2008.
[105] Ip, E., et al. “Coherent detection in optical fiber systems,” OSA Optics Express, Vol. 16,
No. 2, pp. 753, January 2008.
[106] D.S. Millar, et al. “Experimental Comparison of Nonlinear Compensation in Long-Haul
PDM-QPSK Transmission at 42.7 and 85.4 Gb/s,” ECOC2009, Vienna, September,
2009, Paper 9.4.4.
[107] P.J. Winzer, et al. “56-Gbaud PDM-QPSK: Coherent Detection and 2,500-km
Transmission,” ECOC2009, Vienna, September 2009, PD2.7.
[108] J.E. Simsarian, et al., “Fast-Tuning 224-Gb/s Intradyne Receiver for Optical Packet
Networks,” OFC2010, San Diego, March 2010, PDPB5.
[109] Akihide Sano, et al., “240-Gb/s Polarization-Multiplexed 64-QAM Modulation and
Blind Detection Using PLC-LN Hybrid Integrated Modulator and Digital Coherent
Receiver,” ECOC2009, Vienna, September 2009, PD2.2.
[110] J.M. Fabrega, and J. Prat, “New intradyne receiver with electronic-driven phase and
polarization diversity,” OFC2006, Anaheim, March 2006, JThB45.
[111] M. Noda, S. Yoshima, K. Ishii, S. Shirai, M. Nogami, J. Nakagawa, “Dual-rate Optical
Transceiver incorporating Fully Optimized Burst-mode AGC/ATC Functions for
10GEPON Systems” , ECOC 2010, Turin, September 2010, Mo.2.B.2
[112] Gigabit-Capable Passive Optical Networks (GPON): PMD Layer Specifications, ITU-T
Recommendation G.984.2, 2003.
[113] MAC Parameters, Physical Layers and Management Parameters for Subscriber Access
Networks, IEEE Draft P802.3ah, 2003.
[114] C. Su et al., “Theory of Burst-Mode Receivers and its Applications in Optical
Multiaccess Networks,” IEEE Journal of Lightwave Technology, Vol. 15, No. 4, pp.
590-606, 1997.
[115] N. Suzuki, et al.,“Single Platform 10G-EPON 10.3-Gbps/1.25-Gbps Dual-Rate CDR
with Fast Burst-Mode Lock Time Employing 82.5 GS/s Sampling IC and Bit-Rate
Adaptive Decision Logic Circuit,” ECOC 2010, Turin, September 2010, Mo.2.B.3
[116] J. Sugawa, H. Ikeda, D. Mashimo, “10.3Gbps burst-mode receiver capable of upstream
transmission with short overhead for 10G-EPON,” ECOC 2010, Turin, September 2010,
Mo.2.B.4
[117] M. Nakamura, et al., “A 10G burst-mode PIN-TIA module with 10 nsec response for
PON systems,” LEOS, 2007.
[118] A. Li, J. Faucher, and D. Plant, “Burst-Mode Clock and Data Recovery in Optical
Multiaccess Networks using Broad-Band PLLs,” IEEE Photonics Technology Letters,
Vol. 18, Issue 1, pp.73-75, 2006.
[119] T. Ajmal, et al., “Design of a 10Gbps Optical Burst Mode DPSK Receiver for Data and
Clock Recovery,” ECOC 2008, Brussels, September 2008, We.2.F.2.
[120] J. Kwon, et al. “AC-Coupled Burst-Mode OLT SFP Transceiver for Gigabit Ethernet
PON Systems,” IEEE Photonics Technology Letters, Vol. 17, No. 7, July 2005.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 13 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[121] Y. Ohtomo, et al., “A 10 Gb/s burst mode clock and data recovery IC with frequencyadjusting dual gated VCOs,” IEICE Trans. Electron., Vol. E91-C, No 6, 2008.
[122] P. Ossieur, et al., “A 1.25-Gb/s Burst-Mode Receiver for GPON Applications,” IEEE
Journal of Solid-State Circuits, Vol. 40, No. 5, May 2005.
[123] P. Ossieur, et al., “DC-coupled 1.25 Gbit/s burst-mode receiver with automatic offset
compensation,” Electronics Letters, Vol. 40, No. 7, April 2004.
[124] P. Ossieur et al., “A 10 Gb/s Burst-Mode Receiver with Automatic Reset Generation
and Burst Detection for Extended Reach PONs”, OFC2009, San Diego, March 2009,
OWH3.
[125] S Nishihara, et al., “10.3 Gbit/s burst-mode PIN-TIA module with high sensitivity, wide
dynamic range and quick response,” Electronics Letters, Vol. 44, No. 3, 2008.
[126] S. Nishihara, et al., “A burst-mode 3R receiver for 10-Gbit/s PON systems with high
sensitivity, wide dynamic range, and fast response,” Journal of Lightwave Technology,
Vol. 26, No. 1, 2008.
[127] E. Hugues-Salas, et al., “Fast edge-detection burst-mode 2.5Gbit/s receiver for gigabit
passive optical networks,” Journal of Optical Networking, Vol. 6, Issue 5, pp. 482-489,
2007.
[128] Han Sun, et al., “Challenges for VLSI implementation of 100G digital coherent
receivers,” OFC2009, March 2009, OTuN1.
[129] P. Schvan, et al., “A 24GS/s 6b ADC in 90nm CMOS,” ISSCC 2008, February 2008.
[130] J. Lee, et al., “High-Speed Analog-to-Digital Converters in SiGe Technologies”, CSIC
2007, October 2007.
[131] Hiroshi Onaka, “Challenges for VLSI implementation of 100G digital coherent
receivers,” ECOC 2009, Workshop on DSP & FEC, September 2009.
[132] Fujitsu Microelectronics Europe, “Fujitsu Launches Second Generation Ultra-fast
65GSa/s 8-Bit ADC Technology for 100G Optical Transport,” Online :
http://www.fujitsu.com/emea/news/pr/fseu-en_20100913-978.html
[133] I. Dedic, “56Gs/s ADC Enabling 100GbE,” OFC2010, San Diego, March 2010, OThT6.
[134] Maxtek, Online: http://www.maxtek.com/services6.html
[135] Micram Microelectronic GmbH, Online: http://www.micram.de
[136] SP Devices, Online : http://spdevices.com/index.php/products/high-speed-dataacquisition
[137] Mobius Semiconductor, Online: http://www.mobius-semiconductor.com/
[138] IHP, “ADCs and DACs: High-speed ADC and DAC components with 1-50 GS/s in
SiGe BiCMOS technology,” Online: http://www.ihp-microelectronics.com/174.0.html
[139] M. Grözing, et al., “High-Speed ADC Building Blocks in 90 nm CMOS”, Fourth Joint
Symposium on Opto- and Microelectronic Decices and Circuits (SODC 2006),
Duisburg, Germany, September 2006.
[140] National Semiconductor, ADC08B3000, Online:
http://www.national.com/pf/DC/ADC08B3000.html
[141] H. Bülow, F. Buchali and A. Klekamp, “Electronic Dispersion Compensation,” IEEE
IEEE Journal of Lightwave Technology, Vol. 26, Issue 1, pp 158-167, January 2008.
[142] A. Faerbert, et al., “Performance of a 10.7 Gb/s Receiver with Digital Equaliser using
Maximum Likelihood Sequence Estimation,” ECOC2004, Stockholm, PD TH415.
[143] T. Kupfer, “Measurement of the Performance of 16-states MLSE Digital Equalizer with
Different Optical Modulation Formats,” OFC2008, San Diego, February 2008, PDP13.
[144] S. J. Savory, et al., “Digital equalization of 40 Gbit/s perwavelength transmission over
2480 km of standard fibre without optical dispersion compensation,” ECOC2006,
Cannes, September 2006, Th2.5.5.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 14 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[145] L. Wosinski, M. Dainese, H. Fernando and T. Augustsson, “Grating-assisted add-drop
multiplexer realized in silica-on-silicon technology,” Proceedings of the Conference
“Photonics Fabrication Europe”, Brugge, Belgium, Proc. SPIE 4941, pp. 43-50,
October/November 2002.
[146] L. Wosinski, M. Swillo and M. Dainese, "Imprinting of low dispersion Bragg gratings
in planar devices for 40 Gbps DWDM systems," Proc. of International Congress on
Optics and Optoelectronics, Warsaw, Poland, Proc. SPIE 5956, pp. OD1 - OD8,
August/September 2005.
[147] N. Zhu, et al., “Experimental demonstration of a cross-order echelle grating triplexer
based on an amorphous silicon nanowire platform,” Optics Letters, Vol. 34, Issue 3, pp.
383-385, 2009.
[148] L. Wosinski, “Silica-on-Silicon Technology for Photonic Integrated Devices,” ICTON
2004, Wroclaw, Poland, Vol. 2, pp. 274 – 279, July 2004.
[149] Janz, S. et al., “Planar waveguide echelle gratings in silica-on-silicon,” IEEE Photonics
Technology Letters, Vol. 16, Issue 2, pp. 503–505, February 2004.
[150] D-link, Online: http://www.dlinkgreen.com
[151] Zhang Yuanwang, “100G Ethernet Technology and Applications”, ZTE
Communicatinos, Vol. 4, December 2009, Online: http://wwwen.zte.com.cn/endata/
magazine/ztecommunications/2009year/no4/articles/200912/t20091223_178948.html
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 15 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figures and tables
Figure 1 :
Figure 2 :
Figure 3 :
Figure 4 :
Figure 5 :
Figure 6 :
Figure 7 :
Figure 8 :
Figure 9 :
Figure 10 :
Figure 11 :
Figure 12 :
Figure 13 :
Figure 14 :
Figure 15 :
Figure 16 :
Figure 17 :
Figure 18 :
Figure 19 :
Figure 20 :
Figure 21 :
Figure 22 :
Figure 23 :
Figure 24 :
Figure 25 :
Figure 26 :
Figure 27 :
Figure 28 :
Figure 29 :
Figure 30 :
Figure 31 :
A generic and high level activity split between WP3 and WP4. The major areas
of work are in “Network Functions and Scope” and software (SW) and
hardware (HW) design implementation, simulation and emulation. ............... 30
NG-PON Roadmap .......................................................................................... 32
Systems concept tree ........................................................................................ 33
Wavelength plans [6] ....................................................................................... 37
Downstream serial 40G NRZ ........................................................................... 40
Downstream 40G QPSK .................................................................................. 41
Downstream 40G DQPSK ............................................................................... 42
Downstream stacked 4x10G NRZ.................................................................... 43
Generic WDM-PON with (fixed or tunable) laser array in OLT and tunable
lasers in ONUs ................................................................................................. 44
WDM-PON with tunable lasers in ONUs supporting power-split ODN ......... 45
Tuning of tunable ONU via closed-loop control incorporating the OLT and an
ECC .................................................................................................................. 46
Tuning of tunable ONU via closed-loop control incorporating a partial
reflector in the RN ............................................................................................ 46
Cyclic AWG ..................................................................................................... 47
DS and US frequency division at the ONU...................................................... 48
Laser-based WDM-PON with 25 GHz spacing for enhanced channel count .. 48
Laser-based WDM-PON with 25 GHz grid and S-band extension for very high
channel count.................................................................................................... 49
SSMF spectral attenuation. The dashed lines are the boundaries according to
ITU-T G.652A [11]. ......................................................................................... 50
Generic WDM-PON with seeded reflective (REAM) OLT transceiver array
and seeded reflective ONUs (based on RSOAs, REAMs, IL-FP lasers, or
combinations thereof, e.g., REAM-SOA) ........................................................ 51
WDM-PON with seeded reflective (REAM) OLT transceiver array and
tuneable ONUs. Redundant MFLs in the OLT are shared between several
PONs. ............................................................................................................... 51
Increase of spectral efficiency in WDM-PON with seeded ONUs through use
of constant-envelope FSK downstream modulation and intensity upstream
modulation ........................................................................................................ 52
Decrease of seed/upstream Rayleigh crosstalk through end-to-end dual-fibre
working............................................................................................................. 52
Decrease of Rayleigh crosstalk through use of dedicated upstream feeder fibre .
.......................................................................................................................... 53
Decrease of Rayleigh crosstalk through RN-based MFL Seed accommodation
.......................................................................................................................... 53
Super-PON [12] as an example for active-RN-based seed accommodation .... 53
DWDM-TDMA-PON [13] as an example for dedicated upstream feeder fibre ..
.......................................................................................................................... 54
ONU transmitter variants ................................................................................. 54
Simple ONU transmitter................................................................................... 55
FSK generation by means of a directly modulated laser .................................. 56
Frequency to Amplitude Modulation by means of an optical filter ................. 56
Figure RZ/IRZ remodulation scheme .............................................................. 56
OFDM-Transmission on multiple orthogonal subcarriers ............................... 58
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 16 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 32 :
Figure 33 :
Figure 34 :
Figure 35 :
Figure 36 :
Figure 37 :
Figure 38 :
Figure 39 :
Figure 40 :
Figure 41 :
Figure 42 :
Figure 43 :
Figure 44 :
Figure 45 :
Figure 46 :
Figure 47 :
Figure 48 :
Figure 49 :
Figure 50 :
Figure 51 :
Figure 52 :
Figure 53 :
Figure 54 :
Figure 55 :
Figure 56 :
Figure 57 :
Figure 58 :
Figure 59 :
Figure 60 :
Figure 61 :
Figure 62 :
Figure 63 :
Figure 64 :
Figure 65 :
Downstream CO-OFDM [17]: a) direct down/up conversion, b) intermediate
frequency .......................................................................................................... 59
Transmitter and receiver alternatives for DDO-OFDM [18]: a) b) and c) show
different transmitter alternatives and d) a typical receiver. .............................. 60
DDO-OFDMA-PON scheme after [16]. .......................................................... 61
OCDM PON with a power splitter at the remote node .................................... 63
A device with capability of simultaneously encoding/decoding multiple timespread optical codes (OCs) [29]. ...................................................................... 63
(a) Configuration of OCDM-PON system with code splitter at RN and (b)
experimental setups and results: (i) downlink and (ii) uplink [30]. ................. 64
Schematic of the all-optical 2-D OCDM code-drop unit [31] (D: Delay, HPF:
High-pass filter, and TOAD: terahertz optical asymmetric demultiplexer). .... 65
Structure of OCDMA add-drop multiplexer for any coding technology [24]
(TOAD: terahertz optical asymmetric demultiplexer). .................................... 66
Hybrid PON basic setup ................................................................................... 67
Hybrid WDM/TDM-PON ................................................................................ 67
WDM/TDM-PON based on an ONU with one tunable transmitter and one
tunable optical filter, for a general remote node .............................................. 69
WDM/TDM-PON as in Figure 42 with cascaded power splitters for the remote
node. ................................................................................................................. 70
WDM/TDM PON as in Figure 42 with a WDM splitter and power splitter for
the remote node. ............................................................................................... 70
WDM/TDM-PON as in Figure 42 with a wavelength router and power splitter
for the remote node. ......................................................................................... 71
WDM/TDM-PON as in Figure 42 with a wavelength selective switch and
power splitter for the remote node. .................................................................. 72
Hybrid WDM/OCDM PON [27] (OC: optical code, TX: transmitter and RX:
receiver). ........................................................................................................... 73
Hybrid WDM/E-CDM-PON ............................................................................ 74
Hybrid WDM/O-CDM-PON............................................................................ 75
An example of wavelength channel allocation of OCDMA over CWDM using
511-chip SSFBG en/decoder [27]. ................................................................... 76
OCDMA over WDM or hybrid WDM/TDM PONs ........................................ 77
Hybrid WDM/OCDM PON with code splitting at the remote node. ............... 77
Hybrid WDM/O-OFDM-PON ......................................................................... 78
Hybrid WDM/SCM-PON ................................................................................ 79
UDWDM-PON................................................................................................. 80
Hybrid active WDM-PON with active P2P access .......................................... 81
Hybrid active WDM-PON (“PON-in-PON”) .................................................. 82
Reference WDM-PON ..................................................................................... 83
2.5Gb/s WDM-PON employing RSOA-based ONUs and NRZ signaling ...... 96
BER versus received power with multiwavelength source placed at the OLT
(a) and the RN (b)............................................................................................. 96
BER measurements versus received power with multi-channel near RN ........ 97
2.5Gb/s WDM-PON employing RSOA-based ONUs and RZ signaling ......... 97
BER performance for different DS extinction ratio values of 3dB (a) and 5dB
(b) ..................................................................................................................... 98
10Gb/s WDM-PON employing REAM-based ONUs ..................................... 99
US and DS BER performance for REAM WDM-PON ................................... 99
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 17 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 66 :
Figure 67 :
Figure 68 :
Figure 69 :
Figure 70 :
Figure 71 :
Figure 72 :
Figure 73 :
Figure 74 :
Figure 75 :
Figure 76 :
Figure 77 :
Figure 78 :
Figure 79 :
Figure 80 :
Figure 81 :
Figure 82 :
Figure 83 :
Figure 84 :
Figure 85 :
Figure 86 :
Figure 87 :
Figure 88 :
Figure 89 :
Figure 90 :
Figure 91 :
Figure 92 :
Figure 93 :
Figure 94 :
Figure 95 :
Figure 96 :
Figure 97 :
Figure 98 :
Figure 99 :
Figure 100 :
Figure 101 :
Figure 102 :
Figure 103 :
Figure 104 :
Figure 105 :
Figure 106 :
Cut-down (single-user) WDM-PON highlighting REAT-based ONU at CPE.
100
3-section DBR laser ....................................................................................... 103
5-section DS-DBR laser ................................................................................. 103
DS-DBR rear section wavelength comb (measured) ..................................... 104
DS-DBR laser pseudo-wavelength map ......................................................... 104
DS-DBR lasing spectrum – with side-super--modes ..................................... 105
Side mode suppression ratio ........................................................................... 105
DS-DBR reverse-biased SOA as a shutter ..................................................... 106
DS-DBR laser RIN (a), and line width (b) ..................................................... 107
DS-DBR laser temperature-depending output power (as of today) ............... 108
Cross-sectional diagram of a MEM-VCSEL ................................................. 109
VCSEL tuning. (B) shows tuning of an SCC type VCSEL, (A) and (C) for EC
types. .............................................................................................................. 110
Schematic diagram of an ECL ....................................................................... 110
Principle of an FBG-ECL (left), and FBG-ECL from Xponent Photonics, Inc.
(right) .............................................................................................................. 111
Packaging and pin configuration of commercial ChemOptics FBG-ECL ..... 111
Photo of DFB array chip ................................................................................ 112
Photo of DFB array chip [66] ......................................................................... 113
Schematic of tunable TFF with active substrate ............................................ 116
Schematic of tunable TFF for GPON WDM overlay [83] ............................. 116
Packaging of tunable TFF for GPON WDM overlay .................................... 117
Schematic of liquid-crystal filter .................................................................... 118
Schematic of (tunable) FBG ........................................................................... 118
Reflection peak of an FBG ............................................................................. 119
Schematic of tunable Volume Holographic Grating (VHG) .......................... 120
Schematic of MEM-tunable filter .................................................................. 120
Schematic of tunable microcavities................................................................ 121
Coherent detection (basic scheme) ................................................................. 123
Coherent intradyne DP-QPSK transmission .................................................. 124
Two realizations of 90° hybrids in polarization-diverse coherent receivers .. 124
Coherent intradyne receiver: digital realization ............................................. 125
Example of simple FIR filter response (blue: Re, red: Im part) ..................... 125
Phase estimation using Wiener filter .............................................................. 127
Phase estimation in presence of phase noise and combined phase plus
amplitude noise .............................................................................................. 127
Cascaded FOE [103] ...................................................................................... 128
Linear-only or linear plus nonlinear compensation [106] .............................. 129
Homodyne detection without polarization diversity and 90° hybrids ............ 130
Reduced homodyne detector .......................................................................... 131
Digital implementation [110] ......................................................................... 132
Polarization-scrambled heterodyne detection ................................................ 132
Typical Passive Optical Network (PON) scheme with a BMR located in the
OLT ................................................................................................................ 133
a) mass distribution function of differentiated signal level sampled
immediately after the start of each bit b) mass distribution function of
differentiated signal level sampled after passing MLEPW(set to 0.05 of bit
length) after the start of each bit. SNR for both scenarios was set to 9 dB. ... 136
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 18 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 107 :
Figure 108 :
Figure 109 :
Figure 110 :
Figure 111 :
Figure 112 :
Figure 113 :
Figure 114 :
Figure 115 :
Figure 116 :
Figure 117 :
Figure 118 :
Figure 119 :
Figure 120 :
Figure 121 :
Figure 122 :
Figure 123 :
Figure 124 :
Figure 125 :
Figure 126 :
Figure 127 :
Figure 128 :
Figure 129 :
Figure 130 :
Figure 131 :
Figure 132 :
Mass distribution function of differentiated signal level sampled immediately
after the start of each bit. SNR= 7 dB. ........................................................... 137
BER versus the SNR for different values of MLEPW. .................................. 138
a) the differentiated pulse with very large value of RC constant for
differentiator b) the differentiated pulse with small value of RC constant for
differentiator. .................................................................................................. 139
BER versus SNR for a pseudo-random sequence and a sequence of “0 1 0
1…”. ............................................................................................................... 140
The figure shows a set of possible modulation formats for 100 Gb/s systems
and their requirements in terms of ADC performance (green circles) [128].
Also reported is the performance of bipolar and CMOS ADCs from
publications [129], [130]. ............................................................................... 141
Multi-chip (upper figure) and monolithic (lower figure) configuration of
ADC/DSP block in a 100 Gb/s receiver [131]. .............................................. 142
Overview of dispersion compensation methods............................................. 143
Bragg grating assisted optical Add-Drop Multiplexer based on 2x2 MMI
coupler. 30dB extinction ratio between drop and transmitted channels and 3dB
excess loss in the dropped channel have been obtained. ................................ 146
Add-Drop Multiplexer based on balanced Mach-Zehnder interferometer.
Obtained crosstalk -25dB, insertion loss -3dB, switching time 2ms and power
consumption 0.5W have been achieved. ........................................................ 146
Etched diffraction grating demultiplexer: signal from input waveguide with
wavelengths λ1 ,λ2 ,λ3,… is diffracted by the planar concave grating, and
refocused into different output waveguides. .................................................. 147
Schematic of an arrayed waveguide grating: (a) input-output waveguides are
coupled through two slab (FPR) waveguides and an array of curved
waveguides, (b) tilted phase front with an angle , focused at an output
waveguide in the output FPR. ........................................................................ 148
Series of 32 x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm
channel spacing (100 GHz), 25 nm band: a) Fabricated devices, b) AWG
transmission spectrum. ................................................................................... 148
Example GPON OEO..................................................................................... 150
RPT approach ................................................................................................. 151
Comparison of different reach extender technologies for 10G TDM-PON (XGPON): OA based reach extender (L-band EDFA for the 1577 nm downstream
+ SOA for the 1270 nm upstream), O/E/O based reach extender and Remote
Protocol Terminator (RPT)/ mini-OLT approach. ......................................... 152
RE variants ..................................................................................................... 153
Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels. ................. 161
Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels, low loss
series. .............................................................................................................. 162
Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 100GHz ... 162
Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 200GHz ... 163
Oplink’s FTTX triplexer 1310/1490/1550 WDM (1x2) ................................ 163
JDSU Coarse Wavelength Division Multiplexer 4/8 channels ...................... 164
JDSU Dense Wavelength Division Multiplexer 4/8 channels ....................... 164
Oplink’s Dense Wavelength Division Multiplexer 100 GHz. ....................... 165
Oplink’s Dense Wavelength Division Multiplexer 50 GHz. ......................... 166
JDSU Dense Wavelength Division Multiplexer 100 GHz, Narrowband
(Gaussian)....................................................................................................... 166
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 19 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 133 :
Figure 134 :
Figure 135 :
Table 1 :
Table 2 :
Table 3 :
Table 4 :
Table 5 :
Table 6 :
Table 7 :
Table 8.
Table 9 :
Table 10 :
Table 11 :
Table 12 :
Table 13.
Table 14 :
Table 15 :
Table 16 :
Table 17 :
Table 18.
Table 19 :
Table 20.
Table 21 :
Table 22 :
Table 23.
Table 24.
Table 25 :
JDSU Dense Wavelength Division Multiplexer 100 GHz, Wideband (Flat
Top). ............................................................................................................... 167
JDSU Dense Wavelength Division Multiplexer 50 GHz, Wideband (Flat Top).
........................................................................................................................ 167
JDSU Dual Duplexer 1310/1550.................................................................... 168
Key system aspects............................................................................................... 34
Key component/subsystem aspects ...................................................................... 35
Main TDM-PON standards .................................................................................. 38
40G-PON variants for the DS .............................................................................. 39
40G-PON variants for the US .............................................................................. 39
Encoder/decoder technologies.............................................................................. 62
Energy consumption and cost figures for hybrid WDM/TDM-PON................... 68
Energy consumption and cost figures for hybrid WDM/E-CDM-PON............... 74
Energy consumption and cost figures for hybrid WDM/O-CDM-PON .............. 75
Energy consumption and cost figures for hybrid WDM/O-OFDM-PON ............ 78
Energy consumption and cost figures for hybrid WDM/SCMA-PON ................ 79
Energy consumption and cost figures for WDM/UDWDM-PON ....................... 81
Energy consumption and cost figures for hybrid active WDM-PON with active
P2P access ............................................................................................................ 82
Energy consumption and cost figures for hybrid active WDM-PON .................. 82
Energy consumption and cost figures for reference WDM-PON ........................ 84
Selection of Ethernet PHY standards ................................................................... 85
Comparison of relevant parameters of tunable lasers ........................................ 114
Comparison of relevant parameters of tunable filters ........................................ 121
Linewidth requirements for single-polarization modulation at a target BER of
10−3 ..................................................................................................................... 129
Comparison of coherent WDM-PON schemes .................................................. 132
Table summarizing state-of-the-art DACs available. ......................................... 142
Example of supported GPON configurations ..................................................... 150
System concept comparison table ...................................................................... 157
Energy consumption and cost figures of relevant hybrid PON components ..... 160
Relevant optical parameters for hybrid PON performance analysis .................. 160
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 20 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Abbreviations
3R
Reamplification, Reshaping, and Retiming
ADC
Analogue-to-Digital Converter
AFC
Automatic Frequency Control
AM
Amplitude Modulation
AMO-OFDM Adaptively Modulated OFDM
AON
Active Optical Network
AOTF
Acousto-Optic Tunable Filter
APD
Avalanche Photodiode
AR
Anti-Reflection
ARN
Active RN
ASE
Amplified Spontanous Emission
ATM
Asynchronous Transfer Mode
AWG
Arrayed Waveguide Grating
AWGN
Additive White Gaussian Noise
BER
Bit Error Rate
BMR
Burst Mode Receiver
BPF
Band Pass Filter
BPSK
Binary PSK
BS
Base Station
BtB
Back to Back
BTJ
Buried Tunnel Junction
BW
Bandwidth
CD
Chromatic Dispersion
CDM
Code Division Multiplexing
CDMA
Code Division Mutiple Access
CDR
Clock Data Recovery
CFO
Carrier Frequency Offset
CID
Consecutive Identical Digits
CMA
Constant-Modulus Algorithm
CML
Chirp Managed Laser
CO
Central Office
CO-OFDM
Coherent Optical OFDM
CP
Cyclic Prefix
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 21 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
CPE
Customer Premisses Equipment
CPRI
Common Public Radio Interface
CPU
Central Processing Unit
CR
Code Restorer
CRAN
Cloud RAN
CW
Continous Wave
CWDM
Coarse Wavelength Division Multiplexing
DAC
Digital to Analogue Converter
DBA
Dynamic Bandwidth Allocation
DBR
Distributed Bragg Reflector
DCD
Drop Code Decoder
DCF
Dispersion Compensation Fibre
DCU
Dispersion Compensation Unit
DD
Direct Detection
DDO-OFDM Direct Detection Optical OFDM
DFB
Distributed Feedback
DFE
Decision Feedback Equalizer
DIY
Do It Yourself
DL
Downlink
DML
Directly Modulated Laser
DP-QPSK
Dual-Polarized QPSK
DPSK
Differential PSK
DQPSK
Differential Quaternary Phase-Shift Keying
DS
Downstream
DSCA
Dynamic Sub-Carrier Allocation
DS-DBR
Digital Supermode DBR
DSP
Digital Signal Processing
DWDM
Dense Wavelength Division Multiplexing
EAM
Electroabsorption Modulator
EAT
Electroabsorption Transceiver
EC
Extended Cavity
ECC
Embedded Communication Channel
E-CDMA
Electrical CDMA
ECL
External Cavity Laser
E/D
Encoder/Decoder
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 22 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
EDC
Electronic Dispersion Control
EDG
Etched Diffraction Grating
EDGE
Enhanced Data rates for GSM Evolution
EDFA
Erbium-Doped Fibre Amplifier
EG
Echelle grating
EML
Electroabsorption Modulated Laser
EONT
Embedded ONT
EOTF
Electro-Optical Tunable Filter
EPD
Electronic Pre-Distortion
EPON
Ethernet Passive Optical Network
EQ
Equalizer
E-UTRA
Evolved Universal Terrestrial Radio Access
FBG
Fibre Bragg Grating
FBG-ECL
Fibre-Bragg-Grating External-Cavity Laser
FDD
Frequency Division Duplex
FDL
Fibre Delay Line
FE
Fast Ethernet
FEC
Forward Error Correction
FF
Feeder Fibre
FFE
Feed Forward Equalizer
ffs
For Further Study
FFT
Fast Fourier Transform
FIR
Finite-Impulse Response
FM
Frequency Modulation
FOE
Frequency Offset Estimator
FP
Fabry-Perot
FP-LD
Fabry-Perot LD
FPR
Free Propagation Regions
FSAN
Full Service Access Network
FSK
Frequency Shift Keying
FTTB
Fibre-to-the-Building
FTTC
Fibre-to-the-Curb
FTTH
Fibre-to-the-Home
FTTx
Fibre-to-the-x
GbE
Gigabit Ethernet
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 23 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
GCSR
Grating-assisted Co-directional Coupler Laser with Sampled Grating Reflector
GPON
Gigabit-capable Passive Optical Network
GPRS
General Packet RAdio Service
GSM
Global System for Mobile Communications
HFC
Hybrid Fibre-Coaxial
HPF
High Pass Filter
HSPA
High Speed Packet Access
HW
Hardware
IAD
Integrated Access Device
ICI
Inter-Carrier Interference
IEEE
Institute of Electrical and Electronics Engineers
IF
Intermediate Frequency
IFFT
Inverse Fast Fourier Transform
IL
Interleaver
IM
Intensity Modulation
IRZ
Inverse Return to Zero
ISI
Inter-Symbol Interference
LAN
Local Area Network
LCFP
Liquid Crystals FP
LD
Laser Diode
LED
Light Emitting Diode
LMS
Least Mean-Square
LO
Local Oscillator
LTE
Long Term Evolution
LX
Local Exchange
MAC
Media Access Control
MAP
Maximum a Posteriori
MBE
Molecular Beam Epitaxy
MDXM
DWDM Multiplexer/Demultiplexer
MEMS
Micro-Electro-Mechanical Systems
MFL
Multi Frequency Laser
MG-SGC
Matrix-Grating Strongly Gain-Coupled
MIMO
Multiple Input Multiple Output
MMF
Multi Mode Fibre
MMI
Multimode Interference
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 24 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
MMLEPW
Minimum Latch Enable Pulse Width
MMSE
Minimum Mean-Square Error
MQW
Multi-Quantum Well
MSA
Multi Source Agreement
MZ
Mach Zehnder
MZI
MZ Interferometer
MZM
MZ Modulator
NGA
Next Generation Access
NGOA
Next Generation Optical Access
NLC
Nonlinear Compensation
NRZ
Non-Return to Zero
OA
Optical Amplifier
OADM
Optical Add-Drop Multiplexer
OBPF
Optical Bandpass Filter
OBSAI
Open Base Station Architecture Initiative
OC
Optical Code
O-CDMA
Optical CDMA
ODN
Optical Distribution Network
ODL
Optical Delay Line
OE
Optical-Electrical
OEO
Optical-Electrical-Optical
OFDM
Orthogonal Frequency Division Multiplexing
OFDMA
Orthogonal Frequency Division Multiple Access
OLT
Optical Line Terminal
ONT
Optical Network Terminal
ONU
Optical Network Unit
OOC
Optical Orthogonal Codes
OOK
On Off Keying
OPLL
Optical Phase-Locked Loop
OSNR
Optical Signal-to-Noise Ratio
OSSB
Optical Single Sideband
OTN
Optical Transport Network
OXC
Optical Cross-Connect
P2P
Point-to-Point
PAPR
Peak-to-Average Power Ratio
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 25 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
PBC
Polarization Beam Combiner
PBS
Polarization Beam Splitter
PC
Polarization Controller
PCG
Planar Concave Grating
PD
Photo Diode
PDG
Polarization Dependent Gain
PE
Phase Estimation
PHY
Physical Layer
PIC
Photonic Integrated Circuit
PIN
p-i-n receiver
PLC
Planar Lightwave Circuit
PoP
Point Of Presence
PMD
Physical Medium Dependent
PMD
Polarization Mode Dispersion
PON
Passive Optical Network
PRBS
Pseudo Random Bit Sequence
PRN
Passive RN
P/S
Parallel to Serial
PSK
Phase Shift Keying
QAM
Quadrature Amplitude Modulation
QCSE
Quantum Confined Stark Effect
QPSK
Quadrature Phase Shift Keying
QoS
Quality of Service
RAN
Radio Access Network
RAU
Radio Access Unit
RBS
Radio Base Station
REC
Radio Equipment Controllers
RE
Radio Equipment
REAM
Reflective EAM
REAT
Reflective EAT
RF
Radio Frequency
RIN
Relative Intensity Noise
RN
Remote Node
ROADM
Reconfigurable Optical Add-Drop Multiplexer
RoF
Radio over Fibre
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 26 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
RPT
Remote Protocol Terminator
RPR
Resilient Packet Ring
RSOA
Reflective SOA
Rx
Receiver
RXA
Receiver Array
RZ
Return to Zero
SCC
Semiconductor-Coupled Cavity
SCM
Sub-Carrier Multiplexing
SCMA
Sub-Carrier Multiple Access
SFI
System Framer Interface
SG-DBR
Sampled Grating DBR
SOA
Semiconductor Optical Amplifier
SFP
Small Form-factor Pluggable
SFF
Small Form Factor
SFW
Single Fibre Working
SLPM
Spatial Light Phase Modulator
SMF
Single Mode Fibre
SMSR
Side-Mode Suppression Ratio
SNR
Signal to Noise Ratio
S/P
Serial to Parallel
SSFBG
Super Structured FBG
SSG-DBR
Super-Structure Grating DBR
SSMF
Standard Single Mode Fibre
SW
Software
TDM
Time Division Multiplexing
TDMA
Time Division Multiple Access
TEC
Thermo-Electrical Cooler
TF
Tunable Filter
TFF
Thin-Film Filter
TOAD
Terahertz Optical Asymmetric Demultiplexer
TRx
Transceiver
TS-OCDMA Time Spreading OCDMA
TW
Travelling Wave
Tx
Transmitter
TXA
Transmitter Array
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 27 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
TXFP
Tunable XFP
UDWDM
Ultra Dense WDM
UL
Uplink
UMTS
Universal Mobile Telecommunications System
US
Upstream
UTRA
Universal Terrestrial Radio Access
VCSEL
Vertical Cavity Surface Emitting Laser
VHG
Volume Holographic Grating
VMZ
Vertical Mach-Zehnder
WAN
Wide Area Network
WDM
Wavelength Division Multiplexing
WDMA
Wavelength Division Multiple Access
WL
Wavelength
WSS
Wavelength Selective Switch
XFP
10 Gigabit Small Form Factor Pluggable
XG-PON
10 Gigabit-capable PON
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 28 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
1. Introduction
The aim of OASE is the development of Next Generation Optical Access (NGOA)
architecture and system concepts for the “2020” time horizon, based on European
requirements [1]. The focus is on fibre-to-the-home (FTTH) network scenarios. The purpose
of this deliverable is to provide an overview of system concepts/technologies with potential of
meeting the requirements of next-generation optical access (NGOA) as defined in Deliverable
2.1 [2]. Key requirements include residential peak data rates of ≥1 Gbit/s, support for 256 to
1024 customers per feeder fibre, support for 20 to 40 km passive reach, and support for 60 to
90 km extended reach for the protection path.
This document presents a wide survey of proposed system concepts with potential of meeting
these requirements. For each of these next-generation systems, a detailed technical description
is made as well as a coarse investigation considering key aspects such as bandwidth, number
of customers per feeder fibre and reach. In the continuation of the project a reduced number of
concepts will be selected for more detailed techno-economic assessment. WP4 will take input
from WP2 “General Requirements” and WP3 “Architectures” which provide boundary
conditions for work in WP4 with respect to parameters such as:







Power splitter, wavelength splitter configurations
Upstream/downstream bandwidth per subscriber (peak/average),
Number of users per CO (feeder) fibre,
Total fibre distance from subscriber to CO,
Number and possible locations of active equipment in the field
Resilience requirements (availabilities, switch-over times)
System design for minimized OpEx in telecom processes
Other system parameters, such as complexity (which has a strong bearing on the cost analysis
of WP5), and power consumption, will subsequently be analyzed in the assessment of
different system concepts.
2. Task description and scope
This document, D4.1, is the first deliverable of OASE WP4 “System concepts for nextgeneration optical access networks” and is the end-result of task T4.1 as defined in [1]. Task
T4.1 is composed of the following activities:
-
A4.1.1 Survey of current concepts for next-generation optical access
A4.1.2 Survey of evolving components/subsystems
A4.1.1 includes a wide survey of possible next-generation technologies (from conferences,
journal papers, vendors, operators, tradeshows, standardization bodies, etc). For each of the
next-generation technologies, a detailed technical description is made. The technologies are
coarsely investigated/ assessed by filling in a table with all the aforementioned aspects. Out of
the survey of A4.1.1, key evolving components/subsystems are identified and a survey of
possible next-generation technologies is made in A4.1.2 (starting from conferences, journal
papers, vendors, operators, tradeshows, standardization bodies, etc).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 29 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
2.1 RELATIONSHIP TO OTHER WORKPACKAGES
This section aims to clarify the scope of WP4 “System concepts for next-generation optical
access networks” and in particular to define the boundary to WP3 “Architectures”. Although
there is a natural overlap between these two packages the following section aims to define
areas of responsibility for the different work packages.
WP3 focuses on network architectural work that meets the requirements of WP2, as well
as protocol design and non-protocol based network element functionality activities. WP3
includes simulation based work that can be used as a hardware and software design input.
WP4 has an implementation focus. Both hardware design and implementation work will be
part of WP4, while only the implementation part of WP3 designed software will be done in
WP4. This will be transferred into WP4 in task T4.4. WP4 will perform activities in data,
control and management plane software implementation. This implementation can be used in
order to control an emulated or physical data plane system solution.
Figure 1 shows a high level view of the activity split between WP3 and WP4, which allows
for additional input from WP2 on requirements. The figure also states that in both WPs there
can exist activities that focus on a per network as well as a per network element level, i.e. both
WPs can have an end-to-end network scope or “just” a node local scope.
Figure 1 :
A generic and high level activity split between WP3 and WP4. The major areas of work are in
“Network Functions and Scope” and software (SW) and hardware (HW) design
implementation, simulation and emulation.
Figure 1 shows a number of areas in which work will be done in WP3. The “Network
Functions and Scope” area is the main area which basically has the goal of creating the OASE
functional network architecture. I.e. high level requirements supplied from WP2 should be
transformed into corresponding functions needed in different geographical locations and
functional planes of the network elements of the OASE “scoped” network. E.g. if the high
level requirement is “network resilience” then there is need of control plane functions that
disseminates information of resilience node capabilities which allows for path calculation that
considers e.g. protective paths. The control plane should include the possibility to signal this
type of information related to resource allocation. In the management plane the corresponding
functions could be fault identification and localization, which can trigger protection or
restoration actions. All of this depends, as well, on the type of data plane technology and the
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 30 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
network topology of this part of the network.
The other two areas included in WP3 are SW design and simulation. The design part is mainly
work focused on the design of protocols and other software functions that are needed in order
to realize the functions defined in the network architecture, which is created in order to meet
the requirements of WP2. In order to make efficient designs, simulations might be a useful
tool but it can also be used for simulation activities on dynamic bandwidth allocation or how
locality enabled peer-to-peer protocols will influence the access and distribution part of the
network.
Figure 1 shows a number of areas in which work will be done in WP4, as well. The output
from WP2 and WP3 into WP4 is a set of functions that meets a number of requirements. The
work of WP4 is to realize this input through the design and implementation of hardware and
software. As a research and support tool for this realization of the OASE functional network
architecture, hardware simulation or modelling and software emulation can be used. The
output of WP4 is to be used in WP7.
In all of the above interactions between the WPs there are of course feedback loops that
allows for e.g. re-design, re-implementation, added network functions etc.
3. Assumptions and methodology
This chapter outlines the assumptions and methodology for the survey of system concepts.
Section 3.1 describes the rational behind the structure of the survey. Section 3.2 presents an
overview of the relevant assessment parameters and section 3.3 provides a description of the
assumptions for cost and power consumption used in the deliverable and in the continuation
of the OASE project.
3.1 NGOA SYSTEM CONCEPTS
The requirements for the OASE NGOA are defined in Deliverable 2.1 [2]. Two important
trends are reflected in these requirements, i.e. the projected increase in residential and
backhaul bandwidth requirements and the requirements resulting from an operator desire to
reduce number of central office sites. Some of the main technical requirements from [2] are
listed below:
-
FTTH residential peak data rates ≥1 Gb/s
-
Business, backhaul (fixed, mobile) peak date rate: ≥10 Gb/s
-
Average sustainable downstream based on peak-hour service usage of 500 Mb/s per
Optical Network Unit (ONU)/customer
-
Support of more traffic symmetry, with ratio of at least 1:2 between upstream and
downstream
-
Support from 256 to 1024 ONUs/customers per feeder fibre
-
Support of 128 Gbit/s to 500 Gb/s aggregate capacity per feeder fibre
-
Support of 20 to 40 km passive reach option for the working path, depending on the
degree of node consolidation
-
Support of 60 to 90 km extended reach option for the protection path, depending on
the degree of node consolidation
-
Legacy ODN compatibility desirable
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 31 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
NGOA system candidates must therefore support increased customer bandwidth, increased
number of customers and increased reach. Candidates should additionally support smooth
migration from today’s FTTH architectures to future architectures.
In ITU-T and FSAN, work is centred along two main tracks for Next Generation PON (NGPON), one mid-term track (NG-PON1) and one long-term track (NG-PON2):


NG-PON1: NG-PON coexisting on same Optical Distribution Network (ODN) as
GPON based on G.984.5 approach
NG-PON2: “disruptive” NG-PON with no requirement to coexist on same fibres as
GPON
Figure 2 shows the FSAN roadmap for NG–PON1 and NG–PON2, where NG–PON1 is
viewed as a mid-term upgrade and NG-PON2 as a longer-term solution. The timeline is
shown to reflect the nominal expected period for specification and publication of standards for
NG-PON1 (2009-2012) and NG-PON2 (2013-2015).
NG-PON2
Capacity
Component R&D to enable NG-PON2
“Co-existence”
enables gradual
migration in the
same ODN.
G-PON
E.g. Higher-rate TDM,
DWDM, CDM,
OFDM, etc.
NG-PON1 incl.
long-reach option
WDM option to
enable to overlay
multiple XGPONs
XG-PON
(Up: 2.5G to 10G,
Down: 10G)
1G-EPON
Splitter for NG-PON2
(power splitter or
something new)
Power splitter deployed for Giga PON
(no replacement / no addition)
Now
~2010
Figure 2 :
~2015
NG-PON Roadmap
For the OASE NGOA time horizon, emerging and disruptive technologies can be considered
in order to meet the requirements of increased capacity, extended reach and increased number
of subscribers per feeder fibre. Increased system capacity can be achieved for example by
increased transceiver rates, advanced modulation, orthogonal frequency division multiplexing
(OFDM), wave-length division multiplexing (WDM), etc. Multiplexing techniques in future
systems could include TDM, WDM, CDM, SCM, etc. Figure 3 presents an attempt to
systemize NGOA system candidates in a tree like structure. At the top-level systems are
categorized with respect to topology (tree, ring, mesh). The different categories are further
categorized based on fibre infrastructure and whether it is shared or dedicated and whether the
remote node is active or passive. The concepts are further categorized based on remote node
design. The passive tree based concepts can for example be based on power splitters and/or on
wavelength filters. The categorization continues further with technology and realization.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 32 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 3 :
Systems concept tree
The analysis of system concepts for NGOA in this deliverable is based on systems identified
in the system concept tree. These concepts have been grouped into a number of subchapters in
Chapter 4. The TDM-PON section 4.1 covers pure TDM-PON variants as well as stacked
TDM-PON solutions (few wavelengths). The WDM-PON section 4.2 covers different
flavours of WDM PON (tunable ONUs, seeded reflective ONUs and systems based on
wavelength reuse). The OFDM-PON section 4.3 covers different OFDM-PON variants. The
CDM section 4.4 covers various CDM solutions with focus on optical CDM-PON. The
WDM-XXM hybrid section 4.5 considers solutions based on a combination of WDM and
TDM, CDM, OFDM and SCM. It also considers UDWDM which is based on the similar two
stage ODN infrastructure as the hybrid concepts. Section 4.6, on active hybrids, covers
variants with active remote nodes such as active P2P and active WDM-PON remote notes.
The AON section 4.7 considers technologies used for tree, ring and mesh based active
networks. Finally, section 4.8, Radio-over-fibre backhauling, covers both radio-over-fibre
technologies and digital interfaces which are relevant mainly for RAN transport.
3.2 KEY ASPECTS FOR THE ASSESSMENT
In the assessment of different system concepts performed in this document, a number of key
aspects were selected that were considered more crucial for the selection of NGOA system.
These key aspects are summarized in Table 1 and to the extent it is possible each of these key
aspects is discussed for each of the system concepts covered in Chapter 4. The assessment of
the different concepts performed in this document, is naturally complicated by the different
levels of detailed information available for the different concepts. In addition to the list of key
aspects, a list of optional aspects to consider was also made (Table 1).
Based on the identified systems, key emerging components/subsystems are identified, where
component/subsystem evolution is critical for the respective system. Some components are
specific for certain systems, while others are key elements of several systems. The survey of
these evolving components/subsystems is based on a similar list of key aspects (Table 2) as
for the system concepts.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 33 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 1 :
Key system aspects
Key system aspects
Supported topologies (tree, ring, etc.)
-
Architecture, RN design, Technology,
Realization
Reach (km)
-
Without reach extension
-
With reach extension (what
technology)
Power budget (dB)
Optional system aspects
Security (optical encryption…)
Migration options
-
From splitter-based infrastructure
-
For higher bandwidth and/or reach
Protocol complexity (MAC in
time/wavelength domain, resource allocation,
etc.)
Bandwidth efficiency
Bandwidth per sub (peak, sustain)
-
Data rate compared to line rate
No of subscribers per FF
-
Spectral efficiency
Cost [high, medium, low]
-
Dynamic allocation
-
Today
-
Projections to 2020
Power consumption
-
Today
-
Projections to 2020
Support for Re-configurable network
-
Wavelength switching
Temperature ranges
-
Suitable for outdoor, or indoor use
only
Possibility for power saving (sleep mode,
power shedding, etc.)
Support for co-operation
Maturity, time to market (technology
roadmap)
Quantitative cost indication (relative eg to
GPON)
Resilience
Key hardware cost drivers (CapEx)
Supported techniques (1+1, 1:1, shared...)
Support/possibilities for unbundling
Possibility for low OpEx
Footprint, physical size, port density [high, Aggregation solution
medium, low]
Extension/boundary of access network
CoS support
QoS support
Detailed Data plane performance/issues
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
-
Delay
-
Jitter
-
BER
-
Packet loss
Page: 34 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 2 :
Key component/subsystem aspects
Key component/subsystem aspects
Cost [high, medium, low]
-
Today
-
Projections to 2020
-
Optional: Absolute cost
Power consumption
-
Today
-
Projections to 2020
-
Possibility for power saving (sleep mode, power shedding, etc.)
Maturity, time to market (technology roadmap)
Physical size
-
Today
-
Projections to 2020
Temperature ranges
Pluggability
3.3 COST AND POWER ASSUMPTIONS
Cost and power consumption data to be used in the OASE assessment must be adjusted to the
2020 time horizon. Reasonable assumptions on evolution of technology as well as cost and
power consumption must be made. This section presents the basic assumptions to be used
within OASE for compilation of cost and power data.
-
Numbers are the best estimates that can be made at this point in time, based on the
collected experience of the consortium. If new information is found, numbers are
subject to change.
-
Numbers which are to be used in the assessment are aggressive assuming maturity in
the 2020 time horizon and mass-production, ramped up approximately in its third year.
The effect of learning curves is incorporated in the numbers. Part of the data will also
be based on the most advanced components available today (e.g., low-power
consuming SFPs, tunable XFPs, etc.). All data should be considered aggressive with
respect to both cost and power consumption. To the best of our knowledge, different
approaches will be treated equally aggressive.
-
Estimates for critical components are based on internal sources, discussions with
component vendors and discussions within FSAN subject to NDA.
-
Relative differences between different components is based on a complexity analysis
of the components and the manufacturing process. The complexity analysis should
consider the functionality needed (e.g., number of sections in an integrated transmitter,
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 35 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
added XDMA, added DSP, etc.), the bandwidth, and the required power budget (e.g.,
TX-RX budget of 28dB vs. 36dB or even 45dB).
4. System concepts for NGOA
This chapter presents a detailed description of the system concept variants for NGOA
discussed in Chapter 3. The different variants are grouped under the subsections: TDM-PON,
WDM-PON, OFDM-PON, CDM, hybrid WDM/XXM, active hybrids and RoF backhauling.
Within each section the main alternatives are discussed including a short description of each
concept as well as brief discussion on key aspects and optional aspects.
4.1 TDM-PON
The state-of-the-art PON standards are based on TDM/TDMA transmission mechanisms. In
the downstream direction the data allocation to the ONU is realised by TDM controlled by the
OLT. TDMA is used for the upstream direction in order to prevent collisions on the PON.
Gigabit–class PONs have been standardised in ITU-T/FSAN and IEEE. The G-PON (i.e.,
2.488 Gb/s downstream and 1.2 Gb/s upstream) was specified in the ITU-T recommendation
series G.984 [3] and the 1G-EPON (1.25 Gb/s downstream and 1.25 Gb/s upstream, with
8b10b coding giving 1 Gb/s data rate) has been standardised by the IEEE working group
802.3ah in 2004 [4]. Both PON technologies are now being deployed, the G-PON mainly in
Europe and America whereas the 1G-EPON is widely used in Asia.
PON technologies with a downstream data rate of 10 Gb/s are standardised in ITU-T/FSAN
and IEEE. The IEEE working group 802.3av has approved the 10 Gb/s Ethernet Passive
Optical Network (10G-EPON) standard in 09/2009. The downstream rate is 10 Gb/s whereas
two upstream rates are available, 1 Gb/s and 10 Gb/s. In total, three types of EPON ONTs can
be supported:



1G/1G (1G-EPON)
10G/1G
10G/10G
10G EPON and 1G-EPON can coexist on the same ODN. The downstream traffic in 10GEPON is based on broadcast Ethernet frames and the upstream is based on TDMA with each
ONU transmitting in its own timeslot.
XG–PON [5] in Figure 2 represents an NG-PON1 system with 10 Gb/s line rate in the
downstream direction. Two types of XG-PON have been defined in FSAN addressing
different upstream line rates. XG-PON1 has an upstream line rate of 2.5 Gb/s and XG-PON2
allows a symmetric line rate of 10 Gb/s.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 36 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 4 :
Wavelength plans [6]
Figure 4 shows the Wavelength plan of the Gigabit-class and 10 Gigabit-class PON
specifications. Both 10 Gb/s PON specifications are using the same wavelength windows for
the downstream and upstream. In contrast to G-PON and XG-PON the upstream wavelength
windows of 1G-EPON and 10G-EPON are overlapping, i.e. WDM multiplexing in order to
separate the two data rates at the OLT site is not possible. Instead TDMA sharing is performed.
The wavelength window between 1550nm and 1560nm is available for downstream RF video.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 37 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 3 summarizes some key parameters of the main TDM-PON technologies currently
standardized.
Standardization
body
Standardization
Group
PON Family
Name
Standardavailability
Data rate down
[Gb/s]
Table 3 :
IEEE
802.3av
G.984
G.987
Tbd
1G-EPON
10G-EPON
G-PON
NG-PON1
NGPON2
Tbd
1000Base
-PX
2004
10GBasePRX
2009
10GBasePR
2009
1 (Line
rate
1.25G)
1 (Line
rate
1.25G)
1:16/32
10 (Line
rate
10.3G)
1 (Line
rate
1.25G)
1:16/32/64
10 (Line
rate
10.3G)
10 (Line
rate
10.3G)
1:16/32/64
10/20
10/20
10/20
15/20/29
15/20/29
15/20/29
Ethernet
Circuit
emulation
Ethernet
Circuit
emulation
Ethernet
Circuit
emulation
Splitting rate
Reach
extension
Power saving
Supported
topologies
ITU-T (FSAN)
802.3ah
Data rate up
[Gb/s]
Min. reach [km]
Logical reach
[km]
Power budget
[dB]
Frame
Encryption
TDM support
Main TDM-PON standards
Star, tree,
ring
Star, tree,
ring
Star, tree,
ring
-
XG-PON1
XG-PON2
2004
2009/2010
-
2.488
9.953
9.953
Ca.
2015
*
1.244 (2,5)
2.488
~10
*
1:32/64/128
*
*
20
60
1:32/64/12
8
20
60
*
*
*
*
15/20/25/28
/30/32
GEM
AES
GEM
encapsulati
on
G.984.6
29/31/33/3
5
GEM
AES
GEM
encapsulati
on
ffs
*
*
*
*
*
*
*
*
*
*
Star, tree,
ring
ffs
Star, tree,
ring
*
*
*
*
(*) For further study
With the OASE requirements for NGOA of 1 Gb/s per end-user, current TDM-PON standards
are insufficient. Considering that serial NRZ schemes are used for current TDM-PON
standards, the natural starting point for higher capacity (40G-PON) is to investigate higher
rate serial NRZ.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 38 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 4 and Table 5 summarize the 40-G-PON variants covered in this section. In general, all
combinations of mixing any 40G DS variant with 40G or 10G US variant are possible.
Table 4 :
40G-PON variants for the DS
Serial 40G
Serial 40G QPSK
DQPSK
DS 40G
Serial 40G NRZ
Bit rate per WL
40
40
40
OLT Tx
EML+EDFA
DBF+MZ
DBF+precoder + MZ
ONT Rx
Technology
PIN-PD + OA
Coherent receiver
Coherent receiver
Direct detection APD
Disp. Comp.
DCF at OLT side (all
pon), EDC at ONU
side (per drop)
No
No
No
WLs
C/L-band
C/L-band
C/L-band
Tunable Tx/Rx
No
No
No
Link budget
(total
reach+split)
~27 dB
-
-
~26 dB
Long-reach
Optical amp (EDFA),
OEO
Optical amp (EDFA),
OEO
Optical amp (EDFA),
OEO
Optical amp (EDFA),
OEO
OLT- RN
resilience
Dual feeder fibre (2:N
splitter)
Dual feeder fibre (2:N
splitter)
Dual feeder fibre (2:N
splitter)
Dual feeder fibre (2:N
splitter)
Table 5 :
NRZ 4 x 10G
10
EML (XG-PON1,
G.987.2)
C/L-band (XG-PON1
1575-80 nm)
Tunable ONU Rx
(fixed operational
prohibited)
40G-PON variants for the US
Upstream 40
or 10G
Bit rate per
WL
Serial 40G
NRZ
Serial 40G
QPSK
Serial 40G
DQPSK
NRZ 4 x 10G
NRZ 1 x 10G
40
40
40
10
10
ONT Tx
DFB+MZ+OA
DBF+MZ
DBF+precoder +
MZ
DML (802.3av)
DML (802.3av)
Coherent receiver
Coherent receiver
TFF fixed filter +
APD
Direct detection
APD
No
No
No
No
O/C/L-band
O/C/L-band
4 wl in O/C/L-band
O-band
No
No
No
Tunable Tx
No
As XG-PON1 (2931 dB)
-
-
29-31 dB
29-31 dB
Long-reach
Optical amp
(EDFA), OEO
Optical amp
(EDFA), OEO
Optical amp
(EDFA), OEO
Optical amp
(EDFA), OEO
Optical amp
(EDFA), OEO
Resilience
down to RN
Dual feeder fibre
(2:N splitter)
Dual feeder fibre
(2:N splitter)
Dual feeder fibre
(2:N splitter)
Dual feeder fibre
(2:N splitter)
Dual feeder fibre
(2:N splitter)
OLT Rx
Disp. Comp.
WLs
Tunable
Tx/Rx
Link budget
(total
reach+split)
PIN-PD + OA
DCF at OLT side
(all pon), EDC at
ONU side (per
drop)
O/C/L-band
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 39 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
4.1.1 Serial 40G NRZ
The main problem with 40G serial NRZ is the limited reach due to chromatic dispersion
(CD). Using well-known formulas for the dispersion limit [7], the 40G limit is ~4 km in the
C-band and ~30 km in the O-band. The conventional dispersion compensating method is by
DCF, but different methods are indicated in Figure 5. Typical DCF has up to 1 dB/km loss and
1 km DCF compensates 5-10 km of standard fibre in the C-band. In a PON system, each drop
fibre may be of different length and attaching DCF of different lengths to each drop fibre
should be avoided. However, if the difference between the shortest and longest drop fibre
(differential length) is within 4 km (the CD limit of 40G systems), a DCF of the average
length of the PON could be placed at the OLT side to provide dispersion compensation for the
whole PON. If the differential length is >4 km, adaptive dispersion compensation should be
used and preferably low-cost since it has to be used at every ONU. One such solution is
electronic dispersion compensation (EDC). Adaptive EDC can increase the differential length
and/or relax the requirement on the OLT side with DCF being very precise in compensation.
Another alternative for coping with dispersion which eliminates use of DCF is the use of a
chirp-managed laser (CML) for phase correlation between adjacent bits [8]. The transmitter
consists of a conventional 40G directly modulated DFB with a subsequent optical filter which
performs the FM (frequency modulation) to AM (amplitude modulation) conversion. This
transmitter allows for 20km reach without dispersion compensation over a SSMF.
Another problem with 40G NRZ is the high optical power required at the receiver. Reported
results for PIN-SOA receivers [9] states -17 dBm at BER=10-10 in the C-band. However,
similar to XG-PON1, strong FEC can be used which reduces the received BER limit to ~10-3.
In this case, the PIN-SOA 40G receiver sensitivity would be ~-20 dBm. Similar to the XGPON1 link budget class Extended 2, an optical amplifier can be used at the OLT Tx boosting
the optical signal to ~+ 12 dBm. The resulting link budget would be ~27 dB (OLT Tx – DCF
IL at 20 km compensation – CD penalty – ONU Rx sens: 12 dBm – 3 dB – 2 dB – (-20 dBm)
= 27 dB. 27 dB in the C-band can be used for ~25 km ODN reach at 1:32 split.
Due to limited power budget there are clear limitations to increasing reach and/or split ratio
(beyond 1:32) without use of active equipment. Hence, reach as well as number of subscribers
per feeder fibre are clearly restricted. Component availability and maturity is currently quite
low due to the required high rate DFB laser. Cost is impacted mainly by the required high
speed transceivers.
Figure 5 :
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Downstream serial 40G NRZ
Page: 40 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
4.1.2 Serial 40G QPSK
Another alternative for serial transmission is 40G QPSK using coherent detection. Compared
to direct-detection schemes, coherent detection allows for increased power budget with
increase reach and splitting ratio. The main drawback is the high cost associated with the
coherent receiver. The basic setup for the downstream is shown in Figure 6 which includes a
digital coherent receiver based on high speed ADC and DSP. At the transmitter side a DFB in
conjunction with a MZ-modulator is used to generate the optical signal. At the receiver side
the optical signal is detected by a coherent receiver consisting of a LO, an optical phase
hybrid, a photoelectric conversion circuit and AD-converters. A digital signal processing unit
(DSP) manages distortion equalization, carrier frequency offset (CFO) compensation, phase
recovery and symbol detection. For QPSK small laser frequency offset introduces a static
rotation of the constellation. Larger offsets require frequency offset estimation [10]. The CFO
information can be used either as feedback through automatic frequency control (AFC) to
tune the LO or to digitally reverse rotate the constellation in the DSP. Carrier phase estimation
is used in a similar way to reverse rotate the constellation diagram digitally. Decision strategy
(symbol detection) could involve a decision feedback with minor extra complexity.
Practically, electronic dispersion compensation is performed at the receiver in the DSP, hence
avoiding need for costly and bulky optical dispersion compensation. QPSK could also be
combined with Polarization Multiplexing (PM) further reducing the symbol rate. These results
in a 3dB penalty compared to single polarization at twice the rate. The reduced baud rate of
QPSK results in increased CD tolerance but leads to decreased OSNR tolerance. By
employing polarization multiplexing a rate of 10G would be sufficient to achieve 40G.
Long reach, large fan-out and high bandwidth can be achieved with the scheme. The main
drawback is the higher complexity related to coherent detection at the ONU side.
Figure 6 :
Downstream 40G QPSK
4.1.3 Serial 40G DQPSK
An alternative to QPSK modulation is using the differential (D)QPSK format which relieves
the receiver from detecting absolute phase as information is encoded in the changes of the
phase. A schematic system diagram is shown in Figure 7. Synchronization with a local
oscillator (LO) at the receiver is avoided. At the transceiver a precoder is introduced to
provide the correct input to the MZ-modulator based on the current state. At the receiver side,
instead of a LO the signal is multiplied with a version of itself delayed by one symbol period
reducing complexity. Similar to QPSK the DQPSK format transmits 2 bits per symbol
resulting in a baud rate which is half the bit-rate. DQPSK is tolerant to chromatic dispersion
(CD), polarization-mode dispersion (PMD), and has high spectral efficiency. Although the
configuration of a DQPSK system is less complex compared with a QPSK system, large size
and high power consumption of the optical transceivers still pose challenges.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 41 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Long reach and high bandwidth can be achieved. Large fan out can be achieved. The main
drawback is large complexity although lower than for the QPSK scheme. The main cost
contributor for the system is the coherent detector at the ONU although the local oscillator is
avoided compared to the QPSK scheme.
Figure 7 :
Downstream 40G DQPSK
4.1.4 Serial 4x10G NRZ
Higher throughput can also be achieved by stacking multiple 10G TDM-PONs over different
wavelengths. Here consider the downstream stacking of 4x XG-PON in order to realize a
40G-PON. Stacking requires support in the wavelength plan. For the considered system, one
solution is to use the current extended wavelength plan in Figure 4 with 1575-1580 nm
(190.3-187.9 THz) for the downstream using 200 GHz spacing for the four stacked PONs.
The system scheme for the downstream of a stacked PON is illustrated in Figure 8. An EML
(DFB integrated with EAM) per wavelength is used on the OLT side. A MUX/DEMUX is
required on the OLT side for multiplexing the four downstream wavelengths and
demultiplexing the upstream wavelengths if stacking is used also in the upstream. With the
requirement of colourless ONUs, a tuneable filter is used to set the downstream wavelength of
the ONU.
Assuming parameters from XG-PON1 link budget class Nominal 2, with minimum mean
launched transmitter power of +4 dBm and a minimum receiver sensitivity of -28 dBm at
BER reference level ~10-3 (strong FEC), the resulting link budget would be ~26 dB (OLT Tx
– MUX – 2 x diplexer - CD penalty – TF - ONU Rx sens: 4 dBm - 2dB -1 dB – 2 dB – 1dB (-28 dBm) = 26 dB. 26 dB in the C-band can be used for ~21 km ODN reach at 1:32 split
(accounting for 3 splices and 4 connectors). At a 2:32 split the corresponding reach is 20 km.
The additional insertion loss of the MUX/DEMUX and TF reduces reach slightly compared to
10G PON. The existence of a power splitter limits potential split and reach. Increased split or
reach beyond 20 km requires use of active reach extenders. The dispersion limit is 61 km
(1565 at 1dB optical path penalty) which is well beyond the attenuation limit. However for
reach beyond 60 km some form of dispersion compensation may be required.
A principle drawback with stacked PON compared to the other 40G PON alternatives
discussed in this section is that the ONU peak rate is limited to 10G which in principle limits
bandwidth flexibility. Statistical multiplexing across the stacked PONs can to some degree be
exploited, although not to the same extent as in the corresponding 40G TDM-PON system.
Another drawback compared to 40G PON is the need for four times more transceivers at the
OLT.
Component availability and maturity is high since most of the required components have
already been used for 10G PON. The solution depends most critically on the availability of
low-cost tuneable filters and lasers. Stacking could also be used in the upstream direction. For
colourless ONU operation this requires a tuneable laser at the ONU side.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 42 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 8 :
Downstream stacked 4x10G NRZ
4.2 WDM-PON
The potential advantages of using WDM in fibre access networks are well known:




Protocol and service agnostic end-to-end connectivity, on a per-wavelength basis;
Bandwidth scalability, due to bit rate transparency and easy channel upgrade;
Enhanced distance reach, due to the low optical path loss;
Simplified network operation (one solution for business, residential and backhaul)
and maintenance (splitting different services to different wavelengths);
However, the high cost of some of the optical components today still is a major obstacle to the
application of WDM techniques in the access network. This clearly calls for improvements of
(components) technology, at least for some crucial components such as tunable lasers.
In addition, there is a strong dependence between system concept – which in turn depends on
the major components which become available – and the related network architectures. Part of
the network architecture-related advantages listed above (e.g., simplified operations) heavily
depend on the system concepts and the relevant components used, where in turn systemsrelated advantages like long reach depend on architecture prerequisites like the allowance of
filters, rather than power splitters.
As far as the system architecture is concerned, we can distinguish three main classes of
WDM-PONs, all leading to one single, wavelength-agnostic variant of the ONU. This is a
frequent requirement of network operators and service providers in order to simplify the
network operation, reduce the inventory and leverage on high manufacturing volumes.
In the first class, the upstream optical carrier is locally generated, e.g., using tunable lasers
(WDM-PON with Tunable Lasers and Laser-Arrays). In the second class, the optical carriers
are remotely generated and then distributed to the ONUs, where different types of devices
(RSOAs, REAMs or injection-locked FP-LDs) can be used to modulate the upstream signal
(WDM-PON with seeded Reflective Transmitters). For these two first classes, similar
principles are applicable to both, the ONUs and also the OLT (where the OLT would have to
make use of the respective array technology). In the third class, each downstream wavelength
is amplified and re-used to also carry the upstream traffic: this is done on the same fibre using
techniques able to avoid downstream/upstream interference (WDM-PON with  Re-use).
Basically, the third class is a derivative of the seeded reflective approach.
4.2.1 WDM-PON with (Tunable) Lasers and Laser-Arrays
Coming from typical WDM transport systems as they are used in backhaul, metro core and
backbone transport, lasers as transmitters in both, the OLT and the ONUs, are the first natural
choice. However, the requirement for colourless or wavelength-agnostic ONUs leads to the
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 43 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
necessity to use tunable lasers in the ONUs. So far (2010), such lasers are prohibitively
expensive in order to enable WDM-PON to become a widespread – residential – access
solution. This leads to the seeded reflective WDM-PON approaches, and meanwhile there is
work in its way which will also lead to ultra-low-cost tunable lasers which can be used in the
PON access context.
The most basic block diagram of a laser-based WDM-PON is shown in Figure 9. Three major
elements can be identified which are necessary for any efficient WDM-PON solution:



A cyclic AWG in the RN to allow single-fibre working (SFW)
An array technology for the OLT transceivers
An ONU transceiver based on a low-cost wavelength-agnostic transmitter, a diplexer,
and the related receiver
Cyclic AWGs are the most elegant way to provide SFW in WDM-PON. Other filter
technologies like TFF can be used but they either lead to the requirement of additional
diplexers for combining both directions of transmission (downstream, upstream), or they have
higher insertion loss. Here, a C+L-band cyclic AWG is used. Other options exist (both
example red and blue sub-band in the C-band).
RN
L-Band 100 GHz
Tx Array
1
...
Mux
PIC
N
Figure 9 :
Cyclic AWG
1
...
C
C-Band ~100GHz
N
Rx Array
L
Cyclic AWG
OLT
ONU
T-LD
SFF
Rx
Identical Cyclic AWGs
Generic WDM-PON with (fixed or tunable) laser array in OLT and tunable lasers in ONUs
Array technologies for the multiple OLT transceivers are a must. Such photonic-integrated
technologies and their related components (PICs, Photonic Integrated Circuits) are required
for reasons of cost (components’ CapEx), footprint, and energy consumption. An array of
dedicated transceivers (e.g., SFPs, SFFs, XFPs) would not be acceptable as OLT solution
specially with regard to accumulated form factor and energy consumption. This holds
particularly when being benchmarked against EPON/GPON or XG-PON solutions. The OLT
PIC can consist of two arrays, one for the receivers and one for laser transmitters. If the PIC
contains all channels, the lasers need not be tunable, however, tuneability may increase the
production yield of such devices and help decrease the cost of the PICs.
The ONU transceiver needs to be low in cost, low in energy consumption, and wavelengthagnostic (“colourless”). Both, low cost and low energy consumption may lead to the
requirement of a coolerless and wavelength locker-less component. This may lead to special
tuning requirements in the WDM-PON context and is covered later.
A variant of the basic, laser-based WDM-PON is shown in Figure 10. This variant supports
ODNs with power splitters/combiners rather than WDM filters (AWGs).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 44 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OLT
ONU
RN
Tx Array
1
...
Mux
PIC
T-LD
L
C
SFF
TF Rx
Rx Array
1
...
N
Cyclic AWG
N
Figure 10 : WDM-PON with tunable lasers in ONUs supporting power-split ODN
Generally, this chapter focuses on filter-based WDM-PON. WDM support on splitter-based
ODN may become a requirement for reasons of migration into existing (EPON/GPON)
infrastructure. As compared to Figure 9, the basic WDM-PON now requires an additional
wavelength-selective component for the receiver. This function is necessary because now all
downstream channels are broadcasted via the power splitter. Generally, two options exist for a
wavelength-selective receiver, tunable filters or coherent detection using a tunable local
oscillator (i.e., a tunable local laser). As compared to the broadband direct-detection receiver
used in Figure 9, both add cost to the WDM-PON ONU and must hence be seen a critical
component. Considering system performance WDM-PONs should be operated via filters. In
addition, one of the major advantages of WDM-PONs as listed at the beginning of this
chapter also gets lost when using a splitter-based ODN: the long reach performance. Where
AWGs typically have 4-5 dB of insertion loss, the loss of power splitters increases
exponentially with fan-out port count. For example, 1:64 splitters have ~20 dB loss, and
1:1024 splitting would lead to almost 35 dB of insertion loss.
The most important way to make tunable lasers low in cost is to reduce their complexity.
Every sub-component which is not absolutely necessary must be omitted. In addition, most
network operators in FSAN require coolerless transceivers due to both, reduced cost and
reduced energy consumption. Hence, he respective tunables must be coolerless. Tuneability
should not come at significant added cost (e.g., complicated external cavities etc.). Hence,
lasers must be chosen which are inherently tuneable, for example 3-Section DBR lasers.
Further, the onboard wavelength locker should be omitted because of added cost. The result
would be a monolithically integrated (multi-section) laser without locker and cooler, a
component which has true lowest-cost potential (because it is not more complex than grey,
low-cost SFPs or EPON/GPON transceivers). However, since the resulting device will lack its
own wavelength locker, system-wide means for tuning the ONU lasers must be established.
Generally, two ways for ONU tuning in the PON system context are possible, i.e. tuning via
the OLT and tuning via the RN. Both enable implementation of a distributed, shared
wavelength locker, at lower cost than the respective number of dedicated lockers. The
principle of tuning the ONU laser via the OLT is shown in Figure 11.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 45 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OLT
RN
...
L
C
(C-Band) Y GHz
Tune
Embedded
Communications
Channel
ECC
µC2
Rx
with
ECC
1
Data
AWG
Tx Array
...
µC1
N
T-Tx
AWG
AWG
1
ONU
(L-Band) X GHz
Rx Array
N
Figure 11 : Tuning of tunable ONU via closed-loop control incorporating the OLT and an ECC
Tuning is possible because two wavelength-selective elements are part of the optical path –
the AWGs in the RN and in the OLT. Hence, tuning can be based on controlling the ONU
laser in order to receive maximum power at the OLT. The respective control commands can
be sent from the OLT to the ONU via an Embedded Communications Channel (ECC) which
in turn can, e.g., be based on a pilot tone modulation. This ECC can always be established in
the downstream because the ONU can have a broadband receiver.
An alternative tuning scheme is based on autonomously tuning the ONUs with the help of a
partial reflector which is placed in the upstream directly behind the AWG in the RN, refer to
Figure 12. This partial reflector has a small influence on the power budget (<1 dB) Now, the
AWG in the RN is traversed twice. In order to provide a unique tuning criterion, the ONUs
must now add a specific pilot tone to their upstream signal. The respective portion of this pilot
tone can be used in the ONU for exact tuning.
Tx
AWG
AWG
Rx
(L-Band) X GHz
A
C
L
A
(C-Band) Y GHz
RN
ONU
Cyclic AWG
OLT
T-Tx
Tune
Tone
µC
Rx
Partial Reflector
(wavelength-selective)
Data
Figure 12 : Tuning of tunable ONU via closed-loop control incorporating a partial reflector in the RN
Both tuning schemes described herein so far have in common that they do not need dedicated
wavelength lockers per ONU laser. These lockers are replaced by a shared, common locker
mechanism which can be implemented at lower cost. For example, the partial reflector shown
in Figure 12 is part of this shared wavelength locker. This way, the ONU transmitters can be
reduced in complexity and cost. It is worth noting that the same transmitters are useless in
most other applications because of the missing tuning capability (and also because these
transmitters are designed for maximum distances probably not exceeding 100 km).
The DEMUX operation of a cyclic AWG is illustrated in Figure 13 (the MUX functionality is
obtained by inverting the propagation direction): two combs of equally spaced frequencies
enter into the common port of the device; a pair of corresponding wavelengths in the two
combs, separated by a frequency gap , is present at each output port. Each pair of
wavelengths is finally separated or coupled by wideband WDM splitters at the OLT and each
ONT. Typical choices for the two upstream and downstream bands are C (1530 – 1565 nm)
and L band (1565 – 1625 nm) or “blue” (1530 – 1547.5 nm) and “red” (1547.5 nm – 1565
nm) sub-bands in the C band.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 46 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 13 : Cyclic AWG
With a simple, straightforward DWDM-PON, no more than 80…96 bi-directional channels
are possible with SFW (single-fibre working). Since massively more channels may be
required, ways of extending the channel count must be considered. With a straightforward
DWDM-PON involving direct-detection receivers for lowest-possible cost, several such ways
exist, extending the wavelength region, and decreasing the WDM grid to an extent where
simple direct-detection receivers can still be used.
Decreasing the WDM grid down to 25 GHz is considered feasible. This is due to the fact that
only two (multi-stage) filters have to be traversed, one multiplexer and one demultiplexer. In a
PON, there are no effects of multiple cascaded filters (which might have significantly
narrower passbands). Also, tuning to within 25 GHz is still considered possible, as is proper
passive demultiplexing. The opposite is already true for 12.5 GHz grid. Here, all tolerances
decrease to a point where e.g., coherent detection might become necessary.
If bit-rate and frequency stability are sufficiently lower than the AWG bandwidth, a
straightforward technique is to use tunable lasers, to adjust the upstream optical carrier
frequency so that it does not collide with the downstream one but remains close enough to
pass through the same port of an optical AWG. A typical situation is illustrated in following
Figure 14.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 47 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Downstream
optical carrier
Upstream
optical carrier
GbE BW
GbE BW GbE BW
GbE BW
US Frequency stability
DS Frequency stability
1.25 GHz
10 GHz
1.25 GHz 1.25 GHz
10 GHz
1.25 GHz
25 GHz
AWG pass band
Figure 14 : DS and US frequency division at the ONU.
The only drawback of this solution is the cost of the tunable laser that currently limits its use
just to FTTB applications. However, low cost tunable lasers are under study and could extend
the applicability field of this simple and effective scheme.
A simple extension of a WDM-PON to 25 GHz grid is shown in Figure 15. Like before, the
PON is based on highly integrated PICs in the OLT, and low-cost tunables in the ONUs. In
this example, two wavelength grids of 50 GHz each are interleaved to 25 GHz.
OLT
RN 2
C
AWG
L
ONU
RN 1
AWG
L-Band 25 GHz
25 GHz Interleaver
PIC 2
(50 GHz)
25 GHz Interleaver
AWG
AWG
RXA
TXA
PIC 1
(50 GHz)
T-LD
Rx
SFF
C-Band ~25 GHz
FSR = 100 GHz
Cyclic 100 GHz 1:48 AWG
Figure 15 : Laser-based WDM-PON with 25 GHz spacing for enhanced channel count
Different architectures of cascaded filters are possible, which are for further study. Either
interleavers (IL) plus AWGs or cascaded AWGs with suitable FSR can be used. Note that in
the OLT shown in Figure 15, two ILs must be used due to the slightly deviating US/DS grid.
It is also worth noting that the system concept according to Figure 15 is still based on the
same PICs and low-cost tunables as described before. So far, no severe complexity has been
added. Hence similar per-channel cost can be assumed. Further increase of the channel count
is possible by extending the wavelength region. If the low-loss region is considered, only the
S-band and part of the eXtended L-band (XL) are suitable candidates. The XL-band, however,
should be avoided due to the dramatic increase of added fibre-bending loss. This leaves the Sband as the primary contender for wavelength extension.
The block diagram of a DWDM-PON allocating the S-, C- and L-band and also using the 25GHz grid is shown in Figure 16. This WDM-PON is based on band splitters (C/L, S/C, and
also red/blue within the S-band, R/B), interleavers, and also cascaded AWGs. This is one
possible filter architecture, and subject to further investigation.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 48 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OLT
PIC 3 S,
50 GHz
PIC 4 S,
50 GHz
T-LD
Rx
C
L/SR-Band 25 GHz
C
AWG
AWG C/L
25 GHz Interleaver
L
SFF
C
S
S
C/SB-Band ~25 GHz
AWG S
25 GHz IL
PIC 2 C/L,
50 GHz
ONU
RN 2
RN 1
25 GHz IL
25 GHz Interleaver
AWG
AWG
RXA
TXA
PIC 1 C/L,
50 GHz
Cyclic 100 GHz 1:48 AWG
R
B
FSR = 100 GHz
Figure 16 : Laser-based WDM-PON with 25 GHz grid and S-band extension for very high channel count
The PON shown in Figure 16 can still make use of the aforementioned components, and may
thus maintain its inherent cost advantages. However, with such a massive DWDM-PON, first
problems also arise. Due to the multiple splitter / interleaver / filter stages, severe insertion
loss is accumulated. With lowest-cost (PIN) receivers, and without any added means for reach
extension, maximum reach will be limited to something in the range of 5 dB fibre power
budget. Hence, such a PON may requires high-power transmitters (which is generally possible
at least with some of the tunables), pre-amplifiers in the OLT, and/or APD receivers.
Nonetheless, up to a theoretical maximum of 384 channels (96 channels each in C-, L-, S/red
and S/blue) are feasible. Further problems relate to laser safety, and Raman loss / crosstalk.
Laser safety must be carefully considered due to the high channel count. Amongst others,
OLT downstream Raman pumping for extended reach may become impossible due to strict
Laser Safety Class 1M restrictions. Nonetheless, operation with >300 channels over >50 km
is considered possible.
Raman pumping must also be considered carefully due to the broad spectrum covered with
channels. However, it can be shown that the S-band is well-suited to Raman-amplify the C/Lband. This can be used to have the downstream pumped by the upstream. Such a power
transfer may become necessary because typically, the downstream in a PON is weaker than
the upstream. This is due to the fact that the upstream can easily be amplified by means of a
lumped amplifier (e.g., an EDFA) in the OLT. The opposite – a booster amplifier – is not true
due to laser safety restriction.
The spectral loss of a standard single-mode fibre is shown in Figure 17. Also shown are the
lower and upper limits for the loss according to ITU-T G.652 [11]. It can be seen that even
with use of the S-band, the water-peak region is still avoided and maximum fibre loss is
limited to within 0.4 dB/km (assuming correct fibre deployment etc.).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 49 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Attenuation [dB/km]
1.0
S-Band
C
0.5
L
ITU upper Limit
0
1400
1500
1600
Wavelength [nm]
Figure 17 : SSMF spectral attenuation. The dashed lines are the boundaries according to ITU-T G.652A
[11].
Many candidates for tunable lasers have been described in the literature so far. The two most
important classes are DBR (Distributed Bragg Reflector) lasers and external-cavity lasers
(ECL). However, few of these lasers have the potential for true lowest cost. For a complete
overview of tunable laser candidates see chapter 5.3.
In conclusion, the main advantages of a WDM-PON based on tunable laser are:




High distance reach (50 km or more, depending on the cable attenuation), due to
the low optical components loss (~5 dB per AWG, ~0.7 dB per WDM coupler).
The performance is equal for downlink and uplink and can be further increased by
using optical amplification at the OLT;
No penalty from lumped or distributed reflections compared to reflected
approaches (due to wavelength diversity between upstream and downstream) along
the optical path, leading to predictable performance, easy network planning and
minimization of troubleshooting in field, with consequent improvement of
operational costs and customer satisfaction;
10Gb/s and beyond compatibility for future capacity upgrade, in order to allow a
long term profitable investment to the operator, with no need of further
infrastructure changes;
Easy scalability and resilience to failures (channels can be added or removed with
no impact on the other channels, PIC needs less to no cabling, for highest
availability PICs can be duplicated).
4.2.2 WDM-PON with seeded Reflective Transmitters
The basic alternative to using tunable transmitters in a WDM-PON consists of using seeded
reflective transmitters. This approach holds for both, OLT PICs and ONUs, as can be seen
from Figure 18. Unlike older approaches, the light sources used for seeding the reflective
transmitters (transmitter arrays) are based on Multi-Frequency Lasers, MFL. Compared to
spectrally sliced broadband sources like ASE sources or LEDs, MFLs have the advantage of
higher power levels which translates to better reach performance.
Regarding the OLT, compact state of the art components can be used with excellent
performance but with little opportunities for cost reduction. Instead the single chip laser
arrays is the most promising alternative. Arrays bring down costs, in particular those related to
frequency stabilization, that can be shared for a whole WDM comb instead of being
performed on a per channel basis.
Tuning accuracy compatible with the 100 GHz ITU grid is one of major cost sources for
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 50 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
WDM transmitters. This is seen by comparing the price of a gray SFP (about 25 dollars) and a
colored SFP (about 500 dollars). Some suppliers have a low cost solutions in the O band (10
USD for a 16 channels device), based on FPLD based on quantum dots in GaAs. Their
technology, based on quantum dots, is extremely interesting, but difficult to extended to the C
and L bands.
The PON shown in Figure 18 is based on two MFLs, one for seeding a reflective (REAM,
RSOA, or combination thereof) transmitter array in the OLT, and a second MFL for seeding
the ONUs. In a first simple implementation these MFLs can work in two different wavelength
bands, e.g., C-band and L-band.
RN
REAM
Array
L-Band 100 GHz
MFL
C-Band ~100GHz
Cyclic AWG
MFL
Rx
Array
ONU
Cyclic AWG
Cyclic AWG
OLT
Rx
RSOA/
IL-FP/
REAM
Tx Data
Figure 18 : Generic WDM-PON with seeded reflective (REAM) OLT transceiver array and seeded
reflective ONUs (based on RSOAs, REAMs, IL-FP lasers, or combinations thereof, e.g.,
REAM-SOA)
Due to MFL seeding, no tuning of transmitters is required. Seeds and reflected modulated
signals must be separated by means of circulators. It must be noted that the seeded reflective
approach can be implemented for the OLT or the ONUs independently.
Some publications describe that better reach performance can be achieved with ONUs based
on tunables, rather than seeded reflective ONUs. Independent from that, seeded reflective
OLT arrays must be considered an attractive alternative to laser arrays. One reason is that in a
PoP or local exchange which terminates more than one PON, the MFL for seeding the OLT
PIC can be shared between the different PONs, thus decreasing cost. This is shown in Figure
19. Here, a multiple OLT is shown. Two MFLs are used for redundancy, and than shared
(distributed) amongst the PONs. Since the MFL is a contributor to total cost, this sharing
approach helps decreasing per-client cost. Note that the WDM-PON shown in Figure 19 is
further based on (low-cost) tunable ONUs.
Array
L
MFL
2xN
…
L
MFL
WDM-PON N
C
WDM-PON 2
Array
REAM
Array
RN
ONU
C
WDM-PON 1
L
C
Cyclic AWG
2 redundant MFLs
Multi-OLT
T-LD
SFF
Rx
Rx
Array
Figure 19 : WDM-PON with seeded reflective (REAM) OLT transceiver array and tuneable ONUs.
Redundant MFLs in the OLT are shared between several PONs.
With regard to ONU seeding, various approaches have been described in the literature. This
includes different reflective devices – RSOAs, REAMs, IL-FP lasers – as well as different
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 51 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
schemes for providing the seed. The latter applies to both, the fibres used for seeding as well
as the seed wavelengths.
In a WDM-PON using reflective ONUs, upstream and downstream can make use of the same
wavelengths. This increases spectral efficiency, or the number of bi-directional channels. Two
problems arise, reach limitation due to Rayleigh crosstalk (which is unavoidable), and the
necessity to provide a modulated downstream seed which has constant envelope in order to
allow proper re-modulation. The latter can be achieved, e.g., through using FSK modulation
with constant envelope for the downstream, see Figure 20. This modulation is removed in the
RSOA which then performs OOK modulation onto the same wavelength for upstream
transmission. As already mentioned, this approach is severely limited by Rayleigh scattering,
and also by crosstalk due to downstream/upstream modulation.
3dB
Rx
AWG
OOK
ONU
RSOA
Data
A
...
RX
Array
FM Disc.
A
AWG
TX
Array
AWG
FSK
OLT
Figure 20 : Increase of spectral efficiency in WDM-PON with seeded ONUs through use of constantenvelope FSK downstream modulation and intensity upstream modulation
Rx
AWG
TRX
Array
AWG
In order to significantly improve maximum reach of WDM-PONs with seeded ONUs,
dedicated seed fibres have been proposed. These dedicated fibres can run all the way down to
the ONU (end-to-end), see Figure 21, or they can be used between the OLT and the RN only,
see Figure 22. Provisioning of end-to-end seed fibres (Figure 21) has the advantage of
complete elimination of Rayleigh crosstalk. However, it contradicts the common PON
requirement of support of SFW.
SOA
Data
AWG
...
ONU
MFL
Seed
OLT
Figure 21 : Decrease of seed/upstream Rayleigh crosstalk through end-to-end dual-fibre working
Provisioning of dedicated seed fibres between OLT and RN assumes that the majority of total
distance falls into the feeder fibres. In that case, most of the Rayleigh-related penalties are
eliminated, and dual-fibre working on the distribution fibres is avoided. This is the common
implementation used today for long-reach experiments with seeded reflective ONUs.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 52 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Rx
AWG
AWG
Tx
Array
plus
Seed
...
Data
AWG
AWG
Rx
Array
ONU
RSOA
REAM
OLT
Figure 22 : Decrease of Rayleigh crosstalk through use of dedicated upstream feeder fibre
An alternative for reducing Rayleigh losses is to distribute the seed sources more
geographically, refer to Figure 23. Here, an active RN is used to accommodate the seed source
(e.g., an MFL). Such an active RN can still be much lower in complexity as compared to a
fully-blown local exchange, however the obvious disadvantage is that the ODN is not fully
passive anymore. In the case of the PON shown in Figure 23, dedicated seed distribution
fibres are used. One alternative may be the use of band splitters in order to allow SFW. The
variant shown in Figure 23 has long-reach capability. It should be noted that the multiwavelength source does not need to be managed and only needs electrical power supply.
Rx
AWG
AWG
TRX
Array
SOA
...
AWG
MFL
Seed
Data
ONU
OLT
Figure 23 : Decrease of Rayleigh crosstalk through RN-based MFL Seed accommodation
Different combinations of the generic variants for seeded WDM-PONs have been used for
long-reach and high-capacity (user count) demonstrators in the recent past. One such example
is the so-called Super-PON shown in Figure 24 [12]. Basically, this is an application of active
RNs used for seed accommodation, together with a hybrid WDM-PON scheme used for
wavelength sharing. In this early example, 89 km of feeder fibre could be spanned, which
together with another 10 km of distribution fibres allowed for a total distance of 100 km. The
Super-PON is an example for both, active RNs and a hybrid implementation. It is an early
demonstration that the combination of very long reach (100 km) and very high user count may
cost-effectively only be achieved with the help of active components in the ODN.
90 km
Laser
5 km
5 km
R/B
OLT
Active Node
Rx
SOA
ONU
...
CW
Rx
R/B
PN
R/B – Red/Blue-Band Splitter
Figure 24 : Super-PON [12] as an example for active-RN-based seed accommodation
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 53 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Most recently, a highest-performing, seeded-ONU WDM-PON has been described, refer to
Figure 25 [13]. This PON was based on 32 wavelength pairs, each running at symmetric 10
Gb/s. Seeded REAM-SOA combinations were used for the ONUs, and the ODN consisted of
WDM filters and cascaded 2:64 and 1:4 power splitters. Total customer count was 8192, with
each customer getting a symmetric bandwidth of approximately 37 Mb/s. Obviously, with
lower power-split fan-out, guaranteed bandwidth could easily have reached several 100 Mb/s.
PoP / Service Node
Active RN / LX
PRN
PRN
ONU
BC
Seed
APD Rx
BS
REAM-SOA
...
AWG
DS Tx
MDXM
...
Seed
...
DS Tx
2:64
1:4
PRN
PRN
ONU
BC
APD Rx
BS
REAM-SOA
...
AWG
MDXM
BM Rx
...
BM Rx
124 km
10 km
1 km
100 m
Figure 25 : DWDM-TDMA-PON [13] as an example for dedicated upstream feeder fibre
The hybrid WDM/TDMA-PON shown in Figure 25 is another example for a highperformance PON which involved an active RN. Here, this RN was used to accommodate
amplifiers which were necessary in order to compensate the accumulated loss of the ODN.
The implementation of a colourless transmitter at the ONT is a crucial point. RSOAs or FPLD
are viable options for 1.25 Gb/s channels and some manufacturers are investigating if the
RSOA bandwidth (currently about 1.5 GHz) can be used for 2.5 Gb/s.
Future 10Gb/s solutions entirely based on out of the shelf optical components have been
investigated (SOAs, circulators and EAM). Two variants of the ONU transmitter are shown in
Figure 26.
in
out
2
3
1
ONU
SOA
EAM
in
out
SOA
SOA
2
3
EAM
1
10G
ONU
10G
Figure 26 : ONU transmitter variants
For both variants, a CW signal enters into a circulator, is modulated by a conventional EAM
and looped back into the same circulator. The signal can be amplified by an input
bidirectional SOA, or by separated pre- and booster amplifiers within the ring containing the
EAM. In a simpler scheme (Figure 27), a SOA precedes a REAM. An attractive feature is that
the modulator can be properly designed to introduce a chirp which compensates for the fibre
chromatic dispersion, without lossy and expensive external dispersion compensators. Using
the REAM, 50 km of link distance without dispersion compensation at 10 Gb/s has been
observed in the Ericsson Research Lab in Pisa.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 54 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 27 : Simple ONU transmitter
In conclusion, the main advantage of seeded solutions lies in the low cost of implementing
colorless ONU solutions while performance is less strong compared to solutions based on
tunable laser. Performance penalties are mainly due to:


Rayleigh backscattering and the remote distribution of high power non-modulated
channels for the uplink traffic.
The unavoidable trade-off between performance and distance of the remote
channel distribution
4.2.3 WDM-PON with wavelength reuse
WDM-PON systems where downstream wavelengths are reused in the upstream allow for
wavelength bi-directional channels and as a result doubled system capacity as well as number
of users. Such wavelength re-use requires downstream modulation to be cancelled before
applying upstream data and leads to additional penalty to the upstream signal, caused by
residual downstream modulation and optical reflections along the optical link. Different
schemes for wavelength re-use are discussed in the following, and compared in terms of cost
and performance.
The most basic approach is based on RSOA saturation at the ONU. Part of the downstream
signal is tapped and sent to an RSOA. If the power of the tapped signal is sufficiently high (>10 dBm for the best devices available on the market), the RSOA is saturated cleaning the
downstream modulation so that the upstream data can be applied to the RSOA electrical input
with only a small penalty due to the residual modulation. In practice, the input power value
needed to saturate the RSOA limits achievable power budget in practical systems (typical
reach is < 10 km).
An alternative technique is based on FSK/ASK and DPSK/ASK coding. With this technique,
the downstream traffic is FSK (Frequency Shift Keying) modulated by using a directly
modulated laser diode. The DFB laser is biased at a high value and modulated with small
modulation depth. This enables low extinction ratio (1.5 ÷ 2 dB) and high chirp (4.5 ÷ 5
GHz), in order to maximize the frequency shift caused by any variation in the driving current.
At the ONT, the received wavelength is split into two parts: one part is sent to an optical filter
before the downstream receiver while the other part is used as optical carrier by the upstream
data, which is intensity modulated by a RSOA. The function of the filter is to convert the FSK
modulation format into a regular ASK (Amplitude Shift Keying) format. A simple method is
to position the optical carrier frequency on one of the slope edges of the optical filter. The
working principle is illustrated in Figure 29.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 55 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Laser Spectrum
P
1
1
0
0

t
Chirp: 4.5 ÷ 5 GHz
Figure 28 : FSK generation by means of a directly modulated laser
P
filter
input
1
1
0
0
laser
output
λ
t
Figure 29 : Frequency to Amplitude Modulation by means of an optical filter
Colorless ONU operation requires that polarization insensitive periodic filters are used with
free spectral range equal to the ITU-T frequency grid spacing (100 GHz). The main drawback
with this approach is the additional cost associated with the filter, which requires thermal
control or dedicated circuitry to track any drift of the downstream optical carrier. A cost
effective implementation is possible by integrating the filter, RSOA and photodiode in a
single composite device. Another similar scheme is the RZ/IRZ coding technique which
allows for remarkable performance improvement compared to the saturated RSOA approach
without introducing any additional optical components (and related cost). The scheme is
based on a combination of RZ (Return to Zero) and IRZ (Inverse Return to Zero) modulation
formats for upstream and downstream transmission, respectively. The working principle [14]
is illustrated in Figure 30, where the inset shows what happens with all the possible
combinations of “1” and “0” bits in downstream and upstream.
Figure 30 : Figure RZ/IRZ remodulation scheme
The downstream signal is coded using Inverse Return to Zero (IRZ) with 50% duty cycle. The
RZ upstream bit pattern is shifted by half a bit in time with respect to the downstream IRZ
pattern, so that the RZ pulses in upstream always are transmitted when the IRZ signal is at its
high level. When the ONU receives a logical ”1” (a dark pulse) it can either suppress the dark
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 56 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
pulse high rail (to re-transmit a ”0”) or amplify it (to re-transmit a ”1”). If a logical”0” (a
constant power bit) is received, the RSOA will carve a pulse on it (to retransmit a ”1” ) or will
suppress the whole bit (to retransmit a ”0”). This gives four possible 2-bit combinations. The
performance is further improved if the RZ modulation format is applied to a RSOA properly
designed to introduce a large chirp on the upstream signal. Due to the chirp, the frequency of
the upstream signal is made to not coincide with the downstream one in order to mitigate
effects of reflections in the optical link.
Upstream sensitivity at the OLT depends on cross-talk induced by reflections. In absence of
reflections, good sensitivity values can be achieved (about -32 dBm). About 1 dB power
penalty is recorded for a cross talk value of 15 dB. These are exceptional results for this class
of bi-directional systems, where significant outage is expected for a cross talk of about 25 dB.
4.3 OFDM-PON
Orthogonal Frequency Division Multiplexing (OFDM) in optical access networks has been
driven by the necessity to increase transmission rates without increasing bandwidth of
optoelectronic transceivers. OFDM in the downstream could be combined with Orthogonal
Frequency Division Multiple Access (OFDMA) in the upstream. OFDM is a multi-carrier
modulation technique where the data of a single optical channel is transmitted on multiple
narrow bandwidth orthogonal subcarriers each operating on a low data rate. In OFDM the
frequency multiplexed sub-channels allow for transmission of a multi-carrier signal using a
single optical modulator. The sub-channels are orthogonal and overlap in frequency in such a
way that the peak of each sub-channel coincides with nulls of the other sub-channels. Figure
31 shows the principle setup of an OFDM channel. OFDM offers increased spectral efficiency
and reduced baud rate compared to serial OOK. OFDM modulation is very adaptable (bit and
power loading) and scalable to higher order modulation formats (8-QAM, 16-QAM, etc.) on
the same optical hardware due to easy electronic signal processing. Several OFDM channels,
each consisting of several sub-channels, could be carried over the system. An OFDM channel
could be in the range of a few GHz. For downstream transmission a single laser module of
very high bandwidth (10-25 GHz) is used [15]. Assuming 16-QAM modulation and multiple
OFDM channels, aggregate rates of 40-100 Gb/s could be achieved. The bandwidth of the
ONU receiver could typically have the full range of the OLT transmitter. The upstream is
more challenging due to signal to signal optical beat interference (SS-OBI) which occurs if
subcarriers from different ONU transmitters are overlapping in the frequency domain. The
ONU laser bandwidth would most likely be in the range of a few GHz. A wavelength
demultiplexer plus APD or PIN receivers operating at the same bandwidth could be used at
the OLT side. One upstream solution is to operate the ONUs sufficiently far from each other
in order to avoid overlap. This implies use of WDM technology with the associated
challenges of designing a system for colorless operation. Another option is the use of hybrid
ODFMA/TMDA which requires strict synchronization. Ref [16] demonstrates a solution
based on ONU-side carrier suppression.
Orthogonal subcarriers
Guard-band
Optical carrier
10G OFDM-channel
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 57 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 31 : OFDM-Transmission on multiple orthogonal subcarriers
In general, OFDM data transmission is very tolerant with respect to signal distortions like
chromatic dispersion (CD) or polarization mode dispersion (PMD) even at high single
channel data rates. OFDM is also less susceptible to E/O band response limitation and
distortion. This enables longer reach and increase split ratios without need for dispersion
compensation compared to conventional TDM-PON solutions. OFDM is also more robust to
Rayleigh backscattering compared to NRZ in typical PON architectures. OFDM allows for
use of symbol-by-symbol decision for each sub-channel, rather than complex equalization
algorithms to compensate for distortion (e.g chromatic and polarization mode dispersion). In
the presence of selective fading, radio OFDM channels can achieve very good performance.
However, this is not the case for a fibre optic channel, where the amplitude channel response
is essentially flat. An advantage with OFDM is that access to subcarrier granularity enables
flexible resource allocation. Resource allocation is for example implemented by dynamic subcarrier allocation (DSCA) or conventional TDMA approaches. Another potential degree of
adaptability is the modulation format (QAM-levels) depending on channel.
The main challenges of OFDM are the sensitivity to phase noise (PN) and uncompensated
residual carrier frequency offset (CFO) resulting in inter-carrier interference (ICI). Also, the
large peak-to-average power ratio (PAPR) in the OFDM signal implies large sensitivity to
non-linearities in the optical path. OFDM also exhibits various technical drawbacks such as
the requirement of high resolution analogue-to-digital converters (ADC) and digital-toanalogue converters (DAC). DACs with at least 7 - 8 bits are necessary at the transmitter,
whereas ADCs with 8-bit resolution are needed at the receiver. ADCs are now becoming
commercially available while DACs could be a bottleneck, depending on the channel rate.
The number of sub-carriers per OFDM channel is limited by ADC resolution. Requirements
of high resolution ADC may be migitated by increasing the number of QAM-levels. However
this comes at the expense of increased SNR requirements. Other drawbacks are the energy
consumption and spectral overheads needed to correctly recover information. In the frequency
domain, information-less sub-carriers can be used and in the time domain, a cyclic prefix is
used. For 20% of virtual sub-carriers, the speed of ADCs is equivalent to an over-sampling
factor of 1.2 samples per symbol.
In optical communications two main OFDM approaches exist, one being the coherent optical
(CO)-OFDM which involves modulation of the optical field and the other being direct
detection optical (DDO)-OFDM which involves modulation of the optical intensity. COOFDM has the best performance but also requires more components such as optical or
electronic I/Q-Modulators, RF-mixers or coherent receivers. Each of the two approaches
exists in several variants that will be discussed more in detail below.
4.3.1 Serial 40G CO-OFDM
Figure 32 shows different variants of CO-OFDM with different transceiver and receiver
structures [17]. Figure 32a shows a variant based on direct up/down conversion and Figure
32b shows a variant with up/down conversion via an intermediate frequency (IF).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 58 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 32 : Downstream CO-OFDM [17]: a) direct down/up conversion, b) intermediate frequency
In both cases at the transmit end, the original data signal passes through a parallel to serial
converter which divides the channel to a number of parallel bit-streams. The digital time
domain signal is obtained by IFFT and is subsequently inserted with guard interval and
converted into real time waveform through DAC. The guard interval is inserted in order to
prevent inter-symbol-interference (ISI) due to channel dispersion. The sub-carrier channels
can be modulated through different modulation formats such as BPSK, QPSK, 16-QAM, 64QAM, etc. In the direct up-conversion architecture (Figure 32a), the optical transmitter uses
an optical I/Q modulator which consists of two MZMs to up-convert the real/imaginary parts
from the RF domain to the optical domain. The OFDM optical receiver uses two pairs of
balanced receivers and an optical 90◦ hybrid to perform optical I/Q detection. The RF OFDM
receiver performs OFDM base-band processing to recover the data. The main advantages of
direct-conversion are (i) elimination of need for image rejection filter in both the transmitter
and receiver, and (ii) reduction of required electrical bandwidth for both the transmitter and
receiver. In the intermediate frequency (IF) architecture, the OFDM base-band signal is first
up-converted to an intermediate frequency in the electrical domain, and the OFDM IF signal
is further up-converted to the optical domain with one MZM. In the receiver, the optical
OFDM signal is first down-converted to an intermediate frequency and electrical I/Q
detection is performed. The signal is then sampled with an ADC, and demodulated by
performing FFT and baseband signal processing to recover the data. The DSP unit manages
distortion equalization, CFO compensation, phase recovery and symbol detection. With these
two transmitter and two receiver alternatives there are four system design alternatives.
4.3.2 Serial 40G DDO-OFDM
Perhaps the most attractive OFDM solution for PON access, considering cost, is DDOOFDM. DDO-OFDM allows for a simpler receiver structure than CO-OFDM at the cost of
reduced spectral and energy efficiency. The main advantage of DDO-OFDM is the simpler
direct detection receiver. Typically DDO-OFDM is realized by the transmission of an optical
single sideband (OSSB) OFDM signal and a component at the optical carrier frequency. A
frequency guard band separates the OFDM signal from the optical carrier. The signal is
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 59 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
received by detecting the carrier signal mixing products. Three variants of the DDO-OFDM
transmitter (Figure 33) [18] are here discussed. These have different degrees of optical
complexity, but all require a similar type of single photodiode direct-detection photo receiver
(Figure 33).
Figure 33 : Transmitter and receiver alternatives for DDO-OFDM [18]: a) b) and c) show different
transmitter alternatives and d) a typical receiver.
For the different transmitter alternatives the digital sections differ only in the way the input
vector is mapped onto the IFFT input vector. For the first transmitter design, a single input
optical modulator is used to generate a double sideband optical signal and then one sideband
is suppressed using an optical filter. The electrical input to the optical modulator is a real,
baseband signal and only one digital-to-analog converter (DAC) is required. For the second
design, the single DAC of the first design is replaced by two DACs and an electrical RF upconversion stage. This allows the complex baseband OFDM signal to be mixed with an RF
carrier before driving the single input optical modulator. For the third transmitter design, an
optical single sideband signal is generated using a signal and its Hilbert transform to drive an
optical I/Q modulator.
Figure 33 shows a simple direct detection receiver for DDO-OFDM. As a result of the square
law characteristic of the photodiode, the received signal consists of a number of mixing
products. The mixing products are classified into the useful components from which the data
is recovered, unwanted components which fall within band and limit the BER performance,
and unwanted components which fall out of band. The useful components are the difference
terms which result from the mixing of the OFDM sideband and the optical carrier. There is a
frequency dependent phase shift of the optical signal due to chromatic dispersion. For single
sideband optical OFDM systems, with linear field modulation, each OFDM subcarrier is
represented by a single optical frequency, so phase shifts in the optical domain lead to phase
shifts in the demodulated electrical OFDM subcarriers. These phase shifts can be corrected in
the digital section of the receiver by applying a single complex multiplication for each
subcarrier.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 60 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OFDM-PON has been demonstrated by several research groups. In [19], 108 Gb/s
downstream OFDM-PON transmission using polarization multiplexing and direct detection
was demonstrated. In [20] a version of DDO-OFDM referred to as adaptively modulated
optical (AMO)-OFDM was demonstrated at 32 Gb/s over 2 km and 25 Gb/s over 60 km.
Regarding the upstream, the challenge is to avoid the broadband beating noise without
assigning individual wavelengths to each ONU. In [16] a source-free OFDMA-PON
architecture that solves this problem was proposed which uses ONU-side carrier suppression
to achieve simultaneous upstream transmission from multiple ONUs over a single
wavelength. 36 Gb/s upstream transmission over 100km combined with 1:32 optical split was
demonstrated. Two main features of the described solution include the ONU-side carrier
suppression that avoids OLT-side beating noise, and coherent detection at the OLT that
enables full regeneration of the electrical OFDM signal. In Figure 34 the OLT uses two laser
sources, one for downstream optical OFDM transmission, and one as a distributed carrier for
upstream transmission. An electrical OFDM signal generated by the inverse Fast Fourier
Transform (IFFT) is used to drive an intensity modulator (IM), while the upstream continuous
wave (CW) laser is distributed downstream to all ONUs through the same fibre path. At the
ONUs, the downstream OFDM signal is separated from the distributed CW laser source by an
optical filter. The latter is re-used as optical carrier for the upstream of each ONU. For
upstream transmission, the upstream signal is first generated in each ONU and then upconverted to an intermediate RF frequency. Each RF signal modulates an IM driven by the
CW source. The upstream IM at each ONU uses carrier suppression to avoid the beating noise
that would be generated at the OLT due to mixing of multiple optical carriers. As shown in
Figure 34, coherent detection is used at the OLT receiver in order to fully regenerate the
OFDMA signals. Since the upstream optical carrier was originally distributed from the OLT, it
is re-used as the OLT local oscillator (LO).
Figure 34 : DDO-OFDMA-PON scheme after [16].
4.4 CDM
Code division multiplexing (CDM) provides another way of multiplexing data. Code division
multiple access (CDMA) is the use of CDM technology to arbitrate channel access among
multiple network nodes in a distributed fashion. Optical code-division multiplexing (OCDM)
is a promising technique, offering random access to the entire bandwidth along with
advantages such as simplified network control, increased security, and increased flexibility.
OCDM addresses capacity upgrade by adding a code-based dimension to the FTTH system.
In this section, we will review the key technologies for optical systems based on
OCDM/OCDMA.
In an OCDM system, each channel is distinguished by a specific optical code. An encoding
operation optically transforms each data bit before transmission while the reverse decoding
operation is required to recover the original data at the receiver. The main encoder/decoder
technologies are shown in Table 6.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 61 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 6 :
Encoder/decoder technologies
Coding domain
Chip modulation
1D Time
Power
Phase-shift-keying
(PSK)
Wavelength
Required Laser Devices
source
Incoherent
FDL (Fibre delay line)
Coherent
PLC (planar lightwave circuit);
SSFBG (Super structured
Fibre Bragg grating) [21]
Power
Incoherent
PSK
Coherent
Power
Incoherent
AWG; FBG (Fibre bragg
grating)
High resolution phase E/D
[22]; SLPM (Spatial light
phase modulator) [23];
AWG+FDL [24]; FBGs [25]
3D Time/Wavele Power
ngth/Polarizat
ion
Incoherent
AWG+FDL [26];
2D Time/Wavele
ngth
Due to the limited coding space, incoherent 1D optical coding technology (either in time or
wavelength domain) is not feasible for future access networks which are required to support
large numbers of end users. Both 2D and 3D encoding technologies require multiple domains
to realize optical codes. Therefore, it is difficult to smoothly upgrade capacity of an access
network where 2D or 3D encoders/decoders are employed.
1D coherent coding seems to be the only viable technology. They can have a large coding
space. For instance, 511-chip gold code could have in total up to 513 codes (corresponding to
513 ONUs). Paper [21] has already experimentally demonstrated 10-user, 511-chip, 640
Gchip/s (1.25Gb/s) OCDMA system based on 1D coherent coding in the time domain. Paper
[27] proposed an architecture of OCDM over WDM, i.e. hybrid WDM/OCDM PON. In this
way, the capacity could be easily upgraded by expanding the wavelength spectrum. The most
expensive part of the coherent OCDM technology is the transmitter which must be able to
generate ultra-short pulses. Usually, the price of the devices is market-driven. If a large
amount is required by the market, the expected price of ultra-short pulse lasers could become
comparable with tunable lasers which are required in WDM PONs.
The OCDM solution could include a passive power splitter at the remote node. OCDM-PON
using a tree topology with passive power splitters have been widely studied e.g. in [28]
(Figure 35). The maximum number of users that can be supported in this configuration is
limited due to the high insertion loss of the power splitter. The coding space cannot be fully
utilized. In addition, similar as for TDM-PON, reach is influenced by introduction of power
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 62 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
splitters.
Figure 35 : OCDM PON with a power splitter at the remote node
An alternative is to introduce code splitting at the remote node. For coherent time-spreading
(TS-) OCDM, the multi-port OCDM encoder/decoder (E/D) has the unique capability of
simultaneously processing multiple time-spread optical codes (OCs) with a single device
(Figure 36), which makes it a potential cost-effective device to be used in the optical line
terminal (OLT) of an OCDM network in order to reduce the number of E/Ds [29].
Figure 36 :
A device with capability of simultaneously encoding/decoding multiple time-spread
optical codes (OCs) [29].
If such a device is employed at the remote node between the OLT and ONUs, coding splitting
can be realized [30]. In this way, each ONU is codeless, since E/D is not required at the user
side. Furthermore, the insertion loss of this code-splitting device is much lower than the
power splitter and comparable with the AWG. Therefore, without reach extenders, OCDMPON with code splitting can support reach beyond what is achievable in systems based on
power splitting. Paper [30] experimentally demonstrates a 4-user 10Gb/s system (Figure 37)
where both up- and down-link are over 59 km of standard single mode fibre (feeder fibre: 36
km, distribution fibre: 23 km) without inline dispersion compensation.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 63 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
(a)
(b)
Figure 37 : (a) Configuration of OCDM-PON system with code splitter at RN and (b) experimental
setups and results: (i) downlink and (ii) uplink [30].
The multi-port OCDM E/D device has not been commercialized so far. The price will rely on
market demand. Since the fabrication of this device is exactly the same as the AWG, the
expected price is also similar to the AWG if there is large demand.
While an optical access network using OCDM technology usually assumes a tree topology, in
particular for OCDM PON, much research work on ring topology has also been investigated
in recent years. The OCDM add/drop multiplexer is a key element to support the OCDMA
ring.
Paper [31] introduced an all-optical 2D OCDM code-drop unit (Figure 38) and experimentally
demonstrated its use in an OCDM ring. This code-drop unit has three elements: a drop code
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 64 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
decoder (DCD), a terahertz optical asymmetric demultiplexer (TOAD), and a code restorer
(CR). The codes from all other transmitters, whom delays are not matched to the DCD’s
delays, spread in time as a cross correlation of maximum height 1 according to the properties
of the code. The auto- and cross-correlation, after passing the circulator, enters the TOAD by
Port 1. The autocorrelation peak and the control pulse exit the TOAD at Port 2 while the rest
of the signal, outside of the switching window, is reflected back and directed by the circulator
to Port 3. At Port 2, an optical high pass filter separates the sampled autocorrelation peak from
the control signal. The intensity of the dropped code is then detected by a photodiode and
converted into the original 2.5-Gb/s data sequence. At Port 3, the remainder of the signal goes
through a code recover (CR). The CR undoes the time shifts caused by the DCD to the codes
passing through the code-drop unit in order to restore them to their original sequence. The
complete architecture for an OCDM add–drop multiplexer based on this type of code-drop
unit can be realized, by simply adding a 2x1 power combiner and tunable OCDM transmitter
to the output of the node.
Figure 38 : Schematic of the all-optical 2-D OCDM code-drop unit [31] (D: Delay, HPF: High-pass filter,
and TOAD: terahertz optical asymmetric demultiplexer).
However, it is difficult to find a smooth and cost-effective way to upgrade the capacity of a
ring network with this OCDM add/drop multiplexer, which only works for incoherent 2D
optical codes. For either bit-rate upgrading or wavelength spectrum extending, the existing
add/drop multiplexers should be replaced by the new one. To address this problem, paper [24]
introduced an OCDM add–drop multiplexer for any coding technology. The disadvantage is
that the decoders and encoders for all the optical codes used in network are required at each
node. This increases the cost and decreases security.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 65 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 39 : Structure of OCDMA add-drop multiplexer for any coding technology [24] (TOAD:
terahertz optical asymmetric demultiplexer).
Due to the complicated architectures for the OCDM add-drop multiplexers, OCDM ring
might not be suitable for networks which are cost-sensitive.
4.5 WDM+XXM HYBRID
The OASE NGOA system requirements of up to 100 km reach (large power budget) and up to
1000 subscribers per fibre feed (high splitting ratio) provide motivation for hybrid concepts.
The TDM-PON concepts alone cannot fulfil these combined requirements. Pure WDM-PON
is more attractive, although the relatively expensive WDM components such as stable laser
sources prevent residential mass-market solutions. The idea of hybrid concepts is to combine
the advantages of several different concepts. In a hybrid WDM/TDM-PON this would be the
TDM-PON advantages (bandwidth sharing, colourless ONT) with the WDM-PON advantages
(increased number of subscribers and total BW). Hybrid concepts are based on WDM where
several WDM channels are transmitted in the main feeder fibre to a first remote node, which
typically is λ-selective (e.g. AWG wavelength splitter). In the AWG the downstream data
channels are de-multiplexed and each wavelength fed into a second feeder fibre to a second
remote node which depends on the concept which WDM is combined with (Figure 40).
Basically, underlying WDM can be combined with any of the multiplexing and multipleaccess technologies which were listed in Chapter 3. These schemes are TDMA, CDMA and
sub-carrier multiple access (SCMA). In addition, dense or ultra-dense WDMA must be
considered (because it most effectively runs over the same filter-plus-splitter ODN), and
OFDMA may be considered a more recent version of dense SCMA. In addition, CDMA can
be implemented in either the optical (O-CDMA) or electrical (E-CDMA) domain, with
massive impact on cost, and functionality. It is worth noting that in particular OFDMA,
SCMA, and UDWDMA with many densely spaced sub-carriers each do not differ
significantly in their power spectra, nor in the resulting spectral efficiency.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 66 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
λ1
1
λ2
…
ONT
λDS-1
…
ONT
λ1, λ2, λ3 …λN
λN
λUS-1
λDS-1…λDS-N
OLT
…
Passivesplitter
λDS-N λUS-N
…
ONT
…
…
λ3
N
λ-selective
RN (AWG)
λUS-1…λUS-N
ONT
Figure 40 : Hybrid PON basic setup
One important question is which scheme allows an optimization of either cost/performance or
energy-consumption/performance metrics (as two very important evaluation parameters in the
access context). Such an analysis must take the relevant components of each hybrid system
(with regard to cost and/or energy consumption) into account. The related cost and energy
consumption figures as well as ODN assumptions are summarized in Appendix 7.1. In the
comparison, 50 km of fibre were considered. This relates to the fact that almost no PON
system with very high fan-out can span 100 km (or similar distances) without active RNs
(amplifiers, reach extenders). In order to enable passive ODN wherever possible, 50 km
distance was chosen for comparison. Clearly, the active/passive hybrid variants always require
active RNs, but at 100 km distance requirement, almost all other variants would have required
active RNs as well. Such, we avoided considering over-engineered solutions.
4.5.1 Hybrid WDM/TDM-PON
The first hybrid PON analyzed here is the hybrid WDM/TDM-PON. In order to be able to
provide high bandwidth and high customer count, we considered DWDM with 40 bidirectional channels of symmetric 10 Gb/s bandwidth each (extendable to 80 channels). Due
to the combined ODN insertion loss, this requires 10G burst-mode transceivers with 35 dB
power budget. This can be identified as one of the main cost drivers according to Appendix
7.1. A schematic diagram of one WDM/TDM-PON implementation is shown in Figure 41.
Note that the most common solution consists of several TDM PONs embedded in a WDM
PON system.
RN1
Tx/Rx
Array
AWG
TDMA
MDXM
TDMA
MDXM
1xN AWG
OLT
λD λU
RN2
1:k
10G-TDMA
ONUn
APD
FEC
SOA
Data
CLK Rec.
T-LD
Burst M.
FEC
Data
Figure 41 : Hybrid WDM/TDM-PON
Due to the power-budget requirement (35 dB), transceivers need APDs and very likely also
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 67 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
booster amplifiers (SOAs) and added FEC. All these add to the system’s cost and also energy
consumption. The resulting figures (excluding common baseline cost and energy consumption
contribution) are listed in Table 7. Note that both, cost and energy consumption of the
DWDM-tunable 10G 35-dB ONU burst-mode transceiver must be considered as very
aggressive – optimistic! – assumptions. Nonetheless, they are the major contributor as could
be expected from TDM PONs.
Table 7 :
Energy consumption and cost figures for the hybrid WDM/TDM-PON variant shown in
Figure 41, excluding baseline (chapter 7.1)
Cost
217$ total
Energy
3.8 W
AWG/splitter ports
12$
OLT port
0.25 W
OLT port
22$
OLT switching
1.0 W
OLT amplifiers
3$
OLT amplifiers
0.05 W
ONU TRX
175$
ONU
2.5 W
OLT switching
5$
We refer to the WDM-PON section (subsection 4.2.2) for the description of two WDM/TDMPON demonstrator systems: Super-PON (Figure 24, [12]) and DWDM-TDMA-PON (Figure
25, [13]).
To increase the flexibility in terms of dynamic bandwidth allocation (DBA), four different
hybrid WDM/TDM-PON variations are described below. The flexibility to offer any
bandwidth (in the limit of the physical bit rate) to anybody could be a major advantage. A
hybrid WDM/TDM-PON with dynamic wavelength routing can combine the virtues of both
the TDM and WDM solutions. From the OLT multiple TDM-PONs can be set up, each at a
specific wavelength. Each TDM PON serves a set of users, and within this set, capacity is
shared. By means of wavelength selection or routing, the number of users within the set can
be varied, and thus the capacity offered per user can be varied. Hence a flexible hybrid
WDM/TDM-PON can offer capacity-on-demand, and by means of the wavelength selection
or routing flexibility, the congestion probability can be significantly reduced with respect to
the static WDM/TDM-PON configuration. For example, it could be convenient to offer two
different needs: a number of 100 Mb/s connections for subscribers served by a particular set
of wavelengths, another number of 1 Gbps connections for subscribers having more needs on
another set, and why not a full 10 Gbps pipe if really required. So by combining WDM and
TDM, a lot of demands can coexist on a same network infrastructure in a very efficient
manner. This is a strong advantage when compared to other systems.
The flexibility of the WDM/TDM-PON system is increased by changing the implementation
of RN1. The conventional hybrid WDM/TDM-PON scheme presented in Figure 41, which
offers low flexibility in terms of wavelength allocation, is used as reference system. Extra
flexibility is then added, e.g. by replacing the AWG by a power splitter (to increase flexibility
at the cost of power loss and security threats, see Figure 43), by adding extra AWGs used as
wavelength combiners (to increase the number of wavelengths sent to each ONU, see Figure
44) or by adding a more complex switching architecture (to increase flexibility without loss of
security and/or power loss, see Figure 45 and Figure 46). Note that in the different variations,
RN2 consists of a passive power splitter.
In the remainder of this section, we have classified the hybrid WDM/TDM PONs in four main
categories. These four variations are compared in terms of simplicity, flexibility, power loss
and security. We assume that there are Nu uplink and Nd downlink wavelengths while for some
of the architectures Nu = Nd = N. The general OLT and ONU architectures are depicted in
Figure 42 and consist of following components:
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 68 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0

The OLT has Nu uplink line cards and Nd downlink line cards. Each uplink line card is
associated with one photo detector (PD) and similarly each downlink line card is
associated with one DFB LASER for data transmission.

The ONU has a three port circulator for separating uplink and downlink wavelengths.
A WDM splitter can also be used for this purpose. The uplink part has the uplink line
card and the tuneable burst mode transmitter for tuning to any desired wavelength.
The downlink part has one classical photo detector and a tuneable optical filter is used
for selecting the desired wavelength. Note that when using only one tuneable Rx in the
ONU, the OLT has to store the wavelength each ONU was tuned to for the last frame.
If an ONU is tuned to a wrong wavelength or it goes to sleep mode, however, a
solution is needed to indicate to the ONU to which wavelength it can listen to. This
can be resolved by adding a fixed Rx (in addition to the tuneable Rx) to the ONU, or
directly by the MAC protocol, e.g. by indicating one wavelength as fallback.
Central office
OLT
RN1
Wavelength or
power splitter
RN2
Power splitter
User
ONU
TDM-PON
Uplink
line card
PD
PD
ONU control
Uplink control
Uplink
line card
Tunable
Burst Mode
Transmitter
Nu
PD
Downlink control
Remote
Node
Downlink
line card
DFB
PD
Downlink
Tunable
optical filter line card
DFB
Nd
DFB
Figure 42 :
WDM/TDM-PON based on an ONU with one tunable transmitter and one tunable
optical filter, for a general remote node
In Figure 43 the remote node design consists of a two-stage passive power splitter (RN1 and
RN2) and is a simple broadcast and select architecture, where all the wavelengths are
broadcast to all ONUs, and the ONU then selects the desired wavelength. This solution has a
serious security threat as the content of all the wavelengths is available to all ONUs. It also
suffers from high power loss due to the high splitting ratio and therefore requires stringent
power budget. This may further restrict the number of users. We assume here that there are Us
TDM-PON infrastructures and each TDM-PON has s ONUs.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 69 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Central office
OLT
RN1
Power splitter
RN2
Power splitter
User
ONU
Uplink
line card
PD
s
PD
ONU control
Uplink control
Uplink
line card
Tunable
Burst Mode
Transmitter
Nu
PD
Downlink control
Downlink
line card
PD
Us
Downlink
Tunable
optical filter line card
DFB
DFB
Nd
DFB
Figure 43 : WDM/TDM-PON as in Figure 42 with cascaded power splitters for the remote node.
In Figure 44 the remote node design consists of a WDM splitter (and some wavelength
combiners) in RN1 followed by a passive splitter in RN2. This design is proposed in order to
overcome the high power loss from the previous remote node design. Instead of transmitting
all the wavelengths to all TDM-PONs, we provide Nd/Us number of downstream wavelengths
and Nu/Us number of upstream wavelengths to each TDM-PON. Us wavelength combiners are
added to offer multiple wavelengths per TDM-PON. Note that this implementation reduces to
the reference system of Figure 41 if Us = Nd (Nu). With the reduced power loss of the WDM
splitter compared to a passive power splitter, the power budget is increased and more users
can be supported. However, the flexibility is somewhat restricted compared to the previous
remote node design.
Central office
OLT
RN1
Wavelength splitter
PD
Nu/Us
PD
Nd/Us
Uplink
line card
s
Nu
PD
User
ONU
ONU control
Uplink control
Uplink
line card
RN2
Power splitter
Tunable
Burst Mode
Transmitter
Us
Downlink control
Downlink
line card
Us
DFB
PD
Downlink
Tunable
optical filter line card
DFB
Nd
Us
DFB
Figure 44 :
Nu/Us
Nd/Us
WDM/TDM PON as in Figure 42 with a WDM splitter and power splitter for the remote
node.
In Figure 45 the remote node design consists of a wavelength router in RN1 followed by a
passive splitter in RN2. The wavelength router consists of a WDM splitter, a passive splitter
stage, optical switches and a passive combiner stage. For downstream transmission, it first
splits each wavelength channel into Us different paths for each TDM-PON. Each TDM-PON
can get data from all Nd downstream wavelengths. There are Nd optical switches installed to
control which wavelengths to be forwarded to which TDM-PON. Note that each wavelength
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 70 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
can be routed to one or more TDM-PONs, providing a selected and dynamic multicast
environment. The upstream data is carried in a separate path.
This architecture improves the broadcast nature and security concerns of the broadcast and
selects WDM/TDM-PON. However, with slow switches based on micro-electro-mechanical
systems (MEMS), the power loss will be even more for this broadcast and select flavour. On
the other hand, fast-switching semiconductor optical amplifier (SOA) switches will
compensate the high losses due to the couplers and splitters, but at the cost of an expensive
solution. However, the amplified spontaneous emission (ASE) noise might still degrade the
performance which can be addressed by optical suppression of noise.
Central office
OLT
RN1
Wavelength router
RN2
Power splitter
User
ONU
Uplink
line card
PD
Us
PD
s
Nu
ONU control
Uplink control
Uplink
line card
Tunable
Burst Mode
Transmitter
PD
Downlink control
Downlink
line card
PD
Nd
Downlink
Tunable
optical filter line card
DFB
DFB
Us
Nd
Nd
Nd
DFB
Figure 45 :
WDM/TDM-PON as in Figure 42 with a wavelength router and power splitter for the
remote node.
In Figure 46 the remote node design consists of a switching configuration with wavelength
selective switches (WSS) in RN1 followed by a passive splitter in RN2. WSS are generally
implemented in MEMS that provide low insertion loss wavelength switching capabilities.
When a WSS is used as reconfigurable optical demultiplexer, a WSS can steer each optical
channel present on its input common port towards one of its output ports, and multiple
wavelengths can be switched to one output port of the WSS (which increases flexibility, but
again requires a tunable Rx at the ONU). On the other hand, in the reverse direction it can be
used as wavelength blocking device, where it can block some of the wavelengths from each of
the ports to enter to the common port. However, it can be configured in such a manner that it
will allow all the possible wavelengths from each of the output ports to enter into the common
port.
Due to the use of WSS the power budget in the downstream direction is much better than for a
WDM/TDM-PON with a wavelength router using MEMS (Figure 45) or a broadcast and
select WDM/TDM-PON (Figure 43). Compared to an AWG used for the WDM/TDM-PON
with WDM splitter (Figure 44), a WSS enhances the optical reconfigurability of wavelength
distribution among the customers according to their traffic demand, and in this way a partially
flexible RN is built. The WDM/TDM-PON using WSS is a long-term solution that needs
further research in a laboratory environment before it can be considered for commercial use.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 71 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Central office
OLT
RN1
Wavelength switch
RN2
Power splitter
User
ONU
Uplink
line card
PD
WSS
PD
nWSS
Nu
s
ONU control
Uplink control
Uplink
line card
Tunable
Burst Mode
Transmitter
PD
Downlink control
Downlink
line card
NWSS
PD
Downlink
Tunable
optical filter line card
DFB
DFB
Nd
DFB
Figure 46 :
WDM/TDM-PON as in Figure 42 with a wavelength selective switch and power splitter
for the remote node.
4.5.2 Hybrid WDM/CDM-PON
Code division multiple access (CDMA) is one of the alternatives for per-wavelength fan-out
in a hybrid PON. There are different splitting approaches at the remote node that can be
combined with hybrid WDM/OCDM.
The main option (Figure 47) is hybrid WDM/OCDM-PON with a WDM splitter (e.g. arrayed
waveguide grating AWG) and several power splitters each of which connects to one output
port of the WDM splitter [27]. In this system, the WDM splitter and power splitters can be
located separately, e.g. the WDM splitter is at a street cabinet close to the central office while
the power splitters are put at distribution (interconnection) points (see Figure 4.4.4) close to
the users. OCDM channels can be overlaid on WDM wavelength channels. It means multiple
users are individually assigned with different optical codes on each WDM wavelength
channel. We assume in this system that there are in total N wavelength channels and each
channel can accommodate M code channels. The total number of users that can be supported
by this system is N × M. OCDMA over WDM might be viewed in a way that a WDM channel
is shared with M users by equally dividing the bandwidth into different code channels. A clear
differentiation from a conventional TDMA is that this channel division can be realized in an
asynchronous manner without using time slots.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 72 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 47 : Hybrid WDM/OCDM PON [27] (OC: optical code, TX: transmitter and RX: receiver).
Generally, two different versions of CDMA must be considered in fibre-optic transmission,
Electrical CDMA (E-CDMA) and Optical CDMA (O-CDMA). E-CDMA refers to a system
where the code multiplex and the related spectrum spreading (both performed through
multiplication with a fast, client-specific chip or code sequence) is done in the electrical
domain, prior to modulating the lasers, and after photo-diode detection, respectively. For
reasonable code multiplexing, say in the range of 10 multiplexed customers, CDMA typically
already requires, to first approximation, a 100-fold bandwidth. This leads to the fact that in ECDMA PON systems, very high analogue bandwidth is required for the transceiver optoelectronics. Here, we considered a bandwidth of 30 GHz which is sufficient to provide ~500
Mb/s per customer non-oversubscribed.
The schematic diagram of an E-CDMA hybrid PON is shown in Figure 48. The ONU is based
on a tunable (non-reflective) transceiver where the code is applied to the upstream signal, and
the complex conjugate of the code sequence is applied to the downstream receive signal.
Depending on bandwidth and per-wavelength multiple-access, multi-channel transceivers can
be used in the OLT. The multi-channel signal is then constructed in the digital domain.
Depending on the resulting bandwidth and the implementation effort required, such an
approach can help saving cost (CapEx) over an implementation using discrete OLT
transceivers. A multi-channel OLT transceiver design is indicated in Figure 48 on the lefthand side.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 73 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OLT
RN1
AWG
Tx/Rx
Array
RN2
1xN AWG
ECDMA
MDXM
ECDMA
MDXM
10G-ECDMA
ONUn
λD λU
1:k

()*
Code
RF Modulation
ADC
DAC

Digital
SCMA
with
CDMA
S+H
Figure 48 : Hybrid WDM/E-CDM-PON
The resulting cost and energy consumption figures (again without baseline numbers) are
summarized in Table 8. Note that in particular the cost / energy consumption assumption for
an OLT per-client port must be considered very optimistic.
Table 8.
Energy consumption and cost figures for hybrid WDM/E-CDM-PON
excluding baseline (chapter 7.1)
Cost
237$ total
Energy
5.85 W
AWG/splitter ports
14$
OLT port incl. digital Tx
0.75 W
OLT port
22$
OLT switching
1.0 W
OLT amplifiers
6$
OLT amplifiers
0.1 W
ONU TRX
175$
ONU incl. CDMA (1.5 W)
4.0 W
ASICs + switching
20$
According to the numbers in Table 8, a hybrid WDM/E-CDM-PON still has low-cost
potential, although it is more costly than a hybrid WDM/TDM-PON. This must be related to
the fact that the required signal processing (code multiplexing) is performed in the electrical
domain, with respective cost assumptions for the key components. However, the energy
consumption of the E-CDMA hybrid PON as compared to the TDMA hybrid PON is ~2 W
higher, on a per-client basis. In a large-scale deployment scenario, this is an unacceptable
increase of power consumption, irrespective of the question if the difference has to be paid by
the network operator or the end customer.
O-CDMA refers to a fibre-optic implementation of code-based multiple access. Application of
the different Optical Orthogonal Codes (OOCs) and the related spectrum spreading is
performed in the optical domain, using passive components like AWGs and FBGs (Fibre
Bragg Gratings). One possible implementation according to recent literature is shown in
Figure 49. Encoding and decoding (E/D) in the ONUs is performed by two identical chirped
FBGs. Different chirp patterns provide different OOCs. In the OLT, simultaneous multi-port
E/D is performed by an N x N AWG. Both components (FBG, AWG) are fully passive and no
severe cost drivers. In particular, they enable low-energy design since the electronic domain
can be restricted to the range of, say, 1 Gb/s. However, in the design shown in Figure 49, two
circulators are required in the ONU. To date, these devices can not be integrated and are
costly. In addition, the ONUs require a (balanced) delay demodulator.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 74 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
OLT
RN1
AWG
Tx/Rx
Array
RN2
1xN AWG
OCDMA
MDXM
OCDMA
MDXM
λD λU
10G-OCDMA
ONUn
1:k
T
Rx
DPSK
NxN
AWG
Multiport
E/D
DPSK
Encoder /
Decoder
Figure 49 : Hybrid WDM/O-CDM-PON
The resulting numbers for cost and energy consumption are listed in Table 9. Energy
consumption is very low compared to the other hybrid PONs, making the WDM/O-CDMPON an attractive contender for NGA considerations. On the other hand, cost is very high
(mainly due to the use of circulators). This leads to the requirement to find more cost-efficient
O-CDMA implementations in order to allow wide-spread use.
Table 9 :
Energy consumption and cost figures for hybrid WDM/O-CDM-PON
excluding baseline (chapter 7.1)
Cost
394$ total
Energy
2.75 W
AWG/splitter ports
13$
OLT port
0.55 W
OLT port incl. E/D (N x N AWG!)
60$
OLT switching
1.0 W
OLT amplifiers
16$
OLT amplifiers
0.2 W
ONU TRX incl. E/D
300$ (Circulators!)
ONU (DPSK)
1.0 W
OLT switching
5$
In addition, it must be noted that to date, combined WDM and O-CDMA fan-out has been
very restricted, up into the range of a total of 128 customers. This is related to two reasons.
First, all hybrid WDM/O-CDM-PON demonstrators so far were based on a small number of
wavelengths (typically, less than 10). Then, (as with all CDMA) the number of useful OOCs
is limited. Up to 16 per-wavelength customers have been demonstrated, each getting several
100 Mb/s dedicated bandwidth. This translated to optical chip rates in the range of up to 600
GChips/s. Hence, the main technical question is if more wavelengths can be used in an OCDMA scheme.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 75 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 50 :
An example of wavelength channel allocation of OCDMA over CWDM using 511-chip
SSFBG en/decoder [27].
Super-structured Fibre Brag Grating (SSFBG) is a promising device for a long optical code
(OC) because of the relatively low insertion loss. A 511-chip SSFBG enconder/decoder has
been demonstrated. The chip length of the grating and the total length are 0.156 and 80 mm,
respectively, which corresponds to the chip rate of 640 Gchip/s with the duration of the
generated optical code of about 800 ps. The measured peak reflectivity is up to about 92%.
Note that this characteristic is insensitive to the polarization state of the input signal, if the
fabrication progress has guaranteed good cylindrical symmetry of the SSFBG. The notches of
the spectrum spread appear at integer steps of 640 GHz (= 5 nm) from the central peak. If the
neighbouring wavelength channel is allocated at the spectrum notch, WDM interchannel
crosstalk can be almost neglected. Figure 50 shows an example of wavelength channel
allocation where the space between two adjacent wavelength channels is 10 nm (2 x 5 nm) in
hybrid WDM/OCDM-PON using the 511-chip SSFBG encoder/decoder (with 640Gchip/s). It
should be noted that the shorter the length of the code is, the lower chip rate and the smaller
space of neighbouring wavelength channels is required. Figure 51 illustrates two other
possible options of the remote node to support capacity upgrade by adding OCDMA channel
[28]. The original splitter may be replaced by employing WDM multiplexers (e.g. AWG) and
wavelength/waveband selectors (WSs). Here, WS devices mean passive WDM multiplexers
with lower wavelength granularity and are typically used to separate data and signalling
wavebands (e.g. red/blue filter). To maintain broadcast signals in the WDM multiplexer
configurations, the broadcast wavelengths for TDM and/or CDM must be separated from the
point-to-point wavelengths for WDM so as to split their power among all users. WS devices
separate or merge the broadcast signals. Note that the PON configuration in Figure 51 allows
both TDM and WDM ONU to remain untouched and co-exist with the OCDMA upgrade of
the individual ONU.
Furthermore, this approach suffers from high power loss due to the insertion loss of the power
splitter and therefore requires large power budget. This may restrict the number of users and
maximum reach between the OLT and the ONU.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 76 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 51 : OCDMA over WDM or hybrid WDM/TDM PONs
Another option for the remote node design (Figure 52) is a WDM splitter (e.g. arrayed
waveguide grating AWG) combined with several code splitters (e.g. AWG based multiport
E/D [21]) each of which connects to one output port of the WDM splitter. The WDM splitter
and the code splitters can be located in separated places. For instance, the WDM splitter is at
street cabinet close to the central office while the power splitters are close to the users. The
signals can be encoded/decoded at the code splitter so that the ONUs do not need any
encoder/decoder. In addition, the insertion loss of the code splitter is similar as for the WDM
DEMUX which is much smaller than the power splitter. Therefore, this approach can
overcome the high power loss of the previous approach (i.e. WDM splitter and power splitter)
and may have larger number of users and longer reach between the OLT and the ONU.
Figure 52 : Hybrid WDM/OCDM PON with code splitting at the remote node.
4.5.3 Hybrid WDM/OFDM-PON
OFDM has recently gained much attention for optical transmission because of its spectral
efficiency and its flexibility and robustness against dispersion, especially if implemented in
the digital domain. For these reasons, it is now also considered for PONs, both singlewavelength and also hybrid WDM-PONs. Generally, OFDM could be combined with other
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 77 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
multiple-access mechanisms (i.e., TDMA, WDMA, or CDMA). Because OFDM is already
based on multiple sub-carriers, SCMA however seems to be the most natural fit. The
respective SCMA variant is then called OFDMA, and since it is implemented in fibre-optic
systems, it is also referred to as O-OFDMA.
Different implementations of O-OFDM have been described in the literature, including
coherent intradyne, coherent heterodyne, and incoherent direct-detection variants (the latter
with Single-SideBand modulation, SSB). They all have in common that the OFDM signals are
constructed in the digital domain, making use of Inverse Fast Fourier Transform (IFFT) in the
transmitter and FFT in the receiver. Figure 53 shows an implementation of an incoherent
hybrid WDM/O-OFDM-PON. Since OFDMA with potentially very tight sub-carrier spacing
is used, the different ONU transmitters of one nominal upstream wavelength must be tightly
locked to that wavelength in order to avoid interference between different sub-carriers. Either
tunable lasers with very precise tuning / wavelength locking or seeded (reflective) transmitters
must be used. Here, an implementation based on seeded SOAs is shown. In order to allow
SSB modulation, a biased MZM is used. The seed wavelength is de-coupled in a band splitter
and then fed into a non-reflective (standard) SOA which is followed by the MZM. The
modulated upstream signal is then coupled into the distribution fibre by means of a circulator.
This way, the SSB signal is built in the analogue electrical domain. In the OLT, optical SSB
modulation is shown as an alternative. Whether this is more efficient in its implementation
remains for further investigation.
1xN AWG
AWG
cos

RF

SOA
-sin
cos
SSB
Filter
MFL

90°
+
RF Modulation
ADC
Digital
OFDM
DAC
OFDM
MDXM
1:k
DC Bias
Tx/Rx
Array
10G-OFDM
ONUn
λD λU
RN2
RF

sin
ADC
Serial-Parallel
Ext. Removal
FFT
M-ary QAM
RN1
OFDM
MDXM
DAC
Parallel-Serial
Cyclic Extens.
IFFT
M-ary QAM
OLT
Figure 53 : Hybrid WDM/O-OFDM-PON
The cost and energy consumption figures are listed in Table 10, according to the assumptions
stated in 7.1. As compared to the other hybrid WDM-PONs, cost and energy consumption
both increased. In the ONU, this again is related to the cost of the circulator. Hence, other
implementation alternatives should be considered. For both ONUs and OLT, the digital signal
processing adds to cost and energy consumption significantly. This is due to the fact that per
client digital OFDM signal processing at the accumulated per-wavelength bandwidth of 10
GHz must be performed. From today’s perspective, this will be the dominant contributor in
particular to energy consumption, together with the contribution coming from the necessary
high-performance 10G transceivers. In this respect, the numbers stated in Table 10 can not be
regarded as pessimistic. However, they lead to unacceptable energy consumption, and also
comparatively high cost. It may however be useful in cases where variable bandwidth or
bandwidth × reach requirements must be supported. The latter can easily be achieved in
digital OFDMA by adaptation of either the per-client number of sub-carriers, or the per-subcarrier modulation depth (e.g., QPSK vs. 16QAM).
Table 10 :
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Energy consumption and cost figures for hybrid WDM/O-OFDM-PON
excluding baseline (chapter 7.1)
Page: 78 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Cost
362$ total
Energy
7.8 W
AWG/splitter ports
12$
OLT port incl. digital Tx
0.75 W
OLT port
22$
OLT switching
1.0 W
OLT amplifiers
3$
OLT amplifiers
0.05 W
ONU TRX
275$ (Circulator!)
ONU
6.0 W
ASICs + switching
55$
4.5.4 Hybrid WDM/SCM-PON
SCMA is similar to OFDMA in that it assigns different sub-carriers to different clients. In the
OLT, a similar – digital – implementation can be chosen. The main difference is that the
ONUs can be based on narrow-band electrical (RF) transceivers. Only the optical transceivers
(lasers, photo diodes) need to support the accumulated bandwidth (here again 10 GHz). This
has significant impact on potential cost and energy consumption. On the other hand,
flexibility with regard to per-client bandwidth is lower as compared to OFDM.
Different SCMA variants have been proposed, again including coherent and incoherent
versions. Figure 54 shows an incoherent hybrid WDM/SCM-PON where the sub-carrier
modulation is done in the digital domain (similar to OFDM). In the OLT, all sub-carrier
channels of one wavelength can be modulated and de-modulated within a single transceiver.
In the ONUs, only narrow-band sub-carrier processing is required because each ONU-specific
sub-carrier is multiplexed / de-multiplexed to / from the 10-GHz signal by means of an
analogue RF section. Similar to the WDM/O-OFDM-PON shown in Figure 53, the hybrid
WDM/SCM-PON shown in Figure 54 is based on seeded ONU transceivers (in this case
REAM-SOAs). This automatically allows tight wavelength locking. As compared to OFDM,
this leads to much simpler design. Like most other hybrid high-performance PONs,
WDM/SCM also requires high-performance 10G transceivers.
1xN AWG
AWG
Tx/Rx
Array
SCMA
MDXM
RN2
λD λU
10G-SCMA
ONUn
1:k

RF
RF Modulation
ADC
Digital
SCMA
DAC
REAM
SOA

RF
MFL
ADC
Modulation
RN1
SCMA
MDXM
DAC
Modulation
OLT
Figure 54 : Hybrid WDM/SCM-PON
Cost and energy consumption numbers are stated in Table 11. It can be seen that cost and
energy consumption are both significantly lower than the respective O-OFDM numbers. Cost
so far is only undercut by WDM/TDM, but energy consumption is still comparatively high.
Hence, means for reducing energy consumption on WDM/SCM-PONs should be further
investigated.
Table 11 :
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Energy consumption and cost figures for hybrid WDM/SCMA-PON
excluding baseline (chapter 7.1)
Page: 79 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Cost
235$ total
Energy
4.8 W
AWG/splitter ports
12$
OLT port incl. Digital Tx
0.75 W
OLT port
20$
OLT switching
1.0 W
OLT amplifiers
3$
OLT amplifiers
0.05 W
ONU TRX
175$
ONU
3.0 W
ASICs + switching
25$
WDM/O-OFDM and WDM/SCM both provide dedicated per-client sub-carriers over an ODN
which consists of filters (AWGs) and power splitters/combiners. Similar multiple access is
provided by coherent ultra-dense WDM-PONs (UDWDM-PONs), with the exception that
carriers are assigned to customers, not sub-carriers. From that, a comparison of hybrid PONs
with UDWDM-PON makes sense although UDWDM-PON typically is not regarded as
hybrid PON.
4.5.5 UDWDM
UDWDM requires coherent receivers in order to allow wavelength spacing as close as 2 or 3
GHz. Ideally, the local oscillator in the ONU is also re-used for upstream. For SFW, this
requires heterodyne detection, where the RF frequency is also the frequency shift between
downstream and upstream. One possible implementation is shown in Figure 55. Coherent
detection requires alignment of the polarization planes of receive signal and local oscillator.
Typically, polarization diversity is implemented in coherent receivers, which leads to dual
balanced receivers and the related cost. (In homodyne detection, it leads to 4 balanced
receivers due to further split into in-phase and Quadrature components.) Balanced receivers
can be reduced to single-photo-diode receivers at the cost of 3 dB penalty regarding
sensitivity. This penalty may be acceptable for ODN with filters and splitters. It will not be
acceptable for splitter-only ODN because of the very high accumulated insertion loss.
RN1
AWG
Tx/Rx
Array
1xN AWG
UDWDM
MDXM
UDWDM
MDXM
DS
–
+
RF
US
Pol.
Scr.
CLK Rec.
3dB
3dB
3dB
RF Modulation
DAC
ADC
1G-UDWDM
ONUn
1:k
3dB
Digital
SCMA
λD λU
RN2
3dB
OLT
Pol. Scr. /
Diversity
T-LD
Figure 55 : UDWDM-PON
Figure 55 shows an alternative implementation with regard to polarization. It is based on a
polarization scrambler which is clocked at twice the symbol rate. Such, both orthogonal
polarizations of the local laser are superimposed with the receive signal within each symbol
duration. This scheme only requires one balanced detector for heterodyne receivers, however
it leads to 3 dB penalty (because only one half of the local laser signal is aligned with the
receive signal in polarization).
The OLT UDWDM transceiver is based on a broadband, multi-carrier implementation.
Several tightly spaced carriers are multiplexed in the digital domain and then modulated onto
broadband optics. This way, the number of transceivers can be reduced, at the cost of broader
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 80 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
bandwidth required. Note that this approach is similar to the one chosen for WDM/SCMA.
Table 12 lists the cost and energy consumption numbers for UDWDM. Cost is relatively high
which relates to the effort which is required for coherent detection and (OLT) wavelength
stabilization. Energy consumption is comparable to WDM/SCM-PON. No broadband (10
GHz) components are required in the ONUs, but higher effort is required for balanced
receivers, and OLT transmitter stabilization.
Table 12 :
Energy consumption and cost figures for WDM/UDWDM-PON
excluding baseline (chapter 7.1)
Cost
312$ total
Energy
4.5 W
AWG/splitter ports
12$
OLT port incl. digital Tx
1.5 W
OLT port
100$
OLT switching
1.0 W
ONU TRX
175$
ONU
2.0 W
OLT switching + ASICs
25$
Comparing WDM/SCM-PON with UDWDM leads to interesting results. From the spectral
viewpoint, both look very similar, especially if UDWDM wavelengths are grouped in order to
be able to pass through WDM filters. Both are similar with regard to energy consumption.
Both can make use of (partly-) integrated, broadband digital OLT transceivers. Both can
support a very high customer number. A hybrid DWDM/SCM-PON with 96 wavelengths of
10 GHz each and 1:16 per-wavelength split could scale to 1536 customers, with per-customer
bit rates of up to 1 Gb/s which then depends on the sub-carrier modulation depth. UDWDMPONs with 3-GHz spacing can scale into the same region (if no guard bands for WDM filters
have to be used). The main differences relate to cost and reach performance. From today’s
perspective, UDWDM requires higher implementation effort and is more costly. On the other
hand, it offers better sensitivity and hence better reach performance without amplifiers. Over
short distances, it can also support splitter-only ODNs. Since both approaches have similar
reach/cost performance, the most important question relates to the reach requirement. If very
high reach really is required, UDWDM may be the preferable choice. If reach in the 50…60
km range is regarded sufficient, WDM/SCM-PON is the more cost-efficient choice.
4.6 HYBRID ACTIVE WDM
Since OASE aims at access solutions enabling high reach, high fan-out and high (total or perclient) bandwidth, we must also consider passive/active hybrid solutions.
4.6.1 Hybrid active WDM with active P2P access
TXFP
PoP
PRN
TXFP
...
TXFP
SFP
Scalable
Universal
Switch
AWG
TXFP
L2
AWG
...
SFP
Figure 56 shows a combined access/backhaul hybrid active/passive WDM-PON. It is based
on 10G DWDM for backhaul and broadband users, and active P2P fan-out for residential
access. WDM-PON interfaces are based on Tunable XFPs (TXFP).
ONU
SFP
ARN / LX
Figure 56 : Hybrid active WDM-PON with active P2P access
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 81 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
It is obvious that an approach according to Figure 56 can support high reach (easily up to 100
km) and very high fan-out, at the cost of active RNs. However, it can be shown that this
approach has very low cost (CapEx) as well as low energy consumption. This can be seen
from the respective numbers in Table 13.
Table 13.
Energy consumption and cost figures for hybrid active WDM-PON with active P2P access
excluding baseline (chapter 7.1)
Cost
147$ total
Energy
3.3 W
AWG ports
2$
PoP switching
1.0 W
10G TRX
100$ (!)
LX switching
1.0 W
Switch (PoP)
5$
LX TRX (2 x TXFP)
0.3 W
Switch (LX)
10$
CPE TRX (2 x grey SFP)
1.0 W
CPE TRX (2 x grey SFP)
30$
Both very low cost and low energy consumption can be related to the possible use of
cheapest, simplest 1G transceivers for the active fan-out – grey 10-dB SFPs. These
transceivers easily undercut every other transceiver type and lead to lowest cost, even though
integrated backhaul with TXFPs has been considered. In addition, low-energy SFPs can go
down to ~0.4 W which again makes them the most efficient solution for dedicated 1G
services. Also note that energy consumption is comparatively low because for all other
solutions, a first aggregation switch is also included (mostly, in the OLT).
4.6.2 Hybrid active WDM-PON
Scalable
Universal
Switch
AWG
AWG
...
TXFP
TXFP
TXFP
ARN / LX
PoP
ONU
TSFF
AWG
OLT
TXFP
WDM PIC
A derivative of the active/passive hybrid access solution mentioned before is the PON-inPON concept, see Figure 57. It requires the OLT of an access WDM-PON to be
accommodated in an active RN. These OLTs are then backhauled by a WDM-PON running
10G wavelengths. This solution obviously requires other distribution fibres in the ODN, as
compared to the approach with active P2P access. It has very high reach and capacity
performance as well.
PRN2
PRN1
Figure 57 : Hybrid active WDM-PON (“PON-in-PON”)
Cost and energy consumption numbers of the PON-in-PON approach are listed in Table 14.
There is already a significant difference against the approach with active P2P access. Higher
cost and also energy consumption of PON-in-PON can be attributed to the higher
performance of the access WDM-PON, as compared to grey SFPs. The access WDM-PON
already has 25-dB transceivers, which in addition are DWDM-tunable.
Table 14 :
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Energy consumption and cost figures for hybrid active WDM-PON
excluding baseline (chapter 7.1)
Page: 82 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Cost
262$ total
Energy
3.6 W
AWG ports
12$
PoP switching
1.0 W
10G TRX
120$
LX TRX (2 x TXFP)
0.35 W
Switch (PoP)
5$
OLT Array port
0.5 W
WDM-PON OLT port (LX)
50$
ONU TRX
1.0 W
CPE TRX (2 x grey SFP)
75$
OLT Switching
0.75 W
As compared to all other (hybrid, UDWDM) access solutions discussed in this chapter, the
PON-in-PON approach performs very well. Only the hybrid WDM/O-CDM-PON has lower
energy consumption. However, this approach so far only supports up to ~128 customers. Only
WDM/TDM-PON and WDM/E-CDM have lower-cost potential. E-CDM leads to prohibitive
energy consumption, and fully passive WDM/TDM has much lower reach performance.
Hence, the hybrid active/passive PON-in-PON approach must be considered one of the more
powerful alternatives for NGA with high reach / fan—out / bandwidth requirements.
4.6.3 Comparison of active/passive hybrids and WDM-PON
From the discussion given before it becomes clear that active P2P access has advantages with
respect to cost and energy consumption if it can be based on cheapest, lowest-power SFPs.
However, it requires dedicated distribution fibres and has relatively poor foot print in the OLT.
Both can translate to added cost, or may prohibit this approach. It also became clear that any
added per-wavelength fan-out adds to cost and energy consumption, in particular if dedicated
bandwidths in the range of 500 Mb/s to 1 Gb/s are required. It should also be noted that the
cost and energy consumption numbers for the hybrid WDM/TDM-PON to first approximation
also hold for XG-PON (in particular XG-PON2 which is the 10G symmetric variant). Hence,
it is necessary to compare these numbers also to a simple, stringent DWDM-PON which may
come out as the ultimate reference and benchmark with regard to systems CapEx and perclient energy efficiency.
Figure 58 shows the schematic block diagram of a DWDM-PON which is based on OLT PICs
and low-cost tunables. This approach has been discussed in Chapter 4.2. It must be noted that
there are clear design goals for the tunables in the sub-100$ and sub-1-W region. Current
status of the related components research indicates that these numbers are feasible (see the
tunable lasers chapter).
RN
L-Band 100 GHz
Tx Array
1
...
Mux
PIC
N
Cyclic AWG
1
...
C
C-Band ~100GHz
N
Rx Array
L
Cyclic AWG
OLT
ONU
T-LD
SFF
Rx
Identical Cyclic AWGs
Figure 58 : Reference WDM-PON
Table 15 states the relevant cost / energy consumption performance numbers of the WDMPON shown in Figure 58. From these performance figures, it can be seen that a simple WDMPON is superior over all other access concepts with regard to system CapEx and per-client
energy consumption if dedicated broadband services (500 Mb/s and higher) are required.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 83 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 15 :
Energy consumption and cost figures for reference WDM-PON
excluding baseline (chapter 7.1)
Cost (per client)
150$ total
Energy (per client)
2.5 W
AWG ports
20$
OLT array port
0.5 W
OLT array port
50$
OLT switching
1.0 W
ONU TRX
75$
ONU
1.0 W
OLT switching
5$
From all other approaches analyzed in sections 4.5 and 4.6, only the hybrid active/passive
WDM-PON with active P2P access has similar cost (it is actually marginally lower).
However, no other solution has similarly low energy consumption, in particular for dedicated
1-Gb/s services.
It must be noted that there is a single enabler for the cost / energy consumption performance
of a stringent WDM-PON: the possibility to use lowest-cost, 1-Gb/s, small-power-budget
transceivers. Here, small power budget (at 1 Gb/s!) refers to the 25-dB class. A simple WDMPON without any further per-wavelength fan-out is the only approach where such transceivers
can be used. This does include the UDWDM-PON approach. It also runs at 1 Gb/s per
wavelength, however it requires highly stabilized laser sources and coherent receivers. All
other approaches require added effort – on the per-client level! – which also adds to
complexity, cost, and energy consumption. In most cases, high-performance (32…35 dB
power budget) 10G transceivers are required, with additional means for per-wavelength fanout. It is doubtful that these 10G transceivers will ever have the cost and energy consumption
potential of simpler 1G transceivers. It should once again be noted that equally aggressive
cost and energy consumption figures have been assumed for the different transceivers in 7.1.
From the discussion in this chapter it becomes clear that a stringent WDM-PON approach has
good potential to outperform any other access approach with regard to per-client CapEx and
energy consumption. This holds in particular when being compared against single-wavelength
(40G, 100G) next-generation PON approaches. It is extremely doubtful that such approaches
can ever develop low-cost or low-energy potential. In addition to that, a WDM-PON has good
reach performance. The lowest-cost version analyzed herein so far – 25-dB class – can easily
span some 50 km in urban areas. On good (low insertion loss) fibres, it already supports some
60+ km. It is likely that this reach covers the majority of applications and hence avoids an
over-engineered solution. Nonetheless, a WDM-PON can also easily be extended in its
maximum reach be means of OLT-based amplifiers. It can then approach the 80…100 km
reach domain at reasonable added cost and energy consumption (caused by the amplifiers).
There is one obvious disadvantage of the WDM-PON: its maximum client count. A simple
WDM-PON is likely to be restricted to 80, possibly 96, bi-directional channels, using 50-GHz
grid and cyclic C/L-band AWGs. This is an order of magnitude away from OASE’s target fanout number. It may be possible to extend the WDM-PON – refer to Chapter 4.2 – to both, the
25-GHz grid and also into the S-band. This may enable a theoretical maximum of up to 384
channels. The respective system concept will still be relatively simple, low in cost, and low in
energy consumption. However, it will still not have the potential for 1000 customers. Hence,
in order to avoid over-engineered – too costly and energy-consuming – solutions it must be
clarified very carefully where the optimum per-feeder-fibre client number is. This has to
include the consideration that beyond a certain per-feeder-fibre client count, at least the feeder
fibre itself (and possibly also part of the OLT equipment) will require resilience mechanisms
in order to avoid excessively high failure penetration ranges in case of failures where key
components (single points of failure) are affected.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 84 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
4.7 NEXT GENERATION AON: RPR, WDM, ETHERNET
Point-to-point (PtP) optical links provide the basis for active optical networks (AON) that can
support a variety of tree-, ring- or mesh-based FTTx topologies. This section discusses PtP
technologies relevant for next-generation access. Any of the defined IEEE 802.3 physical
layers are relevant for AONs. Table 16 shows a rough outline of some of the relevant PHY
standards that are either finalized or in development.
Table 16 :
Selection of Ethernet PHY standards
Standard
Fibre
Range
Bitrate
100BASE-FX
MMF/SMF
400m-10 km
100 Mb/s
100BASE-BX
SMF
40km
100 Mb/s
1000BASE-SX
MMF
550m
1 Gb/s
1000BASE-LX10
SMF
10 km
1 Gb/s
1000BASE-ZXX
SMF
70 km
1 Gb/s
10GBASE-LR
SMF
10 km
10 Gb/s
10GBASE-ER
SMF
40 km
10 Gb/s
40GBASE-LR4Y
SMF
10 km
40 Gb/s
100GBASE-LR4Y
SMF
10 km
100 Gb/s
100GBASE-ER4Y
SMF
40 km
100 Gb/s
X
: de facto standard, Y: standard not yet finalized
4.7.1 GbE access
Gigabit Ethernet (GbE) is defined in the IEEE 802.3-2008 standard and refers to various
technologies for transmitting Ethernet frames at a rate of gigabit per second. The specification
allows for half-duplex gigabit links connected through hubs, but full-duplex with switches are
more common on the market.
For single-mode fibre, three main physical layer variants are defined. The first is
1000BASE-LX, with 1310 nm for both downstream and upstream transmission, specified to
work over a distance of ≤5 km over 10 µm single-mode fibre. An extended version –
1000BASE-LX10, with a reach of 10 km – was standardized six years after the initial gigabit
fibre versions as part of the Ethernet in the First Mile task group. The extended version is
similar to 1000BASE-LX, but with increased reach due to higher quality optics. In practice
1000BASE-LX10 was already in widespread use before standardization as vendor proprietary
extensions referred to as 1000BASE-LX/LH or 1000BASE-LH. The second variant is the
duplex solution 1000BASE-BX10, with 1490 nm for downstream and 1550 nm for upstream
transmission, specified to work over a distance of ≤10 km over a single strand of single-mode
fibre. Because of the different wavelengths in the downstream/upstream there is an
asymmetry between the terminals on each side of the fibre. A third variant is 1000BASE-ZX,
with 1550 nm for both downstream and upstream transmission, specified to work at a
maximum distance of 70 km to 120 km. 1000BASE-ZX is not an IEEE standard, but rather an
industry-accepted term to refer to gigabit Ethernet transmission using the 1550 nm
wavelength to achieve distances of at least 70 km over single-mode fibre.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 85 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Transceivers for GbE are mature and available on the market (e.g. Oesolution, Delta,
Lasermate, etc.). Transceivers with 1310 nm for both transmission and reception are based on
FPs or DFBs with a range of up to 40 km. Transceivers with 1550 nm for both transmission
and reception can have a reach of up to 120 km. Transceivers using different wavelengths for
transmission and reception, e.g. transmission at 1310nm and reception at 1550nm or vice
versa, have a range up to 60 km. All these transceivers are typically packaged in SFP
packages. The power consumption of a GbE transceiver for a 10 or 20 km link is within 1W.
Cost varies widely, indicatively an SFP will cost from 10 USD to 100 USD (for 100 units
purchases) depending on distance and quality.
4.7.2 10 Gb/s Ethernet access and distribution/aggregation
The 10 Gigabit Ethernet (10GbE) standard defines Ethernet with a nominal data rate of 10
Gb/s. The standard was first published in 2002 as IEEE Std 802.3ae-2002 and over the years
two variants for 10GbE over fibre have been defined: 802.3ae-2002 (fibre -SR, -LR, -ER and
-LX4 PMDs) and 802.3aq-2006 (fibre -LRM PMD with enhanced equalization). The 802.3ae2002 amendment was consolidated into the IEEE 802.3-2005 standard. IEEE 802.3-2005 and
the other amendments have been consolidated into IEEE 802.3-2008. 10 Gigabit Ethernet is
based on full duplex links which can be connected by switches. Half duplex operation and
CSMA/CD (carrier sense multiple access with collision detection) are not supported in
10GbE. The 10 Gigabit Ethernet standard encompasses a number of different physical layer
(PHY) standards. At the time the 10 Gigabit Ethernet standard was developed there was much
interest in 10GbE as a WAN transport which led to the introduction of the concept of the
WAN PHY for 10GbE. This operates at a slightly slower data-rate than the LAN PHY and
adds some extra encapsulation. The WAN PHY and LAN PHY are specified to share the same
PMDs (physical medium dependent) which means that 10GBASE-LR and 10GBASE-LW can
use the same optics. In terms of number of ports shipped, the LAN PHY greatly outsells the
WAN PHY. The different 10GbE PHY standards are summarized in the following.
10GBASE-LR ("long range") uses the IEEE 802.3 Clause 49 64B/66B Physical Coding
Sublayer (PCS) and 1310 nm lasers. It delivers serialized data over single-mode fibre at a line
rate of 10.3125 Gb/s. 10GBASE-LR has a specified reach of 10 kilometres (6.2 mi), but
10GBASE-LR optical modules can often manage distances of up to 25 kilometres (16 mi)
with no data loss. Fabry–Pérot (FP) lasers are commonly used in 10GBASE-LR optical
modules. Fabry–Pérot lasers are more expensive than VCSELs but their high power and
focused beam allow for efficient coupling into the small core of the single mode fibre.
10GBASE-ER ("extended range") uses the IEEE 802.3 Clause 49 64B/66B Physical Coding
Sublayer (PCS) and 1550 nm lasers. It delivers serialized data over single-mode fibre at a line
rate of 10.3125 Gb/s. 10GBASE-ER has a reach of 40 kilometres.
Several manufacturers have introduced 80 km range ER pluggable interfaces under the name
10GBASE-ZR. This 80 km PHY is not specified within the IEEE 802.3ae standard and
manufacturers have created their own specifications based upon the 80 km PHY described in
the OC-192/STM-64 SDH/SONET specifications. The 802.3 standard will not be amended to
cover the ZR PHY.
Optical modules are not specified in 802.3 but by multi-source agreements (MSAs). The
relevant MSAs for 10GbE are XENPAK, X2, XPAK, XFP and SFP+. XENPAK was the first
MSA for 10GE and has the largest form factor. X2 and XPAK which were introduced later
have smaller form factors, but have not been as successful in the market as XENPAK. XFP
which was introduced after X2 and XPAK provides further reduction of the form factor.
The newest module standard, SFP+, developed by the ANSI T11 fibre channel group is
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 86 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
smaller than XFP and allows for reduced power dissipation. SFP+ is now the most popular
socket for 10GbE systems. SFP+ modules only do optical to electrical conversion, without
clock and data recovery, putting higher burden on the host channel equalization. SFP+
modules share a common physical form factor with legacy SFP modules, allowing higher port
density than XFP and re-use of existing designs for 24 or 48 ports in a 19" rack width blade.
Optical modules are connected to a host by either a XAUI, XFI or SFI interface. XFP modules
use a XFI interface and SFP+ modules use a SFI interface. XFI and SFI use a single lane data
channel with the encoding specified in IEEE 802.3 Clause 49.
SFP+ is an upgrade of SFP for 10 Gb/s transmission. SFP transceivers are available with a
variety of different transmitter and receiver types, allowing users to select the appropriate
transceiver for each link to provide the required optical reach over the available optical fibre
type. Optical SFP modules are commonly available in several different categories: 850 nm
550m MMF (SX), 1310 nm 10 km SMF (LX), 1550 nm [40 km (XD), 80 km (ZX), 120 km
(EX or EZX)], and DWDM. SFP dimensions are: 8.5 x 13.4 x 56.5 mm.
The XFP (10 Gigabit Small Form Factor Pluggable) is a hot-swappable, protocol-independent
optical transceiver, typically operating at 850nm, 1310nm or 1550nm, for 10 Gigabit per
second SONET/SDH, Fibre Channel, gigabit Ethernet, 10 gigabit Ethernet and other
applications, including DWDM links. It includes digital diagnostics similar to SFF-8472 but
more extensive, and provides a robust management tool. The XFI electrical interface
specification is a portion of the XFP Multi Source Agreement specification. XFP was
developed by the XFP Multi Source Agreement Group. The physical dimensions of the XFP
are slightly larger than the standard SFP. One of the reasons for the increase in size is to allow
for on board heat sinks for greater cooling. XFP dimensions are: 8.5 x 18.3 x 78 mm.
Transceivers for 10 GbE are mature and available on the market (Brocade, Oesolution, JDSU,
Delta, Lasermate, etc.). They use the same wavelength for transmission and reception, and as
for GbE transceivers, the wavelength used depends upon the target reach: for up to 10km
(standard 10GBASE-LR/LW), 1310nm is used, while transceivers for longer haul (standard
10GBASE-EW/ER for 40km and 10GBASE-ZW/ZR for 80km) use 1550nm. Common form
factors are SFP+ and XFP. The power consumption lies around 2.5W and 3.5W for link
lengths of 10km and 40km, respectively. Cost varies widely, indicatively an SFP will cost
from 100 USD to 1000 USD (for 100 units purchases) on distance and quality.
4.8 RADIO OVER FIBRE BACKHAULING
Radio over Fibre (RoF) has up to now been conventionally considered as an analogue
technology, where an optical wavelength which has been modulated by an RF signal is
transmitted along a fibre. Considering the scope of OASE we do not consider Fibre-to-theAntenna technologies, e.g. spanning the last few metres between a Mobile Access Node
(NodeB, Base Station, etc.) and an antenna by fibre instead of coaxial cable. Instead, we
consider RoF technologies, where analogue RF signals are transmitted from an optical source
located within a CO or other appropriate active site via SMF towards a suitable
environmentally resilient optical-electrical (OE) converter, filter or amplifier unit in the field
within a typical radius of 15 km. In addition to RoF technologies, we also consider digital
interfaces (CPRI, OBSAI) with similar advantages as the RoF technologies but where IQsamples are carried digitally over the system. We also discuss the more general topic of
integration of wireless technologies in next-generation optical access.
The development of RoF systems has generally been motivated by the desire to replace a
central high power antenna with a low power distributed antennas system (DAS). RoF
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 87 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
systems therefore frequently consist of multiple base stations (BS’s) which are connected to a
single central office (CO). A key aspect to RoF development has been the effort to reduce BS
cost and migrate complexity up the network towards the CO [32]. More centralized radio
access architectures are being considered as possible candidates for future radio access
networks (RAN) with concepts such as main-remote and cloud-RAN (CRAN). Typically this
involves partitioning the traditional monolithic base station to a central base band unit serving
several remote radio units over for example RoF or CPRI. For CPRI a maximum reach of at
least 10 km is required. However, maximum reach up to 100 km have been discussed which
would enable even larger degrees of node consolidation in the RAN and potential cost and
energy savings.
In the access context, there are various useful inherent advantages to RoF analogue
transmission and the mentioned digital interfaces. Using analogue transmission avoids the
requirement for complex hardware (e.g. CPU, FPGA, DACs etc.) in the field, and therefore
offers simplified and cheaper operations and maintenance costs. It also offers a simplified
modular concept for hardware in the field which reduces repair times. Fewer spare parts are
required, of which many are universal for a lot of applications especially when it comes to
convergent access networks, so reducing inventory and storage sizes, simplifying logistics,
whilst also optimizing the recycling management of used parts.
Analogue transmission also offers low group-delay, in contrast to the processing delay
associated with the CPU’s of distributed Mobile Access Nodes, allows effective MIMO
implementation i.e. with up to 50 antennas in an array on a single mobile access node located,
for example in the CO. A centralized Mobile Access Node allows the computation of ideal
values for all the connected antennas situated in the field, so as to achieve maximum space
diversity and receiving power aggregation. In this way an optimum carrier-to-interference
ratio can be reached on the fly without the need for comprehensive signalling messages
between several Mobile Access Nodes placed close to each antenna site. In contrast to
femtocells inside the home, a Microcellular Antenna system close to customers can aggregate
many more subscribers (due to the higher coverage). This means that fewer Network
Elements are required and/or need to be managed. Costs for many small antenna sites (i.e.
telephone poles) could be cheaper than one big antenna site with many antennas on a roof-top
(e.g. due to rental fees). Also there are potential aesthetic advantages, since such a small
antenna can possibly be better integrated within a facade. Remote fixing of complex problems
or upgrading the software on-the-fly via a second Ethernet port is possible from “on-site” at
the CO. This means that O&M manpower doesn’t need to go into the field anymore. The
outdoor unit can be mounted on many small places as it can be small in size and cheaper
(fewer parts inside) than classical Mobile Access Nodes with their comprehensive features
and functionalities. As the number of outdoor units builds up in many places, the microcells
coverage area can be made small. With the coverage area being small, the distance to end-user
equipment (i.e. mobile phone or device) is smaller so that less power needs to be transmitted
(e.g. due to field-path attenuation affects). Another advantage is the flexibly installation to
almost any place, so as to avoid natural obstacles such as trees, buildings etc., which cause
high attenuations. In that way there is less interference in the field, so allowing the possibility
of a higher effective throughput. An Optical Line Termination, Mobile Access Node and
Cable Mode Termination System virtualisation on a distributed computing cloud (Blade
Servers in the CO) allows the shutting down of parts of the computing cloud (CPU sleepmode) during times of low traffic intensity and therefore contributes to saving electrical
energy.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 88 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
4.8.1 Radio-over-Fibre (RoF) Technologies
RoF is characterized by up-conversion from the RF to the optical domains, and can be
realized either by direct laser modulation or external modulation methods [33]. Direct
methods have the advantages of simplicity and low cost. However, limitations to this
approach include the relatively limited bandwidth (10 GHz) of low-cost lasers, e.g. VCSELs,
as well their high chirp, non-linear and inter-modal distortion, and an SNR limited by relative
intensity noise (RIN). In contrast, common external modulation methods include:
-
Mach-Zehnder (MZ) interferometer based approaches, but which are still
characterized by limited bandwidths (2-3 GHz), high linearity, low chirp, and high
bias voltages. However, travelling wave configurations of the MZI-based modulator
allow the bandwidth limitations to be overcome.
-
Electroabsorption modulator (EAM) approaches [33], which are characterized by
higher operating bandwidths. EAMs based on the quantum confined Stark effect
(QCSE) in quantum wells can also exhibit advanced performance.
Down-conversion from the optical down to the RF domains can be implemented by PIN and
avalanche photodiode (APD) photodetectors. These are characterised by their relative
simplicity, and typically offer bit-rates of few 10's of Gb/s (e.g. using a surface illuminated
PIN) [34]. Travelling wave (TW) PIN based receivers employing QW technology can also be
used to generate microwave power, with the 3dB frequency response of such a TW
photodetector being approximately 100 GHz.
An electroabsorption transceiver (EAT) can act as a receiver for the downlink direction, as
well as a modulator for the uplink. This approach tends to be most appropriate for mobile and
metro systems, where the cost of the BS needs to be significantly reduced with the complexity
migrated up towards the CO. In this case, an optical amplifier such as an EDFA is often
required so as to compensate the link losses. EATs operate mainly in the 1500 nm wavelength
window, making them compatible only with SMF, Fabry-Perot and/or DFB lasers. However,
EAT technology with its multi-quantum well (MQW) III–V semiconductor active waveguide
exhibits bandwidths of up to 60 GHz, and thus is a highly promising candidate for BS’s in
future broadband wireless access systems [34],[35].
In general, analogue transmission is applicable for the following signal types, including
amongst others:
-
DVB-C (Digital Video Broadcast for Cable TV, also DVB-C2)
-
EuroDOCSIS3.0 (incl. older and future variants and other coax cable RF modulated
signals)
-
GSM, GPRS, EDGE, UMTS, HSPA, LTE and LTE enhanced signals
-
Satellite TV signals (analogue and QAM modulated – Local Satellite Distribution
Service)
-
Analogue TV signals (i.e. PAL or SECAM QAM modulation)
Using a dedicated spectrum (Lambda) for RoF analogue transmissions provides more capacity
for IP-based (e.g. GPON, ATM, Gigabit Ethernet, etc…) access technologies (e.g. out-of-band
transmission, optionally combined on a single fibre) in a highly scalable way. For example,
whereas many subscribers may be satisfied with a low bandwidth capacity (e.g. in the Mb/s
region as provided by EuroDOCSIS) other customers in the same area may prefer high
bandwidth services (e.g. 3D TV) and can get their service via FTTH.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 89 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
4.8.2 CPRI
The Common Public Radio Interface (CPRI) standard [36] defines the interface between
Radio Equipment Controllers (REC) and Radio Equipment (RE) and covers both the physical
and datalink layers. It specifies the necessary items for transport, connectivity and control,
including user plane data, control plane transport mechanism and means for synchronization.
The specification complies with 3GPP UTRA FDD (release 8), WiMax (release 1) and 3GPP
E-UTRA (release 8).
The physical layer must fulfill the requirements relating to clock stability and noise, as well as
the requirements to satisfy a BER < 10-12. Several physical layer line rate options are defined.
These are defined as multiples of 614.4 Mb/s (1x, 2x, 4x, 5x, 8x, 10x). The line rates are
chosen such that the UMTS chip rate (3.84 Mb/s) can be easily recovered. The physical layer
supports both electrical and optical interfaces. Regarding the optical interfaces, for each line
rate a short reach and a long reach option has been defined. The CPRI standard recommends
reuse of optical transceivers from the following standards:
-
Gbit Ethernet: Standard IEEE 802.3-2005 [37] clause 38 (1000BASE-SX/LX)
10 Gbit Ethernet: Standard IEEE 802.3-2005 [37] clause 53 (10GBASE-LX4)
Fibre channel (FC-PI) – Standard ISO/IEC 14165-115 [38]
Fibre channel (FC-PI-4) – INCITS (ANSI) Revision 8, T11/08-138v1 [39]
Infiniband Volume 2 Rel 1.1 (November 2002) [40]
Serial transmission is based on 8B/10B line coding according to IEEE 802.3-2005 [36],
clause 36. The length of a basic frame is 1 TC = 1/fc = 1/3.84 MHz = 260.416667 ns. Each
frame consists of 16 words and the length of each word (8 bits-80 bits) depends on the line
rate.
4.8.3 OBSAI
The Open Base Station Architecture Initiative (OBSAI) defines an open standardized internal
modular structure for the Base Transceiver Station (BTS). Four main blocks are defined: radio
frequency block, baseband, control and clocking block, and transport block. Relevant for
OASE is the definition of the external network interface of the transport block of the BTS. At
this interface according to OBSAI [41] user data and control data should be transported over a
suitable communications path such as T1, E1, DS3, OC1, OC3, Ethernet, or DSL.
For 3GPP systems, the logical interface to the Radio Network Controller (RNC) is designated
by the Iub, described in the 3GPP Series 25.4xx specifications. For GSM/EDGE specification
of the transport interface is given in the Series 3GPP/08.xxx. For 3GPP2 systems, the logical
interface to the Base Station Controller (BSC) is specified by the Abis interface in the
TIA/EIA IS-828-A specifications. For 802.16/WiMax systems, the logical interface is
specified as R6 to the ASN GW (centralized GW) or R3 to the CSN (distributed GW).
The transport block performs interworking functions between the external network interface
and the interfaces as defined by the OBSAI Reference Point 1 Specification and the OBSAI
RP2 Specification.
Regarding upper protocol layers (Radio Network Layer, RNL), these may be RAN/BSS
vendor proprietary or defined by standardization bodies dedicated to the specification of
RAN/BSS network interface functions. OBSAI states that the transport block shall only
support Transport Network Layer (TNL) functions and not any RNL related tasks; it does not
terminate any RNL related protocols. Hence, all RNL traffic (User Plane, Control Plane and
Management Plane) is passed transparently through the transport block, whilst still handled
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 90 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
according to associated QoS attributes. Differing RAN/BSS topologies (e.g. tree, star, and
ring) are also supported.
4.8.4 Hybrid Optical-Wireless PON Architectures
In general, discussion of integration of wireless technologies in next-generation optical access
networking has been a topic of particular research interest, e.g. [42]. In particular, a key aspect
of research is the exploitation of WDM-PON topologies to act as backhaul infrastructures,
e.g. for IEEE 802.16 [43] and IEEE 802.11 [44] networks, where wireless links can also be
considered to form the basis for a robust and resilient multihop wireless networks (mesh
networks). Considering wireless broadband access in particular, favoured technology options
up to now have included the UWB, IEEE 802.16 and IEEE 802.11 technologies, whilst with
regard to the two IEEE standards, the IEEE 802.16m and IEEE 802.11n [45] amendments
have been created to help maximise available end-user data rates to above 100 Mb/s and
beyond. Fixed and/or mobile wireless access is also a key research theme, with significant
recent effort invested to create solution platforms, e.g. WiFi, (802.11x), and WiMax for both
fixed access (802.16-2004 or 802.16d) and mobile provision (802.16e-2005.) Ultra wide band
(UWB) radio with the ECMA-368, ECMA-369, as well as ISO/IEC 26907 standards has also
opened up important new opportunities, with a particular impact on the overall wireless
perspective of future 4G systems also expected.
Associated with these standards is the frequency allocation 3.1 GHz to 10.6 GHz, whilst there
is also the generic “60 GHz” option. For this latter frequency band, the recent release of the
57–64 GHz band in IEEE802.11ad for commercial applications such as wireless HDTV
connectivity is an important development. IEEE802.11n has been created to offer up to
300Mb/s at 40 MHz (with short guard interval), with WiMax offering a somewhat lower
bandwidth of up to 72 Mb/s but potentially over distances of up to 10 km, in comparison to
the 100m distances of WiFi. We note that WiFi 802.11a can also offer a longer range lower
bit-rate option too (802.11-2007 standard), indicating the overall flexibility of the wireless
approach and the possibility of various service platforms.
Integrating a wireless mesh network with an optical backhaul topology has been especially
studied in [46] describing a hybrid architecture concept. Particular aspects considered include:
a solution for a reconfigurable optical backhaul topology so as to load balance capacity
amongst the various optical links in the backhaul side, whilst simultaneously providing an
optimized routing approach from end-users to the RAU across the nodes of the wireless mesh
network. As noted in the study, such a hybrid approach also assists in the deployment of
optical fibre, since cost-sharing among operators of fixed and wireless networks is
intrinsically enabled. In addition, open access topologies are also supported; in contrast to the
separate infrastructures in operation today.
Other recent studies have also analysed hybrid access network architectures consisting of
optical back-end and gateway-routers, which connect end-users using either devices based on
the IEEE 802.11 or IEEE 802.16 standards [47]. In particular, the challenges arising from
deploying these hybrid PON-wireless access networks topologies with appropriate basestation placements and the associated routing issues are also considered. Improved versions of
routing algorithms designed to improve network efficacy are presented; although Quality of
Service (QoS) issues have been neglected in these studies, as has been appropriate Medium
Access Control (MAC) algorithms to optimise the performance of the overall access network.
An important aspect to consider is the avoidance of outage of a large number of end-users in
the event of individual node failure – this can be achieved by appropriate load-balancing
techniques in the WDM and/or TDM domains. For example, traffic prioritization capability is
already inherent to the IEEE 802.16 and IEEE 802.11e/n standards, via a dynamic bandwidth
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 91 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
allocation (DBA) algorithm for application in WDM-PONs. In the context of both
connection-orientated and connectionless data traffic current research issues also include
appropriate prioritisation algorithms and load-balancing within a converged optical-wireless
access network.
Other approaches that have been described which consider the integration of optical
networking with wireless technologies, both within the metro as well as the access contexts,
are considered in [48].
There are several advantages with optical-wireless optical architectures that should be
considered. In areas where it is difficult to bring fibre into a living space, or the acceptance of
a new installation is difficult at least from the customer point of view (e.g. when the final few
metres would otherwise be a coax access solution), an LTE-enhanced or similarly based
access technology provides an important broadband access solution (e.g. Gigabit Level) as
well as significant flexibility in the early years when a FTTH take-up rate may not be
particularly high (e.g. only a 5-20% take-up rate in the first few years.). DIY kits allow
subscribers to plug into existing access (LTE enhanced, coax wall-outlet) provision to allow
them to initially sample a limited broadband service until they decide to go for the full
installation of fibre into their living space. Where a geographical area has completely
migrated from enhanced LTE and HFC (EuroDOCSIS) the active part in the external
hardened box can be shut down (the redundant parts removed) and only the optical power or
colour splitters retained within the box. A single Operations and Maintenance system for all
access technologies (EuroDOCSIS, LTE, GSM, FTTH) provides better failure traceability and
reduces costs. (Naturally, this is addition to a further transport network O&M system.).
5. Survey of evolving components/subsystems
For each of the covered NGOA system concepts in Chapter 4 the system description contains
a presentation of the main system specific components/subsystems. Some of these can be
further identified as key evolving components/subsystems for the NGOA. In this chapter key
components/subsystems are identified. These are then described with respect to a number of
key aspects defined in Chapter 3.
Starting with the WDM based systems including pure WDM-PON, as well as the passive and
active hybrids; these rely on the evolution of a large number of different
components/subsystems depending on system variant. In section 5.1, general WDM
components/subsystems such as wavelength selective components are discussed. Section 5.2
discusses different reflective transmitters used at the ONU in several WDM based PON
concepts. Sections 5.3 and 5.4 cover tunable lasers and wavelength selective receivers which
are basic building blocks for a large number of WDM system concepts. Section 5.6 discusses
analogue-to-digital converters (ADC) and digital-to-analogue converters (DAC) used in
system based on advanced modulation. Some system concepts, and especially 40G TDMPON, require dispersion compensation to meet reach requirement. Different methods for
dispersion compensation are discussed in 5.7. Section 5.8 contains a survey of various passive
wavelength selective devices such as thin film filter-based WDM components, fibre Bragg
grating-based WDM components etched diffraction grating-based WDM components and
arrayed waveguide grating-based WDM components. To meet the largest reach requirements
of ~100 km most systems require some form of reach extension discussed in 5.9. The chapter
is concluded with a discussion on switching components in section 5.10, where evolution in
terms of cost and power consumption will have large impact on AON solutions.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 92 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
5.1 WDM COMPONENTS/SUBSYSTEMS
5.1.1 Wavelength selective components for WDM systems
Wavelength Division Multiplexing (WDM) has been introduced to increase transmission
capacity without having to install new fibres. WDM is a technique that allows signals from
different wavelength channels to be multiplexed into one optical fibre, transmitted together
and demultiplexed.
Today Time Division Multiplexing (TDM) systems are widely used in optical communication
networks. In TDM technology multiplexing and demultiplexing (MUX/DEMUX) are realized
with help of electronics. In each node optical-to-electronic conversion occurs to process the
signal in electronic way and then signals are converted back to optical domain. In these
systems speed is limited by electronics. On the other hand WDM systems are based on
multiplexing and demultiplexing in the ptical domain. Moreover, this treatment is based on
passive wavelength selective devices, that if needed can be used in a massively parallel way
providing high speed and high throughput.
These wavelength selective devices can be designed to have different channel count and
channel spacing depending on their application in the WDM system, such as:
-
WWDM: Wide Wavelength Division Multiplexing uses normally only few channels
with channel spacing ≥ 50 nm,
-
CWDM: Coarse Wavelength Division Multiplexing uses a relatively small channel
count with channel spacing < 50 nm,
-
DWDM: Dense Wavelength Division Multiplexing uses a large number of channels
with channel spacing ≤ 8 nm.
Wide WDMs can be used as triplexers, for example in PONs for FTTx applications. In this
case they are designed to combine or separate three channels at wavelengths of 1310 nm,
1490 nm, and 1550 nm carrying data, video and voice signals onto one single optical fibre.
Coarse WDMs combine usually up to 16 wavelengths onto a single fibre. It is a cheap
solution due to the fact that the large channel separation eliminates expensive stabilized
narrow-band lasers. CWDM technology uses an International Communication Union (ITU-T)
standard 20-nm spacing between the wavelengths, from 1310 nm to 1610 nm.
Dense WDMs combine 40, 80 or 160 wavelengths onto a single fibre. DWDM technology
uses an ITU-T standard channel spacing of 12.5, 25, 50 or 100GHz (0.1, 0.2, 0.4 or 0.8 nm
respectively) arranged in several bands in optical communication window 1500-1600 nm.
This is the most expensive solution, but gives the highest network capacity, suitable for
applications, where there are many users sharing costs. DWDMs are today mainly used for
long haul and metro networks.
There are several kinds of MUX/DEMUX components available. Those based on thin film
filters and fibre bragg gratings consist usually of discrete components for each wavelength.
The second class is Planar Lightwave Circuits (PLCs) also called Photonic Integrated Circuits
(PICs) represented mainly by arrayed waveguide gratings (AWGs), but including also more
seldomly used etched diffraction grating-based multiplexers (EDGs). This is the class that can
manage large numbers of channels and treat them in a parallel way. This class can also be
integrated with other passive and active components on a common platform.
During the last decade, PLC-based components have greatly improved core network solutions
and capacity and are expected to have similar influence on next generation access networks.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 93 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Photonic integration based on silicon platform is now under a very intensive development and
due to its compatibility to standard CMOS electronics promises low cost mass production
allowing for deployment of FTTx infrastructures more cost effectively than ever before.
Possible components to be characterized:
-
thin film filter-based ADD/DROPS
fibre bragg grating-based ADD/DROPS
arrayed waveguide grating-based MUX/DEMUX and more advanced architectures
etched diffraction grating-based MUX/DEMUX, triplexers and other devices
transmitters: FP lasers, DFB/DBR lasers, VCSELs
receivers: p-i-n, avalanche photodiodes
5.1.2 Example of FTTH-PON-based on ASE-injected FP-LDs
WDM technology has been successfully used in backbone network systems for quite a long
time and presently there is a trend to move this technology for application in access networks.
There have been several solutions described in literature.
In [49] the authors proposed a WDM-PON system that employs the ASE-injected Fabry-Perot
Laser Diodes as light transmitters, both in CO (OLT) and ONTs. The system has 32
wavelength channels with channel spacing 0.8nm (100 GHz) and 2 athermal AWGs, one in a
remote node (RN) to distribute and collect signals to and from ONTs (L band was used for
downstream signals and C band for upstream) and one in the OLT for distribution of upstream
signals and multiplex downstream signals. Downstream and upstream signals are separated by
thin-film filters and AWGs, due to their cyclic ability can treat L band and C band
simultaneously.
As light sources, wavelength-locked FP-LD lasers were used, as they are expected to be
economical and practical. In this scheme the broadband light source based on erbium-doped
fibre (ASE) was spectrum sliced by an AWG and injected into the FP-LD. The light output
from the laser was spectral filtered by the same AWG. The upstream (downstream) transmitted
wavelength of (to) each subscriber is determined by the wavelength of the injected light,
sliced by the AWG. In this way a common type of FP-LD for upstream (downstream) data
transmission of (to) all subscribers can be used, and the ONT is “colorless”.
Components used in this architecture:
-
AWG: cyclic, 32 channels, channel spacing 0.8nm (100 GHz), for L and C bands
-
Thin film filter separating L band and C band
-
ASE: amplified spontaneous emission broadband light source based on erbium fibre
amplifiers for C band and L band
-
Multimode Fabry-Perot laser diode with possibility for wavelength locking
-
PD: p-i-n photodiode
-
Optical circulator to distinguish between directions of injected light from ASE
sources: C band for upstream and L band for downstream.
In this or similar architectures, instead of using wavelength locked FP lasers one can apply:
-
tunable DFB laser
-
tunable DBR laser
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 94 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
-
Tunable VCSEL
-
Spectrally sliced superluminescent light-emiting diode
-
Spectrally sliced erbium-doped fibre amplifier
-
External modulator
-
Reflective SOA
The last two devices are used with centralized light source at the OLT and external modulator
or reflective SOA at ONU for upstream modulation. In this case to transmit downstream and
upstream in a full duplex manner in the same wavelength, various data modulation schemes
have been proposed: ASK-ASK using different power offset values, frequency- shift keyingASK, and SCM-SCM.
5.2 ONT BASED ON RSOA, REAM, REAT
5.2.1 ONT based on RSOA
In absence of low cost tunable lasers, other kinds of optical devices must be used to
implement colorless ONTs. The RSOA (Reflective Semiconductor Optical Amplifiers)
appears as one of the most promising solutions, at least for bit-rates equal to or lower than
1.25 Gb/s, due to the capability to amplify and modulate the input signal in the same device.
One of the major limits of the RSOA is the modulation bandwidth. 10Gb/s is not feasible at
the moment. For that reason WDM-PON architectures based on RSOA aim at increasing the
number of subscribers with respect to GPON solutions, rather than offering higher
bandwidths.
The main features of RSOA based WDM-PONs are the possibility to modulate and amplify
the signal in one step with a switching time in the ns range, wide optical bandwidth (more
than 60 nm) which can be centred in a wide range (1200-1550nm) and low noise (down to
6dB).
Reflective SOA with NRZ Signaling
Various system architectures based on RSOAs have been proposed in literature [50].
Compared to a broadband light source [51], the use of a comb of coherent wavelengths to
generate CW carriers allows for longer distances, due to the lower noise induced on the US
signal. Moreover it permits 2.5Gb/s symmetric transmission, which is well beyond the
nominal modulation bandwidth of the device.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 95 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
GPON
OLT
Optical Network Unit
…
Feeder
Fiber
L band
expansion
AWG
…
AWG
Data DL
2.5G
Multi λ
32Ch
ONT
WDM
C/L
L band
expansion
1:8
WDM
Data UL
2.5G
Multi λ CW Carrier
32Ch
Service ONU
AWG
. . .
AWG
. . .
Optical Line Terminal
Multi λ
32Ch
IL
. . .
IL
BERT
RSOA
C/L
SMF
CW Carrier
BERT
IL
ONU
Remote Node
Figure 59 : 2.5Gb/s WDM-PON employing RSOA-based ONUs and NRZ signaling
The system architecture based on RSOA [52] is shown in Figure 59. At the OLT, DS
modulated channels (λ1D,···, λ32D) and CW carriers utilized for US re-modulation are
generated by two different multiwavelength sources each producing 32 100GHz-spaced
channels. The two combs are offset by 50GHz in order to obtain a set of 64 wavelengths
aligned to a 50GHz spaced ITU-T grid. In particular the channel plans are 1534.250nm-to1558.983nm and 1534.643nm-to-1559.389nm for the two multiwavelength sources
respectively. The use of different DS and US carrier wavelengths allow for better tolerance to
back-reflections (caused by fibre connector, splices, etc…), which is one of the most
prominent degradation effects in single feeder fibre schemes.
One of the two generated combs is demultiplexed by a conventional 100GHz AWG and each
wavelength is individually modulated with a 231-1-long PRBS at 2.5Gb/s. The modulation
format is the conventional NRZ-OOK. The 32 modulated channels carrying the DS
information are then recombined in a similar AWG, coupled with the CW comb and sent
downlink through the SMF. A booster amplifier allows fixing the channel power to 5dBm. At
the RN, DS channels and adjacent CW optical carriers are routed to each end users by means
of a 50GHz spaced AWG. Low-cost and low-insertion loss (about 1dB) interleavers (ILs) are
used in order to couple the correct DS signal and CW carrier to the user, keeping a single fibre
connection between the RN and the ONU. The DS signal is detected using an APD receiver,
while the CW carrier is sent to a RSOA for US modulation. The used RSOA is a commercial
device optimized for 1.25Gb/s operation with 20dB small signal gain, 3dBm saturated output
power, low polarization-dependent gain (PDG) (1.5dB), ultra-low front facet reflectivity
(<10-5) and 1.5GHz electrical bandwidth.
The modulated US optical signal (λ1U,···, λ32U) is then transmitted by the ONU to the OLT
across the same feeder fibre as used for the downlink. At the OLT an optical circulator
followed by a 100GHz spaced AWG is used to route the US signal toward the receiver.
To improve the US system performance the CW multiwavelength source can be moved to a
service-ONU (S-ONU) closer to the user terminations and connected to the RN common port
as shown in Figure 59 (dashed line).
Figure 60 : BER versus received power with multiwavelength source placed at the OLT (a) and the
RN (b)
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 96 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Results show that the architecture with the multi-channel source at the RN is less affected by
power budget restrictions. The transmission channel is extended to 35 km with a maximum
uplink penalty of 3 dB. An extension of the transmission channel up to 45.2km reach can be
achieved by using the multi-channel at the RN and an optimum RSOA input power with
maximum uplink penalty of 2.5 dB. BER measurements are shown in Figure 61.
Figure 61 : BER measurements versus received power with multi-channel near RN
Reflective SOA with RZ Signaling
The setup used for experimental tests on a WDM-PON system based on RSOA with RZ
modulation is shown in Figure 62.
GPON
OLT
Optical Network Unit
…
Feeder
Fiber
L band
expansion
AWG
…
ODL
AWG
Data DL
2.5G
Multi λ
32Ch
ONT
WDM
C/L
L band
expansion
1:8
WDM
C/L
BERT
ODL
Data UL
2.5G
. . .
BERT
. . .
AWG
. . .
AWG
Pulse
Generator
Optical Line Terminal
RSOA
SMF
ONU
Remote Node
Figure 62 : 2.5Gb/s WDM-PON employing RSOA-based ONUs and RZ signaling
At the OLT a 32-channels multiwavelength source is used to generate the DS carriers (λ1D,···,
λ32D). A MZ-modulator driven by an electrical pulse generator acts as pulse carver in order
to obtain a train of 70ps-width pulses. The WDM comb is then demultiplexed by a 100GHz
AWG and each wavelength individually OOK modulated by an electrical 2.5Gb/s 231-1
PRBS signal using a MZ-modulator. An optical delay line (ODL) before the AWG is used to
synchronize the RZ pulses with the electrical data pattern. The wavelengths are then
recombined in another AWG and sent into the feeder fibre with a power of 5dBm per channel.
The DS signal extinction ratio is controlled by varying the pulse carver modulator bias
voltage and performance was investigated at 5dB and 3dB of its value. At the receiver of each
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 97 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
ONU the DS channel is divided by a 90%-10% splitter. The smallest part of the signal is
photodetected by a PIN receiver with an optimized threshold. The largest part of the signal is
sent to the RSOA, which erases the DS data, amplifies the recovered pulsed carrier and
generates the US signal. In order to optimally cancel the DS modulation, the RSOA should
operate near to the gain saturation regime. Best performance is obtained by applying a DC
bias current of 80mA, which permits maximizing the RSOA gain, and a RF voltage of 3.3V.
The US data is 2.5Gb/s 231-1-long PRBS, applied to the RSOA RF electrical input by means
of a tee-bias coupling. At the RN all the US channels are multiplexed using a 100GHz AWG
and sent towards the OLT, where they are demultiplexed and detected. The RSOA input
power is - 4.5dBm in the BtB case and decreases when the optical fibre is introduced.
3
BER Evaluation
Downstream Extintion Ratio 3dB
3
DL BtB
UL BtB
DL 20Km
UL 20Km
DL 35.2Km
UL 35.2Km
4
6
7
8
9
10
11
12
DL BtB
UL BtB
DL 20Km
UL 20Km
DL 35.2Km
UL 35.2Km
4
5
-LOG(BER)
-LOG(BER)
5
BER Evaluation
Downstream Extintion Ratio 5dB
6
7
8
9
10
11
12
-26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16 -15
-26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16 -15
Received Power [dBm]
Received Power [dBm]
Figure 63 : BER performance for different DS extinction ratio values of 3dB (a) and 5dB (b)
The US and DS BERs curves are reported in Figure 63(a) and (b) for the 1550.1nm channel in
the case of 3dB and 5dB DS extinction ratio respectively. Performance obtained exploiting
20km and 35.2km of feeder fibre lengths are compared with the BtB case. In particular, for an
extinction ratio of 3dB, the power penalty is less than 1dB up to 20km.
The testing of longer lengths is not possible due to the sensitivity limitations of the PIN
photodiode. A proper comparison of these results with those obtained for the NRZ format,
would require use of an APD detector in the latter case.
Looking at the BtB curves in Figure 60 and Figure 62 one may note that the sensitivity
difference between the NRZ and RZ formats is given by the horizontal distance of the
corresponding curves. Therefore the expected improvement when using an APD is 9.7dB and
8.1dB for an RZ format with 5dB and 3dB of extinction ratio respectively.
It is important to point out that such performance is tolerable if applying conventional
Forward Error Correction techniques to the US signal. It can be also noticed that increasing
the DS modulation depth improves the DS performance but impairs the US.
5.2.2 ONT based on REAM
An example of REAM at the ONT operating at 10Gb/s per wavelength have been reported in
[52]. These devices are less bandwidth-limited as compared to RSOAs and are good
candidates for low cost deployment. Figure 64 shows a WDM-PON system based on
wavelength remote distribution.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 98 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
GPON
OLT
…
Feeder
Fiber
L band
expansion
Optical Network Unit
APD
+
TIA
L band
expansion
1:8
High
Sensitivity
CDR
BERT
IL
AWG
…
Multi λ
32Ch
AWG
Data DL
10G
ONT
WDM
C/L
WDM
C/L
SOA
SMF
Multi λ CW Carrier
32Ch
REAM
Data UL
10G
Multi λ
32Ch
Service ONU
. . .
CW Carrier
. . .
Optical Line Terminal
AWG
. . .
APD
+
TIA
High
Sensitivity
CDR
BERT
AWG
IL
IL
ONU
Remote Node
Figure 64 : 10Gb/s WDM-PON employing REAM-based ONUs
In order to reduce the impairments due to Raleigh back-scattering (RBS), the multiwavelength
source for CW distribution is located at the RN as in the 2.5Gb/s case. DS signals within the
100GHz ITU-T grid, are generated and modulated at the OLT. Conventional NRZ-OOK
modulation is used. The data stream is a 231-1-long PRBS at 10Gb/s. US CW channels have
the same wavelength spacing but they are 50GHz shifted to avoid superposition. A couple of
ILs, at the RN and at the ONU, separate the DS and US channels as in the RSOA-based
scheme. The US data stream is obtained through 10Gb/s NRZ-OOK 231-1-long PRBS used to
directly modulate the REAM. A cascaded SOA is used bidirectionally [53] to compensate for
the REAM losses.
3
BER Evaluation
DL BtB
UL BtB
DL 25km
UL 25km
DL 50km
UL 50km
4
-LOG(BER)
5
6
7
8
9
10
11
12
-32 -31 -30 -29 -28 -27 -26 -25 -24 -23 -22
Received Power [dBm]
Figure 65 : US and DS BER performance for REAM WDM-PON
US and DS BER performance are shown in Figure 65 for above mentioned SOA at optimum
bias current values and for a REAM at a bias voltage of -1.2V and a VRF=4V peak-to-peak.
Results referring to 25km and 50km standard SMF are compared with the BtB case. A power
penalty of 2dB at BER=10-9 has been measured for both US and DS using a 25km-long feeder
fibre link without any dispersion compensation. The same US performance has been obtained
for an up to 50km-long fibre link without any dispersion compensation, due to the chirp
induced by the REAM, whilst a tolerable degradation of 4dB is measured for the DS in the
last case. In fact, whilst the DS degradation can be imputed to the higher bit rate impact on the
fibre transmission, the US performances are strongly related to the ONU configuration.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 99 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
However, this demonstrates the possibility of 10Gb/s upgrade over 2.5Gb/s installed WDMPON.
5.2.3 ONT based on REAT
Another emerging technology offering colourless ONU functionality is that based on the
reflective electro-absorptive transceiver (REAT). This offers similar possibilities as the RSOA
and REAM, but with the additional benefit of requiring a significantly lower component
count. In particular, the REAT located within the ONU acts as an integrated photo-receiver
for the downstream data (at a wavelength in the 1300 nm window), and also as a reflective
modulator for upstream data transmission (with the upstream reflected wavelength within the
1550 nm window.) This has the advantages of not requiring additional splitters, AWGs,
optical circulators or ILs as are required for the RSOA and REAM architectures. However,
while RSOA and REAM devices tend to be used for digital transmission systems, the high
linearity of the REAT allows analogue operation, so that it has currently been most
demonstrated in radio-over-fibre (RoF) systems, such as that developed within the UROOF
project [54], and is currently being studied within the EU FP7 FIVER project [55]. In
addition, the REAT has been optimised for use with DS wavelengths in the 1300 nm region,
while US data is modulated onto wavelengths within the 1550 nm window. This is a
consequence of the REAT acting as a PD at 1300 nm, while acting as a reflector at the longer
wavelengths of 1550 nm. When used as an analogue device, the REAT is employed to
transmit UWB (3.1 – 10.6 GHz RF), LTE (2.4 GHz), and WiMax (3.5 GHz) signals in both
US and DS directions, while the intrinsic bandwidth of the REAT operating in
receive/photodiode (DS) and modulation (US) modes is at least 10 GHz. While the UROOF
project was only able to demonstrate distances between CO and ONU of up to about 1 km, the
FIVER project is aiming at distances of up to 100 km between CO and customer premises.
However, due to the intrinsic losses associated with locating a CW 1550 nm signal source
within the CO, so that light has to travel effectively twice the overall length; as well as the
fact that the downstream light at 1300 nm experiences higher attenuation compared to the
1550 nm window, REAT devices can only realistically allow a maximum distance of up to 20
km between CO and ONU.
Figure 66 : Cut-down (single-user) WDM-PON highlighting REAT-based ONU at CPE.
Figure 66 depicts the generic architecture for a REAT-based RoF access architecture, showing
a 1300 nm source at the central office, with DS data modulation, with a 1550 nm CW source
being multiplexed onto the same fibre via a broadband 1300/1550 nm WDM multiplexer onto
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 100 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
the same fibre. For upstream data transmission, within the CO light at 1500 nm from a CW
laser diode passes through an optical circulator, is subsequently multiplexed with the
downstream 1300 nm data signal, and is passed to the 20 km optical fibre link towards the
client end. Within the CPE, the REAT directly receives the DS data on the 1300 nm
wavelength, but in full duplex mode modulates (with upstream data) the CW 1500 nm light in
reflective mode. The reflected upstream 1550 nm wavelength passes back towards the CO
where the optical circulator routes the upstream 1550 nm signal away from the 1550 nm CW
laser source, and into the photoreceiver where the upstream data is received.
5.3 TUNABLE LASERS
Tunable lasers are one of only two ways to achieve “colourless” ONUs. Here, “colourless”
refers to the requirement that in a mass roll-out scenario, no wavelength-specific ONUs will
be allowed. (Hence, the fixed-wavelength WDM-PON ONU is not seen an alternative.) The
other alternative to tunable laser sources for ONUs are seeded reflective ONUs, based on
reflective components like RSOAs, REAMs, or IL-FP lasers. These are described in the
previous chapter.
There is a number of reasons to not allow wavelength-specific ONUs. These can be split
inventory reduction (i.e., cost reduction), and network operations enhancements (at least
partly again cost reductions).
Inventory reduction refers to two aspects. The first one is supplier- (components vendor-)
driven: tuneability obviously reduces product codes (or variants) and hence ultimately helps
decreasing cost, rather than increasing it. For example, the C-band has some 96 channels
spaced 50 GHz which would translate to 96 different products, with all associated logistic
impact. Tuneability reduces this to one product; given full-band tuneability is achieved (which
is hence the design goal). The second aspect is customer- (service provider-) driven. It relates
to much simpler inventory and spare parts handling. Here, tuneability reduces OpEx. This is
one of the main reasons for service providers to demand tuneability.
Network operations enhancements are obviously driven by the respective service provider.
Here, tuneability drastically eases ONU installation, in particular remote service provisioning
and configuration. For example, the field staff only needs to consider a single type of ONU. In
addition, protection mechanisms based on wavelength protection become simpler, or more
powerful (or even enabled at all). This is complemented by potential advantages when
operation over a flexible (access ROADM) infrastructure or ODN (should this be a serious
alternative to static passive ODNs). These aspects all refer to operations and hence OpEx
savings or optimization.
Obviously, these target advantages of tunables – which hereinafter are referred to as low-cost
tunables – must not come at an increase of cost / price. This would directly contradict the first
inventory aspect, and also the cost / price goal for residential access where cheap EPON
transceivers are the benchmark. Further, tuneability must not compromise on system
performance or reliability. Otherwise, tuneability will be, and stay, a minority interest. This,
amongst other aspects, would translate to a no-profit situation for the components vendors and
consequently to no such developments.
In principle, tuneability of a laser can be achieved by [56]-[60]:
•
•
Electronic Tuning
Thermal tuning
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 101 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
•
Mechanical Tuning
In addition, arrays of DFB (Distributed Feedback) lasers must be considered as a quasitunable transmitter, especially since first mid-price products exist [66]. As a further potential
laser source, Quantum-Dot lasers may be considered. However, at the time being, these lasers
only support the wavelength range of 1050…1320 nm and hence are not considered
hereinafter as a contender for low-cost tunables for WDM-PON.
Electronic tuning refers to the principle of applying an electric field which directly tunes the
laser frequency. Here, several different possibilities exist, with derivatives of the well-known
DBR (Distributed Bragg Reflector) laser being the most important ones:
•
•
•
•
•
•
•
Three-section DBR
Sampled Grating DBR Laser (SG-DBR)
Super-Structure Grating DBR Laser (SSG-DBR)
Digital Supermode DBR Laser (DS-DBR)
Grating-assisted Co-directional Coupler Laser with Sampled Grating Reflector
(GCSR)
Y-Junction laser (Y3 laser, transverse interferometric type)
Vertical Mach-Zehnder Laser (VMZ, vertical interferometric type)
According to [60], only one relevant type of thermally tuned lasers exists, namely the MatrixGrating Strongly Gain-Coupled (MG-SGC) DFB Laser. For mechanical tuning, two relevant
classes can be identified:
•
•
Micro-Electromechanical Tunable Vertical Cavity Surface Emitting Laser (MEMVCSEL)
External Cavity Laser (ECL)
From this list, only very few have strong potential for lowest cost. Here, it must be
remembered that the long-term cost goal for the tunable laser is in the 50$ range. The
respective lasers are a multi-section DBR (namely the DS-DBR and SG-DBR lasers), ECL,
and MEM-VCSEL. In addition, DFB arrays must be considered due to their early availability.
DS-DBR Laser
Multi-section, tunable DBR lasers have been described in various papers, e.g., [56]-[58], [61],
[62]. They are in mass production today. In the form of the DS-DBR laser they are followed
in the low-cost WDM-PON context by Oclaro, Inc. (formerly Bookham, Inc.). The clear
advantage of these lasers is the fact that they are monolithically integrated which directly
translates to low-cost potential. Here, we focus on the concept of the DS-DBR laser because
of its high potential as a lowest-cost source. However, similar considerations hold for other
multi-section DBR lasers as well.
The basis of any tunable, monolithically integrated DBR laser is the so-called 3-section DBR
laser [60]. Its schematic cross-sectional diagram is shown in Figure 67. Not shown here are
two sub-components which are typically necessary for a laser: a Thermo-Electrical Cooler
(TEC, also referred to as Peltier cooler), and a wavelength locker for wavelength tuning or
stabilization.
The important three sections are the Gain, Phase, and Rear sections. The Gain section, as with
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 102 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
every other laser diode, provides the lasing gain. In itself, it is not wavelength-selective and
provides gain in a relatively broad spectrum. The Rear section is able to select a single laser
line by means of the Bragg grating which is etched into the substrate. The Phase section can
provide some degree of fine tuning. Therefore, the 3-section DBR laser can not provide fullband tuning capability. It is, however, relevant as the basis for improved tunable DBR lasers.
p contacts
Igain
Iphase
Irear
p InP
light
output
QW gain region
Tuning Regions
Grating
n InP substrate
Reflection
coating
n contact
AR coating
Figure 67 : 3-section DBR laser
One relevant derivative of the 3-section DBR laser is the so-called Digital-Supermode (DS-)
DBR laser. Its cross-sectional schematic diagram is shown in Figure 67. The DS-DBR laser is
a more recent derivative within the family of monolithically integrated, electronically tuned
lasers. It was developed by Oclaro, Inc., and has been described in various papers, e.g., [61].
Unlike most other tunable lasers, the DS-DBR laser has the potential for uncooled operation,
and such uncooled operation has already been demonstrated, e.g., [62], [63]. Since in a
WDM-PON context the wavelength locker can also be omitted, the DS-DBR laser has unique
potential as a true low-cost tunable source. On the other hand, the lack of TEC and
wavelength locker makes it useless in most other applications so that it will not cannibalize
other laser markets, despite its low cost. Today, it must be seen one of the most promising
lasers for WDM-PON.
Ifront
ISOA
p contacts
Igain
Iphase
Irear
p InP
light
output
QW gain regions
Tuning regions
Grating
n InPsubstrate
AR coating
n contact
AR coating
Figure 68 : 5-section DS-DBR laser
The DS-DBR laser shown in Figure 68 is a 5-section DBR laser. As compared to the 3-section
DBR laser, on section has been added for improved tuning capability. The fifths section is a
monolithically integrated SOA. This section provides both, broadband modulation capability
(up to 10 Gb/s), and high per-channel output power (up to 13 dBm). Due to monolithic
integration, this does not lead to severe cost increase.
Wider tuning range comes from a grating – in the rear section – with many reflection peaks.
This is combined with another grating in the front section for selecting between these peaks.
Partial cancellation between different parts of the rear grating leads to a top-hat like envelope
on the grating comb which is shown in Figure 69.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 103 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
1.0
Reflectivity
0.8
0.6
0.4
0.2
0.0
1500
1520
1540
1560
1580
1600
Wavelength (nm)
Figure 69 : DS-DBR rear section wavelength comb (measured)
Wavelength tuning caused by changes of the front and rear grating currents against their
normalised settings can be depicted in the so-called wavelength map. For the DS-DBR laser,
this wavelength (or tuning) map has wide supermode bands, corresponding to the front
grating settings which are approximately 8nm (~ 1 THz) apart. Each such supermode in a DSDBR laser behaves just like a standard 3-section DBR laser. Since the tuning map contains
supermode jumps, it is also referred to as a pseudo-wavelength map. Additionally the DSDBR has longitude modes which are approx. 37 GHz apart. Within one longitude mode the
frequency of the DS-DBR laser can be fine tuned via the phase section. By setting the
operation point to the middle of a longitude mode a tuning range of approximately +/- 20 GHz
can be achieved. An example for a C-band DS-DBR laser is shown in Figure 70.
Figure 70 : DS-DBR laser pseudo-wavelength map
Shown here are the normalized front and rear grating current settings and, in colour code, the
resulting wavelength. The tuning behaviour is relatively insensitive to actual front grating
currents, which allows a broad tuning range per supermode. If this range is to be exceeded, a
supermode jump becomes necessary. These supermode jumps are indicated in Figure 70
where the thick black lines cross one of the thin, vertical, black lines.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 104 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
The resulting single-mode lasing spectrum of a DS-DBR laser is shown in Figure 71. It can be
seen that output power (per WDM channel, coupled into the fibre) is as high as 15 dBm.
Super-Side-Mode Suppression Ration (SMSR) is in the range of 45 dB (super-side modes can
still be seen in Figure 71). Since a comb with 7 reflection peaks is generated in the rear
section in order to cover a full wavelength band (here the C-band). Hence, they are
additionally suppressed in the AWGs used for a WDM-PON. The corresponding detrimental
effect is negligible. More problematic are although longitude side-modes as they are closer to
the desired frequency. As seen in Figure 72 the SMSR is normally ~ 50 dB.
Figure 71 : DS-DBR lasing spectrum – with side-super--modes
Figure 72 : Side mode suppression ratio
The DS-DBR laser as used today for tunable laser sources and as intended as low-cost source
for WDM-PON is a 5-section DBR laser including a SOA. The SOA provides modulation
bandwidth and high output power. Here, it can also be used as a fast shutter in cases of
switching events or even in cases of the – predictable! – (super-) mode jumps. Then, the SOA
is reverse-biased which is shown in Figure 73.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 105 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
20
Laser Power (dBm)
10
0
-5
-4
-3
-2
-1
1563.33nm
1556.15nm
1549nm
1528.3nm
0
1
-10
-20
-30
-40
-50
SOA Voltage (V)
Figure 73 : DS-DBR reverse-biased SOA as a shutter
Relevant optical characteristics over the entire tuning range (>100 channels in the 50-GHz
grid) are shown in Figure 74. Shown here are the Relative Intensity Noise (RIN, in dB/Hz)
and the laser line width (in MHz). It can be seen that RIN is <-150 dB/Hz for all channels and
that line width is in the range of 500 kHz for all channels. Not displayed here are output
power and SMSR. They are in the range of 15+/-1 dBm (power) and 42…56 dB (SMSR),
respectively.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 106 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 74 : DS-DBR laser RIN (a), and line width (b)
For WDM-PON, the respective transmitter must have lowest-possible power consumption.
For this reason – and in order to reduce CapEx – most service providers require TEC-less
transmitters. Generally, the DS-DBR laser has the potential to work without TEC. However,
temperature drift will not only cause wavelength drift (which can be counter-acted through
grating current tuning) but also variations of the maximum output power. This is
demonstrated in Figure 75. At 25°C, the output power stays above 14 dBm for all channels.
However, at 45°C, output power of some channels drops to slightly below 12 dBm. Hence,
avoiding the TEC may lead to somewhat reduced output power (where the 12-dBm range is
still very good for WDM-PON applications). Other parameters like SMSR are less affected by
temperature drift.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 107 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 75 : DS-DBR laser temperature-depending output power (as of today)
Application of an uncooled DS-DBR laser for WDM-PON has already been demonstrated
[62]. In order to tackle the temperature drift problem, the most recent approach for the DSDBR laser is to be based on new high-temperature materials, e.g., AlGaInAs/InP [64]. This
approach could support uncooled operation well into the 70°C region, possibly supported by a
small low-power heater (rather than a TEC). Target power consumption is in the 1-W range.
The DS-DBR laser has good potential as a low-power-consuming, low-cost transmitter for
WDM-PON. It must be noted that similar (multi-section) DBR laser derivatives with similar
potential exist, for example the Sampled-Grating DBR (SG-DBR) laser or the Super-Structure
Grating DBR (SSG-DBR) laser. More detailed comparisons between these DBR types can be
found in [56], [57].
MEM-VCSEL
Micro-Electro-Mechanically (MEM) tuned VCSELs are a promising alternative the
monolithic (multi-section, DBR) lasers. Generally, VCSELs have low-cost potential, and by
adding MEM means, they can be tuned. Tunable MEM-VCSELs have already been described
for metro-area applications in 2001 [70]. In the last years, they have gained attraction with
regard to WDM-PON due to their cost characteristics.
The basic concept of the MEM-VCSEL has not changed significantly over time. It is based
DBR-VCSEL with an attached micro-mirror which is thermally tunable. (In [70], application
of an electrostatic force to the mirror led to tuning.) A schematic diagram of a more recent
MEM-VCSEL is shown in Figure 76.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 108 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 76 : Cross-sectional diagram of a MEM-VCSEL
MEM-tunable VCSELs are a hybrid two-chip assembly of a MEM system (MEMS) with
concave AlGaAs-GaAs mirror membrane and an InP-based semiconductor cavity with tunnel
junction aperture. Using electro-thermal MEMS actuation, the included air-gap can be
expanded and the cavity resonance be tuned to longer wavelengths. The gold p-contact also
serves as an efficient heat-sink. The active part (according to Figure 76) is formed by 7
compressively strained InGaAlAs quantum wells embedded between an n-type and a thin ptype InGaAlAs confinement layer grown by molecular beam epitaxy. A centrally limited
buried tunnel junction (BTJ) defines the active size (diameter) and reduces the inefficiently
conducting p-type region to a minimum extend forming the junction. The BTJ is overgrown
with a binary n-InP current- and heat-spreading layer in a subsequent MBE process, followed
by an evaporated dielectric backside distributed Bragg reflector (DBR) composed of 3.5 pairs
of calcium fluoride and zinc sulfide with gold coating and an electroplated substrate. Such
back mirror has a limited size, slightly larger that the BTJ. The semiconductor cavity has an
optical length of according to coupled cavity considerations and the corresponding limitation
on the tuning range. The top DBR consists of MBE-grown GaAs-AlGaAs quarter wavelength
layers with refractive index contrast. The inclusion of a graded indium content induces
compressive strain, which is released after membrane processing and shaping, resulting in a
curved mirror membrane with radius of curvature of 2…3 mm, suspended on four arms. The
upward movement is achieved by Joule heating of the upper membrane.
Depending on the cavity configuration, tuning ranges up to 60 nm and single-mode output
power up to 2 mW with high sidemode suppression and stable polarization have been
demonstrated.
Slightly different design schemes for MEM-VCSELs exist, namely the semiconductorcoupled cavity (SCC, type B) and the extended cavity (EC, types A, C) designs. From the
basic SCC design (bottom Bragg mirror cavity air–gap and curved mirror), the EC scheme
can be obtained quite easily by placing an antireflection coating layer (AR-c in Figure 76) at
the semiconductor-air interface of the lower half VCSEL.
By comparing the overall performances of the two schemes, it turns out that an extended
tuning range of EC devices has to be paid at the expense of lower delivered powers and higher
thresholds. The different tuning performances can be compared in Figure 77 by the spectra for
different tuning currents. One can clearly see that the laser can be continuously tuned over
almost the whole free spectral range (FSR) and it stays always single mode with high SMSR.
From such measurements one can also extract the FSR, i.e., the distance between two adjacent
longitudinal modes, which is related to the length of the air-gap.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 109 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 77 : VCSEL tuning. (B) shows tuning of an SCC type VCSEL, (A) and (C) for EC types.
The FSR sets the upper limit for the tuning range. The latter can be limited by the material
gain spectrum, which has to compensate for the wavelength-dependent threshold conditions.
The limitations due to the DBR bandwidth are negligible in both designs.
Tunable ECL
External-Cavity Lasers are an obvious choice for tunable transmitters because of their
inherent capability to be tuned via the external cavity. They are described in [56]-[58]. The
basic ECL consists of a broadband gain section to which an external cavity is coupled via an
Anti-Reflection (AR) coating and a collimating lens. This setup is shown in Figure 78. The
gain section can be a simple Fabry-Perot laser diode. Wavelength tuning is possible by tuning
the external grating. This can be done mechanically or thermally. One contributor to cost is
the optical coupling between the components which typically has to be done via a lens. This is
a disadvantage over monolithically integrated solutions. It also leads to lower output power
due to the loss of the free-space region.
Figure 78 : Schematic diagram of an ECL
There are two basic tuning mechanisms, thermal and mechanical tuning. Mechanical tuning
often is implemented using MEM components. Thermal tuning is often applied to etalons. It
can also be applied to FBGs which in addition offers the chance for better and more costO
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 110 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
effective integration. The corresponding devices are also referred to as FBG-ECLs, a principle
diagram and an early implementation are shown in Figure 79.
+
Lens
FBG
Gain
HR
–
AR
Figure 79 : Principle of an FBG-ECL (left), and FBG-ECL from Xponent Photonics, Inc. (right)
FBGs are also the tuning means for a first mid-price commercial component which is
available [68]. Though not yet being in the price / cost region which is considered to be
necessary for WDM-PON mass roll-out, this product is already cheaper than any other
tunable laser available on the market today. As an early implementation of a tunable laser, it
still has somewhat limited performance specifications. At the time being, maximum tuning
range is ~26 nm, and the device has ~6 W power consumption which is mainly driven by two
TECs, one for the laser (gain) chip, and one for the grating. The device is suitable for up to
2.5 Gb/s bit rate, and it comes in a butterfly packaging, see Figure 80.
Figure 80 : Packaging and pin configuration of commercial ChemOptics FBG-ECL
The per-channel output power is in the range of 6.2+/-0.4 W, and temperature drift over the
range of -10…+70°C is ~0.14 nm. This makes it suitable for the 100-GHz grid, for 50 GHz
some improvements would have to be made.
Alternatives to thermally tuned (FBG-) ECLs are mechanically tuned ECLs. A proposal for a
MEM-tunable ECL which is suitable for UDWDM-PON application has recently been made
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 111 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
[69]. This device is based on a gain chip and an external cavity which consists of a dielectric
filter which is controlled (tuned, turned) by a micro-motor. According to [69], this transmitter
has sufficient availability (failure resistance, having in mind that a motor with moving parts is
used), full-band tuneability with a precision of a few 10 MHz, and a line width of <200 kHz
(the latter being necessary for use in a coherent UDWDM-PON). It is, however, doubtful that
a laser with tuning precision requirements of some 10 MHz – which are necessary in a
coherent receiver – can be implemented without a TEC.
Tunable DFB-Array
DBR lasers, ECL, and (MEM-) VCSELs are the only lasers followed today for WDM tuning.
This also holds for the latest tunable XFP generation which is based on DS-DBR and SGDBR lasers, respectively. The well-known DFB laser in itself is not tunable because gain and
grating sections by definition are not decoupled (unlike, in particular, multi-section DBR
lasers). However, integrated arrays of DFB lasers can be made quasi-tunable and hence are an
alternative to DBR, ECL, and VCSEL. It is also worth noting that first commercial products
already exist [66]-[67]. A photo of the respective device is shown in Figure 81.
Figure 81 : Photo of DFB array chip
The DFB array shown in Figure 81 consists of 12 DFB lasers, spaced 10 µm apart in position
and 450 GHz apart in frequency. The entire chip size is 500 µm by 800 µm, which is
comparable in size to high performance single element devices. It requires four separate
epitaxial growths and two layers of metallization. First, a quaternary InGaAsP grating layer is
grown on an n-type InP substrate, and multiple phase shifted gratings are fabricated by direct
write electron beam lithography and wet etching. The grating is then overgrown with the
strained InGaAsP multi-quantum well active region, and 2 µm wide mesas are formed by wet
etching. A third growth adds the p-n InP blocking regions with selective area growth, and the
fourth growth the top p-cladding and the heavily p-doped InGaAs contact layer. Isolation
regions that are 3 µm wide are then etched between the elements of the array, the sidewalls
are coated with silicon nitride, and a first gold metal line is then deposited and plated on top
of each stripe. An intermediate silicon nitride layer is deposited on the chip. Wafers are then
thinned and polished, n-contact metallization is applied to the back, and the wafers are then
cleaved, coated AR/AR and bonded onto submounts.
A great advantage of this geometry is that each stripe is contacted in the middle, rather than
the end, so that uniform current injection along the stripe can be applied. Furthermore, there is
no real-estate wasted in routing electrical contacts around the array. This results in a compact
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 112 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
geometry. To maintain high yield, the lasers are phase-shifted with AR coating on both sides,
which necessarily implies lower efficiency, since some power is wasted from the backside.
The grating and phase shift were designed such that twice as much power exits the front facet
than the rear.
A chip where the laser elements were varied in frequency by about 450 GHz was soldered in a
butterfly package with a MEMS mirror for selection. One advantage of a widely tunable laser
based on an array is that redundancy can be built in at the edges of the array, so that any error
of a specific laser means that a subset of the entire tuning range can still be used. This way,
the array yield can also be improved.
The optical performance of the DFB array is within the range which can be expected from a
DFB laser. Linewidth is below 500 kHz for all channels, and RIN is better than -153 dB/Hz.
Figure 82 : Photo of DFB array chip [66]
In addition, fibre-coupled output power in the range of +15 dBm and SMSR of >55 dB has
been demonstrated. These optical parameters can be considered sufficient for highperformance WDM-PON applications. A disadvantage is the relatively high power
consumption which is in the range of 5 W for a 12-channel device. Since future improvements
regarding power consumption are expected to be relatively low [56], this parameter is
considered too high for wide-spread deployments.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 113 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Comparison of relevant tunable lasers
From the analysis provided herein so far, it becomes clear that only few promising contenders
for low-cost and low-power-consuming tunable lasers suitable for WDM-PON exist. Here, it
should be noted that focus is on lowest possible cost and energy consumption, rather than best
optical specifications like line width, RIN, or modulation bandwidth. For many WDM-PON
applications, laser output power however will be a relevant parameter. The most relevant
parameters of the lasers discussed so far are listed in Table 17.
Table 17 :
Output Power
[dBm]
Power Dissipation
[W]
Components
in Package
Line Width
RIN
SMSR
Tuning Range
Bandwidth
Future Cost
Potential
Comparison of relevant parameters of tunable lasers
DS-, SG-DBR
>13 CW
>4 10G Tx
<2.5 w/ TEC
<1 w/o TEC
Monolithic
Laser with
integrated
SOA,
no Locker,
no
TEC possible,
plus
Receiver
~500 kHz
<-150 dB/Hz
FBG-ECL
~15 CW
<3.5
Laser,
Lens,
FBG,
TEC?
Receiver
MEM-ECL
~15 CW
<3.5
Laser,
Lens,
MEMS,
TEC?
Receiver
>45 dB
C-/L-band
<100 MHz
<-140
dB/Hz
>30
C-/L-band
>100 kHz
<-150
dB/Hz
>50 dB
C-/L-band
10 Gb/s
Very low
2.5 Gb/s
Low
10 Gb/s
Low
MEM-VCSEL
<3
DFB-Array
>13 CW
<1 W (?)
<4.5
VCSEL,
MEMS,
TEC?
Receiver
8-12
Lasers,
Combiner
plus SOA
or MEMS,
TEC?
Receiver
<100 MHz
<-120
dB/Hz
>30 dB
C-Band
(>40 nm)
2.5 Gb/s
Very low
<500 kHz
<-150
dB/Hz
>50 dB
S-/C-/LBand
10 Gb/s
Low
From Table 17 the most promising candidates for future tunable WDM-PON lasers can be
identified. From viewpoint of lowest power consumption, multi-section DBR lasers without
TEC (uncooled operation of both, the DS-DBR and the SG-DBR laser has been demonstrated
[62], [72]) and MEM-VCSELs are the most promising devices. From viewpoint of RIN,
SMSR, line width and also tuning range, all lasers can be considered as applicable, with only
slight disadvantages for FBG-ECLs. Regarding laser output power, XX-DBR (XX = DS, SG),
ECL and DFB arrays have advantages over VCSELs. This is narrowed down to XX-DBR,
MEM-ECL and DFB arrays in view of future high-speed (10 Gb/s) applications. The last
relevant parameter is the future cost estimation, or the potential for lowest-cost future mass
production. This parameter, to a good degree, correlates with the components which are
necessary within the transceiver package. Here, advantages can be identified for the
monolithically integrated XX-DBR lasers which have the potential for TEC-less design, and
again for VCSELs. From this, we can conclude that XX-DBR lasers are the most promising
transmitter variant for future WDM-PON.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 114 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
5.4 WAVELENGTH SELECTIVE RECEIVERS
Wavelength-selective receivers are necessary whenever specific wavelengths have to be
assigned to specific ONUs in a WDM-PON running via a splitter-based – i.e., broadcast –
ODN. In general, they are not necessary in a WDM-PON with filters. This holds for most of
the hybrid PONs where the splitters are used in order to allow wavelength re-use (or
broadcast). An exception is the UDWDM-PON where in a filter-plus-splitter ODN,
wavelength-selective receivers still are required.
Since splitter-based ODN may become an important migration path for WDM-PON,
wavelength-selective receivers must be considered a potentially important component. An
example of early WDM capabilities in a splitter-based ODN is the stacked-XG-PON
approach. However, at the time being, importance of similar approaches is not fully clear
since, for example, splitters in a WDM-PON eliminate many of the advantages (for example:
security, low insertion loss, and hence part of the potential cost advantages).
In general, two approaches exist for wavelength-specific receivers, receivers with tunable
filters and coherent (heterodyne, intradyne) receivers which make use of a tunable local
oscillator. It must be kept in mind that both approaches are not in massive use in the PON
context today. This has an impact on cost of the respective components which of course will
add to total systems cost.
5.4.1 Tunable Filters
Tunable filters for optical transmission have been investigated since the early WDM days. An
early overview is given by [73]. In general, tuneability of a WDM filter can be based on one
out of the following mechanisms (where WDM in most cases refers to DWDM):










Fabry-Perot (FP) Interferometer
Liquid-Crystal (FP) Filter
MEM-tunable Devices
FBG, temperature- or strain-tuned
AOTF, Acousto-Optic Tunable Filter
EOTF, Electro-Optical Tunable Filter
AWG (thermally tuned)
Mach-Zehnder Interferometer (cascaded, with heaters)
Active Filters incorporating laser diodes
Tunable Ring Resonator filters
From this list, only few technologies have the respective cost, complexity, and also formfactor potential for future applications in (next-generation) PONs. The majority of the
literature [74]-[85] indicates that relevant contenders for low-cost WDM-PON tunable filters
come out of the first four technologies listed above.
Tunable FP-TFF
Most FP interferometers are based on Thin-Film Filter (TFF) technology. TFF is also used for
many fixed DWDM filters (where the alternatives are etched diffraction gratings, AWGs and
FBGs, see § 5.8). A good overview on tunable TFF is given in [79].
Fabry-Perot etalons are comprised of optical glasses with high-reflection coatings, the two
high-reflection surfaces are separated by a certain distance. When light passes through the
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 115 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
etalon, the interference arising from the multi-reflection will let certain wavelengths pass
through and reflect the other wavelengths.
FP interferometers are the basis of most tunable filters. They consist of (TFF, liquid-crystal)
etalons which are either angle-tuned, linearly tuned (i.e., a variable TFF addressed by
translation), or which are tuned with active substrates or active layers. A simplified schematic
diagram of an active-substrate TFF-FP interferometer is shown in Figure 83.
Figure 83 : Schematic of tunable TFF with active substrate
The basic idea of the tuning scheme shown in Figure 83 is to transfer a mechanical strain of
the substrate to the spacer thickness of a DWDM bandpass filter deposited at its surface, and
to achieve a change in the resonant wavelength of this all-dielectric FP structure.
The mechanical strain can be induced either thermally, or via the piezo-electric effect. For
thermal tuning, two effects exist, first the increase of the refractive index of the layers created
by a rising temperature, and secondly the change of their thickness induced by a differential
thermal expansion between the layers and the substrate. These effects can be used to partly
compensate temperature sensitivity or, at the opposite, to maximize thermal sensitivity for
thermal tuning. If temperature is kept constant, a mechanical deformation of the substrate, i.e.
a strain, can also be used to induce a shift in the wavelength: this strain can be produced for
instance by a compressive force applied on a standard substrate, or by an electric field applied
on a piezo-electric substrate.
A first commercial thermally tuned TFF for WDM overlay applications in GPON ODN is
available on the market today [83],[84]. The device is specifically intended for use in stacked
XG-PONs and similar environments, where several (few) wavelengths are added to a standard
GPON. Hence, the device has limited tuning range (~5 nm, or 600 GHz). This corresponds
with the stacked XG-PON approach, where only four bi-directional PONs are intended to be
stacked. An interesting parameter is the maximum tuning power which is <200 mW. A
schematic diagram of this filter is shown in Figure 84.
Figure 84 : Schematic of tunable TFF for GPON WDM overlay [83]
A major advantage, next to low power consumption, of the thermally tunable TFF according
to Figure 84 is its compactness. This is shown in Figure 85.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 116 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 85 : Packaging of tunable TFF for GPON WDM overlay
At the time being, the commercial tunable TFF [83], [84] has limited tuning range, sufficient
for GPON overlay with few wavelengths. If future variants of this product allow broader
tuning ranges is unclear. According to the manufacturer, broader tuning ranges contradict part
of the technology advantages (size, power consumption).
Tunable Liquid-Crystal FP filters
FP etalons based on Liquid Crystals (LCFP) are a promising alternative to TFF [75]. As
compared to other FP filters, main advantages of LCFP filters include low driving voltage
(i.e., low power consumption), broad tuning range, high resolution, and also low insertion
loss. LCFP filters can be used in spectroscopy, laser radar, and in optical networking.
Nematic liquid crystals have optical anisotropy, or birefringence. They exhibit double
refraction, light polarized parallel to the director has a different index of refraction than light
polarized perpendicular to the director. The director is along the same direction as the surface
rubbing directions when both the alignment surfaces are rubbed in the same directions.
Nematic liquid crystals also show dielectric anisotropy, the dielectric constants parallel and
perpendicular to the nematic director are not the same. The dielectric anisotropy introduces
body torque on the molecules in the presence of external field, which in turn gives rise to the
director re-orientation. Under the external field, the director of the liquid crystal with a
positive dielectric anisotropy tends to be aligned parallel to the external field, while the
director of the liquid crystal with a negative dielectric anisotropy tends to be aligned
perpendicular to the external field.
Thus by filling the etalon cavity with the nematic liquid crystal, the refraction index of the
extraordinary component of light will change with the applied voltage. The wavelength of
transmission peak is tunable with different voltages applied. The structure of an LCFP etalon
is shown in Figure 86.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 117 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 86 : Schematic of liquid-crystal filter
Normally, a polarizer needs to be placed parallel to the alignment direction of the liquid
crystal to pass through the extraordinary mode of light and block the ordinary mode that is
mot tunable. For tunable laser and system monitoring, the polarization dependence is not a
key problem. For WDM system, polarization independence is challenging because the state of
polarization of the channel signal may be unknown in the system. Several methods have been
discussed to make polarization-insensitive FP devices. One of them is to use a calcite crystal
to split the light source (polarization direction unknown) into two components with
orthogonal polarization directions, then each component is passing through a tunable LCFP
filter with an alignment direction of the liquid crystal the same as the polarization direction.
After passing through the LCFP filters, the two components are recombined through another
calcite crystal. An alternative is to use two neighbouring LC pixels to match the two polarized
components of one channel. The two pixels need to have alignment layers perpendicular to
each other, and LCD production technology is a cost-effective solution to achieve this.
Tunable FBG
Fibre-Bragg Gratings are an alternative to FP etalons when it comes to tunable filters.
According to [78], they have excellent central-peak transmission and sidelobe suppression,
and they can be designed to cover WDM grids from 200 GHz down to less than 25 GHz. In
addition, they have a certain low-cost potential. This is, however, contradicted by the lack of
integration potential. FBGs consist of periodic refractive-index variations along a piece of
SMF which can be imprinted into the fibre e.g., by means of transversal UV laser radiation.
The periodic grating reflects specific wavelengths, and it can be thermally tuned. The
schematic diagram of an FBG is shown in Figure 87.
Figure 87 : Schematic of (tunable) FBG
In order to obtain a good tuned response, an FBG must present a spectral answer
characterized by narrow bandwidth (BW, to support any DWDM grid), weak side lobe peaks
to avoid BW overlapping, and suitable temperature or strain tuned response, which should not
affect BW or passband amplitude. In case a silica fibre (e.g., a standard SMF) is used, several
parameters must be taken into account: refraction index (core, cladding), index modulation,
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 118 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
and the FBG period Λ as shows in Figure 87. The FBG period is related to the Bragg
wavelength by:
λB = 2neff Λ
For an SMF28, the effective index is neff = 1.447559. The resulting FBG can be tuned
thermally, or via mechanical strain. Heating can be achieved through a metal layer which
covers the fibre and which is traversed by an electric current. The reflected central wavelength
then varies linearly with temperature, and BW and central peak transmission do not change
significantly. For DWDM with grids of 1.6…0.2 nm, the temperatures ranges to be covered
can easily be achieved.
Assuming the assumption that strain is uniform along the FBG, it can be used for tuning as
well. Then, strain does not affect BW and peak transmission of reflected wavelengths. Figure
88 shows the spectral characteristics of an FBG transmission peak.
Figure 88 : Reflection peak of an FBG
Volume Holographic Gratings (VHG) is in many ways similar to FBGs except that the
recording medium for the grating is not a (single-mode) fibre but a volume medium.
Consequently, the incident and diffracted light are not confined to the modes of the fibre, but
can be assigned to any mode that can propagate in the bulk material. This opens new
possibilities for device design and for example allows removal of the circulator, which is
essential for FBGs. VHGs as means for tunable filters have been described in [80]. Wide
tuning range of 1510…1590 nm and low insertion loss of ~1 dB can be achieved.
Figure 89shows a schematic representation of the angle-tunable volume holographic filter.
The Bragg wavelength of a holographic grating is determined by λB = 2neff Λ cos θ, where θ
is the angle of incidence inside the holographic material. By changing θ, the Bragg
wavelength can be tuned continuously. For practical application, the tuning range can easily
cover the entire C band, limited mainly by the higher insertion loss at larger angle θ. The
critical requirement to make this a practical tunable filter is to collect the drop signal into a
fibre during angle tuning without an expensive tracking mechanism or feedback control
system. This is achieved with a self-reflector architecture that recombines the reflective
holographic grating with a wideband IR mirror, as shown in Figure 89.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 119 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 89 : Schematic of tunable Volume Holographic Grating (VHG)
A retroreflector consists of two mirrors with a fixed angle between the mirror surfaces. This
also applies to the structure shown in Figure 89, where one mirror is replaced by the VHG. At
the Bragg wavelength, this VHG works as a reflective mirror with an effective depth into the
material. When rotating the mirror/grating structure around the crossing point of the mirror
and grating surfaces at the effective reflection depth, the drop signal beam is fixed spatially
while the wavelength is tuned by the angle θ. With the addition of a temperature sensor and
suitable control system, the device can be made athermal by compensating θ to offset
temperature-induced variations of the grating period.
MEM-tunable Filters
MEM-tunable devices are the next promising technology for low-cost tunable filters.
Basically, similar principles are applied which are used for the tunable MEM-VCSEL (which
also has low-cost potential). MEM-tunable filters are described e.g., in [74],[85]. Several
MEM-tunable optical filters based on electrostatic actuation or thermal tuning have been
reported. Electrostatic actuation is adequate for low-power operation, but its tuning range is
limited due to the so-called pull-in phenomenon, and its actuation voltage is relatively high.
Also, electrostatic actuation intrinsically has a nonlinear behaviour. On the other hand,
thermal actuation can achieve a wide tuning range but it consumes large power and its
response time is slow. A more recent alternative is magnetic actuation [74]. A schematic
diagram of one such filter is shown in Figure 90.
Figure 90 : Schematic of MEM-tunable filter
The filter consists of a FP cavity with two dielectric Distributed Bragg Reflectors (DBR)
similar to other cavity-type MEM-tunable filters. One DBR is formed on the fixed part of the
filter, and the other is located on the actuator part. The two DBRs are separated by a spacer.
Since the actuator part consists of suspended conductor bridges, it receives Lorentz force by
electrical current flowing through it in a magnetic field. The magnetic field is applied in plane
with the actuator. Displacement of the actuated DBR changes a cavity length and hence
allows tuning. This design allows very wide tuning range in excess of 200 nm and also very
low power consumption in the range of 25 µW. In general it is therefore suitable for future
WDM-PON applications.
Various derivatives of MEM-tunable filters have been described. One of these approaches is
based on tunable microcavities [85]. It is applicable to both, VCSELs and tunable filters and
has low-cost potential. The basic idea is shown in Figure 91.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 120 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 91 : Schematic of tunable microcavities
The device shown Figure 91 is based on a micromachined two-chip concept with a movable
GaAs DBR mirror membrane on one chip and a fixed DBR mirror on a second chip which
together form the microcavity. The mirror membrane is movable due to four flexible
suspension beams. Deflection of the membrane as a consequence of electro-thermal heating is
achieved by injecting a small electrical current (<4 mA, equalling 10 mW of electrical power)
through the suspension beams. The second chip can be a simple dielectric DBR on a plane
glass substrate for filter applications (in case of a tunable VCSEL, it is one half of a VCSEL
consisting of DBR plus active region).
It is possible to fabricate a membrane with rotation-symmetric concave curvature in order to
obtain a stable low-loss cavity. The doped mirror material itself conducts the current for
electro-thermal actuation. There is no need of additional metallization. Monolithic approaches
are preferable in terms of fabrication costs. However, the two-chip concept has two main
advantages. The two chips can be optimized independently which is relevant for complex
devices like VCSELs. It is also possible to implement a quite long cavity. For filter
applications, a longer cavity automatically reduces the bandwidth, which is important for
DWDM with small channel spacing. Usually, a two-chip concept has the disadvantage that
assembly costs are high. Here, assembly is done by placing the membrane chip directly on top
of the second chip. A permanent connection between both chips can then be obtained by
bonding, gluing, or using flip-chip technique.
Comparison of tunable-filter technologies
An overview on the most relevant characteristics of the filter technologies described so far is
provided in Table 18.
Table 18.
Comparison of relevant parameters of tunable filters
Insertion
Loss
TFF-FP
3 dB
LCFP
3 dB
MEM-FP
1 dB
FBG
VHG
0.1 dB
(w/o
circulator)
3 dB
AOTF
4 dB
EOTF
4 dB
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Ch. Spacing
/
Isolation
50 GHz / 30
dB
50 GHz / 30
dB
50 GHz / 30
dB
25 GHz / 22
dB
Bandwidth
(3 dB)
Tuning
Range
Tuning
Speed
Power
Consumption
<0.5 nm
~40
nm
~50
nm
~60
nm
<10
nm
ms
Medium
µs
Low
100 µs
Low
ms
Medium
100 GHz /
25 dB
200 GHz /
30 dB
200 GHz /
25 dB
<0.7 nm
ms?
Low
µs
Medium
ms
Medium
<0.5 nm
<0.5 nm
<0.2 nm
<1.5 nm
<1.5 nm
~40
nm
~60
nm
~50
nm
Page: 121 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Insertion
Loss
Cascaded
MZI
AWG
(tunable)
Ring
Resonator
19 dB
(LiNbO3)
5 dB
3 dB
Ch. Spacing
/
Isolation
25 GHz / 22
dB
50 GHz / 22
dB
50 GHz / 30
dB
Bandwidth
(3 dB)
Tuning
Range
Tuning
Speed
Power
Consumption
<0.2 nm
<10
nm
~40
nm
~25
nm
50 ns
Medium
10 ms
Medium
ms
Medium
<0.2 nm
~0.2 nm
From Table 18, suitable technologies for future WDM-PON applications (WDM running via
splitter-based ODN) can be derived. In this context, low cost, low energy consumption, and
channel spacing and bandwidth suitable for full-band DWDM tuning (C- or L-band) should
be achieved. Compactness is also required, especially since it translates, to a certain degree, to
cost. From this, TFF-FP, LCFP, MEM-FP, and VHG filters can be identified as promising
technologies. Note that a first FBG is commercially available, however with the stated
disadvantage of limited tuning range.
5.4.2 Coherent Receivers
Coherent receivers are the only alternative to (tunable) WDM filters when it comes to
wavelength selectivity. In addition, they provide superior sensitivity (or OSNR performance)
through the use of a local oscillator (laser). This is one of the main reasons why they are now
commonly considered as means for all ultra-high-speed transmission. In the (U) DWDMPON context, coherent receivers may provide the power budgets which are required for
splitter-only ODN with very high splitting ratios.
Direct detection as performed by a slow (as compared to the optical carrier) photo diode leads
to envelope detection where all phase information is lost. In addition, it can only make use of
the responsivity of the (PIN) photo diode, or the added gain and sensitivity of an avalanche
mechanism (in cased of APDs) or an optical pre-amplifier. Coherent detection, on the other
hand, leads to a beat term between received input (in) and Local Oscillator (LO) signals
which can be detected by the photo diode and which also preserves the phase information.
The beat term is centred on the so-called Intermediate Frequency ωIF (sometimes also referred
to as RF which adopts the traditional Radio Frequency view). Regarding ωIF, we must divide
three cases:



Homodyne, ωIF = 0
Intradyne, ωIF ≈ 0 (typically <0.5 GHz)
Heterodyne, ωIF > receive signal bandwidth
Homodyne fibre-optic receivers were heavily investigated in the late 1980s in order to
increase transmission span distances. They require an Optical Phase-Locked Loop (OPLL)
which can track the receive signal carrier’s frequency and phase. An OPLL requires
substantial effort, and the respective receivers have been made redundant with the invention
of the Erbium-Doped Fibre Amplifier (EDFA, also invented in the late 1980s). Today, no
major work is done with regard to the traditional homodyne detection scheme.
Heterodyne detection has been considered in fibre-optic transmission with regard to several
applications. Since it only requires Automatic Frequency Control (AFC) of the local laser –
phase tracking is performed in the electrical IF domain – it is less complex. This advantage
comes at the cost of a 3-dB penalty over the respective homodyne (de-) modulation scheme.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 122 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Today, heterodyne detection is considered a possible solution for coherent UDWDM-PON.
Here, it can provide channel selectivity with regard to ultra-densely (~3 GHz) spaced
channels, together with very good sensitivity which becomes necessary in certain splitter-only
ODN.
Intradyne detection is the most recent implementation of coherent receivers. The principle
was first described in [86]. It is based on tuning the LO such that the intermediate frequency
ωIF becomes close to zero. Then, analogue-to-digital conversion is performed, and exact phase
tracking is done in the digital domain. Advantages include the avoidance of analogue OPLL
and also the possibility to implement polarization-diverse receivers and (linear) dispersion
compensators (equalizers, digital filters) which lead to almost zero penalty with regard to
polarization and (chromatic) dispersion effects. Digital intradyne detection is now the
intended quasi-standard solution for all future ultra-high-speed transmission.
In coherent detection, the detected power consists of the contributions from the input and LO
signals, plus the beat term between the two which is caused by the square-law detection:
Pt   Pin  PLO  2 Pin  PLO  cosRFt  in  LO 
Only the beat term is used for demodulation. It allows coherent phase demodulation (because
the cosine argument contains the input signal’s phase), and high sensitivity (because the term
is weighted with the square root of the LO power). After down-conversion, the (intradyne, IF)
signal is low-pass filtered (LPF) and demodulated according to the modulation scheme used.
A schematic diagram (simplified, since it does not yet consider polarization diversity) is
shown in Figure 92.
I t   R  Pin  PLO   2R  Pin  PLO  cosRFt  in  LO 
E in (t )
+
Photo
Diode
Detection +
Shot Noise
E LO (t )
LPF
Envelope Detector
Homodyne Out
Heterodyne Out
Electrical Amplification + thermal Noise
Figure 92 : Coherent detection (basic scheme)
A coherent (digital) intradyne system usually is more complex than indicated in Figure 92.
This is due to the necessity of a polarization-diverse receiver which takes into account that the
polarization planes of the input (receive) signal usually are not known. Hence, input and LO
signal must be split into orthogonal polarizations which are then mixed and detected in
independent receivers. The resulting system configuration can be found in the literature and is
shown in Figure 93for coherent QPSK transmission [87]-[93].
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 123 of 168
90°
QPSK
Coder
Driver
Filter
LO
PC
90°
90°
Hybr.
0°
Client I/F (CFP)
PBS
90°
FEC, Framing, Monitoring
PBC
PC
CDR, EQ., Carrier Rec., Decoder
PC
PBS
CW
LD
0°
ADC
90°
90°
Hybr.
ADC
Driver
Filter
ADC
FEC, Framing, Monitoring
Client I/F (CFP)
QPSK
Coder
ADC
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 93 : Coherent intradyne DP-QPSK transmission
The system shown in Figure 93 makes use of polarization multiplexing. The transmission
format is hence called Dual-Polarized QPSK, or DP-QPSK. Similar schematic system
diagrams result for other modulation schemes, but today much focus is being put on QPSK
due to its superior OSNR characteristics. For polarization multiplexing, the transmit laser
(CW LD) is split into two orthogonal polarization planes by means of a combined polarization
controller / beam splitter (PC PBS). Both polarization signals are independently modulated
and re-combined by a polarization beam combiner (PBC). At the receiver, the input signal is
first split into orthogonally polarized signals by another PBS. The LO signal is also split and
then combined with the input signal by means of two 90° hybrids (these are basically passive
combinations of 3-dB couplers and 90° phase shifters). Each 90° hybrid has dual output ports
for the respective inphase and quadrature components. In order to avoid OSNR penalties, the
output signals are detected by four balanced receivers, i.e., a total of 8 photo diodes. The
photo diodes are followed by fast Analogue-to-Digital Converters (ADC) which are then
followed by the digital receiver. The digital receiver consists of clock data recovery (CDR),
equalizer (EQ) for dispersion compensation (and possibly nonlinear impairment mitigation),
carrier recovery, and finally decoding (here, QPSK decoding). This is followed by means for
framing, monitoring, and error correction (FEC). The interface towards the application is
provided through the client interface (I/F) which, in the case of high-speed transport, is
implemented as a CFP pluggable. In case of low-speed (residential access) services, the client
interface could consist of an optical or electrical SFP instead.
Input
PBS
90°
LO
PBS
90°
I+(TE)
I-(TE)
Q+(TE)
Q-(TE)
I+(TM)
I-(TM)
Q+(TM)
Q-(TM)
Signal
Ex i  E y j
PIN
PIN
PIN
Elo
PIN

ImE E
ReE E
ImE E




Re Ex Elo*
x
*
lo
y
*
lo
y
*
lo
Figure 94 : Two realizations of 90° hybrids in polarization-diverse coherent receivers
For ultra-high-speed transmission, the intradyne system requires ultra-fast ADC, which is a
significant contributor to cost. In order to decrease the symbol rate and consequently the
sampling rate of the ADC, polarization multiplexing is used which increases the transmitter
complexity and cost. This can be reduced in (coherent UDWDM-PON) access systems.
Therefore, digital intradyne must be considered an attractive way of implementing a costeffective coherent receiver for UDWDM-PON. The digital receiver is described hereinafter in
more detail.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 124 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 95 shows the digital receiver of the transmission system of Figure 93 in detail. It
consists of several stages where different functionalities are implemented [87], [88].
Cubic
Interpolator
M
Y90°
Y0°
 
M
|x|2
56GS/s A/D
X90°
X0°
X
1
arg
2
Y-X
exp(-j2fst)
X-Y
Clock Recovery
Y
Equalization
Decider, Decoder
YI
XQ
1
4
X4
Carrier Recovery
YQ
Adaptation Algorithm
X-X
arg .
c0




c1
c2
c3
c4
Y-Y

Decider
e-j(k)
Delay
XI
Figure 95 : Coherent intradyne receiver: digital realization
After detection and sampling at approximate Nyquist rate, the inphase and quadrature
components have orthogonal but arbitrary planes of polarization each. The four components
are now digitally processed in several succeeding filter stages. In a first stage, clock recovery
is performed. Next, equalization of chromatic and polarization-mode dispersion is provided.
This filter stage also allows adjustment of the arbitrary receive-end polarization planes to
those of the transmit signal, i.e., proper polarization demultiplexing is enabled. The last filter
stage performs carrier recovery, i.e., the digital phase tracking.
The clock recovery is based on a digital filter-and-square timing recovery. It also performs resampling to 2 samples/ symbol. Ideally, the ADC would already sample at an exact multiple
of the Baud rate. More details on digital timing recovery can be found in [89].
It is worth noting that some references state a somewhat different structure of the digital filter
stages, for examples refer to [106], [107], [108].
Equalization is often split into two stages. In the first stage, Chromatic Dispersion (CD)
compensation is done. This stage can trial any other filter stage, including clock recovery. It is
based on a simple n-tap Finite-Impulse Response (FIR) filter. Here, the sampled input data is
convoluted with a vector which represents the inverse of the CD transfer function. An
example of a simple 7-tap FIR filter is shown in Figure 96.
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.3
-0.2
-0.1
0
time (ns)
0.1
0.2
0.3
Figure 96 : Example of simple FIR filter response (blue: Re, red: Im part)
According to [92], the upper bound for the tap number at a Baud rate of B GBd is given by
0.032∙B2 per 1000 ps/nm of chromatic dispersion. At 10.7 GBd (43G DP-QPSK), this
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 125 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
translates to 3.7 taps per 1000 ps/nm which can reduced to 2.2 taps per 1000 ps/nm with
minimal penalty. For 28GBd (112G DP-QPSK), the number of taps increases by a factor of
almost 7 as compared to 42.8 GBd (CD increases with the square of the Baud rate).
PMD equalization and polarization recovery (demultiplexing) is performed by four complexvalued FIR filters. (These are also referred to as Feed-Forward Equalizers, FFE, transversal
filters, or tapped-delay-line filters.) These filters are implemented in a butterfly structure, as
multiple-input/multiple-output adaptive equalizer between the polarization planes, see Figure
95. The filter taps are spaced T/2 in time, where T is the symbol duration. Their tap number
correlates with the target dispersion-compensation capability.
The tap weights have to be optimized, usually using blind equalization techniques in order to
avoid the necessity of training sequences. Blind adaptation can be done, for example, with a
Least Mean-Square (LMS) algorithm. The earliest work on such adaptive filters can be traced
back to the late 1950s. From this early work, the LMS algorithm emerged as a simple, yet
effective, algorithm for the design of adaptive transversal filters. The LMS algorithm was
devised by Widrow and Hoff in 1959 [96]. It is a stochastic gradient algorithm in that it
iterates each tap weight of the transversal filter in the direction of the instantaneous gradient
of the squared error signal with respect to the tap weight in question.
An alternative to LMS is the so-called Constant-Modulus Algorithm (CMA) [93],[94]. One of
the most important features of CMA is that it can equalise constant-modulus signals (such as
QPSK, 8PSK) as well as non-constant-modulus signals (like 16QAM). CMA seeks to
minimize a cost defined by the CM criterion. The CM criterion penalizes deviations in the
modulus (i.e., magnitude) of the equalized signal away from a fixed value. In certain ideal
conditions, minimizing the CM cost can be shown to result in perfect (zero-forcing)
equalization of the received signal.
CMA employs a cost function that does not discriminate between the two equalized signals.
Hence, it is common that this algorithm converges to a tap-weight setup that produces the
same transmitted signal at both equalizer outputs, usually the one that arrived with higher
power at the receiver. The equalizer matrix in this case becomes singular. This problem can be
circumvented by frequently monitoring the equalizer’s matrix determinant and reinitializing
the tap-weights when singularity is approached, a solution that is practical, but could cause
discontinuity issues at the equalizer output. A computationally demanding equalization
algorithm based on the independent component analysis method has also been proposed. A
more recent improvement is multiuser CMA, which is an extension of the conventional CMA
to the multiple input case, such that singularity is avoided. Multiuser CMA employs an
enhanced cost function, penalizing the correlation between the two signals at the equalizer’s
outputs [95].
Frequency and phase offset between local laser and receive signal is corrected by a 4 th-power,
Viterbi-and-Viterbi carrier recovery stage [99]. The approach is to calculate an optimal phase
estimate, and to implement this on a parallel DSP. The best-possible phase estimate is derived
through the Maximum a Posteriori (MAP) estimate where phase (n) and data d(n) are jointly
estimated [101]. The MAP estimate can be calculated by applying a per-survivor method to a
group of symbols, and calculating the phase by successive approximation for each symbol
group instance.
Neglecting higher-order noise terms, small-angle approximation is applied which yields a
phase angle θ = 2 + (additive noise component). Estimation theory says that the best linear
estimate of  is derived if a Wiener filter applied to θ. The Wiener filter solves the signal
estimation problem for stationary (or cyclo-stationary) signals. It was introduced by Norbert
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 126 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Wiener in the 1940s [102]. The filter is optimal in the sense of the Minimum Mean-Square
Error (MMSE). (The Kalman filter solves the corresponding problem in greater generality for
non-stationary signals.) A Wiener phase estimator is shown in Figure 97.
Complex
Signal
d ei+p
(  )2
ei2+p’
arg(  )

Wiener
Filter
÷2
exp(  )
Phase
Estimate
Figure 97 : Phase estimation using Wiener filter
The phase is estimated and applied to the signal before making the 1/0 or multi-level decision.
A smoothing (low-pass) function is needed to reduce effects of additive noise and pass the
actual phase change. Errors in the phase estimate lead to an increase in number of bit errors.
They can lead to so-called cycle-slip errors, i.e., data inversion in case of BPSK or QPSK.
This is the reason why even in coherent intradyne systems differential pre-coding (i.e.,
DQPSK) is applied although it is not necessary from the viewpoint of differential delay
demodulation using self- or incoherent MZM. Figure 98shows the effects of phase noise and
phase plus added amplitude noise on BPSK, respectively.
No Noise
Phase Noise only
Phase + Amplitude Noise
Figure 98 : Phase estimation in presence of phase noise and combined phase plus amplitude noise
One key function is to recover the carrier phase using DSP-based phase estimation (PE) rather
than optical phase-locked loops, thus allowing for a free-running LO laser. Some popular
phase estimations, such as Mth-power [99], require that the frequency offset between
transmitter and local oscillator (LO) laser should be quite small compared to symbol rate. The
frequency offset between transmitter and LO lasers, however, can be as large as ±5 GHz. As a
result, an additional DSP-based Frequency Offset Estimator (FOE) is required to ensure that
subsequent PE algorithms can accurately recover the phase of received signals.
A feed-forward FOE is preferred to avoid performance degradation when being implemented
in parallelism. Mth-power is generally performed to remove data modulation in feed-forward
FOEs. The maximal estimation range is limited to [-Rs/2M, Rs/2M], where Rs refers to the
system symbol rate and M is the number of constellation states. This leads to an estimation
range is ±0.125 Rs for QPSK.
In a standard intradyne receiver, the frequency offset between the receive signal and the local
oscillator lasers should be no more than 10 percent of the symbol rate [98]. For 43G DPQPSK transmission, this leads to ~1 GHz allowable frequency offset. In order to allow larger
offsets, more tolerant Frequency Offset Estimators (FOE) are required. In [103], an ultrawide-range FOE has been described. It is based on a double-stage FOE architecture as shown
in Figure 99.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 127 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Re-sampling to 2
sample/symbol
I (k ) & Q(k )
Resampling
Down-sampling to
1 sample/symbol
 jk fˆcTs
e
e
CMA
↓
Delay
NCO
NCO
MPE
Sweeper
Coarse FOE
Look-Up
Table
Sign
Identifier
 jk fˆ f Ts
fˆc
FFT-based
FOE
fˆ f
1/-1
Fine FOE
Figure 99 : Cascaded FOE [103]
The dual-stage, cascaded FOE consists of a coarse FOE and a fine FOE. In 43G DP-QPSK
systems, the estimation range of the dual-stage FOE has been shown to be 4 times what can
be achieved with the Mth-power algorithm. This is the largest range of FOEs reported so far
in the literatures.
Timing recovery, such as Gardner algorithm, is required to correct for the timing phase error
between the transmitter and receiver clocks in coherent receivers. The simple Gardner
algorithm can be used to generate a phase error output when only two samples per symbol are
available. However, the performance of the Gardner algorithm suffers from performance
degradation in the presence of frequency offset. Such offset makes the Gardner algorithm less
insensitive to the sampling offset in coherent receivers.
The relationship between frequency offset and MPE of the Gardner algorithm can be used as a
measure to estimate the frequency offset. The absolute MPE varies with OSNR and is
normalized to the one at zero frequency offset (FO). Then, the normalized MPE show almost
the same trend within ±9 GHz for different OSNR. A polynomial fit is applied to the MPE/FO
relationship. This polynomial fit can offer a coarse FOE (ΔfC) which is limited to ±1 GHz.
This is within the estimation range of the Mth-power algorithm which is cascaded as fine FOE
in a second step (with ΔfF). The polynomial fit of the MPE/FO relationship can be realized
using a look-up table. Cascaded FOE is capable of estimating a frequency offset up to the
range [-0.5Rs, 0.6Rs], which is approximately 4 times the theoretical limit of single FOE
using Mth-power.
Ultra-narrow linewidth lasers are required to implement a phase-locked loop for carrier
recovery [86], and these are widely believed to be too expensive in today’s cost-sensitive
economic environment. In contrast, a feed-forward carrier recovery scheme relaxes the sum
linewidth requirement to about 0.001…0.0001 times the symbol rate, which is in the reach of
normal, low-cost DFB or DBR lasers.
Phase noise is an important impairment in coherent systems as it impacts carrier
synchronization. In non-coherent detection, the carrier phase is unimportant because the
receiver only measures energy. In DPSK, information is encoded by phase changes, and Δν
only needs to be small enough such that the phase fluctuation over a symbol period is small.
The receive signal is modulated by ejφ(t). In the absence of other impairments, this leads to a
rotation of the received constellation. Carrier synchronization is required to ensure φ(t) is
small so the transmitted symbols can be detected with low power penalty. (φ(t)= φS(t)- φLO(t))
With the phase error variances for FF carrier synchronizer, the power penalty can be
determined. In Table 8, we compare the linewidth requirements for receivers that use FF
carrier synchronizer, assuming a 1-dB power penalty at a target BER of 10−3 [105].
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 128 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Table 19 :
Linewidth requirements for single-polarization modulation at a target BER of 10 −3
QPSK
16-QAM
OSNR per bit
[dB]
7.79
11.52
Max. σφ
Max.
[ΔνTb]
Linewidth
4.91°
1.3×10−4
2.70°
1.5×10−5
In OFDM, laser phase noise destroys the orthogonality of the sub-carriers and causes InterChannel Interference (ICI), which has noise-like characteristics. ICI grows linearly with NC,
with NC the number of sub-carriers. Thus, an important parameter is the linewidth-to-subcarrier-spacing ratio ΔνTC/NC. Phase noise considerations favor a smaller number of subcarriers. At higher phase noise, the carrier phase can no longer be assumed as constant over an
OFDM symbol. The use of an MMSE equalizer which considers ICI and AWGN can improve
system performance [Ip-111]. However, this technique does not estimate or compensate for
phase noise, unlike feed-forward carrier synchronization for single-carrier modulation.
Carrier synchronization in OFDM may require an iterative algorithm. Suppose at the
beginning of an OFDM symbol, an initial estimate of phase kφ ~ is known from the previous
symbol. We can first de-rotate the entire OFDM symbol (NC chips), and then perform the FFT,
equalization and symbol detection. We can multiply the symbol decisions by the channel’s
frequency response and then take the IFFT to compute what the time samples should have
been without phase noise. This allows the receiver to compute estimates of the carrier phase
for each chip period. MMSE filtering can then be employed to find more reliable phase
estimates. We can de-rotate the OFDM symbol again, and a second iteration follows. This
process can be repeated until convergence is achieved. The performance of such an algorithm
has not yet been characterized.
In DSP-based coherent detection, Nonlinear Compensation (NLC) allows the mitigation of
nonlinear intra-channel effects. This is possible since these effects – self-phase modulation –
are deterministic and can hence be compensated. Compensation of inter-channel effects
(XPM, FWM) would require a much more complex MIMO equalizer.
NLC can be implemented, together with chromatic-dispersion compensation, as a fixed
equalizer which trials the adaptive (CMA) equalizer [106]. In order to reduce computational
effort, it is desirable to compensate CD with a frequency-domain filter, followed by timedomain instantaneous NLC which compensates the nonlinear phase shift. In multi-span longhaul transmission, linear and nonlinear intra-channel compensation can be done iteratively,
switching between time and frequency domain by means of FFT and IFFT, respectively, see
Figure 100.
ADC
Resampling
Linear only
Dispersion Compensation
FIR Filter
Fixed Compensation
Disp. Compensation
FOE
IFFT
N Spans
FFT
CMA Equalizer
Carrier Phase Est.
NL Phase Shift
Complex Decisions
Figure 100 :
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Linear and Nonlinear
Linear-only or linear plus nonlinear compensation [106]
Page: 129 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Using intra-channel NLC, the optimum long-haul launch power can be increased by ~1 dB,
and maximum reach for 42.7 Gb/s and 85.4 Gb/s can be increased to more than 11,000 km
and 8000 km, respectively [106].
An alternative homodyne / intradyne detection scheme which avoids both, polarization
diversity and the necessity for 90° hybrids, has been proposed in [110]. The concept was
designed for low-cost applications, for example in coherent ultra-dense WDM-PON systems
but is not restricted to such applications.
Polarization diversity and 90° hybrids are used in coherent detection in order to avoid
polarization control (of both, the receive signal and the local laser) and to provide the Inphase
and Quadrature components of the down-converted signal, respectively. A fully-blown
polarization-diverse I/Q demodulator requires 4 branches with 8 photo diodes if balanced
receivers are used, refer to Figure 93. In the PON context, this leads to prohibitively high cost.
Avoidance of polarization diversity / control and 90°-hybrid-based I/Q demodulation
according to is based on polarization and phase scrambling. The idea is to have the orthogonal
states for phase and polarization in the same bit sequentially. Thus, for the first half of the bit
the signal relative to one orthogonal component (I or H) will appear, whereas for the second
half of the bit the signal relative to the other orthogonal component (Q or V) will be seen. The
principle diagram is shown in Figure 101.
Downstream
+
3dB
Tb
Pol. Scr.
PM
Tb/2
Tb/2
Freq. Doubler
RZ50
CLK Rec.
Upstream
3dB
Q
H
Data
Figure 101 :
LD
t0
Q
V
I
H
I
V
Q
H
t0+Tb
Q
V
I
H
I
V
t0+2Tb t
Data
Homodyne detection without polarization diversity and 90° hybrids
Within each symbol duration Tb, the phase of the local laser is scrambled by means of an
RZ50 (Return-to-Zero with 50% duty cycle) signal which is synchronized with the receiver
clock recovery (CLK Rec.). The RZ50 signal generates, within each symbol interval, two time
slots of duration Tb/2 with fixed 0° to 90° phase modulation. One slot represents the I
component, the second slot represents the phase-shifted Q component. This is demonstrated in
the insert in Figure 101 (together with the sub-slots resulting from polarization scrambling).
By gating the photo receiver output with the data clock and its inverse, the I and the Q
components are obtained separately in the two branches, now with RZ shape. The signal
power fluctuates between the I and the Q branches randomly, due to phase noise, at a rate of
the order of the laser linewidth, and the combination of both outputs can assure its recovery.
This operates like a phase-diversity system. The local laser does not need to be phase coherent
with the incoming optical carrier, although an automatic wavelength controller is convenient
to maintain the two wavelengths close each other. This can be regarded as an intradyne
receiver with near-zero intermediate frequency. As such, dual-fibre working is required in
order to avoid downstream/upstream cross talk.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 130 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Using delay demodulation in the electrical domain, this scheme is suitable for detecting PSK.
(Specially in the PON context, this is no restriction. PSK at 1 Gb/s per channel gives
sufficient bandwidth efficiency to allow for more than 1000 bi-directional channels.)
Electronic signal processing performs differential demodulation of both I and Q components,
with a delay time equal to Tb, and the synchronous combination of the I and Q components.
In Figure 101, polarization is scrambled as well. The polarization scrambler is run at
duplicated symbol rate in order to produce the orthogonal polarization states per I and Q time
slot. This now leads to 4 time slots of duration Tb/4 which contain all combinations of the
orthogonal states of both, phase and polarization. This has been shown in the insert in Figure
101 already. This also means that part of the receive electronics must operate at an increased
bandwidth.
Several derivatives of the sequential phase / polarization diversity demodulator have been
proposed in [110] and further work by the same authors. These mainly aim at providing
different detection schemes with different performance (or penalty against a fully-blown
digital polarization- and phase-diverse intradyne receiver), and at different hardware effort.
One such configuration is shown in Figure 102.
Downstream
+
3dB
Tb/2
Tb
PM
VQ
VI
RZ50
Tb
Upstream
3dB
CLK Rec.
LD
I
Data
t0
Figure 102 :
Q
t0+Tb
I
Q
Data
t0+2Tb t
Reduced homodyne detector
The receiver shown here uses phase diversity by means of scrambling only. This configuration
assumes that polarization switching is performed in the OLT. Given that synchronization
between the polarization scrambler and all downstream channels can be achieved in the OLT,
the polarization scrambling can be done on the optical multiplex section in a single
centralized component, thus reducing cost. Note that the same will not work for the upstream
direction since synchronicity between the upstream channels can not easily be achieved due to
uplink length differences.
For cost efficiency and in order to achieve optimum performance, the configuration shown in
Figure 102will be implemented as a digital intradyne receiver. The resulting configuration is
shown in Figure 103. This is again the reduced variant without dedicated polarization
scrambling. Like the other variants discussed in here, it requires two fibres for downstream
and upstream.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 131 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Downstream
+
3dB
Upstream
3dB
Data
ADC
RZ50
PM
I/Q Proc.
Data
CLK Rec.
LD
Figure 103 :
Digital implementation [110]
The disadvantage of dual-fibre working can be avoided by using a heterodyne detection
scheme instead of (digital) intradyning. Then, the local laser is tuned to an offset against the
receive wavelength which is larger than the payload bandwidth. This offset (the intermediate
frequency in heterodyning) allows to directly re-use the laser wavelength for upstream on he
same fibre. This scheme also does not require 90° hybrids since I/Q processing is done in the
electronic (RF) domain. In addition, it can be combined with polarization scrambling, which
is shown in Figure 104.
Downstream
3dB
Data
+
3dB
RF Detection
Upstream
3dB
Pol.
Scr.
CLK Rec.
LD
Data
Figure 104 :
Polarization-scrambled heterodyne detection
The coherent receivers intended for UDWDM-PON discussed so far can reduce cost, as
compared to a fully-blown intradyne receiver. This cost reduction, however, comes at a
certain signal penalty. Neglecting further implementation penalties (which is valid for digital
realizations), phase scrambling leads to a penalty of 3 dB, as compared to homodyning or
intradyning using 90° hybrids and twice the number of photo diodes. Polarization scrambling
leads to another 3 dB penalty, compared to polarization diversity. Heterodyning also leads to 3
dB penalty as compared to homodyning / intradyning, given the modulation scheme (here:
PSK) stays the same. Altogether, up to 6 dB penalty (for digital implementation) can result.
This is summarized in Table 20.
Table 20.
Comparison of coherent WDM-PON schemes
Phase
Handling
90° Hybrid
90° Hybrid
Scrambling
Heterodyning
Polarization
Handling
Diversity
Scrambling
Scrambling
Scrambling
Sensitivity
Penalty
0 dB
3 dB
6 dB
6 dB
Linewidth
Tolerance
5 MHz
5 MHz
5.4 MHz
5 MHz
Cost
Estimation
High
Medium-High
Medium
Medium
With 6 dB added penalty, a sensitivity of better than -45 dB can easily be achieved, assuming
that PSK is used at a bit rate of 1 G/s. This can be considered enough for PON applications,
even at long access distances and high customer numbers (splitting ratios). The question
however remains if the various coherent receivers reduce complexity and cost sufficiently in
order to allow large-scale access application. Even the polarization and phase scrambling
requires substantial effort. For polarization scrambling, an additional component (the
scrambler) is required. For phase scrambling, an additional phase modulator, together with the
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 132 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
synchronized I and Q delay demodulators running at duplicated bit rate are required.
5.5 BURST MODE RECEIVERS
Burst Mode Receivers (BMRs) are a key component in next-generation optical access packetswitched PON networks. Located within the central office (CO) BMR designs have to
overcome two major challenges: maintaining data recovery with large power fluctuations
between adjacent packets and rapid (ideally instant) clock recovery on packet arrival. These
variations in incoming data packets are due to the different paths that the bursts follow
through a PON network. Moreover, additional issues such as poor extinction ratio, noise
accumulation and random burst arrival times also have to be considered when designing an
appropriate BMR. In addressing all these challenges, future BMR design also needs to
deliver fast-response peak-detection to deal with threshold variations and nanosecondswitching. The system additionally needs to cope with variable gain transimpedance
amplifiers; preset minimum guard time between consecutive packets and a timing–alignment
preamble in front of each packet, so as to adapt the power intensity and bit phase of each
incoming burst to a reasonable level able to be detected. The variation in power between
adjacent packets is characterised by the loud-soft ratio (see Figure 105) with the associated
extinction ratio parameter also indicated in the figure.
1
ONU 1
2
ONU 2
OLT
BMR
n
Loud/soft
ratio
ONU n
1
n
Extinction
ratio
2
Figure 105 :
Typical Passive Optical Network (PON) scheme with a BMR located in the OLT
The maximum dynamic range of the loud/soft ratio is typically up to the order of 20 dB, with
a sensitivity of -28 dBm. Recent BMR designs, e.g. [111] have achieved a receiver sensitivity
of -30.8 dBm at 10G, and -35.5 dBm at a bit rate of 1.25 Gb/s. Conventional ON–OFF keying
(OOK) receivers also experience a sensitivity penalty associated with the adaptive decision
threshold used to handle large inter-packet amplitude variations. Fortunately, differential
coding schemes employing balanced detection, such as differential phase-shift keying (DPSK)
demonstrate a large tolerance to signal power fluctuations, fibre nonlinearities and
polarization mode dispersion [112]- [113] making DPSK coding very attractive for optical
packet/burst switching. Two major issues related to clock recovery in NRZ-DPSK BMR
design are firstly: acquisition of a clock from the NRZ signal which as a rule doesn’t contain a
clock frequency component [114] and secondly, to maintain this clock signal during guard
times between the bursts/packets. For a BMR with adaptive decision thresholds,
determination of the required decision threshold needs to be rapid, i.e. measured in the order
of a maximum of tens of nanoseconds. Recent examples of 10G BMR designs for EPON
applications achieve lock in times of 37 ns at 10.3 Gb/s, and 64 ns at 1.25 Gb/s [115]; an
overhead of 280 ns at a bit-rate of 10.3 Gb/s has been demonstrated in [116], and 10 ns
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 133 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
response time for a 10G BMR has also been reported by Nakamura et al. [117].
Clock Recovery & Timing Extraction
As already discussed above, rapid (as near instant as possible) clock recovery on packet
arrival is essential for successful BMR receiver operation. A phase lock loop (PLL) finds
widespread application in many BMR designs to recover synchronisation from a randomly
arriving packet, e.g. [118] In order to achieve rapid acquisition of timing information and
quick lock-in the PLL must feature a wide bandwidth of operation. However, a wide
bandwidth associated with the various PLL multiple feedback paths (both linear and nonlinear in operation) means that a lot of noise is also admitted into the PLL, so leading to high
jitter levels. This means that PLL-based clock recovery has difficulty in simultaneously
achieving both swift and accurate lock-in times. An alternative method for rapid
synchronisation with a low timing latency is based on the use of passive high-Q resonance
filtering, e.g. [119]. Such a resonant circuit based approach is highly scaleable, being equally
appropriate for clock frequencies from as low as the sub-GHz regime up to the multi-100’s of
GHz range. A high-Q resonant circuit, where the Q value is greater than 1000, means that
1000 clock cycles have to elapse before the amplitude of the resonant response exponentially
falls below the 1/e level, and de-synchronisation occurs. This indicates the stability of the
high-Q resonator approach - in particular, from classical filter and signal theory, the phase
response is closely related to the amplitude response of the resonator (i.e. they form a Hilbert
transform pair) such that phase and amplitude stability are mutually ensured. For a Q>1000
resonator, provided that the intervals between incoming packets don’t exceed, for example,
1000 bits, then once the circuit is already resonating, any phase misalignment of the arriving
packet is passively corrected for within as few as 3 bits, e.g. [119]. The advantages of this
passive resonant high-Q scheme compared to PLL-based schemes are the simplicity and cost
efficiency and the phase acquisition time (at the centre frequency) which are both critical for
operation in a BMR.
Current techniques to BMR design can be divided into two main categories: DC and AC
coupling. Both techniques require a minimum guard time between two consecutive packets
and a preamble for adjustment in front of each packet, which results in a reduction in the
actual efficiency of channel capacity exploitation. It’s also worth noting that the length of the
guard band is ambiguous due to the lack or no knowledge of the arrival of the subsequent
packet. Within the overall concept of AC coupling, is another important emerging BMR
technique based on edge detection. This is emerging as an important technology to achieve
the ever-higher burst mode bandwidths and speeds that are a feature of next-generation optical
access networking. In the following, we consider these three main approaches, discussing
their strengths and weaknesses, as well as their appropriateness to meet the needs of future
NGA optical networking.
5.5.1 AC Coupling
AC coupling technique removes the input DC bias and biases the signal at the average signal
power level, with no feedback and decision level setting requirements required, which makes
the AC coupling technique relatively easy to implement. As such it is a well-established
technology in the field, being cheap to implement, and requiring no feedback or decision level
setting. Reference [120] describes an AC-coupled BMR proposed for Gigabit Ethernet PON
systems. Conventional AC coupled receivers have long time constants (typically of the order
of ms), which, however, is rather too long as bursts become shorter; i.e. a long time constant
exceeds the burst duration. Smaller capacitors at the front end can reduce the time constant,
but tend to filter out low frequency components of the signal, resulting in ISI. As such, line
coding (e.g. 8B/10B) is required to reduce long strings of consecutive identical digits (CIDs).
Scrambling is an alternative to line coding with its associated redundancy, but requires a
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 134 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
longer time constant (and preamble) since the protection against long CIDs is still only on a
statistical basis. In AC coupled BMRs, long CIDs cause level drifting and clock and data
recovery (CDR) malfunction, in which case the time constant needs to be longer than the
longest CID acceptable for a given CDR and BER.
Additional requirements for AC-coupled systems are generally determined by the acceptable
BER for the system. For example, the BMR needs to settle down in less than the preamble in
order to receive the data within the required BER, whilst the signal also needs to be held for
longer than the maximum CID associated with the acceptable BER. By way of example, at
2.5 Gb/s the Guard Time (64 bits) is 25.6 ns, whilst the Preamble (108 bits) lasts 43.2 ns, and
the conventional maximum CID is 72 bits corresponding to 28.8 ns.
Assuming an even mark-space ratio for the data, the AC threshold is conventionally set to the
midpoint. A large change in burst amplitude requires a finite settling time during which data
will not be received. For example, the maximum time constant is determined by the settling
time between loudest and softest bursts, in particular settling to within the upper threshold
level. The minimum time constant is determined by the maximum CID period whilst
remaining within the upper threshold level. Unfortunately reducing the maximum CID period
is difficult, e.g. reducing from a maximum of 72 bits to a maximum of 12 bits is desirable.
That said, maximum CID length is often built into most physical layer formats, e.g. GigE
employs 8B/10B encoding which ensures no than a maximum of 6 CIDs. However, even a
well chosen time constant will still result in some ISI due to the relative filtering effects at
lower frequencies, with the situation worst for the soft to loud case.
Achieving such performance is becoming increasingly difficult in AC-coupled BMRs,
particularly with GPON and LRPON specifications, although a 10G BMR AC-coupled design
has recently been demonstrated by NTT [121] featuring a 10.3 Gb/s data rate, with 1001 bits
of CID tolerance.
5.5.2 DC Coupling
In a DC coupled BMR, the decision threshold changes according to each burst amplitude, so
that the drifting affects associated with AC-coupled schemes due to varying average power
levels are avoided. In this case, the decision threshold level is determined using only the
amplitude of the current bursty traffic, without any memory effects. Unfortunately, although it
avoids the time constant issues of the AC-coupled approach described above, implementation
of the DC-coupled technique within practical systems is more complex proposition compared
to AC coupling. Reference [122] describes a DC-coupled BMR receiver which employs peak
detectors to extract a decision threshold from a sequence of 12 successive non return-to-zero
(NRZ) 1’s and 12 successive NRZ 0’s received at the beginning of each packet, whilst a DCcoupled 1.25 Gb/s burst-mode receiver with automatic offset compensation has been
demonstrated in [123].
A DC-coupled BMR allows a feedforward architecture to be employed in the basic front end
design, with amplitude recovery performed in a differential pre-amplifier setup. Such a
feedforward design allows a faster settling time between bursts, and is important for reset
functionality; however, feedforward is inherently less stable than a feedback-based design and
more prone to oscillation. It also makes the design more complicated and expensive as
compared with the AC-coupled version. A specific design consideration for a DC-coupled
BMR is that the peak detector may need to detect both high and low levels to prevent markspace distortion, especially when the extinction ratio is poor. A fast reset is also required in
order to recover from bursts arriving within the guard period – in this case, a feedforward
design will provide the required speed. For systems employing fixed packet length formats
(e.g. ATM) a fast reset is less of a problem, but variable burst length standards such as
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 135 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Ethernet (which are generally now the preferred option) present more of a problem and
require fast reset functionality.
Particular issues with the DC-coupled BMR approach include the need to reset the threshold
level with each packet, and the need to reset in less than the guard time period. Decision
settling is not assisted by the guard period duration, and must be completed within the
preamble time (e.g. 43.2 ns for a GPON system at 2.5 Gb/s). Feedback control is also still
necessary to reduce pulse width distortion (PWD) which increases the BER due to the
reduced sampling duration (time setting) and may also lead to CDR malfunction since the
CDR assumes equal ones/zeroes pulse widths. A PON compatible DC-coupled BMR
developed in the PIEMAN project has been reported in [124] and is capable of operating at 10
Gb/s, featuring a guard time as short as 25.6 ns, a preamble of only 23.8 ns, and able to
withstand up to 72 CIDs. NTT have recently demonstrated a 10G DC-coupled BMR featuring
a 74 ns lock-in time at a sensitivity of -18 dBm [125][126].
5.5.3 Edge Detection
In an attempt to address the inherent problems associated with DC and conventional ACcoupled schemes, edge detection is emerging as an attractive solution, overcoming many of
the key technical issues in BMR design. Here, advantage is taken of the short time-constant
differentiation process to produce delta functions of alternating polarity at the start and finish
of a bit. Once the edges have been detected, a high-speed comparator discriminates the
received mark/spaces. However, this come at the cost of a still more complex design
compared to the DC and AC coupling approaches, requiring the optimisation of a range of
different sub-system parameters: differentiator constant, comparator holding time, RC
constant and the impact of consecutive bits. Clock acquisition is achieved by introducing a
low-level temporal nonlinearity in the DPSK demodulator, thus generating a spectral
component at the clock frequency. This allows data recovery and clock acquisition to be
achieved from the same data stream. The clock is then extracted in the second stage by using
an elegant high-Q cavity resonator filter technique. Clock recovery using the phase-locked
loop technique has also been demonstrated at a bit rate of 2.5 Gb/s [127]. Experimental results
supported by simulations demonstrate the validity of the edge-detection technique for high
speed (>10 Gb/s) [119] optical PON networks. These results show the advantages of the
edge-detection techniques over the other techniques and its main limitations to be used in
future PON networks.
Figure 106 : a) mass distribution function of differentiated signal level sampled immediately after the start
of each bit b) mass distribution function of differentiated signal level sampled after passing
MLEPW(set to 0.05 of bit length) after the start of each bit. SNR for both scenarios was set to
9 dB.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 136 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Specific issues for edge detector BMRs related to the optimisation of the above-mentioned
system parameters include:
•
The pulse must exceed the comparator threshold voltage prior to the decision time
•
The noise present on the signal must not trigger a false reading following the first
trigger and prior to the decision time
In addition, multiple bits provide a special problem for this receiver design because if bit m
from a sequence of n bits triggers a false level on the comparator then the subsequent n-m bits
will be in error as well (unless another positive error is triggered). However, as analysis
(described below) indicates, the overall BER penalty is actually quite minimal.
The basic principle of hysteretic edge detection is shown in Figure 106 and Figure 107. We
assume that the hysteresis thresholds are set optimally so that as soon as the differentiated
signal exceeds the hysteresis thresholds the BMR receiver responds to it. In practice, there is a
minimum holding time for the signal to be stable so that the receiver doesn’t miss it. Figure
106(a) highlights the well-known base-line variation inherent in conventional AC-coupled
BMRs. It is clear that long sequences of 1’s and/or 0’s will remove the decision reference
threshold. If the AC-coupling time constant is reduced to below 20% of the bit duration, then
inter-pulse influences are reduced asymptotically to zero as shown in Figure 106(b). The
information contained in the original data sequence may be recovered completely by noting
that the short time-constant differentiation process eventually produces delta functions of
alternating polarity at the start and finish of a bit. Standard signal theory shows that an
integrator can then be employed to remove the differentiation, with such functionality
contained in a latest-generation SiGe comparator (e.g. Analog Devices ADCMP580 series).
As shown in Figure 107, a positive impulse sets the comparator output to high with this
condition persisting indefinitely until a negative impulse resets the output to low. By such a
means an exact regenerated copy of the input pulse stream is produced.
Figure 107 : Mass distribution function of differentiated signal level sampled immediately after the start
of each bit. SNR= 7 dB.
Ideally a small value of RC constant for the differentiator should result in improved
performance by avoiding the effects of baseline drift. In other words, a smaller differentiator
RC constant allows the receiver to respond to the changes more quickly, while at the same
time the capacitor can be discharged faster to avoid the possibility of base line drifting.
However the maximum speed of the comparator puts an upper limit to the differentiator RC
constant, i.e. a small RC constant results in short pulses that may not be captured by the
comparator. In this case, the Minimum Latch Enable Pulse Width (MLEPW) is the minimum
time that the latch enabling signal must be high in order to acquire an input signal change. If
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 137 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
the signal level decays below the threshold level before this minimum holding time has
passed, then the comparator may miss the pulse. Such an event has a dramatic effect on BER
performance since if a pulse is missed, provided that the subsequent bit has the same level
(i.e. no value change) the subsequent bit will be also received in error until an edge is
detected. The value of MLEPW depends on the capabilities of the comparator, but a higher
value of MLEPW results in inferior BER performance. Figure 108 shows the graph of BER
versus the signal SNR for different values of MLEPW. The figure clearly shows the effect of
increasing the value of MLEPW. Figure 106 gives a better understanding of the role of
MLEPW. Figure 106(a) shows a distribution of the differentiated signal level immediately
after the beginning of each bit. To show the mass distribution function more clearly, the input
signal level is assumed to be 5 volts for all bursts. The mass distribution function is calculated
over 20,000 samples and the RC constant of the differentiator was set to 10% of the bit length.
Figure 106(b) shows the result of another simulation run with the same setting but the
distribution function is plotted for the differentiated signal level when a time slot equal to the
MLEPW is passed after the start of each bit. In this case, the value of MLEPW was
exaggeratedly set to 5% of the bit length.
0
10
-2
10
-4
BER
10
MLEPW
MLEPW
MLEPW
MLEPW
-6
10
-8
=
=
=
=
0
0.01 * Bit Length
0.02 * Bit Length
0.03 * Bit Length
10
-10
10
2
4
6
8
10
SNR (dB)
12
14
16
18
Figure 108 : BER versus the SNR for different values of MLEPW.
The first graph of Figure 106, can be interpreted as the ideal case where MLEPW=0, this
evidently shows the impact of MLEPW as the distribution in Figure 106(b) is more
concentrated towards the centre as a result of the capacitor discharge during the MLEPW
interval. Clearly this results in a poorer BER performance. The effect of AWGN noise on the
distribution of the differentiated signal levels after start of each bit is illustrated in figure 3.
This shows that a decreasing channel SNR makes the peaks more difficult to be distinguished.
Optimum RC constant of differentiator
The value of the RC constant of the differentiator plays an important role in the BER
performance of the receiver. If the RC constant is too high, this results in very narrow
differentiated pulses which the comparator may not be able to capture. On the other hand, a
high RC constant will result in baseline drift effect. As such, the optimum value of the RC
constant needs to be chosen with regard to the value of the MLEPW. By way of example,
Figure 109 shows a simulation of the differentiated pulses with high and low RC constants.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 138 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
1.5
1.5
1
1
0.5
0.5
0
0
-0.5
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1.5
-0.5
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1.5
1
1
0.5
0.5
0
0
-0.5
-0.5
-1
-1
-1.5
-1.5
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
(a)
(b)
Figure 109 : a) the differentiated pulse with very large value of RC constant for differentiator b) the
differentiated pulse with small value of RC constant for differentiator.
Impact of consecutive bits (CID)
Since the receiver is responding to the bit changes rather than the individual bits, in case of a
wrong decision, if there is no bit change immediately after the erroneous bit, subsequent bits
will therefore tend to be erroneously interpreted. Fortunately, analysis and simulation results
show that this does not severely degrade BMR receiver performance. Figure 110 shows the
graph of BER versus SNR for a pseudo-random sequence and a succession where no
consecutive bits can be found. The successive bits degrade the BER performance of the
receiver in a way that if an erroneous bit is received, then all other subsequent bits are
received incorrectly until a bit-change has occurred. As an example, in the case of two
successive bits, the BER will be doubled. However, the probability of two consecutive bits in
a bit stream is 0.5 (‘00’ or ‘11’). In order to calculate the average BER penalty due to such an
event, the penalty introduced by having specific number of repeated bits needs to be
multiplied by the probability of occurrence of that specific number of bits in the bit-stream
sequence. Using the Gabriel's Staircase series we have:
(1)
In this case r=1/2 meaning that the total BER will change by factor (0.5/0.25)=2 due to the
consecutive bits. Figure 110 shows the BER performance of the PRBS sequence and
compares it with the case where no consecutive bits exist in the bit-stream. The result
satisfactorily verifies the above analytical discussions.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 139 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
0
10
-2
BER
10
-4
10
PRBS
No Consequtive Bits (Analytical)
No Consequtive Bits (Simulation)
-6
10
-8
10
0
5
10
15
SNR (dB)
Figure 110 : BER versus SNR for a pseudo-random sequence and a sequence of “0 1 0 1…”.
Note that the sequence of one and zero is an extreme case where there are no successive 0 or 1
bits and still the performance is not very far from the PRBS sequence. However this can cause
major problem both for AC or DC [114] coupled receivers.
5.6 ADC/DAC
The most advanced uses of Analogue-to-Digital Converters (ADC) today may be found in a
typical 100 Gb/s DP-QPSK coherent receiver implementations. In the receiver, the two
polarizations are split and fed into two balanced 90 degrees hybrid mixers, from which the
output is fed into four parallel 56 GS/s ADCs for decoding. Figure 111 shows the ADC
requirements for different 100G modulation formats and ADC performance that has been
demonstrated and/or published in scientific literature.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 140 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 111 : The figure shows a set of possible modulation formats for 100 Gb/s systems and their
requirements in terms of ADC performance (green circles) [128]. Also reported is the
performance of bipolar and CMOS ADCs from publications [129], [130].
SiGe has been the technology of choice for high speed circuit design. Regarding DAC, there
is however a trend in implementing them in CMOS in order to enable integration with digital
logic, such as DSP functionality, while keeping power consumption low. As an example
Fujitsu and Micram use different technology approaches. While Micram’s analogue front-end
is implemented in SiGe technology which is integrated together with a CMOS interface to
create a multi-chip System-in-Package (SiP), Fujitsu uses CMOS technology which allows for
a monolithic configuration (Figure 112). Table 21 summarizes state-of-the-art DAC available.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 141 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 112 : Multi-chip (upper figure) and monolithic (lower figure) configuration of ADC/DSP block in a
100 Gb/s receiver [131].
Country
Commercial
Research
Sampling rate
Resolution
Bandwidth
Interleaving
Technology
Power
consumption
(typ)
Table 21 : Table summarizing state-of-the-art DACs available.
Fujitsu
Maxtek [134]
Micram [135]
[132][133]
UK
USA
Germany
/ Commercial
Commercial
Commercial
Country
Commercial
/
Research
Sampling rate
Resolution
Bandwidth
Interleaving
Technology
Power
consumption
(typical)
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
55-65 GS/s
8-bit
15 GHz
CMOS 40 nm
1.5 W
Mobious
Semicond.
[137]
USA
Commercial
63 GS/s
12.5 GS/s
SP
Devices
[136]
Sweden
Commercial
30 GS/s
6-bit
20 GHz
Yes, 60 GS/s
SiGe
11.5 W
6.4 GS/s
8-bit
3 GHz
Yes, 12.8 GS/s
Ihp Gmbh [138]
U
Stuttgart
[139]
Germany
Mostly
Research
1-50 GS/s
3 - 12-bit
Germany
Research
National
Semicond.
[140]
USA
Commercial
50 GS/s
3 GS/s
8-bit
3 GHz
SiGe
CMOS 90nm
CMOS
1.6 W
6.25 GHz
Yes, 25-50 GS/s
SiGe
Page: 142 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
5.7 DISPERSION COMPENSATION
Dispersion in optical communications systems leads to inter-symbol interference (ISI) and
limits the transmission distance and/or data rate. In optical single-mode fibres different
dispersion effects occur: group velocity dispersion, also known as chromatic dispersion and
polarization mode dispersion (PMD), a highly dynamic effect. Compensation of chromatic
dispersion is usually required for optical transmission systems operating on bitrates of 10
Gb/s and above, while PMD-compensation is required for 40 Gb/s and beyond. Several
compensation methods exist, Figure 113 shows an overview. The inline dispersion
compensation by dispersion compensating fibres (DCF) along the fibre link was a very
popular approach during the last decade but because it is a static approach, it is now about to
be replaced by more dynamic solutions based on electronic signal processing [141] at the
transmitter or at the receiver.
EDFA
Tx
Rx
SSMF
•Electronic pre-distortion
•Pre-chirping
•PMD-scrambling
•Inline dispersion compensation (DCF)
•Dispersion shifted fibers
•Optical dispersion compensation
•Optical PMD compensation
•Electronic equalization
Figure 113 : Overview of dispersion compensation methods
5.7.1 Compensation at the transmitter
At the transmitter, the electronic pre-distortion (EPD) approach is an effective method to
avoid inline-DCF along the link. The real (I) and imaginary (Q) parts of the data signal are
generated using digital signal processing (DSP) using two FIR-filters (finite impulse
response). The I and the Q part of the optical modulator (e.g. a double Mach-ZehnderModulator) is then driven by the electronic pre-distorted I and Q signals, where the amount of
pre-distortion depends on the dispersion of the link and therefore the fibre length. The goal is
to pre-distort the signal so that the pre-distortion and the fibre dispersion compensate each
other at the target fibre length. Therefore, the signal can only be received at the receiver
placed at the target fibre length and in a small window around, depending on the modulation
format.
There are some other methods for transmitter-sided dispersion compensation. One is prechirping, where the chirp of the MZM is adjusted to increase the transmission length. Another
other is polarization-scrambling, where the input polarization into the fibre is permanently
scrambled which should increase the PMD tolerance of the system.
5.7.2 Compensation at the receiver
At the receiver many approaches for dispersion compensation exist which were intensively
studied during the last years. The simplest method may be to use a DCF at the receiver to
compensate for all link dispersion, but this approach is limited by nonlinear fibre effects and
is also only a static solution. Nowadays the system designers focus on more adaptive
solutions like electronic equalizers. Many types of equalizers were proposed during the last
years.
The first analog equalizers realized for dispersion compensation were FFE and DFE structures
(feed forward and decision feedback equalizer). The FFE typically consists of a tap delay-line
filter with 5 or more forward taps, while the DFE has one tap in the feedback path. Usually
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 143 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
these two equalizers are combined to increase the compensation performance. The FFE/DFE
uses a feedback criterion, e.g. the minimum square error (MSE) to minimize the ISI of the
direct detected received signal. Therefore it operates independently of the signal distortion
and can also be used for joint CD and PMD compensation. For NRZ-modulation the
FFE/DFE can increase the tolerance to chromatic dispersion by about 60%. Practical
implementations of FFE/DFE operating on 40 Gb/s were shown.
The MLSE equalizer (maximum likelihood sequence estimation) is a digital equalizer which
uses the Viterbi-algorithm to calculate the most likely transmitted bit sequence. Since the
processing is done on a bit sequence, the performance of the ISI equalization is getting better
with an increased bit sequence. For optical communication systems MLSE equalizer with 4states and 16-states were experimentally reported for 10 Gb/s [142], [143]. It also operates
independently of the signal distortions and the joint compensation of CD and PMD was
proven. The 4-states MLSE doubles the CD-tolerance and shows a PMD-tolerance of about
1Bit duration. However, the compensation performance is limited by the complexity of the
MLSE at higher Viterbi-states and the clock-recovery at large signal distortions.
The most sophisticated electronic equalization uses coherent reception and digital postprocessing the complex e-field of the signal. In the coherent receiver the optical field is
converted into the electrical domain without loss of information. The e-field of the signal can
then be equalized much better then after direct detection where the phase information is lost.
The equalization of CD is then again done by using FIR-filters. With 128 filter taps
42000ps/nm CD can be compensated [144]. If a polarization diversity receiver is used the
compensation of PMD is also possible.
Other approaches to compensate for dispersion are using multi-carrier modulation formats
such like OFDM (orthogonal frequency domain modulation). In these schemes the signal is
modulated onto multiple low bandwidth and low data rate sub-carriers which are transmitting
only a fraction of the total data. Because of the low data rate of each sub-carrier the
transmission is only slightly influenced by dispersion effects and can also be equalized much
easier.
5.8 PASSIVE WAVELENGTH SELECTIVE DEVICES
5.8.1 Thin film filter-based WDM components
As it was described in § 5.4.1, thin film filters (TFF) or interference filters are made by
depositing of thin alternating layers of two materials with different refractive index on a
transparent substrate and their functioning is based on the interference of multiple reflections.
The transmitted wavelength and filter shape depends on layers thicknesses, refractive indexes,
angle of incidence on the filter (normal in the fibre-based filter) and the numbers of layers in
the stack. Usually the more layers, the finer the resolution, and the narrower the range of
wavelengths selected. There are three types of filters used in WDM devices: line filter, band
filter, and cutoff filter, each with its own transmission characteristic.
Line and band filters either reflect or transmit light in a selected range of wavelengths. If the
range of wavelengths is narrow, we have to do with line filters, for example a filter to select
one 100-GHz optical channel. Filters that select a broader range of wavelengths are called
band filters, for example a filter that selects a 10-nm CWDM channel. Cutoff filters are
designed to make a sharp transition between transmitting band and reflecting band at a certain
wavelength. For example, a filter for separation of optical channels belonging to C-band and
L-band of erbium-fibre amplifier has a cutoff wavelength at 1567 nm.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 144 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Thin film filters can be used to combine and separate wavelengths in WDM systems by taking
light out of the fibre and passing it through a set of filters that sorts the light out by
wavelength. Typically a lens system collimates or focuses the light emerging from the input
fibre, which then passes through one or more filters and focus them into individual output
fibres. Several interference filters can be cascaded to pick off a series of four or eight
wavelengths in such a way that the first filter transmits channel 1 and reflects all other
channels. The remaining channels hit the second filter, which transmits channel 2 and reflects
the remaining channels and so on. In this way n – 1 filters are needed to separate n optical
channels.
Such a demultiplexer works fine for 4 – 8 channels, but the losses increase considerably for
16 channels or more. The solution in this case is to divide optical signals into groups of
channels and treat them with band pass filters before they are then split up individually. This
method does not reduce the total number of filters needed, but the number of filters for each
channel is reduced.
Interference filters are widely used for WDM due to the fact that the technology is well
developed. Interference filters have been known for many years, although the extremely
narrow-line filters used in DWDM systems were developed only recently. Filters can be made
very small, either deposited on the fibre facet or separately with a cross section of a few
millimeters. They have good performance, have modular construction and can be
upgradeable, but it is always need to have as many filters as the number of channels to be
separated. Although for add-drop multiplexers a single optical filter is needed to drop one
channel with the remaining channels reflected and collected for transmission through the rest
of the system. Examples of the devices on the market based on TFF are reported in Appendix
7.2.
5.8.2 Fibre Bragg grating-based WDM components
Fibre (or planar waveguide) Bragg gratings work similarly, but they selectively reflect a
narrow range of wavelengths, while interference filters selectively transmit a narrow range of
wavelengths. In the multi/demultiplexer each reflected channel of light must go through an
optical circulator to be separated from the input light. Using Bragg gratings as optical filters
and optical circulators one can build different architectures with spectral characteristics
depending on the used gratings spectral profiles with channel separations of 100GHz and
50GHz easy obtained and large number of channels, but using the same number of circulators
and at least the same number of gratings.
A single Bragg grating with circulators on both sides can act as an add-drop multiplexer.
Another possibility for this kind of device can be a Bragg grating assisted optical Add-Drop
Multiplexer based on 2x2 MMI coupler shown in Figure 114 [145] or Add-Drop Multiplexer
based on balanced Mach-Zehnder interferometer (MZI) with two identical Bragg gratings
imprinted in both arms of the interferometer shown in Figure 115 [146]. In both cases the
devices were fabricated in planar technology. In the second configuration two 3dB couplers
on both sides of the MZI redirect the reflected wavelength to the correct output, which makes
that circulators are not necessary in this architecture. Moreover, an additional MZI equipped
with a thermal switch and connected to the first one with help of two Bragg grating reflectors
(shown in Figure 115 as Reflector 1 and Reflector 2) changes it to a Switchable Add-Drop
Multiplexer, which is an important component for ring architectures.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 145 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 114 : Bragg grating assisted optical Add-Drop Multiplexer based on 2x2 MMI coupler. 30dB
extinction ratio between drop and transmitted channels and 3dB excess loss in the dropped
channel have been obtained.
Figure 115 : Add-Drop Multiplexer based on balanced Mach-Zehnder interferometer. Obtained crosstalk
-25dB, insertion loss -3dB, switching time 2ms and power consumption 0.5W have been
achieved.
In principle, fibre Bragg gratings (and circulators) can be arranged to obtain different
functionalities with good performance although the cost and complexity of the components
make that they are not commonly used.
The described until now WDM components based on thin film filters and Bragg gratings split
or combine optical signals one at a time. In the following we will describe more advanced
integrated structures that split or combine a large number of wavelength channels in parallel
due to a diffraction grating that spreads out a spectrum of incoming light in such a way that
different wavelengths propagate in different angles. Both, etched diffraction gratings (EDGs)
and arrayed waveguide gratings (AWGs) are usually designed as planar integrated
components and despite advanced architecture can be fabricated in wafer scale allowing for
mass production. Moreover, integration offers the advantages of compactness, reliability and
potential possibility to have light sources, detectors and control electronics on the same chip.
5.8.3 Etched Diffraction Grating-based WDM components
As it was explained above, the two main candidates for integrated multi/demultiplexer in
WDM applications are etched diffraction gratings (called also Echelle gratings) and arrayed
waveguide gratings. Both of them have been fabricated using different material platforms,
including silica-on-silicon, III-V semiconductors-, as well as silicon large core- and nanowire
-based technology.
The principle of operation of an EDG multi/demultiplexer called also echelle grating (EG) or
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 146 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
planar concave grating (PCG) is based on a spectrometer with the Rowland Circle geometry
(Rowland grating). The access waveguides are situated on a circle with a radius R (Rowland
circle). The reflective concave grating with radius 2R, situated such that it is touching the
circle, diffracts light coming from the input waveguide and focuses the reflected light at
different wavelengths back to the points on the Rowland circle, where the output waveguides
are situated (see Figure 116).
Figure 116 :
Etched diffraction grating demultiplexer: signal from input waveguide with wavelengths
λ1 ,λ2 ,λ3,… is diffracted by the planar concave grating, and refocused into different output
waveguides.
The close to Littrow configuration, where the grating facets are almost perpendicular to the
incoming light and reflect light back, almost in this same direction guarantees high efficiency
for a chosen diffraction order.
The first EDG-type demultiplexer was proposed 1979 by G.L. Tangonan and co-workers from
Hughes Res. Labs, where the bulk grating was attached to an edge of circular profiled slab
situated between two glass slides. The first monolithic device in application to DWDM
systems with etched grating in SiO2 glass waveguide on Si was demonstrated in 1992 by P.C.
Clemens et al. from Siemens AG. The main drawback of this construction at that time were
relatively high insertion losses caused mainly by non-perfect verticality of deeply etched
grating facets in a thick slab. Today, using Silicon-on-Insulator material platform the grating
facets verticality is no longer the main constrain for EDGs, where shallow etching depth of
220-250nm is sufficient [147].
There are few companies like Enablance or Kotura making EDG-based PLC’s for commercial
applications. Very recently (July 2010), Intel also announced 4-channel CWDM optical link
using silicon Echelle gratings as MUX/DEMUX.
5.8.4 Arrayed Waveguide Grating-based WDM components
An alternative solution was proposed by M. Smit from Delft Univ. in 1988, where the grating
was replaced with a focusing and dispersive planar component based on phased array (called
PHASAR) known today as arrayed waveguide grating.
An arrayed waveguide grating is composed of input and output waveguides coupled to an
array of waveguides (regularly arranged with increasing lengths) through two focusing slab
waveguides, free propagation regions (FPRs) – where light is not confined laterally and
therefore diverges or converges in the lateral direction, as shown in Figure 117a.
An optical beam entering the 1st free propagation region (input FPR), through the input
waveguide, spreads by diffraction and forms a circular wave front (equal phase) at the end of
FPR. Due to the constant path length difference between adjacent waveguides, after passing
through the array the signals that reach the output slab waveguide have different phase delay,
and due to the mutual interference between the signals from each waveguide, all the wave
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 147 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
fronts are diffracted and for each specific wavelength are focused towards a uniform direction
(Figure 117b). The correct positioning of the output waveguides allows the spatial separation
of the different wavelengths. In this way it demultiplex the input signals from one of the input
channels into different output channels according to the signal wavelength.
Figure 117 : Schematic of an arrayed waveguide grating: (a) input-output waveguides are coupled
through two slab (FPR) waveguides and an array of curved waveguides, (b) tilted phase front
with an angle , focused at an output waveguide in the output FPR.
a)
b)
Figure 118 : Series of 32 x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm channel spacing (100
GHz), 25 nm band: a) Fabricated devices, b) AWG transmission spectrum.
Although it is not straightforward understood, the principle of operation of these two devices
(EDG and AWG) is very similar. In the AWG at the end of the array of waveguides we see the
same diffractive and focusing effect as in the case of concave diffraction grating. Additionally
the front and back profiles of the free propagation regions in the AWG follow Rowland circle
geometry as in the case of EDG.
Silica-based AWG soon became a main multichannel multiplexing device on the market due
to more simple and tolerant technology in comparison to EDG. Fig. 5 shows a series of of 32
x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm channel spacing and their
transmission spectrum [148]. Even devices with high level of integration with cascaded
AWGs and hundreds of channels are possible to fabricate with low loss and crosstalk.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 148 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
5.8.5 Technology constrains for EDG and AWG WDM components
As it was mentioned earlier, when using Silicon-on-Insulator material platform the grating
facets verticality is no longer the main constrain for EDGs, where shallow etching depth of
220-250nm is sufficient and so AWG superiority is no longer valid. In general EDGs offer
instead smaller device size, especially for a large number of channels. When choosing the
type of multiplexer one needs also to consider the parameters that limit the device
performances that are specific to SOI technology.
The devices are usually patterned directly by e-beam lithography or deep UV lithography
using photomasks prepared with e-beam litho. In both cases we have pixelation errors due to
finite step of the e-beam machine (10-50 nm) as well as stitching errors due to limited writing
field (80-120 µm) and usually necessity to put together many writing fields to pattern a whole
component. For EDG case the positions of grating facets are discretized by pixilation and
distorted by stitching. The same errors in AWGs limit the accuracy of positioning of the
beginning and the end of the waveguides in the array causing unexpected phase changes. Both
two errors decrease the efficiency and increase crosstalk.
In AWGs the phase errors can be additionally increased by changes in the effective path
length of the waveguides in the array caused by non-uniform waveguide cross section, local
strain, material composition, temperature gradients and other non-uniformities. In EDGs on
the other hand quality of etching not only decreases the device efficiency, but also affects
scattering from grating facets, increasing crosstalk and increasing overall insertion loss in
comparison to AWGs.
Both, arrayed waveguide grating- and etched diffractive grating-based wavelength selective
devices are of interest for future highly integrated optical communication systems and
computer interconnects. For applications, where a large number of channels should be treated
and compactness is important, EDG-based components should be used, whereas for more
relaxed applications the mature, high quality and low loss AWG devices are preferable.
For example a 256-channel, silica-on silicon-based EDG with 25GHz channel spacing has an
overall size of 20x40 mm [149]. The device has adjacent channel crosstalk of 30 dB, insertion
loss of 10 dB, and a polarization-dependent wavelength shift less than 10 pm. The parameters
are similar to those of AWG devices, but an AWG fabricated on the same material platform
with 256 channels takes up about five times the area of the EDG device. Examples of the
main devices on the market based on AWG produced by Oplink and JDSU are reported in
Appendix.
There are several other actors on the market offering WDM component although in some
cases it is not clear whether they are offering real products or only design of future products.
The companies include GEMFIRE, FITEL, NEL, Neo Photonics, NEC FiberOptech, Hitachi
Cable, SANTEC, IGNIS, Enablance, Kotura, ANDevices, Fiberdyne Labs, Lightwaves2020,
Go!Foton, Optelian, LightGAIN, GIGALIGHT and probably a few others. The number of
companies and their names are changing constantly.
5.9 REACH EXTENDERS
Reach extenders are an option to increase the optical link budget by using mid-span signal
amplification. By using reach extenders, the network planning for an operator is made more
flexible. Basic reach extension can be achieved by either Optical-Electrical-Optical (OEO)
conversion or optical amplification (OA). The latter can be achieved by using either SOA or
fibre amplifiers in the optical distribution network. Basic reach extension technologies are
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 149 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
summarized below.
Optical-Electrical-Optical (OEO) RE
OEO offers the most straight-forward RE solution. It builds on re-use of existing transceiver
technology and offers flexibility and different implementation options. Figure 119 shows an
example implementation using a reset-less OLT transceiver. Additional required building
blocks are the embedded ONT (EONT), optical splitter for EONT, burst-mode clock/data
recovery for upstream data, logic for reset generation, electrical switch for EONT and
protection path and optical switch for protection or 1:2 splitter.
Figure 119 : Example GPON OEO
Measurements on an OEO for GPON show that it is possible to cover the complete logical
reach, with a differential drop reach of 0-24 km and a trunk reach of 0-60 km. As a
consequence the OEO offers simple network planning/deployment. OEO can be used for most
system concepts. Power consumption, cost and reach are related to TRx blocks of each system
concept.
SOA
The use of semiconductor optical amplifier (SOA) for reach extension for up to 50 km reach
with a 1:32 split has been demonstrated for GPON. The configuration allows for a maximum
reach of 37 km on the trunk side and 22 km on the drop side. SOA can be used for multichannel amplification offering co-existence with XGPON1 US (1270 + 1310 nm). However,
analogue amplification makes planning/deployment more difficult than with OEO. As SOA is
currently not a volume product, cost for the technology is high.
Table 22 :
Example of supported GPON configurations
Drop reach (km)
Trunk reach (km)
Total reach (km)
1:32 split
15
37
52
1:64 split
10
34
44
1:128 split
5
28
33
Fibre amplifiers: Raman / EDFA
Most long-reach options require active reach extenders. Raman could potentially offer reach
extension without active equipment in the optical distribution network (ODN). Raman
amplification in PON is however not straightforward. Pump lasers for the upstream would be
in the cut-off region (i.e. < 1200 nm). GPON with XG-PON overlay would result in a
complex wavelength plan with many pumps and signal wavelengths. In addition, Raman
pumps require quite high power (~300 mW range), which stands in contrast to one of the
motivations for PON, i.e. reduced power consumption.
EDFA and similar doped fibre amplifiers could potentially be used for GPON and the XGPON downstream. Remote pump architectures have been shown. However, the upstream
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 150 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
requires complex control to handle burst transmission.
Remote protocol terminator
In addition to basic reach extension one should also consider the possibility of increasing the
complexity of the active RE for performing basic tasks such as protocol termination. The
remote protocol termination (RPT) approach for PON consists of extending the backplane of a
PON OLT (Figure 120). The actual PON port can be separated from the OLT element. As a
consequence, the system is no longer limited by the PON protocol. RPT enables traffic
aggregation unlike the alternative L1 REs (OEO, OA, OTN). As a result conventional
D/CWDM can be used for the uplink which saves fibres if needed. The difference between the
RPT and a “mini-OLT” lies in the reduced management complexity since the RPT is managed
as an OLT line card. There is no difference in the data plane. The network between the RPT
and the OLT can be any packet network and the OLT can be an aggregation switch (AGS).
Figure 120 : RPT approach
Comparison between different reach extender technologies
Figure 121 provides a comparison of different reach extender technologies when used in a
10G TDM-PON system. The total reach is divided into trunk reach (distance between OLT
and reach extender) and access reach (distance between reach extender and ONTs). The
orange shaded areas show achievable distances with 1:32, 1:64 and 1:128 split for a system
without reach extender (Nominal 1 class, 29 dB link budget). Light green areas show
achievable distances with an OA based reach extender (L-band EDFA for the 1577 nm
downstream + SOA for the 1270 nm upstream). Performance is limited by the upstream SOA.
Light blue areas show achievable distances with an O/E/O based reach extender. In dark green
it is shown that the RPT / mini-OLT approach allows for possible extension of the trunk fibre
up to 80 km.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 151 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 121 : Comparison of different reach extender technologies for 10G TDM-PON (XG-PON): OA
based reach extender (L-band EDFA for the 1577 nm downstream + SOA for the 1270 nm
upstream), O/E/O based reach extender and Remote Protocol Terminator (RPT)/ mini-OLT
approach.
With the O/E/O approach is possible to cover longer distances than for the OA approach. For
a 1:64 split the difference is of the order 20 km. When several PONs are routed through the
same remote node, it could make sense to use RPT instead of O/E/O. The benefit would be
traffic aggregation and the possibility use WDM (i.e. CWDM) on the trunk lines.
RE configuration
With potential requirements on resilience there are different RE configurations the offer
different trade-off between cost and resilience. Several potential RE options are displayed in
Figure 122. Regarding the dual RE solution, development of low-power modes for the
standby RE would be of interest. A promising solution is the one with a single RE with dual
trunk. With this architecture it is possible to power off the protection path. For both the dualRE and the single-RE dual-trunk solutions a switch-over procedure needs to be defined. The
dual trunk in the single-RE solution could be implemented either by a 1:2 splitter or dual
uplink ports as illustrated in Figure 3. The exact configuration depends on the relative cost of
the ONT TRx and the OLT TRx for the specific system concept. The implication of using a
1:2 splitter on the optical power budget should also be considered.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 152 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 122 :
RE variants
5.10 SWITCHING COMPONENTS
Ethernet is riding its success from LAN to access and metro markets with service providers
migrating from circuit switched networks to Carrier Ethernet. A number of important reasons
made Ethernet a winner, including low cost, simplicity, flexibility, and scalability. The most
important factor was technological unification, because this guaranteed smooth interworking
without the need for specialized gateways.
Gigabit Ethernet (GbE) is commonly supported in most computer network cards, while in the
access bandwidth is typically not larger than 100Mb/s. However, the evolution of applications
and services such as triple play, HDTV, 3DTV requiring higher bandwidth to the subscriber
drive GbE to be more and more used in the access as well as the home. GbE has been used in
aggregation and backbone networks for many years and its use over either AON or PON has
become a leading architecture for NGOA systems.
The Ethernet switch unit is compulsory for both AON and PON architectures. The foreseeable
major evolution of Ethernet may include: from Fast Ethernet (FE) to GbE by gradual
conversion of low-cost CPEs, IADs, SOHO (small office and home office) switches, and
access switches; deployment of 10G Ethernet (10GbE) in aggregation / distribution network;
and 40GbE or 100GbE in the core network. The transitions are creating new requirement and
research questions related to switching components.
5.10.1 Power consumption
Recently market pressure and legislative action worldwide is demanding improvements in
energy efficiency of networked systems. Reducing power consumption not only provides
good ecological credentials for society, but also improves financial figures for operators since
energy costs are a major component of operating cost. Unfortunately there is a clear
relationship between faster switching speeds and the amount of power consumed. Hence
switches are committed to new designs with lower power requirements. There are known
avenues for lowering power, such as:
-
system level energy management techniques
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 153 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
-
switch fabric design(hardware and software) to improving switch
-
thermal control and fanless cooling systems
-
more efficient power supplies
-
sleeping mode on optical transceivers
-
network level energy management techniques
-
network wide coordinated link and node sleep
-
bandwidth renegotiation
High integration and low power dissipation IAD (Integrated access device) switches or CPEs
are important as well. The largest portion of the per user power consumption is at the CPE.
Several vendors already have “green Ethernet” products on the market [150] where low
energy consumption mode has automatic entry/exit based on actual network traffic and
Ethernet ports that can detect network cable length and unused ports to manage power
consumption.
5.10.2 Challenges on high speed switching process:
10 Gb/s and ultra-high-speed transport Ethernet in aggregation/distribution network requires
high performance switching hardware and software support. The increased interface data rate
leads to a higher threshold for packet processing, storage, system switching, and backplane
technology. Several examples of challenges are listed below [151]:
-
Content-Addressable Memory (CAM) of the network processor. In order to reduce
switching latency, the bandwidth of search interface should be increased.
-
Data bus: There is a bottleneck for the data bus width and rate.
-
Multiple packet processing chipsets solution caused by bus interface conversion, the
board area and power consumption are unacceptable. Single chip processing capability
is limited; the solution based on FPGA-customization still need comprehensive
technologies.
-
Switch fabric chipset: besides the non-blocking full duplex bandwidth capacity, switch
fabric must actually provide additional bandwidth to accommodate cell overhead,
buffering, and congestion-avoidance mechanisms. For supporting 100 Gb/s interface,
the bandwidth of each line card should be upgraded to 200-500 Gb/s bandwidth. The
requirements for backplane design, technique, material, and meeting the bus length are
more critical than before. For the system meeting the carrier-class requirements, the
Virtual Output Queue (VOQ) and Hierarchy Quality of Service (HQoS) should be met.
It requires greater processing bandwidth, queue processing capability, and buffer,
which greatly increases the difficulty of system design.
5.10.3 Network stability
In current network systems all devices are managed by the control and management (C&M)
system for automated monitoring, troubleshooting of network faults, QoS management etc.
Normally the C&M function is implemented in individual devices (e.g. line card) apart from
the switching unit. However it is necessary for switches to have hardware which commits the
connectivity fault monitoring and the assignments from the C&M system.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 154 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
6. Summary
Based on the survey presented in this deliverable a comparison table of potential NGOA
system variants is presented in Table 23. The table covers a coarse comparison of the system
aspects with respect to the main key system aspects defined in Chapter 3 and can be seen as a
summary of the findings in the report.
The focus of Table 23 is to present systems that are more or less realizable today. Cost and
power consumption is given to indicate the order of magnitude today, without consideration
of future potential for the different variants. A more detail assessment is to be provided in
D4.2 and D4.3 [1] which will provide a consolidated view on reasonable technology
evolution as well as cost and power consumption potential for the main candidates in the 2020
time horizon.
The requirements in D2.1 provide the basis for the coarse assessment in Table 23. At current
state of technology, all contenders have problems of fulfilling all requirements
simultaneously, which makes it difficult to compare the systems on equal terms. The system
configuration may be tailored depending on which requirements are prioritized (e.g. reach or
splitting ratio). For this reason, for some systems, ranges have been specified and hence some
care must to be taken when drawing conclusions based on this assessment. Several entries of
Table 23 are listed only as rough estimates.
TDM-PON approaches present significant challenges with respect to simultaneously
increasing bandwidth, reach and splitting ratio, whilst maintaining low cost and power
consumption. The configurations presented in Table 23 comply with reach requirements but
do not meet the requirements on splitting ratio and in some cases bandwidth. Advantages of
WDM-PON include long reach and large per customer sustainable bandwidth. A drawback is
the limited fan-out of traditional WDM-PON approaches, limiting the number of customers
per feeder fibre. OFDM-PON presents similar challenges as TDM-PON concerning limited
total capacity and limited splitting ratio. The particular configuration presented in Table 23
allows for 64 users per feeder fibre but does not fulfill the requirements on sustainable
bandwidth. The high cost and power consumption associated with OFDM and UDWDM in
Table 23 is associated with the immature state of technology. The preliminary assessment of
future cost potential of UDWDM (chapter 4.5.5) shows that it still is a potential candidate for
NGOA despite large cost today. For OFDM the most interesting variant, in terms of meeting
OASE requirements, is the hybrid WDM/OFDM-PON solution which showed slightly worse
cost potential than the other hybrid alternatives (chapter 4.5.3). For OCDM-PON the scaling
of the number of users per feeder fibre presents a serious problem which eliminates it as a
serious contender for NGOA. Hybrid concepts are motivated by the fact that each of the
aforementioned pure system concepts individually may have difficulties in fulfilling the
complete set of NGOA requirements. Hence, hybrid concepts have been proposed that
combine advantages of different concepts. Typically the advantages that are exploited are the
increased overall capacity of WDM and the efficient resource sharing of TDM, OFDM or
CDM. Concepts that involve different types of active remote nodes have also been
considered. A preliminary cost and power consumption analysis of different variants (chapter
4.5 and 4.6) shows that the most promising configurations are hybrid WDM/TDM-PON as
well as various active hybrid variants (WDM-PON with AON access and two stage WDMPON). These concepts have been included in Table 23. The considered WDM/TDM PON
configurations prioritize high splitting ratio at the cost of not fulfilling the sustainable
bandwidth requirement. Concerning AON the main drawback with respect to the OASE
requirements concerns either large port count at the central office or low degree of node
consolidation, depending on AON architecture. Both these factors have implications on total
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 155 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
cost and power consumption.
As an outcome of the survey it is seen that the main candidates for further consideration
within the OASE project based on the posed requirements are different variants of pure
WDM-PON, hybrid WDM/TDM-PON, AON as well as various active hybrid variants based
on WDM-PON and AON.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 156 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
0
1
10
4 PON
ports per
HU in
rack
“Quarter
Pizza
Box”
Start of
deployment
<1
2.5
15
2 PON
ports per
HU in
rack
“Quarter
Pizza
Box”
First
demonstrator
2
8
4-16 slots
of 5HU
25% 100% of
9HU
Shelf
“Quarter
Pizza
Box”
First
demonstrator
3
8
4-16 slots
of 5HU
25% 100% of
9HU
Shelf
“Quarter
Pizza
Box”
First
demonstrator
2
0.04
1.5
32
20
28/30/32
3R
60
50+
20
120
0.5
XG-PON
0.31
10
0.31
10
32
20
29/31/33
/35
3R
60
50+
500
1000
40G serial
TDM-PON
1.25
40
1.25
40
32
20
27
3R
60
50+
5k
2k
WDM-PON
with lambda
reuse
1-10
1
1-10
1
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
1-10
1
1-10
1
40-400
96
15-50
40
5-17
for Fibre
12
-
60-100+
-
22-34
for Fibre
-
~700
700
Page: 157 of 168
~700
350
Time to market [years]
Commercially
available
2.5
4
4
Maturity
Footprint ONU
“Quarter
Pizza
Box”
0.78
EDFA,
Raman
SOA,
TDF
Power consumption today
ONU [W]
Power consumption today
OLT per sub [W]
Cost Today ONU [$]
Cost Today OLT per sub
[$]
Power budget with reach
extension [dB]
Reach with reach
extension [km]
Reach extension –
technology
Power budget without
reach extension [dB]
Reach without reach
extension [km]
Typical No of subscribers
per Feeder fibre
Upstream bandwidth per
sub - peak [Gb/s]
Upstream bandwidth per
sub - sustainable [Gb/s]
Downstream bandwidth
per sub - peak [Gb/s]
8
8 PON
ports per
HU in
rack
GPON
State of the Art
WDM-PON
with tunable
laser
WDM-PON
System concept comparison table
Footprint OLT
TDM-PON
System
concept
Downstream bandwidth
per sub - sustainable
[Gb/s]
Table 23.
Hybrid WDM/TDM
CDM-PON
OFDM-PON
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Serial 40G
OFDM-PON
0.63
40
0.63
20
-
-
-
-
10k
50k
5
20
4-16 slots
of 5HU
25% 100% of
9HU
Shelf
“Quarter
Pizza
Box”
First
demonstrator
>5
“Quarter
Pizza
Box”
Experiment in
lab
>5
4+
64
59
-
-
-
-
700?
700?
-
-
4-16 slots
of 5HU
25% 100% of
9HU
Shelf
OCDM-PON
10
10
10
10
513
(theory)
12
(demons
trated)
Hybrid
WDM/TDM
broadcast &
select
0.31
10
0.31
10
1024
NA/NA/
NA
31/34/37
EDFA,
Raman,
SOA
46/36/26
11/14/17
290
2500
1
10
Half Rack
“Quarter
Pizza
Box”
First
demonstrator
2
Hybrid
WDM/TDM
with lambda
split
0.31
10
0.31
10
1024
20/10/N
A
22/25/28
EDFA,
Raman,
SOA
80/70/60
2/5/8
300
2500
1
10
Half Rack
“Quarter
Pizza
Box”
First
demonstrator
3
Hybrid
WDM/TDM
with lambda
switch
0.31
10
0.31
10
1024
NA/NA/
NA
32/35/38
EDFA,
Raman,
SOA
40/30/20
12/15/18
325
2500
1
10
Half Rack
“Quarter
Pizza
Box”
First
demonstrator
>3
Hybrid
WDM/TDM
with WSS
0.31
10
0.31
10
1024
20/10/N
A
22/25/28
EDFA,
Raman,
SOA
80/70/60
2/5/8
350
2500
1
10
Half Rack
“Quarter
Pizza
Box”
Paper study +
simulation
>5
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 158 of 168
UDWDM
1
1
1
1
5121024
30-100
-
OA
100+
-
10k
10k
10
12
Half Rack
“Quarter
Pizza
Box”
First
demonstrator
>4
AON
GbE access
1
1
1
1
-
70
-
Higher
quality
optics
120
-
15
scalable
1
5
32 ports
per HU
“Quarter
Pizza
Box”
Deployed
0
15+15
EDFA
active
sites
required
8
Main
OLT: Half
Rack
In-Field
OLT: 1
Shelf
“Quarter
Pizza
Box”
Can be
demonstrated
today
3
15+10
EDFA
active
sites
required
8
Main
OLT: Half
Rack
In-Field
AON:
2HU each
“Quarter
Pizza
Box”
Can be
demonstrated
today
1
WDMPON-AON
PON-in-PON
AON
DWDM
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
PON-in-PON
WDM-PON
AON
0.1-1
0.1-1
1
1
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
0.1-1
0.1-1
1
1
1k-10k
23049216
100
60
150+
85+
25+25
25+10
750
200
Page: 159 of 168
800
200
5
5
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
7. Appendix
7.1 COST AND ENERGY CONSUMPTION DATA
This section contains data and assumptions used for the comparison of cost and energy
consumption of different system concepts. System specific components are listed in Table 24.
Table 24.
Energy consumption and cost figures of relevant hybrid PON components
Component
10G base TRX (SFF, coolerless, w/o Locker, 22 dB, not used hereinafter)
10G TXFP (TEC, Locker, 25 dB)
10G Burst-mode TRX, 35 dB (SFF, APD, SOA, FEC, coolerless, w/o Locker)
30GHz TRX (coherent, TEC, Locker, 16 Channels à 1G / 3 GHz)
30GHz TRX (32 dB, coolerless, w/o Locker, single channel)
10G REAM-SOA, incl. Fraction of MFL, 1G 26 dB Rx (!)
10G REAM-SOA, incl. Fraction of MFL, 10G 35 dB Rx (!)
Energy Con.
1.25 W
3.5 W
2.5 W
8W
2.5 W
1W
2W
Cost
100$
1200$
175$
1600$
175$
85$
175$
1G coherent ONU TRX, pol.-diverse or w/ Pol. Scrambler
2W
175$
1G tunable ONU TRX
1W
75$
40x1G Laser/Rx Array
20 W
2000$
40x1G REAM/Rx Array plus MFL and Circulators
20 W
2400$
1G grey SFP, 10 dB
0.5 W
15$
ASIC 1G SCMA ONU
1W
10$
ASIC 10G OFDM ONU
4W
40$
ASIC 10G OFDM / SCMA OLT 16Ch
8W
160$
ASIC 50G UDWDMA OLT 16Ch
16 W
320$
ASIC CDMA OLT 8Ch
4W
120$
OLT EDFA Booster/Preamp Combo
25 W
2000$
Circulator
--
100$
AWG Port / Power Splitter/Combiner Port
--
20$ / 10$
OLT / PoP Switch per 1G
1W
5$
Baseline cost per client (CPE, OLT shelf, motherboards)
5W
100$
For the cost and energy-consumption calculations, common (constant) contributions are not
listed in the following tables. These contributions come from ONU housing (PSU), ONU
higher-layer equipment (i.e., it is assumed that all ONUs can perform the same higher-layer
and subscriber management functionality), and OLT shelf / management (i.e., it is assumed
that to first approximation, all variants require similar OLT cost and energy overhead on a
per-client basis). Hence, the baseline per-client cost and energy consumption as per Table 24
(last row) must always be added to get the respective total cost and energy consumption. The
individually differing numbers are also stated on a per-client, end-to-end basis. This does
include the per-client portions of the RN equipment (filter, splitter ports), which often are not
considered in similar calculations for EPON/GPON.
Further, assumptions regarding the ODN must be made. This applies to the different filters
and splitters which are required, as well as to the fibre per-kilometre insertion loss. Relevant
numbers are summarized in Table 25.
Table 25 :
Relevant optical parameters for hybrid PON performance analysis
1:8 TFF / 1:40 AWG / 1:80 AWG
1:8 / 1:16 / 1:1024 Power Splitter/Combiner
50 km Fiber EOL incl. Patches etc.
3 / 4 / 5 dB
12 / 14 / 35 dB
16 dB
Also note that we have chosen 16 dB per 50 km fibre as insertion loss. This is a trade-off
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 160 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
between minimum insertion loss (which could fall into the 10…12 dB range) and loss of very
poor urban fibres (which would go into the 25 dB range).
7.2 EXAMPLES DEVICES ON THE MARKET BASED ON TFF
Examples of devices on the market based on thin film filters (TFF) are shown below:
Figure 123 : Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 161 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 124 : Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels, low loss series.
Figure 125 : Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 100GHz
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 162 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 126 : Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 200GHz
Figure 127 : Oplink’s FTTX triplexer 1310/1490/1550 WDM (1x2)
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 163 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 128 : JDSU Coarse Wavelength Division Multiplexer 4/8 channels
Figure 129 : JDSU Dense Wavelength Division Multiplexer 4/8 channels
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 164 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
7.3 EXAMPLES DEVICES ON THE MARKET BASED ON AWG
Examples of devices based on AWG are shown below:
Figure 130 : Oplink’s Dense Wavelength Division Multiplexer 100 GHz.
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 165 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 131 : Oplink’s Dense Wavelength Division Multiplexer 50 GHz.
Figure 132 : JDSU Dense Wavelength Division Multiplexer 100 GHz, Narrowband (Gaussian).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 166 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 133 : JDSU Dense Wavelength Division Multiplexer 100 GHz, Wideband (Flat Top).
Figure 134 : JDSU Dense Wavelength Division Multiplexer 50 GHz, Wideband (Flat Top).
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 167 of 168
Survey of Next-Generation Optical Access System Concepts
OASE_D4.1_WP4_EAB_210612_v3.0
Figure 135 : JDSU Dual Duplexer 1310/1550.
---End of Document ---
O
O
A
S
OA
AS
SEEE
FP7 – ICT– GA 249025
Page: 168 of 168
Download