BROADBAND ACCESS
BROADBAND ACCESS
WIRELINE AND WIRELESS –
ALTERNATIVES FOR
INTERNET SERVICES
Steve Gorshe
PMC-Sierra, Inc., USA
Arvind R. Raghavan
Blue Clover Devices, USA
Thomas Starr
Stefano Galli
ASSIA Inc., USA
This edition fi rst published 2014
© 2014 John Wiley & Sons, Ltd
Registered of fice
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom
For details of our global editorial of fices, for customer services and for informa tion about how to apply for permiss ion to
reuse the copyright materi al in this book please see our website at www.wile y.com.
The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright,
Designs and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form
or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright,
Designs and Patents Act 1988, without the prior permission of the publisher.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in
electronic books.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and
product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective
owners. The publisher is not associated with any product or vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this
book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book
and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. It is sold on the
understanding that the publisher is not engaged in rendering professional services and neither the publisher nor the author
shall be liable for damages arising herefrom. If professional advice or other expert assistance is required, the services of a
competent professional should be sought.
Library of Congress Cataloging-in-Publication Data applied for
ISBN: 9780470741801
Set in 9/11pt TimesLTStd-Roman by Thomson Digital, Noida, India
1 2014
To my wife Bonnie Gorshe, and sons Alex and Ian Gorshe; S.D.G.
Steve Gorshe
To the Lord, in the spirit of Karma Yoga.
Arvind Raghavan
To my wife, Marilynn Starr.
Thomas Starr
To Tobey and Hannah.
Stefano Galli
Contents
About the Authors xv
Acknowledgments xvii
List of Abbreviations and Acronyms xix
1 Introduction to Broadband Access Networks and Technologies 1
1.1 Introduction 1
1.2 A Brief History of the Access Network 2
1.3 Digital Subscriber Lines (DSL) 3
1.3.1 DSL Technologies and Their Evolution 3
1.3.2 DSL System Technologies 5
1.4 Hybrid Fiber-Coaxial Cable (HFC) 5
1.5 Power Line Communications (PLC) 6
1.6 Fiber in the Loop (FITL) 7
1.7 Wireless Broadband Access 10
1.8 Direct Point-to-Point Connections 12
Appendix 1.A: Voiceband Modems 12
2 Introduction to Fiber Optic Broadband Access Networks and Technologies 15
2.1 Introduction 15
2.2 A Brief History of Fiber in the Loop (FITL) 16
2.3 Introduction to PON Systems 18
2.3.1 PON System Overview 18
2.3.2 PON Protocol Evolution 19
2.4 FITL Technology Considerations 21
2.4.1 Optical Components 21
2.4.2 Powering the Loop 22
2.4.3 System Power Savings 23
2.4.4 PON Reach Extension 25
2.5 Introduction to PON Network Protection 30
2.5.1 Background on Network Protection 31
2.5.2 PON Facility Protection 31
2.5.3 OLT Function Protection 35
2.5.4 ONU Protection 40
2.5.5 Conclusions Regarding Protection 42
viii Contents
2.6 Conclusions 42
Appendix 2.A: Subscriber Power Considerations 43
References 43
Further Reading 43
3 IEEE Passive Optical Networks 45
3.1 Introduction 45
3.2 IEEE 802.3ah Ethernet-based PON (EPON) 45
3.2.1 EPON Physical Layer 46
3.2.2 Signal Formats 46
3.2.3 MAC Protocol 48
3.2.4 Encryption and Security 49
3.2.5 Forward Error Correction (FEC) 50
3.2.6 ONU Discovery and Activation 51
3.2.7 ONU Ranging Mechanism 52
3.2.8 EPON OAM 52
3.2.9 Dynamic Bandwidth Assignment (DBA) 53
3.3 IEEE 802.3av 10Gbit/s Ethernet-based PON (10G EPON) 54
3.3.1 10G EPON Physical Layer 54
3.3.2 Signal Format 58
3.3.3 MAC Protocol 59
3.3.4 Forward Error Correction 59
3.3.5 ONU Discovery and Activation 61
3.3.6 ONU Ranging Mechanism 61
3.3.7 10G EPON OAM 61
3.3.8 Dynamic Bandwidth Allocation 61
3.4 Summary Comparison of EPON and 10G EPON 61
3.5 Transport of Timing and Synchronization over EPON and 10G EPON 61
3.6 Overview of the IEEE 1904.1 Service Interoperability in Ethernet Passive Optical
Networks (SIEPON) 63
3.6.1 SIEPON MAC Functional Blocks 65
3.6.2 VLAN Support 67
3.6.3 Multicast Service 67
3.6.4 SIEPON Service Management 67
3.6.5 Performance Monitoring and Verification 69
3.6.6 SIEPON Service Availability 70
3.6.7 SIEPON Optical Link Protection 70
3.6.8 SIEPON Power Savings 70
3.6.9 SIEPON Security Mechanisms 71
3.6.10 SIEPON Management 71
3.7 ITU-T G.9801 Ethernet Passive Optical Networks using OMCI 71
3.8 Conclusions 71
Appendix 3.A: 64B/66B Line Code 72
References 75
Further Readings 75
4 ITU-T/FSAN PON Protocols 77
4.1 Introduction 77
4.2 ITU-T G.983 Series B-PON (Broadband PON) 78
Contents ix
4.3 ITU-T G.984 Series G-PON (Gigabit-capable PON) 79
4.3.1 G-PON Physical Layer 79
4.3.2 G-PON Frame Formats 81
4.3.3 G-PON Encapsulation Method (GEM) 87
4.3.4 G-PON Multiplexing 91
4.3.5 Encryption and Security 92
4.3.6 Forward Error Correction 92
4.3.7 Protection Switching 94
4.3.8 ONU Activation 94
4.3.9 Ranging Mechanism 95
4.3.10 Dynamic Bandwidth Assignment (DBA) 96
4.3.11 OAM Communication 97
4.3.12 Time of Day Distribution 97
4.3.13 G-PON Enhancements 101
4.4 Next Generation PON (NG-PON) 101
4.4.1 Introduction to G.987 series XG-PON (NG-PON1 – 10Gbit-capable PON) 102
4.4.2 XG-PON Physical Layer 102
4.4.3 XG-PON Transmission Convergence Layer and Frame Structures 105
4.4.4 Forward Error Correction 108
4.4.5 XG-PON Encapsulation Method (XGEM) 109
4.4.6 XG-PON Management 110
4.4.7 XG-PON Security 110
4.4.8 NG-PON2 40 Gbit/s Capable PON 110
Appendix 4.A: Summary Comparison of EPON and G-PON 112
References 113
Further Readings 114
5 Optical Domain PON Technologies 115
5.1 Introduction 115
5.2 WDMA (Wavelength Division Multiple Access) PON 115
5.2.1 Overview 115
5.2.2 Technologies 116
5.2.3 Applications 120
5.3 CDMA PON 120
5.4 Point-to-Point Ethernet 122
5.5 Subcarrier Multiplexing and OFDM 123
5.5.1 Introduction 123
5.5.2 OFDMA PON 123
5.6 Conclusions 125
References 126
Further Readings 126
6 Hybrid Fiber Access Technologies 127
6.1 Introduction and Background 127
6.2 Evolution of DOCSIS (Data-Over-Cable Service Interface Specification) to Passive
Optical Networks 127
6.2.1 Introduction and Background 127
6.2.2 DOCSIS Provisioning of EPON (DPoE) 128
6.2.3 Conclusions for DPoE 135
x Contents
6.3 Radio and Radio Frequency Signals over Fiber 135
6.3.1 Radio over Fiber (RoF) 136
6.3.2 Baseband Digital Radio Fiber Interfaces 136
6.3.3 Radio Frequency over Glass (RFoG) 138
6.4 IEEE 802.3bn Ethernet Protocol over Coaxial Cable (EPoC) 140
6.5 Conclusions 140
References 141
Further Readings 141
7 DSL Technology – Broadband via Telephone Lines 143
7.1 Introduction to DSL 143
7.2 DSL Compared to Other Access Technologies 144
7.2.1 Security and Reliability 144
7.2.2 Point-to-Point Versus Shared Access 145
7.2.3 Common Facilities for Voice and DSL 146
7.2.4 Bit-rate Capacity 146
7.2.5 Hybrid Access 146
7.2.6 Future Trends for DSL Access 146
7.3 DSL Overview 147
7.3.1 Voice-band Modems 147
7.3.2 The DSL Concept 147
7.3.3 DSL Terminology 149
7.3.4 Introduction to DSL Types 151
7.3.5 DSL Performance Improvement, Repeaters, and Bonding 152
7.3.6 Splitters and Filters for Voice and Data 153
7.3.7 Other Ways to Convey Voice and Data 155
7.4 Transmission Channel and Impairments 156
7.4.1 Signal Attenuation 158
7.4.2 Bridged Taps 159
7.4.3 Loading Coils 162
7.4.4 Return Loss and Insertion Loss 163
7.4.5 Balance 163
7.4.6 Intersymbol Interference (ISI) 163
7.4.7 Noise 164
7.4.8 Transmission Channel Models 170
7.5 DSL Transmission Techniques 170
7.5.1 Duplexing 170
7.5.2 Channel Equalization and Related Techniques 171
7.5.3 Coding 172
References 174
Further Readings 174
8 The Family of DSL Technologies 175
8.1 ADSL 175
8.1.1 G.lite 176
8.1.2 ADSL2 and ADSL2plus 177
8.1.3 ADSL1 and ADSL2plus Performance 178
8.2 VDSL 179
8.2.1 VDSL2 181
8.2.2 VDSL2 Performance 182
Contents xi
8.3 Basic Rate Interface ISDN 184
8.4 HDSL, HDSL2, and HDLS4 185
8.5 SHDSL 185
8.6 G.fast (FTTC DSL) 187
Reference 188
9 Advanced DSL Techniques and Home Networking 189
9.1 Repeaters and Bonding 189
9.2 Dynamic Spectrum Management (DSM) 190
9.3 Vectored Transmission 190
9.4 Home Networking 195
References 195
Further Readings 195
10 DSL Standards 197
10.1 Spectrum Management – ANSI T1.417 197
10.2 G.hs – ITU-T Rec. G.994.1 199
10.3 PLOAM – ITU-T Rec. G.997.1 200
10.4 G.bond – ITU-T Recs. G.998.1, G.998.2, and G.998.3 201
10.5 G.test – ITU-T Rec. G.996.1 202
10.6 G.lt – ITU-T Rec. G.996.2 202
10.7 Broadband Forum DSL Testing Specifications 203
10.8 Broadband Forum TR-069 – Remote Management of CPE 204
References 205
11 The DOCSIS (Data-Over-Cable Service Interface Specification) Protocol 207
11.1 General Introduction 207
11.2 Introduction to MSO Networks 207
11.3 Background on Hybrid Fiber Coax (HFC) Networks 208
11.4 Introduction to DOCSIS 210
11.5 DOCSIS Network Elements 210
11.5.1 CMTS (Cable Modem Terminating System) 211
11.5.2 CM (Cable Modem) 212
11.5.3 FN (Fiber Node) 213
11.5.4 RF Combiner Shelf 213
11.6 Brief History of the DOCSIS Protocol Evolution 213
11.6.1 DOCSIS 1.0 214
11.6.2 DOCSIS 1.1 214
11.6.3 DOCSIS 2.0 214
11.6.4 DOCSIS 3.0 215
11.6.5 Regional History and Considerations 215
11.7 DOCSIS Physical Layer 216
11.7.1 DOCSIS Downstream Transmission 216
11.7.2 DOCSIS Upstream Transmission 218
11.8 Synchronization and Ranging 222
11.8.1 Synchronization 223
11.8.2 Ranging 224
11.9 DOCSIS MAC Sub-Layer 226
11.9.1 Downstream MAC 227
11.9.2 Upstream MAC 228
xii Contents
11.9.3 MAC Management Messages 232
11.9.4 MAC Parameters 233
11.10 CM Provisioning 239
11.11 Security 240
11.12 Introduction to Companion Protocols 242
11.12.1 The PacketCable
11.12.2 The OpenCable
TM
Protocol 242
TM
Protocol 242
11.12.3 PacketCable Multimedia (PCMM) 242
11.13 Conclusions 243
References 243
Further Readings 243
12 Broadband in Gas Line (BIG) 245
12.1 Introduction to BIG 245
12.2 Proposed Technology 245
12.3 Potential Drawbacks for BIG 245
12.4 Broadband Sewage Line 247
Reference 247
13 Power Line Communications 249
13.1 Introduction 249
13.2 The Early Years 250
13.3 Narrowband PLC 251
13.3.1 Overview of NB-PLC Standards 252
13.4 Broadband PLC 253
13.4.1 Overview of BB-PLC Standards 254
13.5 Power Grid Topologies 257
13.5.1 Outdoor Topologies: HV, MV, and LV 257
13.5.2 Indoor Topologies 258
13.6 Outdoor and In-Home Channel Characterization 261
13.6.1 Characteristics of the HV Power Line Channel 262
13.6.2 Characteristics of MV Power Line Channel 262
13.6.3 Characteristics of LV Power Line Channel 263
13.6.4 Power Line Noise Characteristics 263
13.7 Power Line Channel Modeling 269
13.7.1 Recent Results on the Modeling of Wireline Channels: Towards a Unified
Framework 271
13.8 The IEEE 1901 Broadband over Power Line Standard 273
13.8.1 Overview of Technical Features 273
13.8.2 The MAC and the Two PLCPs 274
13.8.3 Access-Specific Features 275
13.9 PLC and the Smart Grid 277
13.9.1 PLC for MV 279
13.9.2 PLC for LV 279
13.10 Conclusions 283
References 284
Further Reading 285
Contents xiii
14 Wireless Broadband Access: Air Interface Fundamentals 287
14.1 Introduction 287
14.2 Duplexing Techniques 287
14.2.1 Frequency-Division Duplex 288
14.2.2 Time-Division Duplex 288
14.3 Physical Layer Concepts 289
14.3.1 The Wireless Channel 289
14.3.2 Diversity 290
14.3.3 Channel Coding 291
14.3.4 Interleaving 291
14.3.5 Multi-Antenna Techniques and Multiple-Input Multiple-Output (MIMO) 291
14.4 Access Technology Concepts 295
14.4.1 Frequency Division Multiple Access (FDMA) 295
14.4.2 Time Division Multiple Access (TDMA) 295
14.4.3 Code Division Multiple Access (CDMA) 295
14.4.4 Orthogonal Frequency Division Multiplexing (OFDM) 297
14.4.5 MAC Protocols 299
14.5 Cross-Layer Algorithms 300
14.5.1 Link Adaptation 300
14.5.2 Channel-Dependent Scheduling 300
14.5.3 Automatic Repeat Request (ARQ) and Hybrid ARQ (HARQ) 302
14.6 Example Application: Satellite Broadband Access 303
14.7 Summary 303
Further Reading 304
15 WiFi: IEEE 802.11 Wireless LAN 305
15.1 Introduction 305
15.2 Technology Basics 306
15.2.1 System Overview 306
15.2.2 MAC Layer 308
15.2.3 Physical Layer 311
15.3 Technology Evolution 312
15.3.1 802.11 b 312
15.3.2 802.11 a/g 313
15.3.3 802.11 n 314
15.3.4 802.11 ac 316
15.4 WLAN Network Architecture 318
15.5 TV White Space and 802.11 af 320
15.6 Summary 320
Further Readings 321
16 UMTS: W-CDMA and HSPA 323
16.1 Introduction 323
16.2 Technology Basics 324
16.2.1 Network Architecture 324
16.2.2 Protocol Architecture 325
16.2.3 Physical Layer (L1) 327
16.2.4 Layer-2 334
16.2.5 Radio Resource Control (RRC) 336
xiv Contents
16.3 UMTS Technology Evolution 338
16.3.1 Release 99 338
16.3.2 Release 5: High-Speed Downlink Packet Access (HSDPA) 339
16.3.3 Release 6: Enhanced Uplink 343
16.3.4 Release 7 347
16.3.5 Release 8 and Beyond 348
16.4 CDMA2000 350
16.5 Summary 351
Further Readings 352
17 Fourth Generation Systems: LTE and LTE-Advanced 353
17.1 Introduction 353
17.1.1 LTE Standardization 353
17.1.2 LTE Requirements 354
17.2 Release 8: The Basics of LTE 355
17.2.1 Network Architecture 355
17.2.2 PDN Connectivity, Bearers, and QoS Architecture 358
17.2.3 Protocol Architecture 360
17.2.4 Layer-1: The Physical Layer 361
17.2.5 Layer-2 and Cross-Layer Algorithms 370
17.2.6 Layer-3: Radio Resource Control (RRC) 380
17.3 Release 9: eMBMS and SON 383
17.3.1 Evolved Multimedia Broadcast Multicast Service (eMBMS) 384
17.3.2 Self-Organizing Networks (SON) 386
17.4 Release 10: LTE-Advanced 386
17.4.1 Carrier Aggregation 388
17.4.2 Heterogeneous Networks with Small Cells 391
17.5 Future of LTE-Advanced: Release 11 and Beyond 395
17.5.1 Cooperative Multi-Point (CoMP) 396
17.5.2 Release 12 and the Future of LTE 398
17.6 IEEE 802.16 and WiMAX Systems 399
17.7 Summary 400
Further Readings 402
18 Conclusions Regarding Broadband Access Networks and Technologies 403
Index 407
About the Authors
Steve Gorshe
Steve Gorshe is a Distinguished Engineer in the CTO organization of PMC-Sierra, Inc., where his work
since 2000 has included technology development and telecommunications standards. He received his
BSEE from the University of Idaho (1980) and both his MSEE (1982) and PhD (2002) degrees from
Oregon State University. Since 1983, he has worked in product development, applied research, and
systems architecture of telecommunications access and transport systems. His standards activity includes
over 300 contributions across six standards bodies, serving as technical editor for nine North American
and international standards, and currently serving as Associate Rapporteur for the Q11 group of ITU-T
Study Group 15.
Steve is a Fellow of the IEEE. His IEEE activities include Communications Magazine Editor-inChief (2010–2012), Associate Editor-in-Chief (2006–2009), and Broadband Access Series co-editor
(1999–2006). He has also served as the IEEE Communications Society Director of Magazines and
Chair of the Transmission, Access and Optical Systems Technical C ommittee.
Steve has 37 patents issued or pending, over 24 published papers, and is co-author of two textbooks and
co-author of chapters in three other textbooks.
Arvind R. Raghavan
Arvind R. Raghavan heads research and development at Blue Clover Devices, where he is involved with
the design and implementation of innovative products for the Internet of Things, with current emphasis on
Bluetooth Low Energy technology. Before joining Blue Clover Devices, he was part of the Radio
Technology and Strategy group at AT&T Labs, where his work focused on the impact of QoS on LTE,
design and analysis of heterogeneous networks, and advanced MIMO techniques for standardization in
3GPP. Prior to joining AT&T Labs, he played a lead role in the Systems Engineering group at
ArrayComm, LLC, where they developed specifications for their multi-antenna signal processing
products, conducted performance analyses, and made contributions to the standardization of WiMAX
systems. Arvind holds MS and PhD degrees in Electrical Engineering from Clemson University.
Thomas Starr
Thomas Starr is a Lead Member of Technical Staff at AT&T Laboratories in Hoffman Estates, Illinois.
Thomas is responsible for the development and standardization of local access and home networking
technologies for AT&T’s network. These technologies include ADSL, HDSL, SHDSL, VDSL and G.hn.
In 2009, Thomas received the prestigious AT&T Science and Technology Medal. He serves as Chairman
of the Broadband Forum and has also served as a member of the Board of Directors since its inception as
the ADSL Forum in 1994. Thomas has been a distinguished fellow of the Broadband Forum From 1988 to
2000, has served as Chairperson of ANSI accredited standards working group T1E1.4, which develops
xvi About the Authors
xDSL standards for the United States, received the Committee T1 Outstanding Leadership Award in
2001, and now serves at ATIS COAST-NAI Chairman. In the ITU-T SG15, Thomas serves as Chairman
of Working Party 1, addressing fiber, DSL, and home networking standards, and participates in the ITU
SG15 Q4 group on xDSL international standards.
Thomas is a co-author of the books DSL Advances, published by Prentice Hall in 2003, and
Understanding Digital Subscriber Line Technology, published by Prentice Hall in 1999. Thomas is
also the author of the Science Fiction novel Virtual Vengeance. Thomas previously worked for 12
years at AT&T Bell Laboratories on ISDN and local telephone switching systems, and twenty US
patents in the field to telecommunications have been issued to him. Thomas holds a MS degree in
Computer Science and a BS degree in Computer Engineering from the University of Illinois in Urbana,
Illinois.
Stefano Galli
Stefano Galli received his MS and PhD degrees in Electrical Engineering from the University of Rome
“La Sapienza” (Italy) in 1994 and 1998, respectively. He is currently the Director of Technology Strategy
of ASSIA – the leading developer of automated management and diagnostics tools for broadband
networks. Prior to this position, he held the role of Director of Energy Solutions R&D for Panasonic
Corporation and Senior Scientist at Bellcore.
Dr. Galli is serving as Chief Information Officer of the IEEE Communications Society (ComSoc),
director of Smart Grid activities for the IEEE ComSoc Technical Committee on Power Line Communications, member of the Energy and Policy Committee of IEEE-USA, and as Editor for the IEEE
Transactions on Communications and the IEEE Communications Magazine. Dr. Galli is also serving as
Rapporteur for the ITU-T Q15/15 “Communications for Smart Grid” standardization group. Past
positions include serving as Co-Chair of the “Communications Technology” Task Force of IEEE
2030 (Smart Grid), Leader of the “Theoretical and Mathematical Models” Group of IEEE 1901
(Broadband over Power Lines standard), Coexistence sub-group Chair of the SGIP/NIST PAP 15,
elected Member-at-Large of the IEEE Communications Society (ComSoc) Board of Governors, and a
variety of other leadership positions in the IEEE. He has also served as Founder and first Chair of the IEEE
ComSoc Technical Committee on Power Line Communications.
Dr. Galli is a Fellow of the IEEE, has received the 2013 IEEE Donald G. Fink Best Paper Award for his
paper on Smart Grid and Power Line Communications, the 2011 IEEE ComSoc Donald W. McLellan
Meritorious Service Award, the 2011 Outstanding Service Award from the IEEE ComSoc Technical
Committee on Power Line Communications, and the 2010 IEEE ISPLC Best Paper Award. He holds
several issued and pending patents, has published over 90 peer-reviewed papers, has co-authored three
book chapters on power line communications, and has made numerous standards contributions to the
IEEE, the ITU-T, the Broadband Forum, and the UK NICC.
Acknowledgments
Thanks are given to the experts who provided assistance for the chapters on DSL technology: George
Ginis, Ken Kerpez, Vladimir Oksman, Craig Schelp, Massimo Sorbara, and Arlynn Wilson. Thanks also
are given to Marilynn Starr for her support and assistance.
Steve would like to thank the following people for their generous help, excellent comments and reviews
for portions of his chapters: Frank Effenberger, Alon Bernstein, Chris Look, Onn Haran, Jeff Mandin,
Lior Khermosh, Bob Murray, Valy Ossman, and Jim Dahl. Steve also wants to thank PMC-Sierra for
allowing some of his white paper material to be adapted for this book.
Arvind would like to acknowledge the significant contributions of his wife, Sanchita Shetty, for
painstakingly generating all the figures in the wireless chapters, and her unwavering support throughout
the writing of this book. He would also like to express his heartfelt gratitude to Paul Chiuchiolo, Rich
Kobylinski, Milap Majmundar, and Tom Novlan, for reviewing the wireless section of the book and
providing excellent feedback for improving the quality and accuracy of the manuscript. Finally, he would
like to thank his family and all his wonderful friends in Austin for their love and encouragement.
List of Abbreviations and Acronyms
2G Second Generation
3G Third Generation
3GPP Third Generation Partnership Project
10GE 10 Gigabit/s Ethernet specified in IEEE 802.3
10G EPON 10 Gbit/s Ethernet Passive Optical Network specified in IEEE 802.3
ABS Almost Blank Subframes
AC Alternating Current
AC Access Category
ACK Acknowledgement
ACM Adaptive Coding and Modulation
ADC Analog-to-Digital Converter
ADSL Asymmetric Digital Subscriber Line specified in ITU-T G.992.1
ADSL2 Asymmetric Digital Subscriber Line 2 specified in ITU-T G.992.3
ADSL2plus Asymmetric Digital Subscriber Line 2plus specified in ITU-T G.992.5
AES Advanced Encryption Standard
AFE Analog Front End
AICH Acquisition Indicator Channel
AM Acknowledged Mode
AMI Advanced Metering Infrastructure
A-MPDU Aggregate MAC Protocol Data Unit
AMPS Advanced Mobile Phone System
AMR Automatic Meter Reading
A-MSDU Aggregate MAC Service Data Unit
ANSI American National Standards Institute
AP Access Point
APD Avalanche Photo Diode
APS Automatic Protection Switching
ARIB Association of Radio Industries and Businesses
ARP Allocation and Retention Priority
ARQ Automatic Repeat Request, Retransmission
AS Access Stratum
ASE Amplified Spontaneous Emission
ASF DOCSIS Aggregated Service Flow
A-TDMA Advanced TDMA (used with DOCSIS)
ATIS Alliance for Telecommunications Industry Solutions
ATM Asynchronous Transfer Mode protocol
xx List of Abbreviations and Acronyms
AWG American Wire Gauge
AWG Arrayed Waveguide Grating WDM filter/multiplexer
BB Broad Band
BCCH Broadcast Control Channel
BCH Broadcast Channel
BE Best Effort service
BEMS Building Energy Management System
BER Bit Error Rate (or Ratio)
BIP Bit Interleaved Parity
BMC Broadcast Multicast Control
BMSC Broadcast Multicast Service Center
B-ONU DPoE Bridge ONU
BPL Broadband over Power Lines
B-PON FSAN/ITU-T Broadband PON protocol specified in the ITU-T G.983 series
BRI-ISDN Basic Rate Integrated Services Digital network
BSS Basic Service Set
BTS Base Transceiver Station (for a wireless network)
CA Carrier Aggregation
CAPEX Capital Expense
CAPWAP Control and Provisioning of Wireless Access Points
CATV Community Access Television
CBR Constant Bit Rate
CBS Committed Burst Size
CC Component Carrier
CCA Clear Channel Assessment
CCCH Common Control Channel
CCK Complementary Code Keying
CCO Capacity and Coverage Optimization
CDD Cyclic-Delay Diversity
CDMA Code Division Multiple Access
CENELEC European Committee for Electotechnical Standardization
CEPCA Consumer Electronics Powerline Alliance
CES Circuit Emulation Service
CFP Contention Free Period
CIF Carrier Indicator Field
CIR Committed Information Rate
CM Cable Modem
CMCI DOCSIS Cable Modem CPE Interface
CMTS DOCSIS Cable Modem Terminating System
CN Core Network
CO Telephone company Central Office
CoMP Cooperative Multi-Point
CP Contention Period
CP Cyclic Prefix
CPC Continuous Packet Connectivity
CPE Customer Premises Equipment
CPICH Common Pilot Channel
List of Abbreviations and Acronyms xxi
CPRI Common Public Radio Interface
CQI Channel Quality Information
CRC Cyclic Redundancy Check
CRE Cell Range Expansion
CRS Cell-specific Reference Signal
CS Circuit Switched
CS Channel Sensing
CSA Carrier Serving Area
CS/CB Coordinated Scheduling/Coordinated Beamforming
CSG Closed Subscriber Group
CSI-RS Channel State Information Reference Signal
CSM Collaborative Spatial Multiplexing
CSMA/CA Carrier Sense Multiple Access with Collision Avoidance
CSO Cell Selection Offset
CTCH Common Traffic Channel
CTS Clear-to-send
CTS Common Technical Specification for G-PON
CV Code Violation
C-VID Customer VLAN Identifier (Ethernet)
CWDM Coarse Wavelength Division Multiplexing
DAC Digital-to-Analog Converter
DAS Distributed Antenna System
dB Decibel, ten times the common logarithm of the ration of two powers
DBA Dynamic Bandwidth Assignment
DBC Dynamic Bonding Change (in DOCSIS 3.0)
DBG Downstream Bonding Group (in DOCSIS 3.0)
DBR Dynamic Bandwidth Report
DC Direct Current
DCCH Dedicated Control Channel
DCF Distributed Coordination Function
DCH Dedicated Channel
DCS Downstream Channel Set (in DOCSIS 3.0)
DELT Dual Ended Line Test
DEMARC Carrier owned Demarcation device between the carrier and the CPE
DER Distributed Energy Resources
DFE Decision Feedback Equalizer
DFT Discrete-time Fourier Transform
DHCP Dynamic Host Configuration Protocol
DIFS Distributed Interframe Spacing
DL Downlink
DLC Digital Loop Carrier
DLL Data Link Layer
DL-SCH Downlink Shared Channel
DM-RS Demodulation Reference Signal
DMT Discrete Multi Tone modulation
DOCSIS Data Over Cable Service Interface Specification
D-ONU DPoE ONU
Downstream Data flowing towards the customer
xxii List of Abbreviations and Acronyms
DPB Dynamic Point Blanking
DPCCH Dedicated Physical Control Channel
DPDCH Dedicated Physical Data Channel
DPoE DOCSIS Protocol over Ethernet protocol
DPS Dynamic Point Selection
DPSK Differential Phase Shift Keying
DQPSK Differential Quadrature Phase Shift Keying
DR Demand Response
DRX Discontinuous Reception
DS Direct Sequence
DS1 Digital Signal level 1 in the North American asynchronous telephone network
hierarchy
DS-CDMA Direct Sequence Code Division Multiple Access
DSCP DiffServ Code Point
DSID Downstream Service ID (in DOCSIS 3.0)
DSL Digital Subscriber Line
DSLAM DSL Access Multiplexer
DSM Dynamic Spectrum Management (in DSL)
DSM Demand Side Management (in Smart Grid)
DSP Digital Signal Processing
DSSS Direct Sequence Spread Spectrum
DTX Discontinuous Transmission
DVB Digital Video Broadcast
DVB-RCS Digital Video Broadcast Return Channel via Satellite
DVB-S2 Digital Video Broadcasting - Satellite - Second generation
DWDM Dense Wavelength Division Multiplexing
E-AGCH Enhanced Absolute Grant Channel
EBS Excess Burst Size
ECH Echo Cancelled Hybrid
eCM embedded Cable Modem
EDCA Enhanced Distributed Channel Access
E-DCH Enhanced Dedicated Channel
EDFA Erbium Doped Fiber Amplifier
EDGE Enhanced Data-rates for GSM Evolution
E-DPCCH Enhanced Dedicated Physical Control Channel
E-DPDCH Enhanced Dedicated Physical Data Channel
E-HICH Enhanced HARQ Indicator Channel
eICIC Enhanced Inter-Cell Interference Coordination
EIR Excess Information Rate
eMBMS Enhanced Multimedia Broadcast and Multicast Service
EMC Electro-Magnetic Compatibility
EMS Element Management System
EO Electrical to Optical signal conversion
eOAM Extended OAM messages used in DPoE
EOC Embedded Operations Channel
EONT Embedded ONT
eNodeB Evolved Node-B
EPC Evolved Packet Core
List of Abbreviations and Acronyms xxiii
EPON Ethernet Passive Optical Network (1 Gbit/s rate)
EPS Evolved Packet System
E-RGCH Enhanced Relative Grant Channel
eSAFE embedded Service/Application Functional Entity
ESP Ethernet Service Path
ESS Extended Service Set
ETSI European Telecommunications Standards Institute
E-UTRAN Evolved UMTS Terrestrial Radio Access Network
EVC Ethernet Virtual Circuit
EVSE Electric Vehicle Supply Equipment
FACH Forward Access Channel
FBI Feedback Information
FCC Federal Communications Commission
FCS Frame Check Sequence
FDD Frequency Division Duplexing
FDM Frequency Division Multiplexing
FDMA Frequency Division Multiple Access
F-DPCH Fractional Dedicated Physical Channel
FEC Forward Error Correction
FeICIC Further Enhanced Inter-Cell Interference Coordination
FEXT Far End crosstalk
FFT Fast Fourier Transform
FH Frequency Hopping
FH-CDMA Frequency Hopping Code Division Multiple Access
FHSS Frequency Hopping Spread Spectrum
FITL Fiber in the Loop
FN Fiber Node (in a HFC network)
FSAN Full Service Access Network industry consortium
FSK Frequency Shift Keying
FTTC Fiber to the Curb
FTTCab Fiber to the Cabinet
FTTCell Fiber to the Cell site
FTTH Fiber to the Home
FTTN Fiber to the Node
FTTO Fiber to the Office
FTTP Fiber to the Premises
G.hn ITU-T G.9960/9961 home networking standard
G.hs ITU-T G.994.1 DSL handshake protocol
G.lite ITU-T G.992.2 reduced complexity ADSL
G.lt ITU-T G.996.2 standard for DSL line test functions
G.test ITU-T G.996.1 standard for testing of DSL modems
GBR Guaranteed Bit Rate
GE Gigabit/s Ethernet
GEM G-PON Encapsulation Method
GERAN GSM Edge Radio Access Network
GFP Generic Framing Procedure specified in ITU-T G.7041
GGSN Gateway GPRS Support Node
xxiv List of Abbreviations and Acronyms
GMSC Gateway Mobile Switching Center
GP Guard Period
G-PON FSAN/ITU-T Gigabit-capable PON protocol specified in the ITU-T G.984 series
GPRS GSM Packet Radio System
gPTP generalized Precision Timing Protocol
GSM Global System for Mobile communications
GTC G-PON Transmission Convergence
HAN Home Area Network
HARQ Hybrid Automatic Repeat Request
HCF Hybrid Coordination Function
HD-PLC High Definition Power Line Communication
HDR High Data Rate
HDSL High bit rate Digital Subscriber Line
HDSL2 High bit rate Digital Subscriber Line, 2 wire version
HDSL4 High bit rate Digital Subscriber Line, 4 wire version
HE Head End
HEC Header Error Check
HEMS Home Energy Management System
HetNet Heterogeneous Network
HF High Frequency
HFC Hybrid Fiber-Coaxial cable network
HLR Home Location Register
HSDPA High Speed Downlink Packet Access
HS-DPCCH High Speed Dedicated Physical Control Channel
HS-DSCH High Speed Downlink Shared Channel
HSPA High Speed Packet Access
HS-PDSCH High Speed Physical Downlink Shared Channel
HS-SCCH High Speed Shared Control Channel
HSS Home Subscriber Server
HSUPA High Speed Uplink Packet Access
HV High Voltage
IAD Integrated Access Device
ICIC Inter-cell Interference Coordination
IEC International Electrotechnical Commission
IED Intelligent Electronic Devices
IEEE Institute of Electrical and Electronic Engineers
IETF Internet Engineering Task Force
IFS Inter-Frame Spacing
IGMP Internet Group Management Protocol
IMT International Mobile Telecommunications
IP Internet Protocol
IP-HSD DOCSIS IP High-Speed Data service
IPP Inter-PHY Protocol
IPTV Television delivered over Internet Protocol
IPv6 Internet Protocol version 6
IR Infra-Red
IR Incremental Redundancy
List of Abbreviations and Acronyms xxv
IRC Interference Rejection Combining
IS-54 A second generation cellular standard
IS-136 A second generation cellular standard, an improvement on IS-54ISI Intersymbol
interference
ISI Inter-symbol Interference
ISM Industrial, Scientific, and Medical
ISO International Organization for Standardization
ISP Internet Service Provider
ISP IEEE 1901 Inter System Protocol
ITU-T International Telecommunication Union – Telecommunication Standardization Sector
JP Joint Processing
JT Joint Transmission
kft kilofeet (length of wire)
L1 Layer-1
L2 Layer-2
L3 Layer-3
LAN Local Area Network
LDPC Low Density Parity Check
LDR Low Data Rate
LED Light Emitting Diode
LF Low Frequency
LLID Ethernet Logical Link Identifier
LOF Loss Of Frame
LoS Line of Sight
LOS Loss Of Signal
LSB Least Significant Bit
LTE Long Term Evolution (mobile telephone standard)
LV Low Voltage
MAC Medium Access Control
MAN Metro Area Network
MBMS-GW Multimedia Broadcast Multicast Service Gateway
MBR Maximum Bit Rate
MBSFN Multicast Broadcast Single Frequency Network
MCCA MCF Controlled Channel Access
MCE Multicell/Multicast Coordination Entity
MCF Mesh Coordination Function
M-CMTS Modular CMTS
MCS Modulation and Coding Scheme
MEF Metro Ethernet Forum
MELT Metallic line test
MF Medium Frequency
MF-TDMA Multi-Frequency Time Division Multiple Access
MIB Management Information Base
MIMO Multiple Input Multiple Out
MLB Mobility Load Balancing
xxvi List of Abbreviations and Acronyms
MLME MAC Layer Management Entity
MME Mobility Management Entity
MMSE Minimum Mean Squared Error
MoCA Multimedia over Coax Alliance
Modem Modulator/Demodulator, a transceiver
MPCPDU Multi-Point Control Protocol PDU
MPDU MAC Protocol Data Unit
MPEG Motion Picture Experts Group video compression standards
MRC Maximal Ratio Combining
MRO Mobility Robustness Optimization
MSB Most Significant Bit
MSC Mobile Switching Center
MSDU MAC Service Data Unit
MSO Multiple System Operator (cable network operator)
MTA Multimedia Terminal Adapter
MTL Multi-Conductor Transmission Line
MU-MIMO Multi-user Multiple Input Multiple Output
MV Medium Voltage
NACK Negative Acknowledgement
NAS Non-Access Stratum
NAV Network Allocation Vector
NB Narrow Band
NE Network Element
NEXT near end crosstalk
NG-PON FSAN/ITU-T Next Generation PON protocol
NI Network Interface
NID Network Interface Device
NMS Network Management System
Node-B Base Station in a third generation cellular system
nrt-PS Non-real-time Poling Service (DOCSIS)
NRZ Non-Return to Zero line code
NSR Non-Status Reporting
NTU Network Termination Units
OAM Operations, Administration and Maintenance
OAM&P Operations, Administration, Maintenance and Provisioning
OBSAI Open Base Station Architecture Initiative
ODN Optical Distribution Network
OE Optical to Electrical signal conversion
OEO Optical to Electrical to Optical signal conversion (repeater)
OFDM Orthogonal Frequency Division Multiplexing
OFDMA Orthogonal Frequency Division Multiple Access
OLT Optical Line Terminal
OLU Optical Line Unit
OMCC ONU Management and Control Channel
OMCI ONU Management and Control Interface
ONT Optical Network Terminal
ONU Optical Network Unit
List of Abbreviations and Acronyms xxvii
OTN Optical Transport Network (ITU-T G.709)
OVSF Orthogonal Variable Spreading Factor
PAM Pulse Amplitude Modulation
PAP Priority Action Plan
PBCH Physical Broadcast Channel
PBR Prioritized Bit Rate
PCB Physical layer Control Block
PCC Primary Component Carrier
PCCH Paging Control Channel
PCCPCH Primary Common Control Physical Channel
PCF Point Coordination Function
PCFICH Physical Control Format Indicator Channel
PCH Paging Channel
PCI Pre-coder Indicator
PCMM Packet Cable Multi-Media protocol
PCRF Policy and Charging Rules Function
PDCCH Physical Downlink Control Channel
PDCP Packet Data Convergence Protocol
PDFA Praseodymium Doped Fiber Amplifier
PDN Packet Data Network
PDN Premises Distribution Network
PDSCH Physical Downlink Shared Channel
PDU Protocol Data Unit
PEIN Prolonged Electrical Impulse Noise
PF Proportionally Fair
P-GW PDN Gateway
PHEV Plug-in (Hybrid) Electric Vehicles
PHICH Physical HARQ Indicator Channel
PHS Payload Header Suppression
PHY Physical Layer
PIFS PCF Inter-Frame Spacing
PIN Photo diode constructed with P-type, Intrinsic, and N-type semiconductor regions
PL Power Line
PLC Power Line Communications
PLCP Physical Layer Convergence Procedure
PLI Payload Length Indicator
PLO Physical Layer Overhead
PLOAM Physical Layer OAM
PMCH Physical Multicast Channel
PMD Physical Medium Dependent sublayer
PMI Precoding Matrix Indicator
PMS-TC Physical media specific transmission convergence sublayer
PON Passive Optical Network
POTS Plain Old Telephone Service
PRACH Physical Random Access Channel
PRB Physical Resource Block
PRIME Powerline Related Intelligent Metering
PS Packet Switched
xxviii List of Abbreviations and Acronyms
PSB Physical Layer Synchronization Block
PSD Power Spectral Density
PSS Primary Synchronization Signal
PSTN Public Switched Telephone Network
PTI Payload Type Indicator
PTP Precision Timing Protocol
PUCCH Physical Uplink Control Channel
PUSCH Physical Uplink Shared Channel
QAM Quadrature Amplitude Modulation
QCI QoS Class Identifier
QoS Quality of Service
RACH Random Access Channel
RAN Radio Access Network
RAT Radio Access Technology
RB Resource Block
RCS Ripple Carrier Signaling
RDI Remote Defect Indication
RE Resource Element
REIN Repetitive Electrical Impulse Noise
RF Radio Frequency
RFI Radio Frequency Interference
RFoG Radio Frequency over Glass
RI Rank Indicator
RIT Radio Interface Technology
RLC Radio Link Control
RMS-DB Root Mean Square - Delay Spread
RNC Radio Network Controller
RoF Radio over Fiber
RoHC Robust Header Compression
R-ONU RFoG Optical Network Unit
RP Repeater
RP Reception Point
RRC Radio Resource Control
RRH Remote Radio Head
RS Reed Solomon
RSOA Reflective Semiconductor Optical Amplifier
RT Remote Terminal
RTD Round Trip Delay
rt-PS Real-time Poling Service (DOCSIS)
RTS Request-to-send
RTT Round Trip Time
SA System Architecture
SAE Society of Automotive Engineers
SAI Serving Area Interface
SCADA Supervisory Control and Data Acquisition
SCB Single Copy Broadcast Ethernet frame
List of Abbreviations and Acronyms xxix
SCC Secondary Component Carrier
SCCPCH Secondary Common Control Physical Channel
SC-FDMA Single-Carrier Frequency Division Multiple Access
SCH Synchronization Channel
SCTE Society of Cable Telecommunications Engineers
SDF Service Data Flow
SDO Standard Development Organization
SDU Service Data Unit
SELT Single Ended Line Test
SES Severely Error Seconds
SF DOCSIS Service Flow
SFBC Space Frequency Block Coding
SFD Ethernet Start of Frame Delimiter
SGSN Serving GPRS Support Node
S-GW Serving Gateway
SHDSL Symmetric High bit rate Digital Subscriber Line, ITU-T G.991.2
SHINE Short High amplitude Impulse Noise Event
SID Service Identifier
SIEPON Standard for Service Interoperability in Ethernet Passive Optical Networks
SIFS Short Inter-Frame Spacing
SIM Subscriber Identity Module
SINR Signal-to-Interference-and-Noise Ratio
SIR Signal-to-Interference Ratio
SLA Service Level Agreement
SLF Super Low Frequency
SMB Small or Medium sized Business
SNMP Simple Network Management Protocol
SNR Signal to Noise Ratio
SOA Semiconductor Optical Amplifier
SON Self-Optimizing Network
S-ONU DPoE Standalone ONU
SPS Semi-Persistent Scheduling
SR Status Reporting
SR Scheduling Request
SRS Sounding Reference Signal
S-SCMA Synchronous CDMA (used with DOCSIS)
SSID Service Set Identifier
SSS Secondary Synchronization Signal
STA Station
STB Set-Top Box
STBC Space Time Block Coding
STM Synchronous Transfer Mode
SU-MIMO Single-user Multiple Input Multiple Output
S-VID Service VLAN Identifier (Ethernet)
T1 Repeatered 1.544 Mbit/s transmission line using Alternate Mark Inversion coding
T1E1.4 United States DSL standards committee now called COAST-NAI
TC Transmission Convergence
TCM Time Compression Multiplexing
xxx List of Abbreviations and Acronyms
TCM Trellis Code Modulation
T-CONT G-PON Transmission Container
TCP Transmission Control Protocol
TC-PAM Trellis Coded Pulse Amplitude Modulation
TDD Time Division Duplexing
TDFA Thulium Doped Fiber Amplifier
TDM Time Division Multiplexing
TDMA Time Division Multiple Access
TD-SCDMA Time Division Synchronous Code Division Multiple Access
TFCI Transport Format Combination Indicator
TFT Traffic Flow Template
TFTP Trivial File Transfer Protocol
TG Task Group
TIA Transimpedance Amplifier
TL Transmission Line
TLV Type-Length-Value field
TM Transparent Mode
TM Transmission Mode
ToD Time of Day
TOS Type of Service
TP Transmission Point
TPC Transmit Power Control
TR-069 Broadband Forum standard for remote management of CPE
TPS-TC Transport protocol specific transmission convergence sublayer
TTI Transmission Time Interval
TWACS Two-Way Automatic Communications System
TWDM Concurrent time and wavelength division multiplexing
UCD DOCSIS Upstream Channel Descriptor
UE User Equipment
UGS Unsolicited Grant Service (DOCSIS)
UGS-AD Unsolicited Grant Service with Activity Detection (DOCSIS)
UL Uplink
ULF Ultra Low Frequency
UL-SCH Uplink Shared Channel
UM Unacknowledged Mode
UMTS Universal Mobile Telecommunication System
UNB Ultra Narrowband
UNI User-Network Interface
U-NII Unlicensed National Information Infrastructure
UPBO Upstream Power Back Off
Upstream Data flowing from the customer
UTRAN UMTS Terrestrial Radio Access Network
VBR Variable Bit Rate
vCM virtual Cable Modem
VCSEL Vertical-Cavity Surface-Emitting Laser
VDSL1 Very high bit rate Digital Subscriber Line 1, ITU-T G.993.1
VDSL2 Very high bit rate Digital Subscriber Line 2, ITU-T G.993.2
List of Abbreviations and Acronyms xxxi
VID VLAN Identifier
VLAN Ethernet Virtual LAN
VLF Very Low Frequency
VoIP Voice over Internet Protocol
VoLTE Voice over Long Term Evolution
VSAT Very Small Aperture Terminal
WAN Wide Area Network
WARC World Administrative Radio Conference
WBF Wavelength Blocking Filter
W-CDMA Wideband Code Division Multiple Access
WDM Wavelength Division Multiplexing
WDMA Wavelength Division Multiple Access
WG Working Group
WiMAX A fourth generation cellular standard based on OFDM/OFDMA
WLAN Wireless Local Area Network
WSD White-Space Device
XGEM XG-PON Encapsulation Method
XG-PON FSAN/ITU-T 10 Gbit/s PON protocol specified in the ITU-T G.987 series
XGTC XG-PON Transmission Convergence
1
Introduction to Broadband Access Networks and Technologies
1.1 Introduction
In the mid-1990s, there were many doubts about the future of broadband access. No one was sure if the
mass market needed or wanted more than 100 kbit/s; what applications would drive that need; what
broadband access would cost to deploy and operate; what customers were willing to pay; whether the
technology could provide reliable service in the real world; or which access technology would “win.”
Government regulation in many countries made it unclear if investment in broadband would yield profits.
It seemed that broadband access would be available only to wealthy businesses. Fortunately, there
were some people who had a vision of a broadband world and who also had the faith to carry on despite
the doubts.
We now live in a world where broadband access is the norm and households without it are the
exception. No one asks today why the average household would need broadband access. The answer is
obvious: we need internet access, with its ever-growing number of applications, and VOD (video on
demand). With more than 600 million customers connected to broadband networks, no one asks if the
technology works or whether it can meet the customer’s willingness to pay.
Furthermore, a growing application of broadband access is the support of femtocells, and small cells in
general. Resorting to small cells has today become the most promising trend pursued for increasing
wireless spectral efficiency, and the key to its success is the availability of a high capacity wired line to the
home. Also, a growing fraction of cellular data is today generated indoors. In addition, it has become clear
that no single broadband access technology will win the entire market, and that the market shares of the
different technologies will change over time.
Each access technology has its strengths and weaknesses. A common constraint is that we can have it
fast, low cost, and everywhere – but not all at the same time. In many cases, the choice of broadband
access technology is driven by the legacy network infrastructure of the network provider. In other cases,
national regulatory considerations are a significant factor. As a result, each access technology has its areas
of dominance in terms of geography, applications, and political domains.
The book is divided into three sections:
The chapters in the first section of the book cover technologies and standard protocols for broadband
•
access over fiber-based access networks.
Broadband Access: Wireline and Wireless – Alternatives for Internet Services, First Edition.
Steven Gorshe, Arvind Raghavan, Thomas Starr, and Stefano Galli.
2014 John Wiley & Sons, Ltd. Published 2014 by John Wiley & Sons, Ltd.
2 Broadband Access
The chapters in the second section cover technologies and standards associated with non-fiber, non-
•
wireless broadband access.
The chapters in the final section of the book address wireless broadband access technology and
•
standards. Some of these technologies have been widely deployed, while others are anticipated to see
deployment soon.
1.2 A Brief History of the Access Network
The traditional access network consisted of point-to-point wireline connections between telephone
subscribers and an electronic multiplexing or switching system. The early access network used a
dedicated pair of wires (referred to as a copper line or “loop”) between the subscriber and the central
office (CO) switch.
many cases to connect subscribers to a remotely located terminal. This remote terminal (RT) would
multiplex calls from multiple subscribers onto a smaller number of wires for the connection to the CO.
Network cost was reduced by having far fewer pairs of wires from the CO to the remote areas. As the
technology evolved from analog frequency domain multiplexing (FDM) to digital time domain multiplexing (TDM), the RT systems became known as digital loop carrier (DLC) systems.
Data access to the telephone network began with the introduction of voiceband modems that could
transmit the data as a modulated signal within the nominally 4 kHz voiceband pass-band frequency. The
shorter lines (loops) allowed by DLCs made increasingly efficient modulation technologies practical.
However, as explained in Appendix 1.A, the maximum data capacity of voiceband modems was limited to
33.6 kbit/s, or 56kbit/s under special circumstances. Modems and their evolution are also discussed
briefly in Appendix 1.A.
As a result, out-of-band technologies were introduced that transmitted signals over the copper line at
frequencies outside the voiceband. Since these technologies sent digital information in the out-of-band
signals, they became known collectively as digital subscriber line (DSL) technology. DSL is discussed
further in Section 1.3 and Chapters 7–10.
Since the subscriber lines are implemented with twisted wire pairs, with multiple lines sharing the
same cable without being shielded from each other, there are limits on the bandwidth that is achievable
with DSL. For this reason, network providers became interested in alternatives to the subscriber line for
providing broadband access. The three main contending technologies are coaxial cable, fiber optic
cable, and wireless radio frequency connections. Each of these technologies is reviewed in later
chapters of this book.
Coaxial cable networks were deployed by community access cable television (CATV) companies to
provide broadcast video distribution. Due to the high bandwidth capabilities of coaxial cables, they had
the potential for offering broadband services to their subscribers. In order to offer broadband data services,
CATV companies evolved their networks to support upstream data transmission, and introduced fiber
optic cables for higher performance in the feeder portion of their networks. As discussed below and in
Chapter 11, coaxial networks have their own challenges as well as advantages.
Telephone network providers responded to the potential broadband advantages of the CATV
companies by deploying additional fiber in their access networks. Telephone companies have deployed
fiber directly to each subscriber’s premises in some areas. Others are deploying fiber to terminals near
enough to the subscribers’ premises that broadband services can be provided by the latest very high-speed
DSL technologies. The most attractive aspect to fiber is its virtually unlimited bandwidth capability. The
primary drawback has been the relatively high cost of the network and its associated optical components.
Wireless access had not originally been a significant contending technology for residential broadband
access. However, as wireless mobile networks have become widely deployed, and new technologies and
1
As the cost of multiplexing technology decreased, it became more economical in
1
Of course, some of the earliest access lines were “party lines” where several subscribers were connected in parallel to the same loop.
Due the inherent lack of privacy and decreased cost of providing access lines, party-lines have become a historical footnote.
Introduction to Broadband Access Networks and Technologies 3
protocols have been developed, wireless broadband access has become increasingly important. It is
especially attractive in regions that lacked a legacy wireline infrastructure capable of evolution to
broadband services. Examples of such regions include developing nations and rural areas. It also offers the
very significant advantage of allowing mobile, ubiquitous service rather than being restricted to service at
the subscriber’s premises.
Since a limited amount of spectrum is available for use in broadband services, the networks to support it
have become increasingly complex. For spectrum efficiency, wireless networks use grids of antenna,
where each subscriber only needs enough signal power to reach the nearest antenna. The region covered
by each antenna is referred to as a cell. The result is that the same frequencies can be used by subscribers in
non-adjacent cells, since their signals should not propagate far enough to interfere with each other. The
signal formats have been optimized in the latest protocols to approach the Shannon limit for data bits
transmitted per Hertz of transmission channel bandwidth. Capacity is further increased by re-use of the
spectrum through smaller cells and smart antenna technologies. Both add cost, and radio signals are
always more vulnerable to various types of interference than wireline technologies. Wireless technologies
are discussed further in Section 1.6 below, and in detail in Chapters 14–17.
1.3 Digital Subscriber Lines (DSL)
1.3.1 DSL Technologies and Their Evolution
DSL operates over a copper line at frequencies outside the voiceband, sending digital data directly from
the subscriber, and thus avoiding the need for an analog to digital conversion. Since the telephone lines
were designed to provide good quality for voiceband signals, they are often not particularly well suited for
higher rate data signals. Reflections become a significant problem in the electrical domain at rates beyond
the voiceband. One of the worst sources of reflections in North American networks is bridge taps. When
the feeder cables are installed from the CO into the loop area, they go to splice boxes where the wires
going to the subscribers are connected. When service is disconnected to a subscriber (e.g., due to the
homeowner moving), a second pair of wires may be connected to the feeder cable to serve a different
subscriber without removing the other line. The result is a bridge tap, and it is possible to have bridge taps
at more than one location along the connection to a subscriber. The unterminated end of the unused line(s)
causes electrical reflections of the DSL signals, and these reflections can cause destructive interference for
certain frequencies (any impedance mismatch along the copper connection to the CO to the subscriber can
cause harmful reflections, but the bridge taps are especially bad).
The first widely deployed services using a digital subscriber line were the Digital Data Service (DDS)
from AT&T. DDS used baseband signals over the line and offered data rates including 2400, 4800 and
9600 bits/s, and 56 kbit/s. The lower rate signals were sometimes converted to analog signals at the CO
and then mapped into a voiceband channel, thus avoiding any noise or distortion from the subscriber line.
DDS required the end-to-end service be synchronized to a common atomic clock. DDS circuits also
usually required that the line be groomed to remove impairments such as bridge taps. While DDS circuits
were very valuable for some customers (e.g., banks using them for connections to ATM machines), they
were too expensive to deploy to residential subscribers or even to many business subscribers.
The first serious attempt to provide higher data rates to subscribers was the basic rate interface of the
Integrated Services Digital Network (ISDN-BRI). ISDN-BRI used baseband signals
line to offer bidirectional data rates of 144 kbit/s. ISDN-BRI was designed to operate over most subscriber
lines of up to 18 000 feet without having to remove impairments such as bridge taps from the lines. The
144 kbit/s signal was typically divided into two 64 kbit/s bearer (B) channels and a 16 kbit/s data (D)
channel. The B channels could be used for voice or data, while the D channel carried the connection
signaling information, with its leftover bandwidth available to carry subscriber data packets. It was also
2
over the subscriber
2
Specifically ISDN BRI used the 2B1Q line code, which mapped two input data bits to a quaternary symbol (i.e., a symbol that has four
possible amplitude values).
4 Broadband Access
possible to merge the two B channels or merge the Bs and D channel into a single 144 kbit/s channel. The
cost of ISDN-BRI was relatively expensive, however, and there were no driving subscriber applications to
generate high demand. ISDN also required that the connection signaling protocol be processed by the CO
switch, which meant a major upgrade to the switches. By the time that Internet connectivity became a
driving application, much higher rates were practical for DSL.
bandwidth, too late, with too much network complexity.
3
In effect, ISDN BRI provided too little
4
DSL modems that were dedicated to data services began to be widely deployed instead of ISDN-BRI.
Initially, there were two broad categories of DSL. The first was a high speed DSL (HDSL) that provided
bidirectional symmetric service at half the DS1 rate over a single pair,
two pairs (half on each pair). Although it would seem that HDSL had no advantage over T1
5
or symmetric full DS1 rate over
6
service,
which also uses two pairs, HDSL was capable of operating over much longer line lengths than T1, and it
could do so without requiring repeaters. The total cost of HDSL was less than half of T1 lines, mainly due
to eliminating most of the labor needed to install repeaters and remove bridged taps. It became common
for carriers to use HDSL as the primary technology for providing DS1 connections to business customers.
The current generation of HDSL is HDSL2, which allows bidirectional symmetric transmission of up to
2.048 Mbit/s payloads over a single wire pair.
The second category is the DSL lines optimized for residential subscriber access. The first generation
was called ADSL (asymmetric DSL) due to its use of asymmetric data rates in the upstream and
downstream directions. Since residential subscriber are typically downloading more information than they
are providing to the network, they typically require much higher data rates from the network (downstream) than they do for upstream. This asymmetry in the desired data rates per direction was exploited to
achieve the higher downstream rates. The service rate for ADSL is affected by several factors, but line
length is the primary one. Over the past 25 years, telephone companies have tried to limit the line lengths
to 12 000 feet.
7
Rates of 768 kbit/s downstream with 384 kbit/s upstream are possible over most of these
lines. The actual rate is often determined adaptively as the system uses feedback to determine the
frequency response of the line. In addition to higher data rates, another advantage of these DSL systems
over ISDN-BRI was that they left the voiceband frequencies available for voice signals. This allowed
analog POTS (Plain Old Telephone Service) signals to “ride underneath” the DSL data in its native
format, which kept the voice and data signals separate within the network and allowed subscribers to use
their existing telephone sets without conversion to digital signals at the subscriber premises.
ADSL rates and signal formats have been standardized by the ADSL Forum (now the Broadband
Forum), by T1E1 (now ATIS COAST-NAI) and by the ITU-T SG15. SG15 is the primary body
developing the current generation of DSL standards. The latest generation of ADSL is specified in the
ITU-T G.992.5 standard for ADSL2plus which enables up to 20 Mb/s, with 12 Mb/s possible at 3000 feet.
Video delivery will require rates of 10–50 Mbit/s, depending on the service. For these rates, very highspeed DSL (VDSL) is required. VDSL requires lines lengths limited to 5000 feet. ITU-T SG15 has
developed the VDSL2 standard, whose specifications are provided in ITU-T G.993.2 which enables rates
up to 100 Mb/s upstream and downstream with 25 Mb/s downstream possible at 3000 feet. The ITU-T is
developing the G.fast standard which promises to achieve bit rates up to 1 Gb/s over short copper lines.
3
While DDS and ISDN BRI are digital subscriber loop technologies, it is most common to use the term DSL to refer to their successors
that operated at higher data rates.
4
Note thatanother early application of digital subscriber loops was to provide a second voice line over the same loop. This application is
known as “pair gain” since it provides multiple voice channels over the same pair. Some carriers used ISDN technology to provide the
simple pair gain service.
5
When discussing subscriber loops, the term “pair” means a twisted pair of wires used for differential signal transmission.
6
The term “T1” is commonly misused as being equivalent to a DS1. Strictly speaking, DS1 refers to the 1.544 Mbit/s signal and frame
format, while T1 refers to a specific AT&T carrier system that transmits DS1 signals over 4-wire repeatered copper pairs.
7
In the Bell System, the 12 000 ft. range was known as the Carrier Serving Area (CSA). Independent companies like GTE, who served
more rural areas, specified their loop limits at 18 000 feet. Beyond 18 000 feet, inductive load coils need to be added to the loops to
compensate their frequency response. DSL technology typically cannot operate through these load coils.
Introduction to Broadband Access Networks and Technologies 5
Both ADSL2plus and VDSL2 support transmission of packet transport mode (PTM), asynchronous
transport mode, and synchronous transport mode (STM). ITU-T G.997.1 specifies management parameters for ADSL2plus and VDSL2.
1.3.2 DSL System Technologies
The first generation of DSL equipment connected DSL modems at the subscriber premises to DSL access
multiplexers (DSLAMs) located in the central office. DSLAMs were next deployed in remote locations
that were often co-located with DLC RTs. If the DLC RT was served by a SONET fiber connection, the
DSLAM traffic would be multiplexed onto the same SONET signal as the DLC voice traffic. One of the
challenges of co-locating the DSLAM and RT is that the DSLAMs require much more power per line than
DLC equipment. This leads to heat dissipation issues when they shared the same cabinet, which can
restrict the number of DSL lines that can be served. The DSLAM is not connected to the RTs backup
batteries, however, since there is no requirement to maintain DSL service during a power outage.
DSL was developed at a time when Asynchronous Transfer Mode (ATM) appeared to the preferred
multiplexing technology for next generation networks. ATM provided adaptation techniques to carry a
wide variety of packet-oriented data and constant bit rate (CBR) traffic such as voice signals. Hence, ATM
was a natural choice for the encapsulation technology over the DSL line and for the multiplexing
technology within the DSLAM. ATM allowed some statistical multiplexing for more efficient bandwidth
utilization on the trunk from the remote DSLAM to the CO, or within the network.
There are two drawbacks to ATM, however. The first is that it adds at least five bytes of overhead to
each 53-byte cell, causing a roughly 10% bandwidth overhead penalty. The bandwidth penalty is
sometimes referred to as the ATM “cell tax.” The other drawback to ATM is that it typically uses a rather
complex signaling protocol that is overkill for purposes such as carrying connections to the Internet. Since
most of the data going over DSL systems uses the Internet Protocol (IP) for Layer 3, it makes sense to use
lower layer protocols that are more efficient with IP packets. Consequently, the emerging generation of
DSLAMs is IP-based and uses Ethernet for the Layer 2 protocol instead ATM. These are commonly
referred to as IP-DSLAMs.
8
1.4 Hybrid Fiber-Coaxial Cable (HFC)
While the telephone companies have focused on DSL, the CATV companies have deployed a network
that is optimized for broadband broadcast traffic. As the demand for internet connectivity increased and
the regulations allowed competition for providing telephone service, CATV companies have upgraded
their networks to allow upstream data transmission from their subscribers.
As illustrated in Figure 1.1, the CATV network uses a shared coaxial copper cable medium to connect
to its subscribers. The coaxial segments are connected to remote equipment that provide the conversions
to/from a fiber connection with a head-end office. These networks are called hybrid fiber coax (HFC). The
bandwidth of the shared coaxial cable is divided into frequency bands, with one or more frequency bands
being allocated for upstream transmission. Individual subscribers compete for the shared upstream
bandwidth through a medium access control (MAC) protocol.
The most popular protocol for providing voice and data access is the Data Over Cable Service Interface
Specification (DOCSIS
TM
) protocol developed by CableLabs, a laboratory that is jointly funded by
multiple cable network providers. The downstream data is modulated into the RF channel slots that would
otherwise have been used for carrying video signals. The upstream data is similarly transmitted using RF
8
The assumption is that if the power is out at the DSLAM location, the power is also out at the subscriber premises, and hence there is no
subscriber equipment operational to use the DSL line.
9
An early challenge for most CATV systems was that their signal repeaters only worked in the downstream direction, and required
upgrades to support upstream traffic. The deployment of fiber reduced the number of repeaters requiring upgrade, since the fiber systems
were designed to support bidirectional traffic.
9
6 Broadband Access
Local
Satellite
feed
Headend
Internet
Remote
Video
optical amplifiers
and transmitters
Secondary
Hubs
Video
Server
coax
amplifier
fiber
node (FN)
Master
Headend
Regional
Content
Servers
PSTN
Primary
Hubs
Core Network
Secondary
Hubs
Aggregation Networks
Distribution
Hub
Figure 1.1 CATV network illustration
modulation into dedicated upstream frequency slots. The DOCSIS protocol assigns the frequency bands
that are used by the cable modems at each customer’s premises, and uses a shared-medium MAC protocol
to determine the time slots in which cable modems can transmit their upstream data. The DOCSIS
protocol also addresses the security issues associated with having a transmission shared among multiple
subscribers where each can see the others’ signals. The DOCSIS protocol is described in detail in
Chapter 11.
Having been optimized for delivering video, the CATV networks are much better suited for delivering
broadband video, including high-definition TV (HDTV) than are the telephone networks. A coaxial
segment has typically been shared among several hundred subscribers. This degree of sharing inherently
limits a CATV network’s ability to provide high per-subscriber upstream bandwidth, and it also limits the
number of video-on-demand (VOD) channels that can be provided. Reducing the coax sharing obviously
increases their flexibility but also increases the cost of the CATV network. This is the primary tradeoff
faced by the CATV providers in offering broadband access. To win in the market place, the cost and
performance of the HFC networks must compete with the telephone company DSL networks and fiber to
the home/curb networks (FTTH/C). One advantage HFC has over FTTH systems is that the copper
coaxial cable allows a CATV company to provide power to the home telephone in the same manner that
the telephone company does today.
1.5 Power Line Communications (PLC)
The use of power lines as a communication medium has been around for at least 100 years. This
technology is generally referred to as Power Line Communications (PLC) and has sometimes enjoyed
some degree of success over the years depending on the application it was used for.
The attractive feature of PLC is the high penetration of electrical infrastructure in the world, which in
many areas is much higher than any other telecommunication infrastructure. As access to the internet
Introduction to Broadband Access Networks and Technologies 7
today is becoming as indispensable as access to electrical power, and since devices that access the internet
are normally plugged into an electrical outlet, the unification of these two networks always appeared to be
a compelling option, despite the various technical challenges. As virtually every line-powered device can
become the target of value-added services, PLC may be considered as the technological enabler of a
variety of future applications that would probably not be available otherwise.
Among the various applications, today’s interest for PLC spans several important applications:
broadband internet access; indoor wired LAN for residential and business premises; in-vehicle data
communications; smart grid applications (advanced metering and control, peak shaving, mains monitoring); and also municipal applications, such as traffic lights and lighting control and security.
In particular, smart grid applications have been and continue to be today a successful and promising
area for PLC. Similarly, the interest in using PLC for home networking is increasing rapidly and, despite
today’s low penetration, many believe that home networking will be one of the most important areas of
success for this technology. On the other hand, the great interest in the late 1990s for using PLC for
providing broadband access to households has encountered many disappointments over the last two
decades. Higher than anticipated costs in deploying PLC, growing EMC (Electro-Magnetic Compatibility) issues for the interference caused to radio services in the HF bands, its smaller capacity compared to
DSL and cable, and the availability of other (and often cheaper) means to provide broadband access to
consumers have made the initial enthusiasm in PLC for broadband access greatly diminish if not vanish.
There are very few PLC deployments in the world for broadband access and its use in industrialized
countries, where the availability of other broadband access technology is abundant and cheaper and has
made PLC a marginal technology. Perhaps the area where broadband access via PLC may still have some
possibility of success is in third world countries, where access to the internet is essential to economic
growth but there is no or very little telecom infrastructure. Similarly, rural areas in industrialized countries
where it is very uneconomical to provide broadband services at competitive prices could also benefit from
the deployment of PLC as most of these areas lack traditional telecom infrastructure but nevertheless have
access to power.
Despite its failure to become a successful technology for broadband access, PLC will be addressed in
Chapter 13. Because of its widespread use as a Smart Grid technology, the use of PLC in the power grid
will also be addressed and its unique benefits for this application will be highlighted.
1.6 Fiber in the Loop (FITL)
Telephone company revenue from plain old telephone service (POTS) is declining as the result of losing
some of their POTS customers to mobile phones and CATV companies. In order to increase their future
revenue potential, the telephone companies believe that they need to be able to offer the best “triple-play”
services, consisting of telephone, data (especially internet access), and video. At one time, telephone
companies considered the idea of deploying the same type of HFC networks used by CATV providers.
One drawback to this approach is that the coax networks typically have inferior reliability for voice
service. The other main drawback is that they would only be “me-too” for video and data, thus providing
no advantage over the CATV companies.
The model now preferred by telephone companies is based on their traditional approach of either
avoiding fully shared media or limiting the amount of sharing. For non-shared subscriber medium, tripleplay services typically will be provided through a fiber to the node (FTTN) architecture. With FTTN, as
shown in Figure 1.2, a high-speed fiber connection
enough to the subscriber to allow individual VDSL connectivity to each subscriber served by that node.
Variations on FTTN are Fiber to the Curb/Cabinet/Building (FTTC/FTTCab/FTTB). FTTN is very
attractive when many subscribers are close enough together to be reached by VDSL (e.g., dense housing
10
The high-speed fiber connection can either be a SONET/SDH link or a Gbit Ethernet link.
10
exists between the CO and a remote node that is close
8 Broadband Access
ONT
FTTN ONT
POTS
IF
xDSL IF
& Circuits
NI
Buff.
& TM
Buff. = Buffer
CO = Central Office
FTTC = Fiber to the Curb
ONT = Optical Network Terminal
IF = Interface
TM = Traffic Management
CO
Figure 1.2 FTTN network illustration
neighborhoods or multi-tenant buildings). When subscribers are spaced further apart or require very high
upstream bandwidth, fiber to the premises/home (FTTP/FTTH) becomes more attractive.
The passive optic network (PON) is the most attractive technology for FTTH/FTTP. PON systems
share the fiber medium among a limited number of subscribers. Due to the directional nature of fiber optic
transmission, only the downstream signals are visible to all subscribers on that PON. This simplifies the
encryption processes required to ensure privacy relative to those required for shared coaxial cable or
wireless networks.
Due to the relatively high cost of optical components (especially lasers and optical receivers), it is not
cost effective to give each subscriber a separate fiber connection to the CO. The best way to reduce the
number of optical components, as well as reducing the amount of fiber, is to have multiple subscribers
share the same passive fiber network for their connection to the optical line terminal (OLT)
11
in the CO.
The PON is illustrated in Figure 1.3. The terminal at the subscriber premises is typically called an optical
network unit (ONU) or optical network terminal (ONT). Different generations of PON technology allow
different numbers of ONUs to be connected to an individual PON, but 16 and 32 are typical numbers, with
some systems connecting up to 64 and future systems being capable of higher numbers. Since passive
optical splitters are used to divide (and merge) the optical signal among the ONTs, the number of ONTs
connected to a fiber is often called the split ratio (e.g., 32-to-1).
PON systems typically transmit both upstream and downstream data over the same fiber. In some cases,
only directional couplers are used to separate the upstream and downstream traffic, but higher speed
systems typically use different wavelengths in each direction. The most common is course wave division
multiplexing (CWDM), in which 1490 nm is used for the downstream direction and 1310 nm for the
upstream. This wavelength assignment has the advantage of putting the less expensive 1310 nm lasers at
the ONTs.
In the downstream direction, the OLT broadcasts the data for all ONUs. This downstream signal is
comprised of the downstream data for all the ONTs and synchronization information for the upstream
transmissions. The ONTs extract their downstream data based on either time slots or cell/packet address
information.
In the upstream direction, the ONUs need a medium access control (MAC) protocol to share the PON.
The most common MAC protocol is time domain multiple access (TDMA), which is similar to the
protocols used by broadcast television satellites. With TDMA, the nodes are granted time slots in which to
transmit their upstream data. In basic PON systems, each ONT is preassigned a fixed portion of the
upstream bandwidth, and transmits its data at the appropriate time. In order to achieve greater efficiency,
11
Another popular name for the OLT was host digital terminal (HDT). The OLT can either be located in the CO or at a remote (RT) site.
Introduction to Broadband Access Networks and Technologies 9
ONT
CO
FTTH ONT
POTS
IF
Ethernet
Circuits
PON
IF
Buff.
& TM
OLT
NE Control
OLU (PON IF)
OLU (PON IF)
OLU (PON IF)
Switch Fabric
ONT
FTTC ONT
POTS
IF
xDSL IF
& Circuits
PON
IF
Buff.
& TM
Buff. = Buffer
CO = Central Office
FTTC = Fiber to the Curb
FTTH = Fiber to the Home
OLT = Optical Line Terminal
OLU = Optical Line Unit
ONT = Optical Network Terminal
IF = Interface
NE = Network Element
TM = Traffic Management
Figure 1.3 Illustration of a PON
PON systems now typically allow dynamic bandwidth allocation (DBA) among the ONTs. With DBA,
each ONT uses part of its upstream transmission to inform the OLT of its bandwidth requirements.
For example, this information could be based its input queue fill level, including the levels for data in
different classes of service. The OLT evaluates the requests from the ONTs, and assigns the bandwidth for
the next upstream transmission frame. This bandwidth is typically communicated as a transmission start
time and either a stop time or transmission duration time within the upstream frame. These bandwidth
assignments are sent in the downstream transmission frame. The information used by the OLT in
determining the appropriate bandwidth allocations can include the service level agreements (SLAs)
associated with the ONT data flows. In some systems, the ONT is responsible for determining how to
accommodate the relative priorities of its transmit data within the granted upstream transmission slot. The
most popular TDMA PON protocols are described in detail in Chapters 3 and 4.
One alternative to TDMA is wavelength division multiple access (WDMA) in which each ONU has its
own upstream and downstream wavelength for communication with the OLT. In other words, the separate
wavelengths allow each ONU to have a point-to-point connection to the OLT over the shared PON fiber.
The main drawback to WDM is that each ONU needs a unique wavelength, which would be very hard to
administer if subscribers are allowed to buy their own ONUs. Tunable lasers would alleviate this problem,
but they are currently too expensive. Other frequency selective technologies are being researched and
developed for use at ONUs, but to date they have not been cost effective relative to TDMA technologies.
Another alternative is code division multiple access (CDMA). CDMA uses a spread spectrum approach
where the subscriber bit stream modulates a code sequence, essentially in the same manner as is used for
mobile phones. CDMA is very attractive since it can be implemented with entirely passive components at
the transmitter and receiver. A further advantage of CDMA is that each subscriber can use a different
native client interface. CDMA circuits, however, typically require optical amplifiers and precision
10 Broadband Access
receiver discriminator circuits to achieve the required signal to noise ratio. Other optical-domain medium
access methods are also possible, but WDMA appears to be the most likely long-term approach. These
optical domain technologies are described in Chapter 5.
There are also technology combinations that use a PON infrastructure in combination with a different
technology, such as carrying the radio-frequency modulated CATV signal over a PON. These hybrid PON
protocols and technologies are covered in Chapter 6.
1.7 Wireless Broadband Access
The mobile computing paradigm has seen phenomenal growth in the first decade of this century. Most
services traditionally accessed on desktop PCs or dedicated networked hardware are being augmented
with, or completely supplanted by, mobile access on tablets and smartphones. Mobile devices are clearly
being powered by wireless technologies. However, before delving into the different wireless technology
options, one must first establish a clear understanding of the role played by both wired and wireless
technologies in delivering broadband access to the untethered end user.
Let us take a simplistic view of wireless access technologies initially, and divide wireless technologies
into “long range” and “short range”. Long-range wireless links (such as those used by cellular
technologies) can serve users over a widely distributed geographical area, and can therefore be seen
as a true alternative to the wired access options introduced in the previous sections. On the other hand,
short-range wireless links only cover a small area such as a home or an office. Short-range wireless
technologies therefore need to be augmented by wired backhaul access technologies in order to provide a
complete solution for broadband access to the end user.
It is important to understand in this configuration that the “speed” of the broadband connection is
actually determined by the smaller of the access rates of the wireless portion and the backhaul portion. To
give a concrete example, a WiFi installation at a cafe may use the latest and fastest version of the standard,
providing several hundred megabits of throughput. However, the cafe may use a DSL backhaul
connection providing only a few tens of megabits of throughput, due to cost or availability considerations.
In this example, the end user experience will be limited by the backhaul speed. The converse is also
possible, where the wireless access speeds can limit the overall user experience, as we shall see.
Another possible distinction between wireless access technologies can be based on whether they
provide “fixed” access or “mobile” access. In the early 2000s, several vendors developed systems based
on a DOCSIS-like protocol for fixed access, where equipment would be installed at customers’ premises
and provide the long-range backhaul for Ethernet-based LANs. Early on, these systems were proprietary,
but the need for a common standard soon became apparent. This led to the development of the IEEE
802.16 standard to provide long-range, fixed wireless access under the title “Wireless Metropolitan Area
Networks” (Wireless-MAN).
However, fixed wireless systems, whether they were proprietary, or based on 802.16, were mainly
restricted to smaller deployments in low-density population centers where the cost of installing wired
access was seen as expensive for the corresponding revenue potential. Moreover, it also became apparent
that a single technology, developed for both fixed and mobile access, would result in a more robust
ecosystem with more applications and adoption potential. As a result, the 802.16 standard evolved to
provide mobile access, but due to the deployment of competing cellular technologies, 802.16-based
systems have not been able to gain any significant market share.
The two types of technologies that are most popular for providing broadband access today are the IEEE
802.11-based Wireless LAN (WLAN) standard, popularly known as WiFi, and the third and fourth
generations of cellular technology. WLANs use unlicensed spectrum with restrictions on the transmit
power and are, therefore, mostly used as a short-range technology. In addition, WiFi is expected to coexist
with other systems in an unregulated environment, so it has been designed to be able to coexist and
be robust in the presence of interference. Furthermore, the technology is designed to use a simple
architecture that is easy to configure and install. The basic topology consists of an access point providing
Introduction to Broadband Access Networks and Technologies 11
the broadband connectivity to several associated wireless clients that represent end users. It is this
combination of the use of unlicensed spectrum, ease of installation and interference-robustness that led to
the rapid adoption of this technology. It is ubiquitous today as the predominant access technology in lowmobility environments such as homes, offices, campuses, and other public spaces.
In contrast, cellular technologies took a very different evolution path to becoming an alternative for
broadband access. While WiFi was designed from the start to provide access to data networking services,
cellular technologies were initially designed with the sole goal of providing mobile voice service. The
requirement to support seamless mobility for voice via handovers resulted in a more complex and
expensive system. Furthermore, they were deployed in licensed spectrum to ensure that interference can
be managed in a regulated fashion, thereby ensuring high reliability for voice services. As such, these
systems are owned and operated by service providers, not by individuals or small enterprises. With
transmit power not severely restricted, as in the case of unlicensed spectrum, cellular systems can cover
much larger areas.
The basic topology of a cellular system is based on the concept of cells and frequency reuse. Early
systems were not designed to be robust to interference, and therefore needed to use frequency separation
and cells to manage interference, as shown in Figure 1.4. The figure shows a frequency reuse factor of
three, because a separate carrier is needed in each cell in the repeating cluster of three cells in order to
maintain a minimum interference separation. Other reuse factors are possible, with varying degrees of
interference separation. Early cellular technologies, in the first and second generations required frequency
reuse and supported mainly voice services. However, as the need for mobile data services grew with the
growth in internet traffic, combination of scarce licensed spectrum resources, and the greater capacity
needed for data services led to the design of reuse-one systems where all cells could use the same
frequency, and the interference mitigation was carried out by a more robust physical layer designed to
operate at lower signal-to-noise ratios.
In the wireless section of this book, comprising of Chapters 14–17, we take an in-depth look at the three
most widely deployed technologies for mobile broadband access. Before delving into the details of the
technologies, we first try to establish, in Chapter 14, the fundamental concepts that apply to all wireless
systems. In addition, the various basic building blocks that are part of the air-interface of any broadband
access technology are explained. Next, in Chapter 15, we discuss WiFi based on the IEEE 802.11
standard. Lastly, in Chapters 16 and 17, we discuss third and fourth generation cellular technologies. In
Chapter 16, we focus on the technology based on Wideband Code Division Multiple Access (W-CDMA),
and briefly discuss how it contrasts with the other third generation system based on CDMA-2000. In
Repeating
cluster
Frequency #1
Frequency #2
Frequency #3
Figure 1.4 Cellular system with frequency-reuse factor 3
12 Broadband Access
Chapter 17, we discuss the fourth generation systems based on LTE and LTE-Advanced that are expected
to be widely deployed, with a brief mention of WiMAX which is based on the IEEE 802.16.
All of these technologies are developed and implemented on the basis of specifications set forth by
standards development organizations. As such the evolution of these technologies can be tracked by
examining each new release of the specification in sequence. In discussing these technologies, we will first
discuss the baseline features, network and protocol architecture of the technology. Next we will see how
the technology evolved by taking a detailed look at each significant release of the standard and pointing
out the key features and capabilities that were introduced in that release. Each chapter concludes with a
summary that condenses all the material covered in the chapter into a few paragraphs to provide a quick
review of the most noteworthy elements of the technology.
1.8 Direct Point-to-Point Connections
While direct point-to-point connections are not cost effective for residential subscribers, they will
continue to be used for large corporate subscribers. Copper wireline connections can be DS1, E1, DS3,
or Ethernet. Fiber connections include SONET/SDH, 1G, 10G, or 40G Ethernet, dark fiber, or a
WDM wavelength. Wireless point-to-point connections are typically microwave radio links. The
primary advantages of these direct connects are guaranteed bandwidth and security (since there is no
shared medium).
While direct fiber connections are often not available to enterprise subscribers, DS1/E1 connection
availability is ubiquitous. In North America, the regulatory environment can also create a price advantage
for services providers to lease DS1/DS3 connections rather than fiber connections through the local
exchange carrier networks. With the addition of virtual concatenation support for DS1/E1/DS3/E3
signals, copper connections through the traditional telecommunications infrastructure have become much
more flexible. GFP then provides the transparent mapping for packet data services (see PMC-Sierra white
paper PMC2041096). Previously, providing copper connections between the DS1 and DS3 rates required
fractional DS3 or some relatively inflexible or inefficient method of combining DS1 s. These methods
included inverse multiplexing with ATM (IMA), packet-specific techniques such as the IETF Multi-Link
Point-to-Point Protocol (ML-PPP), or proprietary solutions.
1GE and 10GE fi ber connections are becoming increasingly important as the UNI to enterprise
subscribers. The telecommunications network provider may, in turn, use WDM for increased
utilization, or map this data into its SONET or OTN infrastructure where TDM multiplexing allows
even greater fiber utilization.
Appendix 1.A: Voiceband Modems
Voiceband modems began by using dual-tone frequency-shift key modulation for rates of 300 bit/s. As
technology advanced, it became practical to us phase-shift key modulation and combinations of the
amplitude modulation and phase modulation such as quadrature amplitude modulation (QAM) for
greater efficiency. The capacity of any information channel is determined by the Shannon channel
capacity theorem:
C Blog
The capacity limits on the data rates for voiceband modems are primarily determined by the analogto-digital conversion that takes place when the modem signal from the subscriber reaches the telephone
network equipment (DLC or central office switch). Specifically, the 8 kHz sampling rate, and the
quantization noise introduced when converting a voiceband signal to a 64 kbit/s digital signal determine
the channel bandwidth (B) and the noise (N) terms of the Shannon capacity equation. The modem signal
power (S) is limited by both the dynamic range of the analog-to-digital conversion, regulation, and the
need to avoid crosstalk into other subscriber loops in the cable. The resulting capacity limit (C) for a
1 S=N
2
Introduction to Broadband Access Networks and Technologies 13
voiceband modem is approximately 34 kbit/s, considering data transmission over voiceband channel with
additive white Gaussian noise and assuming a nominal bandwidth of about 3.5 kHz and a signal-to-noise
ratio of about 30 dB. Using efficient modulation techniques and error correction technologies, such as
trellis coding, allowed standard voiceband modems to approach this limit with 33.6 kbit/s.
However, the value of 33.6 kbit/s was still far from the theoretically possible DS0 data rate of 64 kbit/s
that could have been achieved with the same bandwidth but higher signal-to-noise ratio. The 64 kbit/s
maximum value depended on the use of a 8ksample/s sampling rate and of 8 bits/sample in the analog-todigital conversion.
In some circumstances, modems are indeed capable of approaching the theoretical maximum of 64 kbit/
s if certain conditions of low quantization noise are met, for example, when a subscriber is connected via
an analog line to a switched digital network and thus only one analog-to-digital conversion takes place. In
some cases, the source of the data sent to a subscriber has a digital connection to the network (e.g., a DS1/
T1 link) rather than a modem connection. Examples of such data sources include internet service
providers. The digital-to-analog converter connecting to the subscriber loop creates a downstream signal
that has none of the quantization noise that would have been created by an analog-to-digital conversion. If
the other noise sources affecting that subscriber loop are small enough, then the channel capacity of the
loop can be approached. In these circumstances, and sill using a sampling rate of 8ksample/s but encoding
with data only seven bits of the 8-bit word in the analog-to-digital conversion,
12
then modems can achieve
56 kbit/s downstream rates. Since the upstream signal from the subscriber must go through the telephone
network equipment’s analog-to-digital conversion, the upstream signal rate of these modems is still
limited to the standard 28.8 or 36.6 kbit/s rates.
12
To improve the probability of error, only 128 PCM values are used.
2
Introduction to Fiber Optic Broadband Access Networks and Technologies
2.1 Introduction
Telephone companies and community access television (CATV) providers (also called “cable” providers)
are competing to offer subscribers the triple play services of voice, video, and high-speed data access.
Historically, both telephone and CATV networks have relied on copper cables to connect through the last
mile to their subscribers, but a coaxial cable of the CATV companies has superior bandwidth capabilities
relative to the twisted pair wiring from telephone companies. However, the coaxial cable must be shared
by many subscribers in order to be economical. Clearly, the most flexible and future-proof medium is fiber
optic, with its virtually unlimited bandwidth availability. For telephone network providers, fiber
connections are attractive as the key to leapfrogging the capabilities of CATV providers. In response,
CATV providers are also beginning to deploy all-fiber networks for enterprise customers and are
considering it for residential customers.
Because providing a direct optical connection between the telephone company central office (CO)
each subscriber is cost prohibitive in terms of cost, most optical access systems share a passive optical
network (PON) among multiple subscribers. PON standards and technology are the focus of this section
of the book. The section begins with a brief history of fiber optics in access systems, including early PON
systems. This chapter also includes discussions of general PON topics and technologies that are largely
independent of the specific PON protocol. These topics include an introduction to PONs, technology
challenges, system powering issues, and protection for survivability. The remaining chapters of this
section cover the different major families of PON protocols.
There have been two standards bodies developing protocols. Chapter 3 covers the IEEE PON protocols,
which are the IEEE 802.3ah Ethernet PON (EPON) and 802.3av 10Gigibit/s EPON (10G EPON)
standards. Chapter 4 covers the protocols developed by the Full Service Access Network (FSAN)
consortium in conjunction with the International Telecommunications Union – Telecommunications
1
and
1
Telephone network terminology is typically used in this chapter, since the telephone companies have driven much of the PON
standards development and have deploye d the majority of the PON networks. The equivalent cable provider terminology is used when
appropriate.
Broadband Access: Wireline and Wireless – Alternatives for Internet Services, First Edition.
Steven Gorshe, Arvind Raghavan, Thomas Starr and Stefano Galli.
2014 John Wiley & Sons, Ltd. Published 2014 by John Wiley & Sons, Ltd.
16 Broadband Access
Standards Sector (ITU-T). The FSAN consortium defines the requirements for these standards, which are
then fully developed and published by the ITU-T. The FSAN/ITU-T standards include the G.983 series
Broadband PON (B-PON), G.984 series Gigabit PON (G-PON), and the emerging G.987 XG-PON
standards. FSAN and ITU-T are also beginning work on next generation (NG-PON) protocols that go
beyond XG-PON.
Both the current IEEE and FSAN/ITU-T protocols are primarily based on time division multiple access
(TDMA) for their medium access control (MAC) layer. Chapter 5 covers protocols that primarily used
optical domain MAC techniques. These protocols include wave division multiple access (WDMA) and
optical code division multiple access (CDMA) techniques, and some frequency-domain multiplexing
techniques.
2.2 A Brief History of Fiber in the Loop (FITL)
FITL began with connecting remote terminals of a digital loop carrier (DLC) to the central office (CO)
with a fiber instead of T1 lines
1980s, as the telephone companies gained experience with Integrated Service Digital Network (ISDN)
wideband services to subscribers. Rapid advances in the technology of optical transmitters, receivers and
fibers made FTTH appear to be potentially just over the horizon. However, cost and powering issues with
FTTH led to various fiber to the curb or cabinet (FTTC, FTTCab) systems as an alternative. The following
discussion considers both the FTTH and FTTC/Cab technologies.
The first generation of FTTH systems attempted to replace the copper line (loop) directly with fiber. An
optical network terminal (ONT)
company side of the fiber was terminated on a line card in an optical line terminal (OLT) or a traditional
DLC. The topology with such OLTs is called an active star, since there is an active fiber transceiver in the
telephone company CO for each fiber radiating out to the subscriber. When DLCs were used, the topology
was called an active double star since fibers from the CO connected to multiple DLC remote terminals
(RTs) which, in turn, had active fiber transceivers for the connections to the subscribers. Most large
equipment manufacturers built prototype or field trial versions of this type of system, for example, [1].
Typical bandwidth over the fiber to the subscriber on these systems ranged between the ISDN Basic Rate
(160 kbit/s) to a DS1 or E1 signal.
Passive optical networks (PONs) were also explored as a way to reduce the cost per subscriber by
reducing the number of optical transceivers and fibers. As illustrated in Figure 2.1, a PON system uses a
single optical transceiver at the OLT to serve multiple subscribers over a fiber tree constructed with
passive optical signal splitters. The first serious trial PON system was developed and deployed by British
Telecom [2]. First generation PON systems were also developed by both major equipment vendors and
startup companies. NTT wrote its own standard for such PON systems for deployment in Japan. The
evolution of PON systems is discussed in the next section.
Another approach to reducing the cost of FITL systems was to serve multiple subscribers from the same
ONT. These systems were commonly called fiber to the curb (FTTC) systems, with 4–12 subscribers
typically served from the same ONT. FTTC provided three major cost benefits. First, it reduced the
2
. The first serious interest in fiber to the home (FTTH) began in the late
3
was installed at (or near) the subscriber’s premises. The telephone
4
2
Since the telephone connection to a subscriber is a pair of wires, the two-wire connection is often referred to as a “loop” or “subscriber
loop.”
3
The EPON standard uses the term Optical Network Unit (ONU) exclusively to refer to the optical terminal closest to the subscriber.
The GPON standards regard the ONU to be the more general term, with ONT referring to an ONU that serves a single user. From a PON
protocol standpoint, there is no difference between ONUs serving single and multiple users. Consequently, the GPON standards use the
terms OLU and ONT interchangeably.
4
In the US regulatory environment of that time, the incumbent carriers were restricted regarding what data or video services they could
offer. Consequently, the state public utility commissions required carriers to justify the cost of any new access technology on the basis of
its cost for basic telephone service.
Introduction to Fiber Optic Broadband Access Networks and Technologies 17
ONT
CO
FTTH ONT
POTS
IF
Ethernet
Circuits
PON
IF
Buff.
& TM
OLT
NE Control
OLU (PON IF)
OLU (PON IF)
OLU (PON IF)
Switch Fabric
ONT
FTTC ONT
POTS
IF
xDSL IF
& Circuits
PON
IF
Buff.
& TM
Buff. = Buffer
CO = Central Office
FTTC = Fiber to the Curb
FTTH = Fiber to the Home
OLT = Optical Line Terminal
OLU = Optical Line Unit
ONT = Optical Network Terminal
IF = Interface
NE = Network Element
TM = Traffic Management
Figure 2.1 PON Network Example
number of optical components relative to FTTH (FTTC can be deployed as either an active star/double
star or with a PON. Active stars were initially more common since serving multiple subscribers per ONT
would quickly exhaust the capacity of early PON systems). The second cost advantage of FTTC is that it
preserved the copper line connections from the “curb” over the last 1000 feet or so to the home. Installing
fibers on this final subscriber drop is very expensive and opportunities for sharing costs on this portion are
minimal. Since the short line length allows a high-speed DSL connection, most of the same services could
potentially be delivered. The third advantage of FTTC was that it simplified the means for the network
provider to continue to provide the power for the subscriber’s phone (see Section 2.4.2 and the Appendix
for further discussion of subscriber power issues).
Variations on the FTTC theme include fiber to the cabinet (FTTCab), where the cabinet serves more
subscribers than a typical curb unit, and fiber to the premises (FTTP), where the premises are a multitenant building. FTTN (Fiber to the Node) has become a popular term for FTTC/P/Cab systems in the US.
FTTC, while VDSL is the current preferred plan for some European and US carriers.
The cost-effectiveness of FTTC/FTTCab/FTTP systems depends on numerous factors, including:
the relative cost of the number of ONTs served per OLT optical transceiver;
•
the fiber and its installation cost;
•
the cost of the DSL transceivers at the ONT and subscriber premise;
•
the overall cost of powering the ONT;
•
the real estate cost of placing the ONT.
•
FTTC/FTTCab/FTTP systems are clearly less flexible for high bandwidth services than FTTH systems,
since there is much more network equipment impactwhenever the subscriber wants a different service rate.
18 Broadband Access
The first deployments of commercial PON systems targeted business customers. This initial application
had a relatively small market, since it was uncommon to have a cluster of business customers wanting
access bandwidth greater than DS1/E1 who were all reachable by the same PON. The circumstance that
greatly accelerated PON deployment, beginning in the early 2000s, was the demand for high-speed
internet access by residential customers. The cost of FTTH PON systems was still cost-prohibitive, so
telephone companies relied on DSL technology to reach residential subscribers.
DSL rates were a good fit for the residential subscriber applications of the late 1990s and early 2000s.
However, in order to make DSL ubiquitously available, the telephone companies needed to remove the T1
signals from the copper bundles due to spectral compatibility issues with DSL. As a result, PON became
the most attractive option for serving business customers, freeing the copper cables for residential DSL
service. In recent years, the cost of PON technology has decreased and the bandwidth requirements of
residential customer applications have increased to the point where PON has finally become attractive for
FTTH applications.
2.3 Introduction to PON Systems
2.3.1 PON System Overview
As discussed above and illustrated in Figure 2.1, a PON system consists of a passive fiber tree/bus network
that connects multiple ONTs to a single OLT optical transceiver. For FTTH, the ONT is at the customer’s
premises, either mounted on an outside wall by the network interface or inside the house. The FTTH ONT
optionally provides the POTS (“plain old telephone service”) interface to the subscriber and an Ethernet
interface for data services. For non-FTTH ONTs (e.g., FTTC), the ONT provides the final copper drop to
the subscriber. This interface will typically use a DSL technology for the data service, with analog POTS
sharing the copper drop in its native frequency range. With FTTC, the line lengths are typically less than
1000 feet, making VDSL very practical for video delivery. For G-PON, a mechanism exists to map the
POTS channel directly into the G-PON frame. Alternatively, VoIP can be used to eliminate the need to
carry the POTS signal separately from the data signal. VoIP is the method to provide the voice service
with EPON systems.
As shown in Figure 2.1, the OLT consists of a number of PON interface units, a switch fabric for the
data services (and potentially a simple switch fabric or multiplexer for the voice channels), and a NE
controller. The ONTs are ultimately also managed by the OLT NE controller, which is responsible for all
ONT provisioning and OAM&P reporting. The OLT and ONTs together form the PON system, enabling
it to function logically as a single NE. In some ways, the fiber interconnection can be thought of as an
extended backplane.
The OLT transmits the data for all ONTs in the downstream direction. This downstream signal
comprises the downstream subscriber data for all the ONTs, the overhead for OAM&P, and the
synchronization information for the upstream transmissions. The ONTs extract their downstream data
based on either time slots, cell/packet frame address information, or wavelength.
In the upstream direction, the ONTs need a medium access control (MAC) protocol to share the PON.
The most common MAC protocol is time domain multiple access (TDMA), which is similar to the
protocols used by broadcast television satellites. With TDMA, the ONTs are each granted a time slot in
which to transmit their upstream data. In basic PON systems, each ONT is pre-assigned a fixed portion of
the upstream bandwidth and transmits its data at the appropriate time. More typically, in order to improve
the PON bandwidth efficiency, the OLT dynamically grants the ONTs time slots in which to transmit their
upstream data, and it communicates these bandwidth grants within the downstream signal.
As noted in Section 2.4.1, a guardband time is required between the upstream burst transmissions of the
ONTs so that their transmissions do not overlap at the Optical Line Unit (OLU) receiver in the OLT. The
ONT signals propagate through the fiber at the speed of light divided by the index of refraction of the fiber
(approximately 2 10
trip delay for just the fiber would be 200 μsec, which corresponds to 200 kbits for 1 Gbit/s and 2 Mbits for
8
m/s). With 20 km of fiber, which is a common maximum PON length, the round
Introduction to Fiber Optic Broadband Access Networks and Technologies 19
10 Gbit/s line rates. Thus, to minimize the length of the guardband time, the relative fiber length from each
ONT to the OLU should be taken into account. Beginning with second generation PON systems, most
have a ranging protocol to measure this delay so that the ONT burst times can be assigned to allow a
minimum guardband time between bursts arriving at the OLT.
As noted, PON systems now commonly allow dynamic bandwidth allocation (DBA) among the ONTs
in order to achieve greater upstream bandwidth efficiency. With DBA, each ONT communicates its
bandwidth requirements to the OLT. For example, this information could be based on its input queue fill
level, including the levels for data in different classes of service. The OLT evaluates the requests from the
ONTs and assigns the bandwidth for the next upstream transmission frame. The information used by the
OLT in determining the appropriate bandwidth allocations can include the service level agreements
(SLAs) pertaining to the data flows associated with the ONTs. These bandwidth assignments are sent in
the overhead of the downstream transmission frame, and they are typically communicated as a
transmission start and stop time within the upstream frame. In some systems, the ONT is responsible
for determining how to accommodate the relative priorities of its transmit data within the granted upstream
transmission slot. Allowing the ONT to decide how to fill its upstream bandwidth allocation allows it to
minimize latency for higher priority traffic that has arrived at the ONT since it made its bandwidth request.
It also distributes some of the processing load between the OLT and the ONTs.
In contrast to TDMA PON systems, wavelength domain multiple access (WDMA) systems use
separate wavelengths to create virtual point-to-point connections between the OLT and each ONU.
Optical filters within the PON are used to connect the appropriate wavelength(s) to each ONU. Each of
these connections operates at the desired rate without the need for dynamic bandwidth assignment. While
the bandwidth allocation is simpler with WDMA, the additional WDM optical components have made
these systems more expensive than TDMA systems. This point will be discussed further in Chapter 5.
PON systems typically transmit both upstream and downstream data over the same fiber. In some cases,
the same wavelength is used for both directions with only directional couplers to separate the upstream
and downstream traffic. Higher speed systems typically use different wavelengths in each direction. The
most common implementation is coarse wave division multiplexing (CWDM), in which the 1550 nm
region is used for the downstream direction and the 1310 nm region for the upstream. This wavelength
assignment has the advantage of putting the less expensive 1310 nm lasers at the ONTs.
Note that some PON systems use 1490 nm for the downstream PON signal, with analog video
optionally overlaid and transmitted downstream at 1550 nm. Using WDM for video overlay provides a
simple upgrade to existing deployments and increases the downstream capacity. Using a separate
wavelength for analog video transmission also avoids such problems as lack of digital content and
regulations involving digital content.
2.3.2 PON Protocol Evolution
As noted above, the first generation of PONs was based on TDM signals such as DS1/E1 signals. The
downstream frame was a TDM frame, where each ONT’s data was placed into time slots reserved for that
ONT. With any TDMA protocol, the data transmitted upstream must be broken up into blocks that can be
transmitted in bursts. These first generation PONs collected the data from their upstream TDM time slots
and transmitted them at a higher rate during their assigned upstream burst time slot. For voice signals, this
corresponded to a number of voice samples. For packet data, it was simply the number of bytes of the
packet that would have been transmitted during that frame in a corresponding point-to-point TDM signal.
The second generation of PONs was based on Asynchronous Transfer Mode (ATM), which provided a
convenient protocol for chopping the upstream data into blocks for the upstream transmission bursts.
ATM supplied the mechanism for carrying TDM traffic, fragmenting large packets, and assisting QoS
support. Also at this time, ATM was regarded as the likely basis for next generation networks and was
already being used for broadband access in DSL systems. The upstream burst time slot allocated by the
OLT to the ONT was simply the number of ATM cells it was allowed to send in that burst. NTT specified
20 Broadband Access
such an ATM-based PON (APON) system for use in their network. APON was also chosen by the FullService Access Network (FSAN) consortium, which publishes its standards through the ITU-T. The ITUT G.983 series covers APON systems that are commonly referred to as Broadband PON (B-PON)
systems. B-PON is described in Chapter 4.
With IP packets comprising more of the subscriber data, and Ethernet providing the typical connection
to the CPE, it made sense to avoid ATM adaptation and to use packet technology (e.g., IP) for the packet
routing. Consequently, the third generation of PON systems are based on, or optimized for, carrying
Ethernet frames. The two primary high-speed third generation PON standards are Ethernet PON (EPON)
from the IEEE (802.3ah) and Gigabit PON (G-PON) from the ITU-T (G.984 series). With EPON
(described in detail in Chapter 3) the upstream transmission is a burst of one or more Ethernet frames.
G-PON, which was developed after EPON, is much more flexible. As described in Chapter 4, a G-PON
upstream burst can contain whole Ethernet frames, fragmented Ethernet frames, or TDM traffic bursts.
Third generation systems thus avoid the protocol complexities and the bandwidth overhead associated
with ATM adaptation.
The emerging generation of PON protocols uses a combination of higher speed TDMA and wave
division multiplexing (WDM). With the 10 Gbit/s IEEE 10G EPON, the WDM is primarily used to allow
EPON and 10G EPON ONUs to share the same physical PON fiber infrastructure. The 10G EPON OLT
communicates with both types of ONUs. As discussed in Chapter 4, FSAN/ITU took the same approach
with XG-PON in developing its next generation of PON protocols.
Future generations of PON protocols are expected to make more use of WDM to increase the PON
capacity in terms both of the number of subscribers per PON and of the bandwidth per subscriber. As
discussed in Chapter 5, the current optical technology does not make pure WDM PONs cost-competitive
with TDMA PONs, and this situation is expected to persist for many years to come. Until WDM PON
components become more cost-effective, high-capacity PONs will continue to use a mix of TDMA
and WDM.
Some carriers envision PON technology as the means not only to provide new broadband services, but
also as a means to reduce the cost of their metro and access networks. BT was the first carrier to announce
such a vision as part of the CN21
most subscribers
5
and DLC technology extended the reach, reducing the need for new COs. PON offers
TM
plan. Central offices were originally built within less than 4–6kmof
the potential to serve subscribers from fewer COs located much further away from the subscribers. To be
economically feasible, the PON reach would need up to about 60 km, and each PON would need to
connect to around 1000 subscribers. One carrier estimates that such a PON-based network would allow
them to reduce the number of COs by a factor of ten. The financial gain from selling the COs that are no
longer needed, combined with the greatly reduced network operating expenses, could easily justify the
cost of building the new PON network.
Achieving the longer reaches, especially with higher split ratios, will likely require either optical
amplifiers or repeaters on the PON. The ITU-T has defined the framework for this type of reach extension
for G-PON systems [3]. An experimental “Super-PON” research system is described in Chapter 5 as one
potential architecture for achieving the 1000 ONU capacity over 100 km of fiber.
Achieving 1000 subscribers per PON will require a mix of WDM and higher-speed TDMA. One of
the advantages to WDM is that WDM optical splitters introduce much less loss than other passive
optical splitters. A symmetric passive optical power splitter sends half the optical energy from all
wavelengths down each of the two legs of the splitter, hence introducing about 3 dB of optical loss.
Since this power split occurs for every 1 : 2 splitter, the ten stages of splitting (2
10
) needed to serve 1000
subscribers would have 30 dB of optical loss. In contrast, a WDM splitter separates (preferentially
directs) the energy of the different wavelengths into different legs of the WDM splitter with very little
loss for each wavelength.
5
Depending on the carrier, in North America this distance was typically limited to either 12,000 ft or 18,000 ft.
Introduction to Fiber Optic Broadband Access Networks and Technologies 21
2.4 FITL Technology Considerations
2.4.1 Optical Components
FITL systems have always faced challenges related to the cost of the optical components that provided the
desired capabilities.
generation of FTTx systems uses a combination of lasers in the 1260–1360 nm range for upstream and in
the 1490–1590 nm range for downstream.
Fused optical power splitters may be produced relatively inexpensively. They are constructed by fusing
fibers together so that their core areas are close to each other for some distance, with the fused area
becoming an optical signal mixing region. As an optical signal passes through the mixing region, a portion
of its light couples into the other fibers, going in the same direction. The fraction of the light that couples
into each of the other fibers is determined by the construction of the splitter.
A common implementation for a 2 2 splitter is to have half the optical power coupled into the other
fiber so that same optical power appears on each of the two output ports. N N splitters may also be
constructed, with the power on each output fiber now being 1/N of the input power ( 10log(N). In the
case of a tree structured PON network, one of the ports on the OLT side of the splitter is not used (recall
that the splitter is still a 2 2 device, so one still loses half the power to the ‘unused’ port). The result is that
while the device is constructed as a 2 2, it is normally used as a 1 2 splitter, in which half of the
downstream signal is transmitted over each output branch toward the ONUs. In the upstream direction
through the same splitter, half of the upstream optical signal is coupled into the fiber towards the OLT.
Consequently, the optical loss through the 1 2 splitter is at least 3 dB in each direction. In practice, power
splitters introduce a fraction of a dB additional loss. Note that effectively the same type of splitters can be
constructed with planar-integrated optics using close-proximity wave guides instead of fused fibers for the
mixing region. The loss vs. number of ports behavior remains unchanged, however.
Wavelength filters are a primary component for WDM PON, but they are also used with TDMA PON
systems to separate different signals (e.g., the downstream data signal and broadcast video signal). A
wavelength filter routes input signals to different output ports based on their wavelength. Since there is no
power splitting for any given wavelength, the filter introduces a smaller amount of insertion loss. The
wavelength routing can be accomplished either by refraction (e.g., with a prism) or by using interference
between multiple beams of light. An example of the latter is the Arrayed Waveguide Grating discussed in
Chapter 5.
With TDMA-based PON systems, the various issues surrounding switching the upstream transmissions
between the different ONUs are a challenge. One set of such issues involves the guard time that is required
between transmission bursts of different ONUs. The factors that determine the guard time are often
independent of the upstream data, so these factors have an increasing impact as the upstream data rate
increases, since the burst transmission rate must increase more rapidly in order to achieve the desired
6
Single mode fiber has been chosen for its higher bandwidth capabilities. The current
6
First-generation FITL systems identified a number of technology challenges. Although multi-mode fiber was less expensive for the
fiber drop to the subscriber, single-mode fiber was preferred, due to its superior bandwidth capabilities. Single-mode fiber, however,
requires laser transmitters, since the core diameter is too small to couple adequate optical energy from incoherent light sources such as
LED transmitters. There was initially some thought of using the inexpensive lasers used in CD players and CD-ROM drives, but the
problem with these lasers is that their wavelengths (750 nm and 810 nm are typical) still propagate as multi-mode in glass single-mode
fibers. The least expensive lasers that allowed single-mode transmission used 1310 nm, and these lasers cost several hundred dollars in
the 1990 time frame. Fiber transceivers are still not cost-effective for direct point-to-point connections to each subscriber as of this
writing. The development of fused fiber splitters dramatically reduced their cost, making PON more attractive, but any type of passive
splitter divides the optical energy between the branches of the splitter. As the number of splitters between the OLT and the ONT
increases, the power decreases quickly. For this reason, split ratios of ONTs per OLT transceiver are limited to between 16 to 1 and 64 to
1. Higher split ratios would require optical amplifiers. Since the cost of the optical amplifier can be shared among multiple subscribers,
this can still be cost-effective. The other drawback of higher split ratios is that the bandwidth is shared among more subscribers, limiting
the bandwidth that any one user can send. For this reason, split ratios higher than 64 to 1 will probably be used primarily on higher rate
PON systems (e.g., future 10 Gbit/s PON systems).
22 Broadband Access
overall data rate. One such factor had been the speed at which a light source can be turned on and turned
off to the point where it has negligible output power, so that the collection of ONUs on the PON do not add
too much interference to the signal from the ONU authorized to transmit. This is typically no longer a
significant issue, since the current manufacturing techniques have reduced this time to a few nanoseconds.
Another factor in the guard times is the time required for the OLT to adjust its receiver to the difference
in received power levels between ONUs that are at significantly different differences from the OLT. This
adaptation to the different bursts includes the time required for OLT to achieve clock and data recovery
synchronization. The OLT upstream receivers become even more challenging when different ONUs on
the PON have different upstream data rates. For example, as discussed in Chapter 3, 10G EPON allows a
mix of ONUs with some transmitting upstream at 1 Gbit/s and others at 10 Gbit/s. This requires the OLT
burst mode receiver not only to adjust to the received power level, but also to adjust its received signal
equalization and clock recovery circuits for the data rate of each burst.
The technology choices for dealing with mixed rate burst mode receivers can impact the effective
optical link budget. For example, splitting the different rate signals in the optical domain introduces the
3 dB penalty of the optical splitter, while some circuits for handling the different signal rates in the
electrical domain result in reduced receiver sensitivity.
As will be discussed further in Chapter 5, WDMA introduces additional challenges. From an operations
perspective, it is not practical to build PONs with each ONU having a different fixed wavelength laser.
Carriers would need to keep the optical modules for each ONU wavelength in their inventory and track all
the different ONUs on each PON to ensure that none used the same wavelength. Building a cost-effective
“colorless” ONU while maintaining adequate PON reach is a significant challenge.
Optical amplifiers can be used to overcome the optical loss associated with greater reach, higher split
ratios, or other impairments. Reach extension techniques, including optical amplifiers, are discussed in
Section 2.4.4.
2.4.2 Powering the Loop
The single biggest obstacle to FITL systems in many countries, however, has been not the cost of the
optical components, but the challenge of providing reliable power to the subscriber telephone. As noted in
the FTTC description, telephone companies have traditionally supplied power to the subscribers’
telephones through a –48 Vdc power feed. The network-provided power virtually guarantees that
telephone service remains available during the loss of utility service power. This high service availability
is often referred to as ‘lifeline POTS’ since subscribers can count on the service being available for
emergencies. A telephone company typically provides batteries to back up their equipment for about eight
hours of typical usage in the event of a power outage from the power utility company.
uses batteries in the CO, and DLC RTs – which are powered by a local connection to the utility company –
have batteries co-located at the RT. In the event that a major disaster (e.g., a flood or a hurricane) disrupts
the power utility for more than eight hours, generators are used at the CO to provide the power needed for
charging their batteries. The RT batteries can be charged by portable generators on trucks that are driven
among the RT sites.
FTTH and FTTC complicate the power situation for several reasons. First, there are many more
locations with active components (e.g., potentially 100–2000 more ONTs than DLC RTs to reach that
many subscribers). Clearly, it is not possible to send trucks with portable generators to each ONT during
prolonged power outages. More significantly, batteries have life spans of 5–10 years and thus must be
replaced regularly, and replacing batteries is very costly and manpower-intensive. If the ONT is located
within the subscriber premises, the telephone company must obtain permission to enter the home to install
the replacement.
7
While this lifeline POTS capability is mandated in the US by the FCC, other countries (e.g., Japan) do not require it.
7
The CO equipment
Introduction to Fiber Optic Broadband Access Networks and Technologies 23
A second factor is the number of connections to the power utility company. DLC RTs each have their
own metered power utility connection. FTTC can also use per-ONT metered power utility connections,
but the cost per subscriber increases. One alternative for FTTC is to use a separate ‘power pedestal’ that
powers multiple ONTs and contains their back-up batteries. For FTTH, it is cost-prohibitive to provide
per-ONT metered power. Power pedestals are one option, although more pedestals are required. The other
option is to have the subscribers provide their own power, with the FTTH service cost effectively
discounted to cover the subscribers’ expense. More details of the power requirements and restrictions are
provided in Appendix 2.A.
Supplying and backing up the power has been attacked on a number of fronts. Better batteries (e.g.,
longer life span, higher efficiency over the temperature range, higher capacity per unit size and cost)
would mitigate many of the problems. Battery research (also motivated by the automotive industry) has
brought some improvements, but has not solved the problems for FTTH/C. Solar power has been explored
for the FTTH/C and power pedestals, but it is geographically limited to regions that have adequate reliable
sunshine. More power-efficient electronic components have been developed, but the power consumption
of the electronics was always a relatively small percentage of the overall power.
So, while power might seem like a mundane topic in the glamorous world of lasers, fiber, and highspeed data access, it is a substantial problem with no easy solution. A more recent phenomenon may
have turned the tables, however. With the rapidly increasing use of cellular/mobile telephones, subscribers
have become accustomed to providing their own power and maintaining their own batteries for their
phone service. This situation may lead customers who desire broadband services over FTTH to be very
willing to take responsibility for the power and battery back-up of their ONTs. Also, it is highly likely that
an FTTH subscriber is already a mobile phone subscriber and can use that phone for lifeline service
instead of the landline phone connection. This phenomenon, combined with the growing desire for
broadband services, has opened the door for residential PON deployment.
2.4.3 System Power Savings
Another important extension to existing and future PON protocols is power savings. In one study, BT
estimated that without some type of power savings capability, the additional power required for moving
from current DSL to an ONU at each home would be equal to the output of a typical British power utility
plant. An important driver for ONU power savings during normal operation is the European Union’s Code
of Conduct on Energy Consumption of Broadband Equipment (BBCoC). While BBCoC compliance is
voluntary, the European carriers plan to work toward it. Another critical driver for ONU power
consumption is the North American requirement, discussed above, for the ONU to continue to provide
emergency POTS service for up to eight hours after the loss of utility power.
While power savings during normal operation and during battery backup operation are related, the
criteria for battery operation are somewhat different. Battery backup operation implies that the subscriber
has lost utility power and hence will typically be able only to use telephone service and not require video
or data services.
an active telephone connection. The more the power requirements during battery operation can be
reduced, the smaller the battery the ONU requires.
While improvements in technology can reduce the power consumption of the ONU components, the
expected increase in bandwidth and services will tend to increase the overall power requirements for the
ONU. Achieving more substantial power savings will require shutting down those portions of the ONU
that are not currently active. The ITU-T has been studying the different approaches to ONU power savings
and has created a supplement document on the subject [4]. It groups the ONU power savings techniques
into the three categories of ONU power shedding, dozing, and sleeping.
8
With the increased use of laptop computers, subscribers may begin to request premium battery-backed services that allow a period of
continued data service during power outages. In these instances, the router or gateway device would also require battery backup.
8
This situation allows the ONU to activate only those circuits associated with supporting
24 Broadband Access
ONU power shedding: The most universal approach to ONU power savings is to use system-on-chip
design techniques that idle portions of the chip such as a microcontroller when they are not in use.
This type of ONU power savings is referred to as “power shedding.” The subscriber interfaces and
their associated circuits can also be shut down as long as there is a sure way to determine whether the
subscriber is actively using them. For the telephone service, this is straightforward if the ONT provides
the interface to the subscriber’s telephones. For data services, however, including VoIP, it can be difficult
to determine whether the interface is truly inactive or only experiencing a pause in the subscriber’s use
of a service.
ONU dozing: The ONU upstream interface circuits consume a significant portion of the ONU’s power.
Consequently, additional power can be saved if the ONU ceases its upstream transmissions when the
ONU is not proving active services to the subscriber. This type of power savings approach requires
coordination with the OLT so that it will not decide that the ONU has been removed from the PON. “ONU
dozing” refers to the technique of shutting down the ONU’s upstream interface during inactive periods,
but always keeping the downstream PON interface awake. The characteristics of ONU dozing can be
summarized as:
The ONU continues to listen to the downstream PON signal.
•
The OLT continues to send small upstream bandwidth allocations to the dozing ONU so that it can
•
quickly signal a request for upstream bandwidth when the subscriber initiates a service. The ONU
ignores these bandwidth grants if it has no information to send.
The ONU wakes up and sends upstream information when either a subscriber service requiring
•
upstream bandwidth becomes active or the OLT requests a response from the ONU. The requested
responses are typically OAM&P-related reports.
An advantage to dozing is that it allows an ONU to respond almost immediately when a subscriber
service becomes active. The ONU uses ongoing OLT upstream bandwidth grants to request appropriate
upstream bandwidth quickly for the newly activated service. Another advantage is that it can work
concurrently with services such as video that are only active in the downstream direction. A disadvantage
to the dozing is that some downstream and upstream bandwidth is wasted due to the ongoing unused
allocations to the dozing ONUs.
ONU sleeping: The power savings methods that save the most power are those that put all non-essential
ONU functions into a sleep or standby state for periods of time when it is not carrying active traffic. The
essential functions include timers and monitoring for the activation of subscriber services. Sleeping
requires additional coordination between the ONU and OLT. For example, if another subscriber initiates a
service connection to that ONU (e.g., dials that subscriber’s telephone), then the OLT must buffer the
downstream information until the ONU wakes up. Also, the OLT must schedule its OAM&P message
exchanges with the ONUs so that they are awake to receive the messages. The characteristics of ONU
sleeping can be characterized as:
The OLT sends a periodic broadcast message to initiate the sleep cycle for all inactive ONUs on that
•
PON.
ONUs wake up after a pre-determined time when a local ONU timer expires. All ONUs have the same
•
sleep time period, which results in all ONUs on the PON using the sleeping function go to sleep and to
wake up at effectively the same time.
The OLT buffers any new downstream service or OAM&P information for a sleeping ONU until it
•
enters the awake portion of the sleep cycle.
When the ONU detects a newly active service, it wakes up to buffer the information and sends the
•
appropriate upstream information after its wake-up timer expires.
When it is time for the sleeping ONUs to wake up, the OLT sends them upstream bandwidth grants so
•
that they can make any necessary bandwidth requests for services that have become active since the
ONU last entered its sleep state.
Introduction to Fiber Optic Broadband Access Networks and Technologies 25
When the ONUs are awake, the OLT sends them any information that it had buffered during their sleep
•
period.
Synchronizing the sleep times of all the ONUs allows the OLT to maintain a single sleep cycle state
table rather than per-ONU state tables.
The main advantage to ONU sleeping is that it provides the most power savings. Also, there is no
downstream and upstream bandwidth lost due to the type of periodic bandwidth grants that the OLT must
make to a dozing ONU. The main disadvantage to ONU sleeping is that there can be some lag in activating
a new service with a sleeping ONU. It also requires some additional buffering at the OLT and ONU and
additional OLT state tracking.
ONU sleeping can be further divided into “deep sleep” and “fast sleep” modes. The fast sleep mode
operates in the manner described above. The OLT creates periodic sleep cycles, which are periods of
activity followed by sleep periods. The periodic initiation of a sleeping period allows inactive ONUs to
enter a sleep state soon after determining that they are inactive. The OLT can continue to send upstream
allocations to sleeping ONUs in order either to recover from a potential mismatch between the ONU and
OLT sleep state views or, potentially, to allow an ONU to awake early due to it detecting new subscriber
activity.
A deep sleep period can be initiated by the ONU, for example in response to the customer powering
down the ONU. It is critical that the ONU inform the OLT so that it does not generate alarms due to the
subsequent loss of upstream information from that ONU. The original deep sleep specification had the
ONU inform the OLT of its entry into deep sleep by sending a “dying gasp” message. It is also possible to
use a handshake to communicate entry into the deep sleep mode. The OLT will periodically send upstream
allocations to a deep-sleeping ONU, so that it will be able to communicate its exit from deep sleep.
2.4.4 PON Reach Extension
There are two general approaches to PON reach extension. One approach is to remain in the optical
domain, using optical amplifiers, while the other is to use some type of repeater that operates in the
electrical domain. There are tradeoffs to each method. Both methods are described in ITU-T Recommendation G.984.6 [3].
Since both the amplifier and repeater are active electronic elements that are subject to failure and have
a view of the network status that is not visible to the OLT, it is important for the OLT to have a
management communications link with them. G.984.6 refers to this function as an EONT (Embedded
Optical Network Termination for management of the extender). The EONT function essentially behaves
as another ONT on the PON except that its only bandwidth requirements are for the management
communications.
2.4.4.1 Optical Domain Reach Extension
The typical target distance for PON reach extension has been 60 km, which corresponds to the logical
delay limit supported by the GPON protocol for a purely passive optical distribution network (ODN)
(see Chapter 4). Increasing the fiber from the typical 20 km limit of the current generation of PON
protocols to 60 km adds ≈14.5 dB additional loss at 1310 nm and ≈12 dB at 1550 nm.
The primary advantage of optical domain techniques such as optical amplifiers is that they are virtually
agnostic to the bit rates and formats of the signals they carry. This flexibility allows the potential to
upgrade PON systems to carry future protocols with no changes to the ODN. However, there are also
challenges in using optical amplifiers in PON systems. One general challenge is that optical amplifiers
inherently generate noise in the form of amplified spontaneous emission (ASE). If the signal-to-noiseratio (SNR) of the optical signal becomes too low prior to amplification, the ASE can significantly reduce
the signal’s SNR. It has been demonstrated that 10 Gbit/s transmission over 100 km with 1024-way splits
is achievable using a cascade of optical amplifiers along the fiber [5].
26 Broadband Access
1310nm
(PDFA / SOA-O)
ONU
Analog video
(1555nm)
ONU
ONU
1490nm
(TDFA / SOA-S)
a) Discrete amplifiers per wavelength band
ONU
ONU
ONU
b) Raman amplifier for PON upstream and downstream
1555nm
(EDFA / SOA-C)
WDM Combiner
US Raman Pump
DS Raman Pump
Optical
Repeater
Figure 2.2 Illustration of reach extension with optical amplifiers
OLT
Diplexer
OLT
Since the TDMA PON systems typically use different wavelength bands for downstream, upstream,
and broadcast video signals, as illustrated in Figure 2.2a, multiple different types of amplifiers would be
required for each PON. The different wavelength bands and corresponding optical amplifier technologies
are illustrated in Figure 2.3. The EPON and G-PON protocols would require separate amplifier types for
the upstream signal (1310 nm, O-band) and downstream signal (1490 nm, S-band), and an additional
amplifier type if broadcast video signals are being transmitted over the same PON (1555 nm, C-band). The
EDFA (erbium doped fiber amplifier) amplifiers commonly used in telecommunications applications only
cover the 1555 nm region. Special amplifiers are required for the O and S-bands.
The PDFA (praseodymium doped fiber amplifier) and TDFA (thulium doped fiber amplifier)
technologies are similar to EDFA. One advantage to EDFA-type amplifiers is that they produce low
noise and high gain. All doped fiber amplifiers operate by coupling an optical pump signal into a doped
section of fiber, through which the data signal also passes. The pump signal wavelength excites the doped
fiber, raising its energy level. Light of the data signal’s wavelength causes the doped fiber to release
its energy, which results in amplification of the data signal. The reliability of TDFA and PDFA has not
been proven.
Another challenge for fiber doped amplifiers in the upstream direction is the “bursty” nature of the
upstream signal, with gaps between the bursts. Also, the burst power level depends on the distance to the
Introduction to Fiber Optic Broadband Access Networks and Technologies 27
0.28
0.26
0.24
0.22
Loss (dB/km)
0.20
0.18
O-band:
PDFA,
SOA
S+
Loss curve for longer
wavelengths
14601310
Wavelength (nm)
S-band:
TDFA,
SOA
SOA = semiconductor optical amplifierEDFA = erbium doped fiber amplifier
TDFA = thulium doped fiber amplifierPDFA = praseodymium doped fiber amplifier
C-band:
EDFA,
SOA
L-band:
Gain-shifted
EDFA,
SOA
1600155015001450
Figure 2.3 Illustration of wavelength bands and the associated optical amplifier technology
ONU and the wavelength varies slightly between ONUs. Input signal power variations occurring in
timeframes of 10
Also, if a burst arrives at the EDFA after a relatively long period with no upstream activity, its gain would
be too low and could take up to 10
4
to 103seconds can cause the EDFA gain to react such that the burst can be distorted.
3
seconds to reach the correct gain. Techniques exist to compensate for
this, but they add cost and complexity.
An alternative optical amplifier technology is the semi-conductor optical amplifier (SOA), which uses a
semiconductor for the gain region rather than a fiber section, and is electrically pumped rather than
optically pumped. These are constructed similarly to Fabry-Perot lasers, except that they have nonreflective cavity endfaces that prevent them from operating as a laser.
The advantages to SOA amplifiers include being much smaller, using less power, and being able to be
manufactured relatively easily for different wavelength ranges. SOAs are capable of operating in the
0.85–1.6 nm range and they also have faster gain dynamics than doped fiber amplifiers, making it easier
for them to handle bursty upstream traffic.
The disadvantages of SOA are that they are typically 2–3 dB noisier than doped fiber amplifiers and
have more limited gain (currently <13 dBm). The gain reacts rapidly to changes in the data signal power
or the electrical pump power, and the gain changes cause phase changes that can distort the signal. This
nonlinearity is the greatest performance challenge in this application. Like TDFA and PDFA, SOA have
unproven field reliability. However, published data from vendors suggest that SOAs should have
appropriate reliability. It is reasonable to expect that if there is sufficient demand, optical amplifiers
at the typical PON wavelength bands can reach a deployable state. Another drawback to SOAs, though,
remains their high cost.
An alternative optical amplifier arrangement makes use of Raman amplifiers [3]. This arrangement is
illustrated in Figure 2.2b. The Raman pump output at the appropriate wavelength is coupled into the ODN
to give reverse-pumped Raman gain for the upstream signal. For the downstream direction, either a
separate optical amplifier can be used, or a Raman pump at a different wavelength can be used to give
forward-pumped distributed Raman gain for the downstream signal. The downstream signals and Raman
pump wavelengths are combined with a WDM combiner, which also separates the upstream signals. ASE
noise can also be removed by the WDM combiner when it is used as an optical bandpass filter. The generic
parameters for Raman amplifiers can be found in [6]. Some typical specific parameters for PON systems
28 Broadband Access
use a Raman pump laser wavelength of 1240(±0.5)nm, an upstream passband of 1300–1320 nm, and a
downstream passband of 1480–1500 nm (Amendment 2 of G.984.6).
Amplifiers need to be located on the shared portion of the fiber, since placing amplifiers at the ONUs
would be cost prohibitive. In order to keep the PON truly passive between the CO and subscriber, the
amplifiers would need to be co-located with the OLT in the CO. However, having the amplifier at the OLT
results in more noise power at the receiver than if it is located at the splitter. In the upstream direction,
locating the amplifier at the OLT means it can only function essentially as a pre-amplifier, where it would
not be as effective as if it were a transmit signal amplifier. The downstream amplification is limited by the
physics of the fiber. When the optical signal power becomes too high, non-linear transmission effects
occur in the fiber. There are also safety concerns about craftspeople working with systems with that
intensity of light, since the Class M1 safety limit is 21.3 dBm.
Another potential reach extension technology is using WDM filters instead of power splitters, in order
to reduce the optical loss due to the splitting. For example, a 1 : 32 split with power splitting introduces
18.4 dB of loss vs. 3.5 dB loss through a 1 : 32 wavelength multiplexer. This 15 dB difference corresponds
to an additional ≈40 km reach. While this technology may be attractive for future new ODN deployment,
it has two drawbacks: the first is that it is not compatible with current ODNs that are built with power
splitters; while the second is that WDM PON requires colorless ONUs, which is currently a technical
challenge. See Chapters 4 and 5 for additional discussion of WDM.
2.4.4.2 OEO (Optical-to-Electrical-to-Optical) PON Repeater Reach Extension
Converting the optical signals back into the electrical domain for regeneration avoids many of the cost and
technology issues of all-optical reach extension. The primary drawback, however, is the lack of upgrade
flexibility since the repeater must be implemented to operate at the bit rates of the PON. OEO repeaters
also do not help with the analog video overlay signals, which would better suited for optical-domain
amplification. Basic OEO repeater-based reach extension is illustrated in Figure 2.4. Note that the OLT
provides the reference clock for the ONUs and the repeater.
A further complication of OEO repeater reach extension is handling the burst-mode nature of the
upstream signal in TDMA protocols. For the basic repeater of Figure 2.4, the added complexity is
primarily the need for a burst mode receiver similar to that at the OLT. It needs to be able to recover the
clock and data for each ONU upstream burst, including adapting to the different signal levels that can exist
for each due to their different distances from the repeater. A mechanism is required to report these signal
levels to the OLT, so that the OLT can set the ONU transmit signal levels such that they are roughly the
same when they arrive at the repeater. This received signal level reporting could be handled through the
EONT function.
Since some of the upstream burst preamble may be effectively lost as the repeater attempts to achieve
clock and data recovery for the burst, it may be necessary to use a longer preamble with repeater-type
reach extension so that an adequate preamble remains when the signal arrives at the OLT.
The upstream repeater may also insert some bit pattern between the bursts rather than transmitting
nothing (i.e., a steady string of 0 s). This pattern is typically under the control of the OLT, and it is chosen
ONU
Rx Tx
ONU
RxTx
clock
Figure 2.4 Basic OEO repeater PON reach extension illustration
OLTONU
Introduction to Fiber Optic Broadband Access Networks and Technologies 29
ONU
ONU
OEO
OEO &
burst ->
CBR
remote
interface
OTN
Terminal
OTN link
OTN
Terminal
local
interface
OLTONU
Figure 2.5 Illustration of an OTN link used for reach extension
to contain a balance of 0 s and 1 s and not resemble the preamble that begins a burst. The OLT can
determine the arrival of a burst by looking for the preamble. Also, since the OLT schedules the upstream
transmission explicitly, it knows when the burst is supposed to arrive.
Another variation on the repeater-based reach extension is to implement the repeater in a distributed
manner by using a transmission link. The primary example of this is the use of an OTN transport link, as
specified in G.984.6 for GPON and illustrated in Figure 2.5. Although G.984.6 only covers G-PON, the
same approach could be adapted for Ethernet PON systems. The important differences between a basic
repeater and a transmission link such as OTN include:
The upstream signal must be converted to a continuous bit rate (CBR) signal by the remote burst mode
•
receiver in order for it to be mapped into the transport channel protocol.
Since transport channels are typically defined to have symmetrical bandwidth, the upstream burst to
•
CBR conversion process may need to encode the signal to adapt it to the same effective bandwidth as
the downstream signal.
The transport transmission channel is typically a two-fiber interface rather than a single fiber.
•
The transport signal can potentially be carried over portions of a carrier’s metro transport network,
•
sharing it with other signals in a consistent manner.
The transmission link has its own OAM&P. In the case of OTN, this includes the capability to manage
•
different wavelengths.
The coding conversion of the upstream CBR signal to match the downstream signal need not be strictly
necessary if the burst-to-CBR converter is integrated into the transport node and if the upstream rate is
an integer divisor of the downstream rate. Under these conditions, the upstream signal can effectively
be over-sampled to create the upstream signal within the transport system. This condition would hold for
G-PON, where the upstream rate of 1.24 416 Gbit/s is exactly half the 2.48 832 Gbit/s downstream rate.
However, when a separate burst-to-CBR converter is used, the upstream interface labeled as “remote
interface” in Figure 2.5 must have sufficient transition density for the OTN terminal input to lock onto the
signal. Consequently, the Manchester code was chosen for the upstream CBR signal to provide a reliable
input signal to the OTN terminal and to conveniently exactly double the bit rate of the upstream signal to
match the downstream signal.
The use of OTN could be particularly advantageous for carriers who plan to use PON to reduce the
number of COs.
9
The Manchester line code, used with several other interfaces, including some 10BASE and 100BASE Ethernet, uses 01 to encode a
data 0 and 10 to encode a data 1.
10
BT, in their CN21TMplan, was the first carrier to announce a plan to use PON for a combination of higher bandwidth per subscriber
and reducing the number of manned COs in maintains. Other carriers have also expressed interest in this approach.
10
The OLTs can be pulled further into the network to a smaller number of COs, with their
9
30 Broadband Access
metro OTN networks carrying the PON signals to a repeater location closer to the subscribers. In addition
to allowing use of their metro networks, this approach has two distinct advantages. The first is that OTN
provides a TDM mechanism for combining multiple PON signals onto a single wavelength. For example,
four G-PON signals can be multiplexed onto a single 10 Gbit/s OTN link (ODU2). The second is that
OTN is well suited for the future increasing use of WDM in PON networks since it was designed for
efficient management of WDM networks.
2.4.4.3 Reach Extension through a Remote OLT
Although it is not reach extension is the strict sense of extending the ODN, the same effect can be achieved
by moving (or keeping) the OLT closer to the subscriber. Using a remote OLT allows the ODN to be kept
short enough to support the desired number of ONUs. While this may, at first, seem to be contrary to the
desire for a passive outside plant, all of the reach extension technologies discussed above rely on some
type of active electronics. The remote OLT is not substantially different than using a repeater in this sense,
although the OLT circuitry is somewhat more complex. For example, the OLT could replace the burst-toCBR converter in Figure 2.5, with the signal carried over the OTN being the OLT’s network uplink.
Another potential variation on the remote OLT approach is to use cascaded PONs. For example, a
10 Gbit/s PON with its OLT in a CO could serve multiple remote OLTs that are attached to its 10 Gbit/s
OLUs. These remote OLTs then connect to the subscriber ONUs, using 1 or 2.5 Gbit/s PONs.
2.5 Introduction to PON Network Protection
Telephone network providers and cable television providers both employ degrees of redundancy in their
networks in order to protect their services against failures. Network redundancy strategies are chosen as a
function of the cost of providing the redundancy relative to the amount of traffic, the number of
subscribers, and the types of services affected by the potential failures. PON technology enables
delivering an increasing number of services over the access portion of the network, with increasing
bandwidth. PON is also increasingly used for connections to enterprise customers and wireless base
stations. Consequently, the deployment of PONs is expected to increase the need and desire for resiliency
in the access network. This section examines a reasonably comprehensive range of redundancy and
protection options for different portions of the PON network. Some of the options are not expected to see
substantial deployment, but they are presented here for completeness.
There are several different parts of the PON system to consider for protection. These parts and the
extent of their impact are summarized in Table 2.1.
Table 2.1 PON system components and the impact of a fault on the component
PON System component Sub-unit Extent of the fault’s impact
PON facility Fiber and passive components The feeder fiber between the OLT and
Active components (e.g.,
extenders such as amplifiers, if used)
OLT Network connection (uplink) All subscribers served by the OLT
Optical Line Units (OLUs) that
connect to the PON
Common units/functions Potentially all subscribers served by the OLT
Entire OLT/CO All subscribers served by the OLT
ONU Optical interface The subscriber(s) connected to the ONU
Other ONU functions Potentially the subscriber(s) connected to the
splitters affects all subscribers on that
PON. The drop fiber between the final
splitter and an ONU affects just that ONU.
All subscribers on the PON(s) connected to
that OLU
ONU
Introduction to Fiber Optic Broadband Access Networks and Technologies 31
After a brief background on protection, the remainder of this section is organized to discuss
each of these aspects of PON protection outlined in Table 2.1. There are multiple options for protecting
the different components. This section begins with the options that are the most complex and provide
the highest reliability. The discussions move progressively to the least complex options that
provide the lowest network reliability.
2.5.1 Background on Network Protection
The level of reliability in the telephone network has traditionally been a function of the number of
subscribers affected by a single fault, the level of service to which the customer subscribes, and the
expected failure rates of the network and equipment. For example, for economic reasons, no protection
was typically provided against a fault that affected 48 or fewer residential telephone customers. Since
enterprise customers typically pay a premium price for high reliability service, equipment and, in some
cases, route redundancy is provided on their connections to protect against single faults.
Within the telephone network, constraints are placed on the time to detect a network fault and to restore
the service through the redundant equipment and/or facilities. The maximum fault detection time is
typically 10 ms and the subsequent maximum restoration time is typically 50 ms. These times, which are
driven by legacy network equipment, insure that no telephone connections will be dropped due to the
fault. The telephone network providers expect packet-based networks to offer similar high-speed
protection in order to maintain traditional service quality when voice is carried as VoIP.
The ITU-T Study Group 15 (SG15) defined facility and optical interface protection architectures in Rec.
G.983.5 [7]. The ITU-T documented further PON protection considerations in the informative supplement
G.Sup51 [8]. The discussions in this section will refer to the G.983.5 and G.Sup51 architectures whenever
appropriate. Note that G.Sup51 covers failure rate and availability calculations for the protection architecturesthat it describes. TheIEEE SIEPONstandard described inChapter 3specifies whichof theseprotection
modes should be used with EPON. Additional information on PON protection can be found in [9,10].
2.5.2 PON Facility Protection
The PON facility includes the fiber and any passive optical components such as splitters, filters and
connectors. In the event that active optical amplifiers or repeaters are required to achieve the desired fiber
split ratios and/or distance, the amplifiers are also considered as part of the PON facility. A fault on the
feeder portion of the PON facility (between the OLT and the first splitter) affects all the subscribers
connected to that PON, while faults on the final drop portion (between the last splitter and the subscriber)
affect only that subscriber.
Note that most small and medium businesses (SMB) do not currently have redundancy on their physical
access links to the carrier network. A typical SMB user network interface (UNI) is a single fiber or copper
connection. Redundancy is typically used only for the portion of the access network shared by several
subscribers, such as using rings to protect the interconnection between access network nodes and the CO.
Furthermore, labor costs typically dominate the ONU installation. Fortunately, some of these labor costs
could be shared for a redundant PON, and hence they may not be doubled to enable protection. This type
of study would need to be performed by the carrier based on their own practices and cost structures for
pulling fiber cables, terminating fibers, etc. The cost of protection installation must also take into account
the potential additional revenue that could be derived by offering protection as a premium service.
2.5.2.1 Option 1 – Connect Each ONU to Two PONs, with Each of the Two PONs Connected
to a Separate OLT
This option, illustrated in Figure 2.6, provides the highest reliability since it protects against failures of the
entire optical path including the OLT, PON facility, and the ONU optical interface. The OLTs in this
option may even be located in separate COs in order to protect against a failure affecting one of the COs.
32 Broadband Access
Control communications
between the OLTs to
coordinate the switch over
ONUj
OI
OI
ONUk
OI
OI
ONUi
Router
Network
Access/Metro
NI
(W)
NI (P)
NI
(W)
NI (P)
W
P
OLT1
W
P
OLT2
OLU
OLU
OLU
OLU
OLU
OLU
OLU
OLU
Figure 2.6 Illustration of protecting an ONU and the optical connection with two separate PONs
and OLTs
This option can enable protection at either Layer 1 or Layer 2. Layer 1 protection would have only one
PON active at a time (i.e., the working PON). Layer 2 protection would have both PONs active with the
protection switching performed at Layer 2 outside the PON, and the potential need for communication
between the two OLTs.
Two ONU protection examples for this option are illustrated in Figure 2.6. ONU protection is discussed
in Section 2.5.4.
No OLU protection is required in this option, since it inherently locates the redundant OLU in a
different OLT. This approach is more expensive than using OLU protection within the same OLT if it is
used for all ONUs, but it can be a good option for an ONU carrying critical data. This option essentially
enhances the Type C protection architecture of G.983.5 [7] and G.Sup51 [8] by adding dual OLT
parenting. Fault recovery is complex, since it affects both the physical layer and the Layer 2 routing from
the OLTs (or subscriber router) into the metro network.
2.5.2.2 Option 2 – Connect Each ONU to Two PONs that are Both Connected to the Same OLT
This option is similar to the first, except that both PONs are connected to the same OLT (see Figure 2.7).
Assuming that the OLT contains protection for its common function and network interface units, and that
the two PONs have diverse fiber routing, this option provides virtually the same reliability as the first
option. The only fault it will not cover is the total failure of the OLT (e.g., due to a catastrophe that disables
the CO where the OLT is located). Here, the ONU is registered only on one PON at a time. G.983.5 refers
to this protection option as Type C protection architecture.
The recovery is much simpler with this option than with Option 1. Here, the protection only affects the
PON side of the OLT and has no Layer 2 effect on the network side of the OLT.
2.5.2.3 Option 3 – A Single PON with Redundant OLT Equipment and Feeder Fiber
A compromise PON protection architecture is illustrated in Figure 2.8. This architecture protects the
OLUs and the feeder portion of the PON between the OLUs and the distribution splitter, and is referred to
in G.Sup51 as Type B protection architecture. The OLT equipment redundancy can be either through
redundant OLUs in the same OLT, or through separate OLTs. G.Sup51 refers to the configuration with
Introduction to Fiber Optic Broadband Access Networks and Technologies 33
OI
OI
ONU
Network
Access/Metro
NI
(W)
NI (P)
W
P
OLT
OLU
OLU
OLU
OLU
Figure 2.7 Illustration of protecting an ONU with separate PONs and a common OLT
Control communications
between the OLTs to
coordinate the switch over
W
P
OLT1
W
P
OLT2
OLU
OLU
OLU
OLU
OLU
OLU
OLU
OLU
OI
ONU
2:N
Splitter
Network
Access/Metro
NI
(W)
NI (P)
NI
(W)
NI (P)
Figure 2.8 Illustration of protecting the OLTs and feeder portion of the PON (Type B protection
architecture)
separate OLTs as “dual-parented” Type B protection. A primary advantage of this protection architecture
is that it is less expensive than fully redundant PONs and yet, with the exception of the splitter, it protects
all of the elements that are shared by multiple ONUs.
11
This architecture also exploits the fact that the
optical loss through a 2 : N splitter is essentially the same as that through a 1 : N splitter.
Each ONU protected by dual-parenting is registered on both OLTs. This allows the protection OLT
quickly to replicate the logical state of the working OLT.
The proponents of this architecture have typically preferred to dual-home the feeder legs of the PON to
separate OLTs in separate locations, as illustrated in Figure 2.8.
The most difficult aspect of the dual-parented version of this protection option is the communication
that is required between the working and protection OLTs. Since the OLTs only see the upstream signals,
the protection OLT cannot determine directly the reason for a lack of upstream traffic. For example, there
can be periods with no upstream traffic when the working OLT opens a discovery window for new ONUs
11
If the carrier locates the splitters close to the subscribers, as is typical in many carrier networks, this architecture protects most of the
PON fiber.
34 Broadband Access
OI
OLU
OLU
W
OLU
OLU
P
Spare fiber
2:N
Splitter
ONU
Network
Access/Metro
NI
(W)
NI (P)
Figure 2.9 Illustration of protecting the feeder portion of the PON (Type A architecture)
to announce themselves. If the protection OLT begins to transmit while the working OLT is still
transmitting, it would create interference and potentially overwhelm or cause harm to the ONU receivers.
In the example of Figure 2.8, the fault status communication connection between the OLTs goes through
the access/metro Ethernet network. Another alternative is for the Network Management System (NMS) to
control the switchover. The NMS would use the alarm reports from the OLTs as the protection trigger and
would implement the switch by commands to the OLTs. Switch times are a potential issue for any
approach using separate OLTs, especially if they are controlled by the NMS.
Note that it is also possible to implement the Type B architecture by connecting the redundant feeder
fibers of multiple PON ODNs to a centralized optical switch that can connect one of these ODNs to an
OLT that provides shared protection for this group of ODNs. This is effectively the same configuration as
the 1 : N OLU protection illustrated in Figure 2.13, except that here it protects the entire OLT. While this
1:N configuration reduces the amount of OLT protection equipment, the optical switch offsets some of
the cost savings and brings significant additional control complexity.
Note that it is still possible to support ONU redundancy with this protection architecture by connecting
two of the PON’s drop fibers to the same subscriber.
Note also that there are additional considerations when the Type B protection is applied to protocols
that use dynamic wavelength or frequency assignment. Examples of such PON protocols include the NGPON2 protocol described in Chapter 4 and OFDMA protocols described in Chapter 5. The ONU
parameters are constrained by the need for them to remain constant during the switch.
2.5.2.4 Option 4 – Protecting Just the Feeder Portion of the PON
This architecture, illustrated in Figure 2.9, is similar to the previously described architecture, except that
the only the feeder fiber is protected. The OLT equipment is not protected, so therefore no additional OLT
port is required. This architecture is useful in applications where the feeder fiber is vulnerable to failures.
For example, it could be used if the feeder fiber is deployed in an aerial configuration, where it is
potentially vulnerable to falling branches or trees during storms.
With this configuration, referred to as a Type A protection architecture in G.Sup51, the protection can
either be performed by an optical switch on the line card, or through a manual intervention.
2.5.2.5 Option 5 – Use a Single Unprotected PON
While this approach provides no PON redundancy, the PON is highly reliable and should not typically
need protection.
12
The fiber and passive splitters have very low failure rates. Using optical amplifiers to increase the reach or split ratio of the PON could
potentially increase the failure rate enough to make redundant PONs more desirable.
12
A PON will typically only fail due to a fiber cut. The other redundant PON options are
Introduction to Fiber Optic Broadband Access Networks and Technologies 35
W
P
OLT
OLU
OLU
OLU
OLU
NI
(W)
NI (P)
Figure 2.10 Illustration OLT network interface protection architecture
only more reliable if the two PONs are each routed in physically separate cables so that a cable cut cannot
affect both PONs.
A variation on this option is to have redundant optical interfaces at the OLT that are connected to the
single PON through a passive splitter. G.983.5 refers to this option as a variation on the Type B protection
architecture discussed above. This option is covered in the section on OLT OLU protection.
2.5.3 OLT Function Protection
OLT protection includes the OLUs that connect to the PON, the network interface, and the common
functions such as systems clocks, management and switch fabrics. Each of these has multiple protection
options.
2.5.3.1 Network Interface (Uplink)
The OLT network interface (NI) connects it to the metro network or WAN. Each of the protection options
uses a pair of NI units, each connected to a separate bidirectional facility. EPON systems use Ethernet
links for NI. For G-PON systems, the NI can use either SONET/SDH or Ethernet for the physical layer. If
SONET/SDH interfaces are used, any of the SONET/SDH protection mechanisms are available to protect
the uplink. However, Ethernet interfaces are becoming more common for G-PON.
The Ethernet interface is typically GE, multiple GEs, or 10GE. Since this interface carries all the OLT
data, it is assume that it will be redundant, with redundant interface units, each connected to separate fiber
pairs. See Figure 2.10.
For Ethernet interfaces, the following options are possible:
13
1 :1 unit protection (bidirectional). Only one of the NI units is active at a time. If a failure is detected on
•
this unit or the facility to which it is connected, all traffic will be transferred to the other unit. The failure
is detected as the inability to communicate with the node at the other end of the link. The spanning tree
protocol can be used to select the active interface.
Ethernet Link Aggregation (LAG). With this option, both the working and protection NI units and their
•
associated physical links carry traffic under control of Ethernet LAG. If one of the interface units or
physical links fails, the network interface continues to function with half the bandwidth over the
remaining link and NI unit.
Note that Multiple Spanning tree (as a static method to control bandwidth between links) usually
replaces LAG on fast links.
IEEE 801.17 Resilient Packet Ring (RPR). RPR is protocol for a ring topology network that provides
•
fair access to the ring’s bandwidth. See [11] or [12] for a full tutorial on RPR. As the name implies,
13
When SONET/SDH is used for Layer 1, the payload of the SONET/SDH signal is an Ethernet packet stream. Therefore, the Ethernet
uplink protection methods discussed here are still applicable if they are used instead of the SONET/SDH layer protection mechanisms.
36 Broadband Access
RPR was designed to use the inherent route diversity of the ring topology in order to provide protection
for the traffic on the ring. Although RPR is independent of the Layer 2 protocol being carried, it is
optimized for Ethernet transport.
One of the primary advantages of RPR is that i ts fairness mechanisms can be used to guarantee the
QoS for the traffic on the ring. This is especially useful if the ring is used to backhaul the data from
multiple OLTs. In this application, the bridging capability of RPR allows logical connections for
direct data exchange between OLTs rather than performing the routing at a centralized location
further into the network. The main disadvantage to RPR is that it has not seen wide deployment
to date.
ITU-T G.8031 [13] Ethernet Linear Protection or G.8032 [14] Ethernet Ring Protection. G.8031
•
specifies a mechanism for fast protection of point-to-point VLAN-based Ethernet network links that is
modeled on traditional telecommunications network protection. G.8032 expands the protection
mechanism to include ring topologies. One of the motivations for G.8031 and G.8032 is to protect
VoIP connections fast enough to minimize interworking issues with traditional voice equipment in the
network. The principle behind this mechanism is that all nodes periodically transmit a continuity check
messages (CCM) to their neighbor node(s) and use a pre-determined protection path to route around a
failure. The CCM allows a fast detection of a failure, and the pre-determined protection path allows
immediate re-routing of the data without the need for running a spanning tree protocol. G.8031
supports both 1 1 and 1 : 1 protection architectures with both unidirectional and bidirectional
switching. Both revertive and non-revertive switching are also supported. The 1 : 1 architecture allows
the protection path to be used for preemptible traffic when no failures exist. G.8031 also supports
traditional manual protection switching operation.
IEEE 802.3ag Connectivity Fault Management. Similar to the CCM of G.8031, 802.3ag specifies
•
periodic messages to quickly detect connectivity faults in the network.
2.5.3.2 OLT Common Unit/Function Protection
Since a switch fabric failure would affect all traffic in the OLT, it should use 1 : 1 unit protection.
Depending on the implementation of the control plane interaction with the fabric, control plane
processing redundancy should also be considered.
2.5.3.3 Optical Line Unit (OLU) Protection
The OLU module typically contains the 1490/1550 nm laser and drivers, the optical receiver with its
support circuitry, clock and data recovery circuits, and the PON MAC functions. There are multiple
mechanisms available for protecting the OLUs. Since an OLU carries traffic for a limited number of
subscribers, OLU protection is much more cost-sensitive than NI protection. This section discusses some
of the potential OLU protection schemes. The order in which they are presented ranges from the most
robust and simplest to schemes that add complexity in order to reduce the overall system cost. Note that
OLU protection also provides a mechanism for upgrading an OLU without taking down all its PONs
during the upgrade. The OLU protection options are:
1 :1 OLU Protection. With 1 : 1 OLU protection, a protection OLU is dedicated to each working
•
OLU. Only one of the two redundant OLUs (the working OLU) is transmitting data. If it fails, the
other OLU (the protection OLU) transmits all the data. This option is referred to as Type B
protection in [7]. Note that when an OLU contains multiple PON interfaces (each to a different set
of ONUs), all PONs must be transferred to the redundant OLU in order to replace the OLU with
the failure.
Introduction to Fiber Optic Broadband Access Networks and Technologies 37
W
P
OLT
OLU
OLU
OLU
OLU
Optical switch
Passive splitter
NI
(W)
NI (P)
Figure 2.11 Illustration of 1 : 1 OLU protection implementations
The Type C protection illustrated in Figure 2.7, with separate working and protection PONs, assumes
1 :1 OLU protection.
14
Figure 2.11 illustrates two different implementations of 1 : 1 OLU protection
when both the working and protection OLUs are connected to the same PON. The tradeoffs between
these implementations are discussed below.
1 :1 OLU protection using a passive splitter. One option, illustrated in the upper portion of
Figure 2.11, is to use a passive splitter to connect the working and protection OLUs to the
PON. The advantages of this approach are that the passive splitter is extremely reliable, and no
optical switch control is required to transfer service between the working and protection OLU. The
primary drawback is that the passive splitter increases the optical loss by around 3 dB. If optical
splitters are used at both the OLT and ONU, FEC may be necessary; however, it is well within the
range for the EPON and G-PON standard FEC. See Section 2.4.1 for a discussion of the implications
of this additional optical loss.
1 :1 OLU protection using an optical switch. A second option, illustrated in the lower portion of
Figure 2.11, is to use a 1 2 optical switch to connect the two OLUs to the PON. Since the optical
switch is not dividing power between the branches, it only increases the optical loss by <1 dB. The
main drawback to using an optical switch is that a control mechanism is required in order to activate
the switch to connect the protection OLU to the PON.
1:N OLU protection. In order to reduce the number of OLUs, it is possible to use a single redundant
•
OLU to protect multiple working OLUs. The cost-effectiveness of this approach is determined by the
cost of the optical switch matrix and its associated control and mechanical packaging relative to the cost
of the additional OLUs and associated components for 1 : 1 OLU protection. The switch control is
complex relative to 1 : 1 OLU protection. Depending on the implementation, the optical switch may
need to be protected with a redundant switch unit. 1:N switches have been less common since there has
never been a good application to drive them. Some potential implementations and their tradeoffs are
discussed in this section.
Protecting N OLUs with an N N optical switch. OLU protection using an N N optical matrix is
illustrated in Figure 2.12 for N 8. The optical switch matrix functions as a bank of 1 8 switches
that connect either the appropriate working OLU or the protection OLU to each PON. While optical
switch fabrics are quite reliable, a redundant switch fabric may be required in order to achieve the
desired system reliability. If a redundant switch fabric is used, additional 1 2 optical switches are
14
With Type C PON protection, it would be possible for the protection OLU to also be transmitting data to the ONUs. For example,
Ethernet Link Aggregation could be used merge the bandwidth of the two PONs. In event of a PON or OLU failure, only one of the
OLUs and PONs would be available to carry data.
38 Broadband Access
OLU 1
OLU 2
OLT
OLU 3
OLU 7
OLU P
8 x 8
optical
switch
PONs
Figure 2.12 1 : N OLU protection with an N N optical switch fabric
required on either side of the switch fabrics in order to connect the online fabric to the OLUs and the
PONs.
Protecting N OLUs with a 1 N optical switch. An alternative approach for 1 :N protection is to use
a1N optical switch that is connected to the PONs and working OLUs with passive optical
splitters. This approach is illustrated in Figure 2.13. The advantage of using the passive coupling is
that a switch fabric failure should not affect normal operation of the working OLUs. The
disadvantage is the 3 dB additional splitter loss, as discussed in Section 2.4.1.
Protecting N OLUs with a M protection OLUs and an MxN optical switch. This option, illustrated
in Figure 2.14, is a variation on the 1 : N protection that allows protecting multiple OLUs
simultaneously. Consequently, its reliability and cost would be between that for 1 : 1 and 1 : N
protection if N is less than M times the number of OLUs that could reliably be protected with 1 : N
protection. However, the additional reliability inherent in an M : N protection architecture allows
using a larger number of working units per protection unit than could be used with 1 : N
protection.
General Discussion about OLU Protection. Using passive splitters in order to connect the working and
•
protection OLUs to the PON introduces a little over 3 dB additional optical link loss. This additional
optical loss may necessitate FEC. Both the EPON and G-PON standards specify a shortened ReedSolomon RS(255,239) FEC code that provides a coding gain of approximately 2.5 dB. This additional
gain will be adequate for most links if the passive splitters are used only on the OLT side of the PON. If
passive splitters are used for both the OLU and the optical interface at the ONU, then the RS(255,239)
will not be adequate to compensate for additional optical loss.
OLU 1
OLU 2
OLT
OLU 3
OLU 7
OLU P
Passive splitter
1 x 7
optical switch
PONs
Figure 2.13 1 : N OLU protection with a 1 N optical switch fabric and passive coupling
Introduction to Fiber Optic Broadband Access Networks and Technologies 39
OLU 1
OLU 2
OLT
OLU 3
PONs
OLU n
OLU P1
OLU Pm
Passive splitter
m x n
optical switch
Figure 2.14 M : N OLU protection with a M N optical switch fabric and passive coupling
There are multiple options for handling the additional loss if optical splitters are used at
both the OLT and ONUs. These options are summarized in Table 2.2, with their respective
pros and cons.
Using optical switches can reduce the loss to less than 1 dB. Although they are not as reliable as
passive splitters, optical switches are very reliable, and complex switch components have been
qualified for use in telecommunications networks for some time. Non-latching optical switches have
the advantage that they default to a known position when no power is applied to them. The optical
switch requirements are: (a) an integrated device so that it is simple, physically small, and cheap; (b)
low power (latching has zero power draw in normal operation); and (c) reliable.
The additional optical components for protection (splitters, switches, and/or amplifiers) can be
integrated into the OLT or located in a separate shelf. Using a separate shelf would allow the optical
shelf to serve multiple OLTs and would also simplify the fiber management at the OLT. For
applications using smaller OLTs, it could be more advantageous to integrate the additional optical
components into the OLT.
Unprotected OLUs: The final option is not to protect the OLUs. If the OLU failure rate is
•
low enough, and the subscribers can tolerate the OLU repair time, this is clearly the least expensive
option. Consequently, it is the most common option on currently deployed PONs serving residential
customers. For business customers, however, it may not always be an acceptable option.
2.5.3.4 Protection of the Whole OLT or CO
Failures of an entire OLT require a redundant OLT. Full OLT protection would typically be implemented
to protect against CO failures, in which case the two OLTs are located in different COs. The network
topologies for redundant OLTs are illustrated in Figure 2.6 and Figure 2.8. The first illustrates protection
of the entire PON, including the ONU optical interfaces. The second saves cost by protecting only the
feeder portion of the PON, up to the 2 : N splitter.
For the second approach, as discussed above in Section 2.5.2.3, some type of control channel is
required between the two OLTs so that only one is active at a time. In the case of CO failures, the
backup OLT will need to be able to detect the failure and to become active autonomously. The CO or
OLT failure could be detected by lack of communication (e.g., through periodic messages) from the
working OLT or by monitoring the downstream optical signal. Monitoring the downstream signal at
the backup OLT requires reflecting some of the downstream signal to it. It is difficult for the ONUs to
40 Broadband Access
Table 2.2 Options for accommodating additional optical splitter loss for protection
Option Pros Cons
Restrict the PON in
terms of distance and/
or split ratio
Employ a stronger FEC • Allow fulldistance andsplit
Use APD receivers • Increases the optical budget
Use optical amplifiers
(note)
Note: one option is to place the optical amplifiers on the PON side of the splitter that connects the working and
protection OLU to the PON. This option requires one set of amplifiers per PON, and is a good solution with 1 :1 OLU
protection. A more attractive option with 1 : N OLU protection is to place the optical amplifiers only at the input/output
of the protection OLU. This option requires only one amplifier set for every N PONs. Since passive optical splitters can
be manufactured with asymmetric loss when the power is split between the legs, the splitters here would provide a low
loss (<1 dB) coupling to the working OLU, with a higher loss coupling to the protection OLU. The optical amplifiers
compensate for the higher loss.
• Passive solution
• No impact on current OLU
and ONT equipment
ratio capability
• Works well with the RS
FEC
• This option has been chosen with some standard
protocols
• Allow full and potentially
increased distance and split
ratio capability
• No impact on current OLU
and ONT equipment
• Can increase cost by requiring additional
PONs and OLTs, and reducing the sharing of
equipment and facilities among fewer
subscribers
• Strong FEC is not standard
• Stronger FEC typically requires additional
overhead bandwidth, reducing the overall
PON capacity
• Adds expense to all ONTs
• Not backward compatible
• Expense of optical amplifiers and associated
optical components. A circulator is requiredto
separate the upstream anddownstream signals
for amplification.
• The different upstream and downstream
wavelengths necessitates different amplifier
technology for each direction.
report the failure, since they are only allowed to transmit when the OLT grants them upstream
bandwidth.
For either approach, the physical layer protection is implemented by the OLTs. Protecting the data
coming from the OLTs into the network can be performed at Layer 1 through a dual-homed ring.
Alternatively, the network connection protection can be performed at Layer 2, for example, through an
Ethernet protection mechanism.
2.5.4 ONU Protection
There are four main options for protecting the ONUs. These options are presented here in order of the most
expensive and complex to the least.
2.5.4.1 Completely Redundant ONUs, Each Connected to a Separate PON
This option, illustrated with ONUj and ONUk in Figure 2.6, is clearly the most robust, and it may be
desirable for some enterprise or military customers. Each of these ONUs is connected to a separate PON
and is active on that PON. A customer router/switch implements the protection at Layer 2. This option also
adds cost to the CPE, since it must have a separate interface to each ONU and a means for switching
between them during a failure.
Introduction to Fiber Optic Broadband Access Networks and Technologies 41
2.5.4.2 Connect the ONU to Two PONs with Separate Optical Interfaces
This option, illustrated in Figure 2.6 (ONUi) and Figure 2.7, corresponds to the Type C protection in [7].
With 1 :1 protection, the ONU is only active on one PON at a time. The ONU switches to using the
other PON when the one it is currently using fails.
The Figure 2.7 architecture also supports 11 protection. With 1 1, the ONU is active on both PONs,
with one of the PONs serving as the primary connection to the OLT. Similar to SONET/SDH, the OLT
and ONU can use the bandwidth of the other (protection) PON for “Extra Traffic”, i.e., data that uses the
additional bandwidth available on the protection PON, as long as no failures exist on the primary PON.
When a failure occurs, the ONU uses the protection PON, preempting any use of that PON for Extra
Traffic. The Extra Traffic is typically regarded as lower priority, and the OLT may restrict the amount of
Extra Traffic that the other ONUs can send when one or more ONUs have switched to the protection PON.
An approach similar to 1 1 protection is to use Ethernet Link Aggregation across the two PONs. As
described above for OLT NI protection, Link Aggregation combines multiple Ethernet physical links to
create a single higher bandwidth channel for the Ethernet frames. If one of the physical links fails, Link
Aggregation automatically scales back its transmission rate to use just the remaining healthy link(s). The
process of falling back to the healthy links is known as “fail-over”. Since the bandwidth is reduced during
fail-over, Link Aggregation provides service restoration for services that can tolerate the reduced
bandwidth rather than providing full Layer 1 protection.
With either the 1 1 protection or Ethernet Link Aggregation methods, the ONU would be active on
both PONs simultaneously. Consequently, the network management and traffic management becomes
much more complicated. In the case of Link Aggregation, the division of the traffic flow between the two
PONs would be handled as part of the standard Ethernet Link Aggregation processing. The dynamic
bandwidth assignment (DBA) algorithms would become significantly more complex since they need to be
aware of whether one PON or two are available for the connection to each ONU. The 1 1 protection
architecture sends the identical data stream over both fibers, with the receiver only actively taking data
from one of the streams. For the 1 1 PON case, the ONU must be able to take initialization, management,
and provisioning data from both streams since it is logically behaving as a separate ONU for each PON.
2.5.4.3 Connect the ONU to a Single PON Using Redundant Optical Interfaces
As illustrated in Figure 2.15, this option uses redundant optical interfaces on the ONU that are connected
to the fiber through either a passive splitter or an optical switch.
OI
OI
ONU
OI
OI
ONU
Network
Access/Metro
NI
(W)
NI (P)
W
P
OLT
OLU
OLU
OLU
OLU
Optical switch
Passive splitter
Figure 2.15 Illustration of redundant ONU optical interfaces to a single PON
42 Broadband Access
As with the OLU protection case, if a passive splitter is used, then the optical budget must be able to
handle the additional 3 dB loss of the splitter. This additional loss could potentially necessitate using
FEC. If the passive splitter approach is used for 1 :1 protection at both the OLT and ONU, the total 6 dB
loss may be an issue. See the discussion of passive splitter loss in Section 2.4.1 and 24.4.1.
Using an optical switch substantially reduces the optical loss, but it requires a control interface for the
switch. Ideally, a non-latching switch should be used. See the discussion in Section 2.4.1.
2.5.4.4 Use a Single Optical Interface to a Single PON
This option is the cheapest and simplest, since it provides no protection. For cost reasons, it is typically
used for residential customers.
With this option, it is very beneficial if the PON system can perform diagnostics during normal
operation in order to detect impending failures in the optics before they occur. Detecting the degradation
allows the problem to be repaired with a limited maintenance outage rather than a much longer outage due
to the failure. PON devices from PMC-Sierra include integrated diagnostic features for this purpose.
2.5.5 Conclusions Regarding Protection
A wide range of protection options is available for PON systems. At one extreme, the most costly and
complex of these provide full redundancy for each part of the equipment and network, including route
diversity for the PON fibers. However, this level of redundancy is typically too expensive unless
the subscriber requires it and pays for such a premium service level. At the other extreme, redundancy can
be omitted altogether. This option may be practical for cost-sensitive non-critical services such as some
video services.
PON service to residential customers typically uses redundancy for only those OLT functions that
are shared by multiple PONs. Examples include the uplink from the OLT to the metro/core network
and, potentially, some of the control functions. Different levels of redundancy for the PON and the
interfaces to the PON may become important as differentiators for premium business services. There
are a number of cost and technical issues that must be taken into account when providing this level of
redundancy.
2.6 Conclusions
Due to its very high bandwidth capability, fiber is the most flexible medium for broadband service
delivery to the home. After years of being a promising “next generation” technology, FTTH has finally
become an economically viable option for providing residential triple-play services. The various technical
and operational hurdles that have slowed large-scale deployment of FTTH have largely been resolved.
PON is the most cost-effective approach to providing FTTH broadband services. By providing a highly
flexible platform for different services, and by eliminating the active electronics from the access plant,
PON provides carriers with substantial ongoing OAM savings over copper-based technologies such as
DSL or coaxial cable with cable modems.
Different PON protocols are favored by different carriers and different regions. The most popular PON
protocols currently being deployed are the IEEE 802.3ah Gbit/s Ethernet the most popular PON and the
ITU-T G.984 2.5 Gbit/s G-PON standard, which is favored by the North American and European carriers.
The IEEE 802.3av 10G EPON equipment began field trials in 2010. These PON protocols are the subject
of the next chapters.
Optical domain PON protocols, discussed in Chapter 5, can be used alone or in conjunction with the
IEEE or ITU-T protocols. Optical domain protocols have not been cost-competitive, and are not expected
to be so until possibly around the 2015 timeframe. Since optical domain techniques promise maximum
flexibility for carrying different subscriber signals, and they allow the highest overall per-subscriber data
rates, they will be a primary topic for further research.
Introduction to Fiber Optic Broadband Access Networks and Technologies 43
Appendix 2.A: Subscriber Power Considerations
A primary factor in providing power is subscriber usage statistics. In providing power from the CO to
subscribers, it can be assumed that only some fraction of the subscribers will be active (‘off-hook’)ata
time, and that a much smaller percentage will have their phones ringing. For example, traditionally for
CO-delivered voice service, fewer than 10% of subscribers could be safely assumed to be active. The
probabilities associated with a large number of subscribers allows using a much smaller power and battery
source than would be required if all the subscribers were off-hook or had ringing phones. Since DLC RTs
connect to fewer subscribers, they must make more conservative assumptions regarding the number of
simultaneously active subscribers. Here, it is only safe to assume no more than 30–50% of the subscribers
are off-hook. As a result, more battery capacity is required per subscriber. With FTTC (and FTTH), it is no
longer safe to assume that all of the subtended subscribers will not be simultaneously off-hook, because of
data and video service usage in addition to voice.
In addition to the average power, it is also possible that multiple subscribers on a FTTC ONT will have
their phones ringing simultaneously, creating a high peak power demand. In summary, the fewer
subscribers that share a battery resource, the more battery capacity must be used per subscriber to provide
the same level of reliable back up.
Finally, power delivery is more economical if it can re-use the existing copper wires with the existing
voltages. These copper wires are small gauge (26–22 AWG). Since the loss is I
2
R, these wires have
substantial loss with higher currents. Higher voltages increase efficiency by lowering the current, but they
quickly become considered “hazardous” and require a different type of craftsperson certification than the
traditional –48 Vdc. The limit is determined again not by the average per-subscriber power, but by the
peak power requirements during ringing.
References
1. Hasegawa T, Kuritani K, Makin, K, et al.Optical customer access based on digital loop carrier. Proc. IEEE ICC’90.
1990; 341.3.1–341.3.5.
2. Rowbotham T, Ritchie B, Hoppit, C. Plans for the Bishops Stortford (UK) fibre to the home trials. Proc. IEEE
Globecom’89. 1989; 1320–1325.
3. ITU-TG.984.6. Gigabit-capable Passive Optical Networks (GPON): Reach Extension and its amendments; 2008.
4. ITU-TG.Sup45. GPON Power Conservation; 2009.
5. Shea D.P., MitchellJE. A 10Gb/s 1024-Way Split 100-km long reach optical access network. Journal of Lightwave
Technology. 2007; 25(3): 685–693.
6. ITU-TG.665. Generic characteristics of Raman amplifiers and Raman amplified subsystems; 2005.
7. ITU-TG.983.5. A broadband optical access system with enhanced survivability; 2002.
8. ITU-TG.Sup51. PON Protection Considerations; 2012.
9. Gorshe S. Protection Strategies and Mechanisms for PON Systems, PMC-Sierra white paper, PMC-2080622;
2007.
10. IEEE1904.1. Service Interoperability in Ethernet Passive Optical Networks (SIEPON); 2013.
11. Gorshe S. Resilient Packet Ring Technology White Paper, PMC-Sierra white paper, PMC-2041096; 2005.
12. Gorshe S. Resilient Packet Ring (RPR). China Communications 2005; 2(4): 91–103.
13. ITU-T G.8031. Ethernet linear protection switching, 2011.
14. ITU-T G.8032. Ethernet ring protection switching, 2012.
Further Reading
1. Van de Voorde I. et.al.The SuperPON demonstrator: an exploration of possible evolution paths for optical access
networks. IEEE Communications Magazine. 2000; 38(2): 74–82.
3
IEEE Passive Optical Networks
3.1 Introduction
IEEE 802.3 has developed two PON (point-to-multi-point) protocols based on their point-to-point
protocols of the same rate. These include the Ethernet PON (EPON) protocol based on 1 Gigabit/s
Ethernet and the 10G EPON protocol based on 10 Gbit/s Ethernet. EPON has seen extensive use,
especially in Asia, with Japan taking the lead role in deploying it. Its re-use of Ethernet technology has
given it some significant benefits, and 10G EPON is expected similarly to benefit from 10 Gbit/s Ethernet
technology.
EPON uses 1 Gbit/s rates in both the upstream and downstream directions, and 10G EPON uses a
10 Gbit/s downstream rate with both 1 and 10 Gbit/s supported in the upstream direction. The downstream
directions of both protocols are essentially the same as for point-to-point Ethernet streams of those rates,
with some changes to the Ethernet frame overhead and additional management frames defined in order to
support the point-to-multipoint operation. The upstream direction uses a TDMA protocol in which
the ONU upstream transmissions are bursts compromised of Ethernet frames. No frame fragmentation
is allowed.
In order to maximize backward compatibility and to allow co-existence on the same PON, 10G EPON
is largely an extension of the EPON protocol (which is described in detail in the first section of this
chapter). The second section describes 10G EPON protocol primarily in terms of how it differs from
EPON. The differences between the two protocols are summarized in Table 3.2 at the end of the chapter.
Although both protocols are contained within the 2012 version of IEEE 802.3 [1] and its amendment IEEE
802.3bk [2], this chapter refers to the outputs of the projects under which they were developed, namely
[3] and [5].
1
3.2 IEEE 802.3ah Ethernet-based PON (EPON)
The IEEE 802.3ah PON standard [3] was developed after the ITU-T B-PON and before the ITU-T G-PON
protocol (see Chapter 4), although there was overlap in the development of the three. The EPON standard,
which was developed as part of the IEEE Ethernetin the First Mile (EFM)project, was motivated bya desire
1
At the time this manuscript was submitted to the publisher, the IEEE Communications Society was developing a new “Standard for
Service Interoperability in Ethernet Passive Optical Networks (SIEPON)” project. The scope of the IEEE 802 standards is Layers 1 and
2. SIEPON addresses other functional aspects that are required for multi-vendor interoperability. Specifically, the scope includes
“equipment functionality, traffic engineering, and service-level QoS/CoS mechanisms,” and “management specifications covering:
equipment management, service management, and power utilization.”
Broadband Access: Wireline and Wireless – Alternatives for Internet Services, First Edition.
Steven Gorshe, Arvind Raghavan, Thomas Starr and Stefano Galli.
2014 John Wiley & Sons, Ltd. Published 2014 by John Wiley & Sons, Ltd.
46 Broadband Access
to leverage the traditional advantages of Ethernet. These advantages included the ubiquitous presence of
Ethernet at the customer premises, the relatively low cost of Ethernet UNIs, and Ethernet’s potential to
providea lowercost Layer 2 technology thanATM, which wasused byB-PON. EPONwas developed touse
the same transmission rate as the Gbit/s Ethernet interface that had recently been standardized. The
B-PON protocol,in contrast, provided less bandwidth both in terms of transmission rate and the substantial
ATM cell overhead. ATM is also a heavyweight Layer 2 technology that added more complexity than was
required for Ethernet private line and LAN extensions through the MAN/WAN.
In contrast to the ITU-T PON protocols described in Chapter 4, which are based on a TDM frame, the
EPON payload consists of Ethernet MAC frames in the upstream and downstream directions. As a
consequence, EPON lacks a 125μs reference for use with voice or other potential TDM clients and,
instead, relies on VoIP for carrying voice traffic and circuit emulation service (CES) for carrying other
TDM clients. While this adds complexity for TDM traffic, it reduces some complexity within the PON
protocol. It also fits well with the general move to VoIP among many carriers. EPON supports autodiscovery of ONUs and FEC.
3.2.1 EPON Physical Layer
The MAC data rate of an EPON system is 1 Gbit/s. The data is encoded with the 8B/10B block code for
transmission, resulting in a 1.25 Gbit/s signal transmission rate.
The data is transmitted over a single PON fiber. While a split ratio of 1 : 16 (i.e., 16 ONUs on a single
PON connecting to one OLT interface) is shown in 802.3ah, actual deployments commonly use 1 : 32,
with several using 1 : 64.
wavelength window, while the upstream transmission uses the 1310 nm window to take advantage of
less expensive lasers.
There are two distance options specified for EPON. One supports up to 10 km as the maximum OLT to
ONU distance and the other allows up to 20 km.
These types are summarized as follows:
2
The downstream signal is transmitted with a laser using the 1490 nm
3
PX refers to the optical interface options for EPON.
PX10 specifies an optical channel insertion loss of 20 dB for 10 km reach with at least 1 : 16
•
split ratio
PX20 specifies an optical channel insertion loss of 24dB for 20 km reach with at least 1 :16 split
•
ratio
PX30 specifies an optical channel insertion loss of 29 dB for 20 km reach with at least 1 : 32
•
split ratio
PX40 specifies an optical channel insertion loss of 33 dB for 20 km reach with 1 : 64 split ratio.
•
3.2.2 Signal Formats
The downstream signal is simply a stream of Ethernet frames and Idle characters, as with a point-to-point
Gbit/s Ethernet signal. The upstream signal is transmitted in bursts, like other TDMA protocols.
The preamble and start of frame delimiter (SFD) are modified for EPON from their normal values for
Ethernet. Specifically, where the normal 8B/10B-encoded Ethernet 8-character preamble/SFD consists of
/S/, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, and 0xd5, the EPON preamble/SFD consists of 0x55, 0x55,
SLD, 0x55, 0x55, 2-octet LLID plus MODE bit, and CRC-8. The SLD is the Start of LLID Delimiter and
has the value 0xd5. The LLID is the two-octet logical_link_ID field that uniquely identifies the ONU
MAC. (Note that IEEE1904.1 allows multiple unicast LLIDs per ONU.) As discussed in Section 3.2.5, the
2
The limits on the split ratio are a combination of the functions of the optical parameters (e.g., loss budget) and the desired per-ONU
bandwidth.
3
For the both options, the minimum distance between an OLT and ONU is specified as 0.5 km.
IEEE Passive Optical Networks 47
Octets
6
6
2
2
4
40
4
Destination Address
Source Address
Length/Type = 88-08
Opcode
Timestamp
Data/Reserved/Pad
FCS
MPCPDU frame
= 00-02 for GATE
= 00-03 for REPORT
Number of grants/Flags
Grant #1 Start time
Grant #1 Length
Grant #4 Start time
Grant #4 Length
Sync time
Pad/Reserved
GATE information
Number of queue sets
Report bitmap
Queue #0 Report
Queue #7 Report
Pad/Reserved
REPORT information
Octets
1
0/4
0/2
0/4
0/2
0/2
13-39
1
1
0/2
0/2
0-39
Figure 3.1 MPCPDU illustration, including the GATE and REPORT MPCPDU information fields
LLID is assigned to the ONU by the OLT during the registration phase of the discovery process. The
CRC-8 covers the SLD through the LLID octets, and this uses the generator polynomial x
8x2
x1.
In addition to the one Logical Link ID (LLID) that is unique to each ONU, all ONUs respond to the
Single Copy Broadcast (SCB) LLID. The SCB provides an efficient mechanism for the OLT to broadcast
information to all ONUs without having to duplicate it for each. For example, the SCB is used when the
OLT invites new ONUs to make their presence known during the discovery process. Multicast trafficis
transmitted using the SCB LLID. An ONU can use standard L2 networking processing, such as VLAN
filtering and IGMP snooping, to narrow the amount of received multicast traffic and receive only the
designated multicast traffic.
Multi-Point Control Protocol PDUs (MPCPDUs) are control frames used by the ONUs to make their
requests for bandwidth, and also by the OLT to assign it. As illustrated in Figure 3.1, the MPCPDU
4
frame
is a basic 802.3 MAC control frame containing a four-byte timestamp and a 40-byte field filled with data
and padding as needed. MPCPDU messages are also used for the discovery and ranging processes, as
discussed in Sections 3.2.5 and 3.2.6. MPCPDUs are layered below the data interface, and they have
higher priority than any data packet. This ensures that the bandwidth requests and grants are sent in a
timely manner.
4
The MPCPDU Ethertype is 0x88-08. The opcodes are assigned between 00-02 and 00-06.
48 Broadband Access
Downstream traffic
from ONUx
Bandwidth request report
Bandwidth grant to ONUx
Payload from ONUx
Periodic grant to ONUx
Upstream traffic
LEGEND:
MPCPDU
Ethernet payload frame
Ethernet Idle frame
Traffic to/from ONUx
from ONUx
Bandwidth request report
Bandwidth grant to ONUx
Figure 3.2 Ethernet PON MAC operation example
Note that EPON supports the use of PAUSE frames for flow control. When ONUs are a long way from
the OLT, however, the delay makes PAUSE inefficient.
3.2.3 MAC Protocol
The EPON MAC uses upstream bandwidth requests from the ONUs and upstream bandwidth transmission grants from the OLT. The protocol makes use of local timers at each ONU that are synchronized to
the OLTs local timer. The MAC operation discussed in this section is illustrated in Figure 3.2 with the
downstream and upstream data flows.
3.2.3.1 GATE Messages for Upstream Bandwidth Grants
The OLT grants bandwidth to an ONU in the GATE message. Gating is the function that controls when
the ONUs are allowed to transmit upstream data. The Gating function relies on a local timer that is
synchronized to the OLT timer (see the ranging protocol description in Section 3.2.6 for a further
discussion of the timers). The GATE message specifies the ONU upstream start time and transmission
length relative to the ONUs local timer. The bandwidth grants from the OLT are always made for at least
1024 time quanta into the future so that the ONU has time to process the GATE message and be ready to
transmit. An ONU turns its laser on when its local time matches the start time specified in the GATE
message. The length field gives the length of time the ONU is allowed to transmit in that burst
5
The transmission time is specified with respect to the number of periods of the ONU’s local clock.
5
. The start
IEEE Passive Optical Networks 49
time is a 32-bit number (the same length as the local timer counter) and the length field is a 16-bit number.
The OLT includes the time required to turn the ONU laser on and off and the time to send the upstream
synchronization patterns when it assigns the grant time.
The ONU includes the inter-frame gap and FEC bit times in its requests for bandwidth. The ONU is
responsible for ending its transmission long enough before the end of its time to allow its laser to turn fully
off, but the grant times are specified by the OLT to be long enough to accommodate the laser on time, off
time, and synchronization time.
The OLT sends GATE messages to each ONU periodically so that they can report their upstream
bandwidth needs. The first (left-most) grant in Figure 3.2 is an example of such a periodic grant. The
ONUs have watchdog timers that are reset whenever a GATE message is received.
6
Up to four upstream transmission grants can be made to a given ONU in a single GATE message. As
discussed in Section 3.2.5, the ONU advertises the number of outstanding grants it can accept during the
discovery process. The first payload transmission from ONUx in Figure 3.2 illustrates data from two
grants. The GATE message can also request reports from the ONUs corresponding to those grants. In
practice, however, using multiple grants per GATE message adds considerable complexity to the OLT
bandwidth assignment process. Sending a single grant in the GATE message gives much finer resolution
and faster response for the upstream bandwidth assignments, and it has only a small impact on the
downstream overhead bandwidth. The preferred approach, especially for per-flow dynamic bandwidth
assignment (DBA), is for the OLT to send each ONU a single grant in each GATE message to service its
bandwidth requests and to let the ONU decide which data should be sent in that upstream grant.
3.2.3.2 REPORT Messages for Upstream Bandwidth Requests
The ONUs communicate their upstream bandwidth requirements by sending REPORT MPCPDU
messages. The OLT grants the upstream bandwidth for these REPORT messages in its GATE messages.
In addition to the timestamp, the REPORT message consists of a summary of its requests for upstream
bandwidth and the specific amount of bandwidth it needs. EPON supports the eight queue priority levels
defined in IEEE 802.1Q. The summary field of the REPORT message indicates how many and which, if
any, of these queues have data to send. The summary is followed by binary numbers to indicate the
specific number of bits to be transmitted from each queue. The bit count is a 16-bit number, and this
includes any required overhead bits, Inter-Packet Gap (IPG) characters, and FEC bits that it needs to send
in the transmission.
Each ONU sends REPORT messages periodically, even if it has no data waiting for transmission, in
order to reset a watchdog timer at the OLT. If the watchdog timer expires, the OLT deregisters that ONU
from the network.
3.2.4 Encryption and Security
The EPON standard did not address the security of the downstream transmissions (i.e., protecting against
ONUs listening to traffic other than their own). EPON can potentially use the 802.1ae and 802.1af link
encryption standards that were subsequently developed. In the meantime, regional specifications have
become common.
The encryption method, described in [4], is a symmetric block code using the 128-bit key AES. A
counter (CTR) mode is used, in which the output of pseudorandom counters is exclusive OR’ed with the
original data to generate the cipher text. This encryption method can be implemented either for just
6
For the purposes of keeping the ONU’s watchdog timer alive, an OLT can also periodically send empty GATE messages when it has
no pending bandwidth requests for that ONU.
7
An encryption protocol developed by PMC-Sierra in conjunction with NTT became an early de facto standard in some regions [11]. As
part of PMC-Sierra’s agreement with NTT, no royalty fees are charged to vendors that sell to NTT carriers, and several other equipment
and silicon vendors have since implemented this encryption protocol.
7
50 Broadband Access
IPG
S_FEC symbol
Data portion of the
first FEC code word
Data
239
bytes
Data
Data
Ethernet frame data bytes
16-byte check symbol of
the first FEC code word
Data portion of the
last FEC code word
Data
Data
Data
239
bytes
F
F
E
E
C
C
FEC
check bytes
T_FEC symbol
F
F
F
E
C
F
E
E
E
C
C
C
IPG
Figure 3.3 EPON FEC illustration
the downstream traffic (as is done in G-PON) or for both the upstream and downstream directions. The
primary purpose of upstream encryption is to provide message authentication. If the OLT receives a
transmission with the incorrect encryption key for that upstream bandwidth grant, it will discard the data.
While upstream encryption is not mandatory, it is being demanded by an increasing number of carriers.
3.2.5 Forward Error Correction (FEC)
EPON allows the optional use of FEC.8The FEC code is the ITU-T G.975 systematic RS(255, 239, 8)
code. The downstream data packets are divided into 239-octet blocks, to which 16 check code bytes are
appended. The upstream bursts are similarly divided into data blocks and error check code bytes. The first
upstream FEC code word of a burst begins with the /S/ character, and the last FEC code word of the burst
contains the /T/ character. The additional bandwidth for the FEC check bits is taken into account in the
bandwidth requests that the ONU makes to the OLT.
For both the downstream and upstream directions, the FEC is packet-based rather than stream-based.
The EPON FEC code arrangement is illustrated in Figure 3.3. The FEC check bytes for all the FEC code
words associated with an Ethernet frame are placed in order at the end of that Ethernet frame rather than
immediately following the data portion of each code word. Since the Ethernet frame length will generally
not be an integer multiple of 239 octets, typically there will be a shorter block at the end of the packet. The
16 error check bytes for the last block will be calculated over a logical block that includes the X actual data
bytes followed by 239-X “0” padding bytes. The “0” padding bytes are not transmitted, but are re-inserted
by the receiver when it performs its error check calculation for this block.
When the receiver FEC decoder is unable to correct a character, it replaces the uncorrectable character
with a /V/ character.
In an environment noisy enough to require FEC, the preamble/SFD and end of frame (EOF) delimiters
are vulnerable to transmission errors. In order to provide the desired robustness when FEC is enabled, an
additional new SFD is added to the beginning of each MAC frame and an additional new EOF is added to
the end. The new SFD, designated as S_FEC, is five octets long and consists of the K28.5/D6.4/K28.5/
D6.4/S/character set. The new EOF is designated as T_FEC, and it has separate versions for even and odd
alignment at the end of the frame. The T_FEC character sets are T_FEC /T/R/K28.5/D10.1/T/R/ (for
even alignment, positive running disparity), /T/R/K28.5/D29.5/T/R (even alignment, negative running
disparity), or /T/T/R/I/T/R/(for odd alignment). The receiver is required to recognize the S_FEC or
T_FEC even if they contain up to five bit errors.
8
EPON FEC has the ability to be selectively activated on a per-ONU basis, thus reducing the overall FEC overhead on the PON.
IEEE Passive Optical Networks 51
OLT ONU
Discovery GATE
Discovery
window
Grant start time
REGISTER_REQ
REGISTER
(Including LLID assignment)
GATE
(to grant bandwidth for the REGISTER_ACK)
REGISTER_ACK
Discovery handshake complete
Random
delay
Figure 3.4 ONU discovery handshake
3.2.6 ONU Discovery and Activation
The OLT periodically opens a Discovery Time Window in order to allow new ONUs to announce
themselves.
OLT opens the window by transmitting a Discovery GATE message, which includes the length of the
Window and its start time.
The unregistered ONUs respond to the discovery GATE message by transmitting a REGISTER_REQ
message. The REGISTER_REQ message includes the ONU’s MAC address and the number of outstanding grants that it can accept (see Section 3.2.3 regarding multiple grants). A contention algorithm is
used in order to minimize the chance of collision when multiple ONUs are attempting to register during
the same Discovery Time Window. The contention algorithm operates by having each ONU delay its
transmission by a random time relative to the beginning of the Discovery Time Window.
already registered ignore the discovery GATE message.
When the OLT receives the REGISTER_REQ, it assigns an LLID to the ONU, bonding the LLID to the
ONU’s MAC address. The OLT then sends a Register message to the ONU in order to communicate the
ONU’s LLID and the required OLT synchronization time, and to echo the maximum number of pending
grants that the ONU can accept. The synchronization time is the amount of time the OLT will require in
order to synchronize reliably to the ONU’s upstream transmission burst. The synchronization time is
9
The discovery message, registration, and handshake flow are illustrated in Figure 3.4. The
10
ONUs that are
9
The period of the Discovery window is implementation-dependent.
10
The upper bound requirement on this hold-off time is that it must be short enough that the ONU can transmit its entire
REGISTER_REQ before the end of the Discovery Time Window.
52 Broadband Access
specified in multiples of 16-bit data patterns that the ONU sends as IDLE code pairs at the beginning of
the burst.
After the ONU has processed the Register message, it sends a REGISTER_ACK message to the OLT in
response to a standard GATE message from the OLT.
Note that the Discovery GATE, REGISTER_REQ, and REGISTER messages are sent on t he
broadcast channel since ONU doesn’t know its LLID until it receives the REGISTER message. After
the ONU receives its LLID, the remaining GATE and REGISTER_ACK messages are sent on the
unicast channel.
Mechanisms exist in the protocol to deregister an ONU (e.g., if a watchdog timer expires) and to
re-register.
3.2.7 ONU Ranging Mechanism
In order to prevent overlap i n the different ONU upstream transmission bursts when they arrive at
the OLT, the OLT assigns the upstream burst transmission times with enough guard band time between
the bursts of successive ONUs. This guard time allows for the laser of one ONU to turn off and for the
laser of the next ONU to turn on, and to cover any uncertainty in their relative signal propagation
delays to the OLT.
a crude approach would be to have this uncertainty window include the round trip propagation
delay difference between an ONU at the shortest fiber distance and an ONU at the longest anticipated
fiber distance from the OLT. With the desired maximum ONU distance of 20 km and the 2.045 m/s
speed of light through a fiber, the resulting differential delay is 200 μs. This would clearly be very
bandwidth-inefficient, since only a few bit periods would be required if the ONUs were equidistant
from the OLT.
The solution to this problem is to use a ranging mechanism that allows the OLT to determine the
relative distances of the ONUs. The OLT can then take this range into account and assign the upstream
burst times with a minimum of guard band time.
The ranging mechanism for EPON is based on local clocks and counters that are maintained at the OLT
and each ONU. A counter has 32 bits and is incremented once every 16 ns.
The OLT counter is the PON master. When the OLT transmits a MPCPDU message, it loads its current
counter value into the message’s 32-bit timestamp field. When the ONU receives a MPCPDU, it resets its
own local counter to the value contained in the MPCPDU timestamp field. When the ONU sends a
MPCPDU to the OLT, the ONU loads its updated counter value into the timestamp field. The OLT then
compares the offset between its current count and the value it receives in the MPCPDU timestamp field,
with the difference being the round trip time (RTT) associated with that ONU. The RTT is then used to
establish the ONU’s range, which is taken into account when the OLT assigns the start times for upstream
bandwidth grants.
Some drift may occur in the RTT over time. When the drift exceeds a provisioned threshold, a
timestamp drift error condition is declared. Either the ONU or OLT can detect this condition as an offset
between the expected value received in the MPCPDU and the one actually received.
11
Since the ONUs can be located at different physical distances from the OLT,
3.2.8 EPON OAM
Since EPON lacks an outer transport frame structure like those used in SONET/SDH and G-PON, it has
no dedicated overhead frame bits to communicate OAM data. The lack of link OAM was addressed as part
of the same IEEE 802.3ah project that developed EPON for application to individual links. The 802.3ah
11
If the ONUs do not turn their lasers off when they are not transmitting, spontaneous emission noise from ONUs closer to the OLT
would interfere with data transmissions from ONUs further from the OLT.
IEEE Passive Optical Networks 53
Table 3.1 OAMPDU types
OAMPDU code OAMPDU type Comment
00 Information To communication remote and local OAM information
01 Event notification Alerts the remote Ethernet node of link events
02 Variable request Request for MIB variable(s)
03 Variable response Return of MIB variable(s)
04 Loopback control
05-FD Reserved
FE Organization-specific Reserved for organization-specific extensions, identified by the
FF Reserved
Organizationally Unique Identifier
group defined several Ethernet frames to communicate the link OAM information. The types of OAM
information are categorized as follows:
1. Remote failure indication: Indication sent in the reverse direction by an Ethernet terminal node to
indicate that it cannot properly receive messages on that link
2. Remote loopback
3. Link monitoring: Messages to support link performance notifications for diagnostic and performance
monitoring purposes
4. Miscellaneous: Mechanism to provide additional OAM functions such as OAM capability discovery,
or to support higher layer management applications.
All OAM PDUs share the common frame format illustrated in Figure 3.5. The different OAMPDU
types are listed in Table 3.1.
3.2.9 Dynamic Bandwidth Assignment (DBA)
The most basic method of allocating upstream bandwidth is to distribute it equally among the ONUs. This
method is very inefficient, especially with packet traffic, since the bandwidth needs of the ONUs will
Octets
6
6
2
1
2
1
42-1496
4
Dest. Addr. = 01-80-c2-00-00-02
Source Addr.
Length/Type = 88-09 (Slow Protocols)
Subtype = 0x03 (OAM)
Flags
Code
Data/Pad
FCS
Figure 3.5 OAMPDU frame format
Fixed
common
header
for all
OAMPDUs
54 Broadband Access
rarely be equal at each instant in time. Considerable overall bandwidth utilization gains can be made if the
upstream bandwidth is allocated dynamically according to the current needs of the ONUs.
12
The ITU-T
addressed DBA in its G.983.4 recommendation; while not specifying a particular DBA algorithm,
G.983.4 specifies the framework and mechanism to implement DBA in B-PON and G-PON systems, and
it is equally applicable to EPON systems. G.983.4 is discussed further in Chapter 4. Some DBA comments
relevant to EPON are provided in this section.
As described above, the ONU REPORT messages inform the OLT of their current bandwidth needs.
Their bandwidth requests are reported in terms of the number of characters they have in the different
priority queues awaiting upstream transmission. The OLT can also take into account the service level
agreements (SLAs) that have been specified for the service flows associated with an ONU. For example,
an ONU with an active VoIP service will need a fixed amount of bandwidth on a regular basis.
Consequently, the OLT can regularly grant the upstream bandwidth for this service flow so that the ONU
does not need to waste upstream bandwidth reporting bandwidth requests for it. As another example, if the
OLT receives upstream bandwidth requests from multiple ONUs, it can grant more bandwidth to ONUs
that recently have been consistently requesting more bandwidth than it does to ONUs that have made few
recent requests. In other words, the OLT attempts to reduce latency by anticipating the needs of the ONUs.
In this example, however, the DBA algorithm needs to ensure that nodes with fewer bandwidth requests
do not become starved or encounter high latency while the ONUs with more bandwidth requests
are serviced.
EPON DBA has the flexibility to customize EPON network behavior to meet various carrier needs. Its
flexible nature allows quick adaptation to possible carrier challenges, making the EPON infrastructure
compliant with the ever-growing, ever-changing carriers’ requirements. It is possible to map both user and
service flows into specific containers that are managed by the DBA and to provide the QoS that is needed
for every customer and service. Two straightforward adjustable parameters related to EPON DBA are
latency and total system performance (upstream bandwidth utilization).
3.3 IEEE 802.3av 10Gbit/s Ethernet-based PON (10G EPON)
The IEEE 802.3av PON standard [5] was developed to increase the data rate of EPON systems from
1 Gbit/s to 10 Gbit/s, in keeping with the 10 Gbit/s Ethernet interface. 10G EPON shares much of its
protocol with EPON. A combination of coarse wave division multiplexing (CWDM) and time division
multiplexing (TDM) is used in order to allow EPON and 10G EPON systems to co-exist on the same
PON. As with EPON, 10G EPON relies on VoIP for carrying voice traffic and circuit emulation service
(CES) for carrying other TDM clients.
3.3.1 10G EPON Physical Layer
The downstream data rate of 10G EPON is 10 Gbit/s, and both 1 Gbit/s and 10 Gbit/s rates are supported in
the upstream direction. The 64B/66B block line code, described in Appendix 3.A, is used for all of the
10 Gbit/s signals with a resulting signal line rate of 10.3125 Gbit/s. The 1 Gbit/s upstream uses the same
8B/10B block line code as EPON, giving a line rate of 1.25 Gbit/s.
The downstream and upstream data is transmitted over a single PON fiber, using WDM to separate the
upstream and downstream signals. The wavelengths used by the different upstream and downstream
signals are shown in Figure 3.6. As noted above for EPON, because there are many ONUs on the PON and
only a single OLT, the wavelength bands were chosen to allow the use of less expensive lasers at
the ONUs.
12
DBA can be used for, or can be considered as, a form of statistical multiplexing.
IEEE Passive Optical Networks 55
EPON PX
PR (10Gbit/s)
Upstream
[1270 +/-10 nm]
PRX (1Gbit/s) and
EPON PX Upstream
1270 nm
1310 nm
1300 nm 1400 nm 1500 nm 1600 nm
Downstream
[1310 +/-50nm]
PR and PRX (10Gbit/s)
Downstream
1490 nm
[1577 -2,+3nm]
1555 nm
Video
Overlay
1577 nm
1590 nm
Figure 3.6 EPON and 10G EPON optical spectrum allocation
For 1 Gbit/s upstream operation, 10G EPON uses the same 1310 nm wavelength as the EPON upstream
signal. This allows the OLT to use the same receiver for all 1 Gbit/s signals. The dynamic bandwidth
allocation algorithm (see Section 3.3.8) allocates the bandwidth of the 1 Gbit/s upstream signal between
the EPON and 10G EPON ONUs.
The 10 Gbit/s upstream signals use a separate wavelength band, but it overlaps with the 1 Gbit/s
upstream wavelength band. When an OLT supports both 1 Gbit/s and 10 Gbit/s operation on the same
PON it is referred to as a dual rate mode. The dual rate OLT can either separate the 10 Gbit/s and 1 Gbit/s
upstream signals by dividing the signal in the optical domain or in the electrical domain. Dual-rate
receiver considerations are discussed below in this section.
The advantages to allowing 10G EPON to operate over the same PON optical distribution network as
EPON include:
allowing customers to use the most cost-effective ONU for the desired service;
•
allowing a network to migrate from EPON to 10G EPON by upgrading the OLT then migrating the
•
ONUs as needed;
continued operation of the existing network and services during the upgrade of the network.
•
Figure 3.7 illustrates a network where an OLT supports a mix of EPON ONUs, ONUs with 10 Gbit/s
downstream and 1 Gbit/s upstream, and ONUs with 10 Gbit/s upstream and downstream. For
convenience, the wavelength color key in Figure 3.7 is consistent with the key for Figure 3.6.
Note that WDM is used to separate the 1 Gbit/s and 10 Gbit/s traffic in the downstream direction with
the filters at the ONUs, and a combination of WDM and TDM is used in the upstream direction. The
discovery and other protocol extensions to support the co-existence of EPON and 10G EPON ONUs
are discussed in the appropriate sections below.
As its reference for the optical link loss budgets, the 802.3av specification uses a split ratio of either
1 :16 (i.e., 16 ONUs on a single PON connecting to one OLT interface) or 1 : 32. In practice, larger split
ratios such as 1 : 64 or 1 : 128 can be used if the other optical losses (e.g., the length of the fiber) are
constrained to offset the additional 3 dB loss that is incurred when the split ratio is doubled. All of the
interfaces are specified to operate at an uncorrected bit error rate no worse than (BER) 10
3
. After FEC
56 Broadband Access
10G/10G
ONU
1G
ONU
10G/1G
ONU
10G/10G
ONU
NOTE – Each ONU receives its
downstream signal from
the correct wavelength
1G
ONU
1G EPON downstream
1G EPON upstream
10G EPON downstream
10G EPON upstream
OLT
Figure 3.7 Illustration of EPON and 10G EPON ONUs sharing the same PON
correction, the bit error rate will be no worse than 10
12
. The nomenclature adopted to identify the
different optical interface options may be summarized as follows:
PRX interfaces use 10 Gbit/s downstream and 1 Gbit/s upstream transmission
•
PR interfaces use 10 Gbit/s for both downstream and upstream transmission
•
PR-Dn and PRX-Dn (n 10, 20, 30) refer to the OLT optical interface specification
•
PR-Un and PRX-Un (n 10, 20, 30) refer to the ONU optical interface specification
•
PR10 and PRX10 specifies an optical channel insertion loss of 20 dB for 10 km reach with 1 : 16
•
split ratio
PR20 and PRX20 specifies an optical channel insertion loss of 24 dB for 20km reach with a 1 : 16
•
split ratio or 10 km reach with a 1 : 32 split ratio
PR30 and PRX30 specifies an optical channel insertion loss of 29 dB for 20km reach with a 1 : 32
•
split ratio
PR40 and PRX40 specifies an optical channel with a reach of at least 20km and a split ratio of at least
•
1 :64.
As with EPON, the 1550– 1560 nm wavelength band is reserved for downstream video transmission.
Following the same approach as EPON, the upstream burst timing is relaxed for 10G EPON in
order to allow the use of existing off-the-shelf components. The standard has mechanisms to
allow for future tighter timing to be implemented with better components for increased bandwidth
efficiency.
Dual-rate operation refers to an OLT that simultaneously receives upstream signals from ONUs
using 1 Gbit/s and 10 Gbit/s rates. As illustrated in Figure 3.8, the received 1 Gbit/s and 10 Gbit/s
streams can either be split in the optical domain or electrical domain. Since both signals time-share the
IEEE Passive Optical Networks 57
Input from Upstream
PON Channel
(1260 – 1300 nm)
Input from Upstream
PON Channel
(1260 – 1300 nm)
Optical
Amplifier
(Optional)
1:2
splitter
PMD
a) Optical domain rate splitting
Dual-rate
detector & TIA
PMD
b) Electrical domain rate splitting
10G detector
1G detector
& TIA
& TIA
10G line
amplifier
Output to
10G PMA
1G line
amplifier
Output to
1G PMA
10G line
amplifier
Output to
10G PMA
1G line
amplifier
Output to
1G PMA
Figure 3.8 Dual rate receiver option illustration
same upstream wavelength, it is not possible to use WDM filters to separate them in the optical
domain.
Splitting the signals in the optical domain involves using a 1 :2 optical splitter. Each of the two splitter
outputs goes to its own photodetector followed by an electrical receiver with a filter optimized for its
bandwidth in order to maximize the receiver’s sensitivity. The drawback with this approach is the 3dB
additional optical loss introduced by the 1 : 2 optical splitter. If this additional loss cannot be tolerated, a
low-gain optical amplifier must be used in the receiver.
Splitting in the electrical domain allows using a single photodetector and introduces no additional
optical signal loss. In the electrical domain, one approach is to design the receiver filter as a compromise
that allows reception of both the 1 Gbit/s and 10 Gbit/s signals. This means that the receiver sensitivity is
not optimal for either signal, lowering it by about 1 dB for each. Alternatively, the OLT can adjust (switch)
the transimpedance of its transimpedance amplifier (TIA) filter for that burst’s rate. The APD bias can
either be set to a compromise value or switched along with the transimpedance.
13
While the performance
of an adaptable receiver is optimum, its additional complexity impacts the receiver cost. Detecting the rate
of the current incoming burst must be performed fast enough to switch the receiver. The burst rate could
be detected by looking for spectral energy that would only be present for a 10 Gbit/s burst. Alternatively,
the OLT could exploit its knowledge of which upstream burst is scheduled to arrive. However, since this
knowledge is in the MAC layer and not the PMD, requiring it would be a violation of layer stack
restriction.
13
Using a compromise APD bias results in a loss of around 1dB receiver sensitivity, which is 1dB better than using the compromise
transimpedance value.
58 Broadband Access
Burst Delimiter =
0x 6B F8 D8 12
D8 58 E4 AB
SP
Laser
synchronization
pattern
SP
BD
/I/ /I/
Preamble/SFD =
0x55, SLD, 0x55, 0x55,
LLID (2-octets), CRC-8
Data
Data
p/SFD
First FEC block
Upstream burst
Data
Parity
Parity
Parity
Parity
Burst terminator pattern
(alternating 1010 …
pattern, i.e., Flag = 10,
Payload = 0x55 55 … 55)
Data
Data
0 0 0
Parity
Figure 3.9 10 Gbit/s upstream burst transmission illustration
3.3.2 Signal Format
With the exception of the added forward error correction (FEC) coding, the downstream signal is simply a
stream of Ethernet frames and Idle characters, as with a point-to-point 10 Gbit/s Ethernet signal.
upstream signal is also essentially an Ethernet stream except that, as discussed above, a TDMA burst
format is used. The upstream signal also uses FEC.
The beginning of an upstream burst is illustrated in Figure 3.9. The synchronization patterns at the
beginning of an upstream transmission burst allow the OLT to synchronize its receiver to new burst from
an ONU. The Burst Delimiter pattern is used by the OLT to determine the start of 66B block transmission
and the FEC codeword alignment. The 66-bit value of the Burst Delimiter is 0x 6B F8 D8 12 D8 58 E4 AB
(which results in a transmission bit sequence of 01 1101 0110 0001 1111 0001 1011 0100 1000 0001 1011
0001 1010 0010 0111 1101 0101, since the characters are transmitted LSB first). The FEC codeword
alignment can be achieved in the presence of transmission errors. This burst delimiter is followed by two
66-bit blocks containing Idle characters. These Idle characters allow the OLT to synchronize its
descrambler and delineate the start of the actual data frame. The first two blocks of Idle characters
are included in the initial FEC codeword.
As discussed above in Section 3.2.2 with EPON, the preamble and start of frame delimiter (SFD) are
modified for EPON and 10G EPON from their normal values for Ethernet. Specifically, the preamble
bytes are replaced by the transmitting MAC ’s MODE and LLID variables. While the Ethernet 8-character
preamble/SFD consists of /S/, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, and 0xd5, the EPON and 10G EPON
preamble/SFD consists of 0x55, 0x55, SLD, 0x55, 0x55, 2-octet LLID, and CRC-8. The SLD is the Start
of LLID Delimiter, and it has the value 0xd5. The LLID is the two-octet logical_link_ID field that
uniquely identifies the ONU MAC. The MSB of the two octets that contain the LLID is the MODE
indication bit. As discussed in Section 3.3.5, the LLID is assigned to the ONU by the OLT during the
registration phase of the discovery process. The CRC-8 covers the SLD through the LLID octets, and it
uses the generator polynomial x
8x2
x1.
The upstream transmission ends with a burst terminator pattern comprised of three 66-bit blocks of
alternating zeros and ones (1010 . . . 10) after the last FEC codeword of the burst. The ONU turns off its
laser at the beginning of the burst terminator pattern, which ensures that it will be completely off by the
end of the burst.
Each ONU has one unique Logical Link Identifier (LLID) that the OLT associates to the ONU for
unicast traffic. In other words, these MAC instances are used to emulate a point-to-point connection
14
The
14
As explained in Section 3.3.4, both the upstream and downstream streams are encoded into FEC blocks in a manner that preserves the
64B/66B block stream format.
IEEE Passive Optical Networks 59
between and ONU and the OLT over the PON. (Note that IEEE1904.1 allows multiple unicast LLIDs per
ONU.) Additionally, the OLT has two Single Copy Broadcast (SCB) MAC instances that are used as an
efficient mechanism to broadcast downstream traffic to the ONUs. Such a broadcast is used for broadcast
data or for when the OLT must communicate with unregistered ONUs. In the upstream direction, an SCB
MAC is only used for client registration. The LLID value of 7F-FF is associated with the SCB MAC for
1 Gbit/s downstream operation and the LLID value of 7F-FE is associated with the SCB MAC for 10 Gbit/
s downstream operation. An ONU can use higher layer networking processing, such as VLAN filtering
and IGMP snooping, to narrow the amount of received multicast traffic that is passed to applications. It is
possible that these higher layers may require addition multicast MAC instances at the OLT, in which case
an OLT can have more MACS than two plus the number of ONUs.
As with EPON, MPCPDU control frames (see Figure 3.1) are used by the ONUs to make their requests
for bandwidth, by the OLT to assign bandwidth, and by both ONUs and OLT during the discovery and
ranging processes.
3.3.3 MAC Protocol
The 10G EPON MAC-layer control protocol is based on the protocol for EPON and includes enhancements for management of 10G FEC and inter-burst overhead. This MAC protocol operates on the basis of
the ONUs informing the OLT of their upstream bandwidth requirements, and the OLT scheduling and
granting bandwidth to the ONUs to transmit their upstream data (as described in Section 3.2.3 above). The
details of the MAC protocol specific to 10G EPON are described in this section.
For 10G EPON, a “Sync Time” field in the GATE MPCPDU is used by the OLT to communicate to the
ONU the amount of time the OLT needs at the beginning of the upstream transmission burst to
synchronize its receiver to the new burst. As illustrated in Figure 3.9, each burst begins with a
synchronization pattern, followed by a Burst Delimiter pattern, followed by two blocks of Idle characters.
The ONU transmits the 66-bit synchronization pattern repeatedly, and then transmits the Burst Delimiter
so that the duration of the entire sequence is the same as the Sync Time requested by the OLT.
Like EPON, 10G EPON supports the eight queue priority levels defined in IEEE 802.1Q. The summary
field of the REPORT message indicates how many and which, if any, of these queues have data to send.
Unlike EPON, the bandwidth value carried by the 10G EPON REPORT does not include burst overhead
or FEC overhead. The OLT already knows this information and takes it into account.
3.3.4 Forward Error Correction
EPON and 10G EPON use different FEC approaches. FEC allows a link to function with a higher line bit
error rate at the receiver. Consequently, FEC effectively increases the optical link budget, which in turn
allows increased distance or split ratios. FEC becomes increasingly important as bit rate increases and, for
this reason, it is mandatory in 10G EPON. Additionally, the 10G EPON FEC differs in two ways from
EPON. First, 10G EPON uses a more powerful RS(255, 223) code for error correction of 16 symbols
rather than the 8 symbols that can be corrected with the optional RS(255,239) code specified for EPON.
Second, the 10G EPON FEC is applied to fixed-length sequences of streaming data rather than Ethernet
frames as illustrated in Figure 3.10. Figure 3.10 illustrates the downstream transmission direction, which
is a continuous stream of FEC codewords that includes the Ethernet frames and all inter-packet
information such as IPG and Ordered Set data. The upstream transmission is similar except that, as
illustrated in Figure 3.9, the first FEC codeword of an upstream burst is aligned with the beginning of the
burst in order to allow the OLT FEC decoder immediate codeword synchronization for each burst.
Note that when the RS(255,223) FEC parity is taken into account, the effective data rate of a 10G EPON
link is approximately 8.7 Gbit/s.
15
The generator polynomial G(x) x8x4x3x21 is specified for the 10G EPON RS(255,223).
15
60 Broadband Access
802.3 frame FEC
F
Data
E
C
29 zero
pad
Data
Block 1
321
IPG
802.3 frame FEC
a) FEC overhead illustration for 1G EPON
D
F
E
C
D
F
I
a
a
P
t
a
E
t
G
C
a
Data
F
E
C
Data
F
Data
E
C
b) FEC overhead illustration for 10G EPON
Figure 3.10 FEC overhead locations (downstream)
Block 27Block 2
6
6
32
321
6
6
6
6
32
6
6
321
6
6
32
6
6
IPG
F
Data
E
C
Input 66B code blocks
Remove the redundant 66B
flag bit and pad the FEC
code word with leading 0s
Parity (check symbol)
D
F
a
E
C
P
t
G
a
F
I
E
C
321
Block 1
Full code word: 223 data bytes (including the zero pad) plus 32 parity bytes
Create a string of 31 valid 66B blocks (the original 27 data blocks plus
Transmitted data
6
6
the parity bytes sent within 4 valid 66B blocks)
321
Block 2
6
6
321 21 Parity 21 Parity
6
6
Block 31Block 28Block 27
Figure 3.11 10G EPON FEC code block formatting and transmission
Oneof thechallengesin addingFEC to the 10GE PONstream is extendingthe 64B/66B blockcodeformat
so that a 10GbE receiver can receive and synchronize to the stream that now includes FEC parity data. The
methodusedis illustrated inFigure3.11,where each FECcodewordcoversa groupof27 64B/66Bblocks. As
alsoshown inFigure 3.11, the firststep ofFEC encoding is removing thefirst flag bit of the64B/66B block.
16
Since the two leading flag bits of the 64B/66B block are intentionally redundant, only one of them needs to be protected by the FEC.
16
IEEE Passive Optical Networks 61
The resulting 27 65 1755-bit blockis padded with 29 leadingzeros to get a total of 1784 bits (223 bytes).
The RS(255,223) encoding produces 32 parity bytes.
In the final stage, the zero pad bits are removed, the original 27 64B/66B blocks are restored, and the
parity bytes are converted into a sequence of 64B/66B blocks for transmission. Specifically, the 32 FEC
parity bytes are treated a four groups of 64 bits. Each of these 64-bit parity groups is then given a pair of
leading header bits in order to create 64B/66B blocks. In order to create a recognizable header pattern, the
header bits for the four parity blocks are 00, 11, 11, and 00, respectively.
characters is then transmitted.
The receiver can then synchronize to the 64B/66B character stream and extract the original data through
the reverse process, performing error correction as it decodes the FEC blocks.
17
The string of 31 64B/66B
3.3.5 ONU Discovery and Activation
The ONU Discovery protocol for 10G EPON is the same as for EPON with the following exception. With
10G EPON, a Discovery GATE MPCPDU includes a Discovery information field that communicates to
the ONUs whether the OLT is capable of receiving 1 Gbit/s upstream signals, capable of receiving
10 Gbit/s upstream signals, and whether the Discovery Window being opened is for 1Gbit/s or 10 Gbit/s
upstream signals from the ONUs. Also, as described in Section 3.3.2, the OLT uses a separate SCB LLID
for Discovery messages associated with 1 Gbit/s and 10 Gbit/s upstream discovery invitations. These
additions allow the OLT to communicate with and register ONUs of both upstream rate capabilities, and
also allow an ONU that supports both rates to determine which upstream rate it should use.
3.3.6 ONU Ranging Mechanism
The 10G EPON ranging mechanism is identical to the EPON ranging mechanism. See Section 3.2.6 above.
3.3.7 10G EPON OAM
10G EPON also uses the 802.3ah Link OAM. See Section 3.2.7 above.
3.3.8 Dynamic Bandwidth Allocation
10G EPON DBA is similar to EPON DBA. The primary difference is that the OLT must schedule
upstream traffic for both 1 and 10 Gbit/s ONUs if both types are present on the PON. As noted above,
ONUs that use 10 Gbit/s upstream assume that the OLT already takes into account the required overhead
bits rather than depending on the ONU to include them in its bandwidth request.
One benefit of the 10G EPON system is the ability to overcome system bottlenecks via adjustments in
the EPON DBA algorithm. The DBA cycle length and bandwidth allocation per ONU can be adjusted so
that the total OLT upstream transmission going into the switch will be smoother, less bursty in nature,
allowing carriers to overcome blocking elements in their network topology (e.g., assigning more
bandwidth to the OLT ports than the uplink ports in the switch connected to the OLT to save CAPEX).
While this is also true for EPON, the higher bandwidth of 10G EPON allows additional flexibility.
3.4 Summary Comparison of EPON and 10G EPON
The essential aspects of the EPON and 10G EPON protocols are summarized in Table 3.2.
3.5 Transport of Timing and Synchronization over EPON and 10G EPON
Both EPON and 10G EPON can use the IEEE 802.1AS [6] protocol. While 802.1AS is a generalized
precision timing protocol (gPTP) for use with all Ethernet applications, it includes a clause that specifies
how it can be used with the Ethernet PON protocols. The 802.1AS protocol is based on a modified version
17
As shown in the appendix of this chapter, the normal allowed header bit patterns are 01 and 10.
IEEE Passive Optical Networks 63
of the IEEE 1588 [7] precision timing protocol (PTP). While a detailed description of the IEEE 1588
protocol is beyond the scope of this book, the manner in which it is used for Ethernet PON networks is
summarized in this section.
Recall that EPON uses 32-bit local counters that are incremented every 16 ns (i.e., they use a time
quantum of 16 ns). These counters are used for the ranging and upstream transmission synchronization
processes (see Section 3.2.6). These counters are also used in the EPON timing synchronization process.
Specifically, the 32-bit counter is the LocalClock entity of the time-aware system. The OLT is the clock
master, and it is assumed to have an accurate synchronization time derived from a grandmaster clock
source. The associated ONUs are clock slaves. A time-aware system consists of no more than one ONU,
which is a clock slave to that EPON link but may contain multiple OLTs, since the ONU may have EPON
links to multiple OLTs.
The PON application is different from other Ethernet links, in that the upstream direction uses a TDMA
protocol that results in asymmetry between the downstream and upstream delays. (See Section 3.2.3). The
use of different wavelengths for upstream and downstream transmission also impacts the directional delay
asymmetry, since the fiber’s index of refraction, and hence the propagation speed, are wavelengthdependent.
The 802.1AS protocol works as follows for EPON. The OLT (clock master) communicates to an ONU
(the clock slave) the accurate synchronization time at the point in time when the ONU’s local counter
reaches a certain value. This information is communicated using an Ethernet Organization Specific Slow
Protocol (OSSP) message. The specific process, which accommodates the asymmetry between the
downstream and upstream transmissions delays, can be summarized as follows:
The OLT and ONU each compute their local latency factors. The ONU latency factor is the difference
between the ONU’s ingress latency and the scaled sum of its ingress and egress latencies. The OLT
latency factor is the difference between its egress latency and the scaled sum of its ingress and egress
latency. For both the ONU and OLT, there are two scaling factors. The first is the ratio of the downstream
index of refraction to the sum of the upstream and downstream indices of fraction, where the effective (i.e.,
wavelength-dependent) index values are used. In other words, it is the ratio of the upstream propagation
speed to the sum of the upstream and downstream propagation speeds. The second scaling factor is the
rateRatio, which is the ratio of the grandmaster clock frequency to the local clock frequency. As part of
that computation, the 802.1AS standard provides a mechanism by which the OLT clock master can
measure the rateRatio.
The OLT clock master selects a timing reference that is a future value X for its local MPCP counter. The
value of X is arbitrary as long as it is adequately far in the future to be communicated to the ONUs in time
and is within the current MPCP counter epoch. The clock master then calculates the value of
the synchronized ToD when the ONU slave MPCP counter will reach X. This time value at the
ONU will be ToD at count X at the OLT clock master, plus the difference between the OLT and
ONU latency factors, plus the scaled RTT (the RTT scaling factors are, once again, the rateRatio and the
ratio of the upstream propagation speed to the sum of the upstream and downstream propagation speeds).
The OLT then uses the TIMESYNC message to send the ONU the value X and the adjusted ToD value that
its local counter should have when it reaches a count of X.
3.6 Overview of the IEEE 1904.1 Service Interoperability in Ethernet Passive
Optical Networks (SIEPON)
IEEE 802.3 specified the Layer 1 and Layer 2 aspects of EPON and 10G EPON. However, additional
specifications are required in order to allow equipment from multiple vendors to interoperate in a network.
The ITU-T standardizes these areas for G-PON and XG-PON, but they are outside the scope of the IEEE
802 activities, so the IEEE Communications Society consequently launched the P1904.1 project to
address them [8]. The reference architecture for 1904.1 is shown in Figure 3.12, which illustrates the
relative scope of the 802.3 and 1904.1 standards. Service-specific functions are optional on either an OLT
64 Broadband Access
Covered by IEEE Standard 1904.1
Covered by IEEE Standard 802.3
OLT
Servicespecific
functions
Service
OLT
802.3
clients
Client
OLT
LEGEND:
OLT_MDI/ONU_MDI = Medium Independent Interface
OLT_LI/ONU_LI = Interface between the L-OLT and C-OLT / L-ONU and C-ONU
OLT_CI/ONU_CI = Client Interface
ODN = Optical Distribution Network
802.3
stack
Line
OLT
ODN
OLT_MDIOLT_LIOLT_CINNI
802.3
stack
Line
ONU
clients
Client
ONU
ONU
802.3
Service-
specific
functions
Service
ONU
UNIONU_CIONU_LIONU_MDI
Figure 3.12 SIEPON reference architecture
or ONU. Since the SIEPON primarily addresses topics at higher layers than those covered in this book,
this section will be restricted to an overview.
The major technical features of SIEPON include [9]:
Management.
•
QoS guarantees.
•
Multicast service delivery over EPON.
•
Power saving.
•
VLAN modes and tunneling.
•
Protection switching, including optical link monitoring.
•
Data encryption.
•
ONU authentication.
•
ONU discovery and maintenance.
•
Behavior of the MAC, MAC control and OAM clients.
•
Note that VLAN modes can be provisioned either for the entire ONU, or on a per-port basis.
The specifications are defined in terms of three service packages, each defining the required features for
that package. The SIEPON reference model is unified to all packages but all specific features are packagespecific. The feature requirements common to all packages are listed in Table 3.3, and the feature
requirements that are different among the packages are shown in Table 3.4.
The ONU is further broken down into logical elements.
Line-ONU (L-ONU), which represents the functions covered in IEEE 802.3/802.3av.
•
Client-ONU (C-ONU), which represents a logical layer comprised of at least one L-ONU function,
•
along with the associated clients (including the MAC Control, MAC, and OAM clients) that are
required for proper network operation per 802.3/802.3av.
The Service-ONU (S-ONU), which is comprised of a C-ONU, at least one UNI, and optional additional
•
functionalities.
The OLT is similarly broken down into corresponding L-OLT, C-OLT and S-OLT elements. The
additional functions that may be supported by the S-OLT include switching, POTS, and service initiation
IEEE Passive Optical Networks 65
Table 3.3 Features Required for Package A, B and C
Required feature
REPORT MPCP format
Report queue length calculation
Queue service disciple
ONU authentication (including secure provisioning)
Management (eOAM-based)
Device and capability discovery
Software update
Management entities
Power saving
VLAN support by ONU and OLT
Multicast connectivity
Multicast coexistence
Note: While these feature requirements are common to all three packages, there may be
variations in how they are defined or implemented.
protocols to support delivering specific services to subscribers. Such services and solutions are typically
outside the scope of 1904.1.
3.6.1 SIEPON MAC Functional Blocks
The MAC functional blocks specified by SIEPON include the Input, Classifier, Modifier, Policer/Shaper,
Cross-Connect, Queue, Scheduler and Output blocks. Together, these blocks describe a unified data path
architecture, which allows uniform provisioning and interoperability.
The Input block is the ingress port that receives frames from the S-ONU or S-OLT (e.g., UNI, NNI, or
•
MAC service frames).
The Classifier function examines the frame headers in order to identify all frames, the EPON Service
•
Path (ESP) to which they belong, what actions are required for that frame, and which queue should
forward that frame. The Classifier operates on a set of rules that is composed of provisionable elements.
The Classifier output vector specifies the actions of the Modifier, Policer/Shaper, and Cross-connect.
The Modifer operates on the VLAN TAG information. Specifically, it is allowed to pass the tags, to add
•
or remove tags, replace/alter tag fields of the outermost one or two tags, or take no action. The fields that
it may modify are the TPID, PCP, CFI, DEI, or VID. The Modifier is also able to alter the IEEE 802.1ah
fields.
The Policer/Shaper enforces SLA conformance of the ESPs on a per-flow basis. It operates using a
•
token bucket mechanism based on four parameters: rate, burst, action-on-conformant-frames, and
action-on-non-conformant-frames. When functioning as a Policer, it deals with the coloring and
discard-eligibility of frames and delaying non-conformant frames. When functioning as a Shaper, it
manages the appropriate frame transmission delays.
The Cross-connect routes each frame to the appropriate queue. In the case of multicast or broadcast
•
flows, it replicates the associated packets and maps them to the appropriate set of queues.
The Queue holds frames until they are polled by the Scheduler so that they can be transmitted. In
•
addition to the data frames, the Queue inputs include control and coloring information. Its outputs
include unmarked data frame, alarms, and statistics.
A Schedule instance provides the multiplexing function for the frames stored within the subset of
•
Queue block queues that are provisioned to it. In the case of the ONU upstream transmission, the OLT
DBA controls its Scheduler through the mechanisms described earlier in this chapter. The Scheduler
Loading...