GIGABIT-CAPABLEPASSIVE OPTICAL

NETWORKS

GIGABIT-CAPABLEPASSIVE OPTICAL

NETWORKS

Dave HoodElmar Trojer

Copyright � 2012 by John Wiley & Sons. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey

Published simultaneously in Canada

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, scanning, or otherwise,

except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either

the prior written permission of the Publisher, or authorization through payment of the appropriate

per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,

978-750-8400, fax 978-750-4470, or on the web at www.copyright.com. Requests to the Publisher

for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc.,

111 River Street, Hoboken, NJ 07030, 201-748-6011, fax 201-748-6008, or online at

http://www.wiley.com/go/permission.

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. No warranty may be created or extended by sales

representatives or written sales materials. The advice and strategies contained herein may not be

suitable for your situation. You should consult with a professional where appropriate. Neither the publisher

nor author shall be liable for any loss of profit or any other commercial damages, including

but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact

our Customer Care Department within the United States at 877-762-2974, outside the United States

at 317-572-3993 or fax 317-572-4002.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print

may not be available in electronic formats. For more information about Wiley products, visit our

web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Hood, Dave, 1945-

Gigabit-capable passive optical networks / Dave Hood, Elmar Trojer.

p. cm.

Includes bibliographical references and index.

ISBN 978-0-470-93687-0 (cloth)

1. Passive optical networks. 2. Gigabit communications. I. Trojer, Elmar. II. Title.

TK5103.592.P38H66 2011

621.380275–dc232011028223

Printed in the United States of America

10 9 8 7 6 5 4 3 2 1

http://www.copyright.comhttp://www.wiley.com/go/permissionhttp://www.wiley.com

To Kent McCammon

CONTENTS

Acknowledgments xi

1 Introduction 1

1.1 Target Audience / 3

1.2 Evolution of G-PON Technology and Standards / 3

2 System Requirements 9

2.1 G-PON Operation / 10

2.2 ONU Types / 13

2.3 Network Considerations / 16

2.4 OLT Variations and Reach Extenders / 26

2.5 ONU Powering / 33

2.6 Technology Requirements / 42

2.7 Management Requirements / 44

2.8 Maintenance / 46

3 Optical Layer 49

3.1 Introduction / 50

3.2 Optical Fiber / 54

3.3 Connectors and Splices / 61

3.4 WDM Devices and Optical Filters / 64

3.5 Passive Optical Splitters / 67

3.6 Power Budget / 71

3.7 Coexistence / 77

3.8 Optical Transmitters / 82

3.9 Optical Receivers / 91

3.10 G-PON Transceiver Modules / 106

vii

3.11 Optical Amplifiers / 113

3.12 Reach Extension / 118

4 Transmission Convergence Layer 127

4.1 Framing / 129

4.2 ONU Activation / 166

4.3 ONU Transmission Timing and Equalization Delay / 177

4.4 ONU Registration / 187

4.5 ONU Energy Conservation / 190

4.6 Security / 200

5 Management 219

5.1 The Toolkit / 219

5.2 Equipment Management / 240

5.3 Reach Extender Management / 249

5.4 PON Maintenance / 252

5.5 Obsolete Fragments of Information Model / 254

6 Services 255

6.1 Basic Ethernet Management / 256

6.2 Multicast / 275

6.3 Quality of Service / 285

6.4 IP Services / 303

6.5 POTS / 306

6.6 Pseudowires / 317

6.7 Digital Subscriber Line UNIs / 333

6.8 RF Video / 337

7 Other Technologies 339

7.1 Ethernet PON, EPON / 340

7.2 Wireless Broadband / 352

7.3 Copper / 353

7.4 Ethernet, Point to Point / 354

7.5 WDM PON / 357

7.6 Access Migration / 360

viii CONTENTS

Appendix I – FEC and HEC in G-PON 363

I.1 Redundancy and Error Correction / 363

I.2 Forward Error Correction / 364

I.3 Hybrid Error Correction / 374

Appendix II – PLOAM Messages 375

II.1. PLOAM Messages in G.987 XG-PON / 375

II.2. PLOAM Messages in G.984 G-PON / 388

References 403

International Telecommunications Union, Telecommunication

Standardization Sector / 403

Broadband Forum / 405

Institute of Electrical and Electronics Engineers / 405

Internet Engineering Task Force / 406

Other / 407

Acronyms 409

Index 423

CONTENTS ix

ACKNOWLEDGMENTS

We wish to thank the many who encouraged our effort, in particular, Dave Allan,

Dave Ayer, Chen Ling, Dave Cleary, Jack Cotton, Lou De Fonzo, Jacky Hood, Einar

In de Betou, Denis Khotimsky, Lynn Lu, KentMcCammon, DonMcCullough, Derek

Nesset, Peter €Ohlen, Dave Piehler, Albert Rafel, Bj€orn Skubic, and Mara Williamsfor their help in reviewing various parts of the manuscript and the premanuscript

studies. Tom Anschutz, Paul Feldman, Richard Goodson, David Sinicrope, and

Zheng Ruobin helped to clarify issues that came up in the course of the work.

We would especially like to recognize Dewi Williams, an ideal match for our

target reader profile, who was willing to provide intelligent and thoughtful comment

and discussion well beyond the call of duty.

Needless to say, the inevitable remaining errors and inconsistencies are entirely

our own responsibility.

Finally, special thanks to Jacky and Antonia for their support and patience through

the process.

PUBLISHER’S BRIEF REVIEW

Although thoroughly grounded in the G-PON standards, this book is far more than

just a rehash of the standards. Two experts in G-PON technology explain G-PON in a

way that is approachable without being superficial. As well as thorough coverage of

all aspects of G-PON and its 10Gb/s evolution into XG-PON, this book describes

the alternatives and the reasons for the choices that were made, the history and

the tradeoffs.

xi

1

INTRODUCTION

Fiber optic access networks have been a dream for at least 30 years. As speeds

increase, as the disparate networks of the past converge on Ethernet and IP (Internet

protocol), as the technology and business case improve, that dream is becoming

reality.

The access network is that part of the telecommunications network that connects

directly to subscriber endpoints. This book details one of the technologies for fiber in

the loop (FITL), namely gigabit-capable passive optical network (G-PON) technol-

ogy, along with its 10-Gb sibling XG-PON. Figure 1.1 shows how G-PON and XG-

PON fit into the telecommunications network hierarchy. This book is about the

G-PON family.

For quick reference, Table 1.1 summarizes the common PON technologies,

including both the G-PON and EPON families. The G-PON family is standardized

by the International Telecommunications Union—Telecommunication Standardiza-

tion Sector (ITU-T), while EPON comes from the Institute of Electrical and

Electronic Engineers (IEEE). Chapter 7 includes a comparison of G-PON and

EPON.

Because they share many properties, this book uses the term G-PON generically

to refer to either ITU-T G.984 or G.987 systems unless otherwise stated. Where a

distinction needs to be made, we make it explicit: G.984 G-PON or G.987 XG-PON.

Figure 1.2 illustrates the fundamental components of a PON. The head end is

called the optical line terminal (OLT). It usually resides in a central office and usually

Gigabit-capable Passive Optical Networks, First Edition. Dave Hood and Elmar Trojer.� 2012 John Wiley & Sons, Inc. Published 2012 by John Wiley & Sons, Inc.

1

serves more than one PON.* The PON contains a trunk fiber feeding an optical power

splitter, or often a tree of splitters. From the splitter, a separate drop fiber goes to each

subscriber, where it terminates on an optical network unit (ONU). ONUs of various

kinds offer a full panoply of telecommunications services to the subscriber.

Telecommunications networks

Access networks

Wireline access networks

Optical access networks

Passive optical networks: PONs

G-PON and XG-PON

1G-EPON and 10G-

EPON

Point-to-point fiberand WDM

PON

Twisted pair copper networks

Co-axial cable

networks

Wireless networks

Core and aggregation networks

Figure 1.1 G-PON taxonomy.

* The term OLT is therefore ambiguous: it may refer only to the terminating optoelectronics and MAC

functionality of a single PON or it may mean the entire access node, terminating a number of PONs and

forwarding traffic to and from an aggregation network.

TABLE 1.1 PON Family Values

Technology Standard Speed

G-PON ITU-T G.984 2.5Gb/s downstream, 1.25Gb/s upstream

XG-PON ITU-T G.987 10Gb/s downstream, 2.5Gb/s upstream

EPON IEEE 802.3 1Gb/s symmetric

10G-EPON IEEE 802.3 10Gb/s downstream, 1 or 10Gb/s upstream

OLT

ONU

Optical power splitter

Opticaldrop

Optical trunkline OTL

Access network

Optical distribution network (ODN)PON

ONU

Downstream Upstream

Bidirectional single mode optical fiber

To aggregation and core network

Subscriber premises network

Figure 1.2 G-PON terminology.

2 INTRODUCTION

A single-fiber connection is used for both directions, through specification of

separate wavelengths for each direction. As we shall see, wavelength separation

also allows for coexistence of other technologies on the same optical distribution

network (ODN).

After a brief overview and history in this chapter, Chapter 2 outlines the require-

ments and constraints of a G-PON access network. Chapter 3 explains the optical layer

of the network. Moving up the stack, Chapter 4 covers the transmission convergence

layer, the home of most of the features that uniquely distinguish G-PON from other

access technologies. Chapter 5 introduces the management model in the context of

equipment and software management, while Chapter 6 shows how the management

model is used to construct telecommunications services. Finally, Chapter 7 describes

current and future alternatives, competitors, and partners of G-PON.

Two appendices, a list of references, a guide to acronyms and abbreviations, and

an index appear at the end of the book.

1.1 TARGET AUDIENCE

This book is written for the experienced telecommunications or data communica-

tions professional whose knowledge base does not yet extend into the domain of PON

or, in particular, G-PON and XG-PON. We also address this book to the advanced

student, who cannot be expected to have a grounding in the ancient and forgotten lore

of telecoms. Hoping to strike a balance against excessive redundancy, we neverthe-

less include a certain amount of background material, for example, on DS1 and E1

TDM services, and we always try to indicate where to find additional information.

We argue that a simple restatement of the standards adds no value. Accordingly,

we structure this book with a view toward explaining and comparing the standards,

rather than simply paraphrasing them. This is most evident in the frequent side-by-

side comparisons of G-PON and XG-PON. This book also addresses many important

aspects of real-world access networks that lie beyond the scope of the standards.

Disclaimers The complete and authoritative specifications are in the standards

themselves. While we make every attempt to be accurate, we have necessarily elided

any number of secondary details, especially in the peripheral standards. We trust that

the reader who ventures into the formal standards will find few surprises. Although

both authors are employed by Ericsson, we should also state that this book is not

sponsored by Ericsson, and the views expressed do not necessarily represent Ericsson

positions.

1.2 EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS

PON technology began in the 1980s with the idea of a fiber ring dropping service to

each subscriber. Ring topology was abandoned early for reasons that will be apparent

EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 3

upon reflection, and subsequent PONs have been based on optical trees. In the early

days, several companies* developed products around the integrated services digital

network (ISDN) standards. The systems delivered plain old telephone service

(POTS), but offered few advantages over copper-fed POTS. Although there were

some deployments, the technology (cost) and the market need (revenue) were too far

apart to justify a realistic business case.

By the 1990s, optical communications were starting to mature in the long-haul

network, speeds were increasing, and the industry was starting down the cost curve

on the technology side. At the same time, a market for Internet access was

developing, and subscribers were increasingly frustrated with modems running at

9.6 or even at the once-impressive speed of 56 kb/s. The PON industry tried again.

The first generation of what might be termed modern PON was based on

asynchronous transfer mode (ATM), originally designated A-PON. Significant

commercial deployments of ATM PON occurred under the moniker B-PON (broad-

band PON). B-PON was standardized and rolled out in the last years of the twentieth

century.

Like G-PON, B-PON is defined by the ITU-T, in the G.983 series of recom-

mendations. Several data rates are standardized. Early deployments delivered data

services only, at aggregate bit rates of 155Mb/s, both upstream and down. This is one

of the bit rates used by the synchronous digital hierarchy (SDH), an optical transport

technology developed during the 1980s and 1990s and widely deployed today,

albeit having evolved over the years to the higher rates of 2.5 and 10Gb/s and beyond.

B-PON specified the same bit rates as SDH, with the intention to reuse component

technology and also parts of the SDH standards for specifications such as jitter.

In the early years of the millennium, B-PON matured in several ways:

. Service definitions expanded beyond best-efforts data, most notably to include

POTS and voice over Internet protocol (VoIP).

. The original downstream wavelength spectrum was redefined into two bands, a

basic band for use by the B-PON protocols and an enhancement band intended

for radio frequency (RF) content such as broadcast video. Spectrum was also

identified for use by independent dense wavelength division multiplex

(DWDM) access, coexisting on the same fiber plant.

. To improve the utilization of upstream capacity, dynamic bandwidth allocation

(DBA) was defined and standardized.

. Although the common upstream rate remained at 155Mb/s, the technology,

cost, and market requirements had evolved to the point that 622Mb/s—another

SDH speed—became the expected downstream rate.

*One of the original companies—DSC-Optilink—can be tied to today’s Alcatel-Lucent; another—

Raynet—can be linked to Ericsson.

4 INTRODUCTION

Reasonable B-PON deployment volume was achieved at these levels. At the time of

writing, in 2011, B-PON was still being installed to fill out empty slots in existing

chassis.

Although the B-PON standards are ITU-T recommendations, a group called

FSAN (Full Service Access Network) guided the requirements and recommenda-

tions and continues to guide its successors to this day. FSAN is an informal

organization of telecommunications operators founded in 1995. Unlike formal

standards development organizations (SDOs) such as ITU-T and IEEE, and non-

SDOs such as Broadband Forum (BBF), FSAN has no membership fees and no staff.

Equipment and component vendors are members by invitation only. While FSAN

emphasizes that it is not an SDO, the same companies and the same people carry their

discussions from FSAN into ITU-T for the formal standardization work, often on

successive days of the same meeting. In the early days, the text of the recommenda-

tions was actually developed under the FSAN umbrella, then passed to ITU-T for

formal review and consent.

Around the turn of the millennium, digital video began to come out of the lab. The

bandwidth limitations of B-PON became a concern—622Mb/s distributed across 32

subscribers is only (!) 20Mb/s average rate per subscriber, much of which would be

consumed by a single contemporary high-definition digital video stream—while the

cost of technology had continued to improve. Standardization discussions began on a

new generation of PON, this one known as gigabit-capable PON, or G-PON. The first

G-PON standards, the ITU-T G.984 series, were published in 2003. As with B-PON,

the G-PON standards recognize several data rates, but the only rate of practical

interest runs at 2.488Gb/s downstream, with 1.244Gb/s in the upstream direction,

capacity shared among the ONUs. These are also SDH bit rates.

The initial versions of the G.984 series recognized the ATM of B-PON, but ATM

was subsequently deprecated as a fading legacy technology. Another capability that

was initially standardized and later deprecated was provision for G-PON to directly

carry TDM (time division multiplex) traffic. This might have been useful for services

such as DS1 (digital signal level 1) or SDH, but detailed mappings were never

defined, and it was overtaken by the development of standards for pseudowires, about

which we shall learn in Chapter 6.

The only form of payload transport that remains in G.984 today is the G-PON

encapsulation method (GEM), usually encapsulating Ethernet frames.* It is perfectly

accurate to think of G-PON as an Ethernet transport network, notwithstanding

marketing claims from the EPON competition that G-PON is not real Ethernet.

Another important evolutionary step from B-PON to G-PON is accommodation

of the operators’ requirement for incremental upgrade of already deployed installa-

tions. The B-PONwavelength plan did not provide for the coexistence of B-PON and

G-PON on the same optical network. The eventual need to upgrade access network

technology thus presented a dilemma:

*Other mappings are also defined in G.984 G-PON; some were recognized to be of no market interest

and were not carried forward into G.987 XG-PON. The only additional mapping in G.987 XG-PON is

multiprotocal label switching (MPLS) over GEM. Time will tell whether it is useful.

EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 5

. It was usually not economical to install a new optical distribution network,

particularly the distribution and drop segments, in parallel with an existing one.

. It was not feasible to replace all ONUs on a PON at the same instant. Imagine an

army of 32 service technicians calling on 32 subscribers at precisely 10 AM

next Tuesday morning—or at any other time for that matter!

. It was unacceptable to shut down telecommunications service to a group of

subscribers for several hours or days to allow for a realistic number of service

technicians to schedule realistic service appointments with subscribers.

. And maybe only 1 of those 32 subscribers was willing to pay for upgraded

service anyway.

The upshot of this consideration was a requirement for G-PON networks to reserve

wavelengths to allow incremental upgrade to the next generation of PON technology,

whatever that might be, and to include the necessary wavelength blocking filters in

ONUs. Chapters 2 and 3 discuss this in further detail.

G-PON began to be deployed in substantial volume in 2008 and 2009.

Once a standard is implemented and deployed, it is natural to want to confirm that

everyone has the same interpretation and that the various implementations will

interwork. This led to a series of interoperability test events, beginning with the basic

ability of an OLT to discover and activate an ONU on the PON. The first G-PON

plugfest occurred in January 2006, and there have been two to four events per year

since then. Today’s G-PON equipment is largely interoperable, although the final

proof remains to be seen: there have not yet been widespread live deployments of

multivendor access networks.

As testing moved further up the stack, it became apparent that the flexibility of the

ONU management and control interface (OMCI) was not an unmixed blessing.

Different vendors supported given features in different ways. If the OLT tried to

provision the feature in oneway and the ONU supported only some different way, the

pair would not interoperate.

Interoperability was a primary motivation for standardization. FSAN therefore

created the OMCI implementation study group (OISG) with the charter to develop

best practices, recommendations for the preferred ways to implement various

features. OISG was and is a vendors-only association, theoretically free to discuss

implementation considerations under mutual nondisclosure agreements.

In 2009, OISG released an implementers’ guide of OMCI best practices, originally

published as a supplement to the OMCI specification ITU-TG.984.4. As G.984.4 was

migrated into G.988, the implementers’ guide material was incorporated into G.988,

where it resides today.

OMCI best practices continue to evolve as minor questions arise, but the issues

that spawned OISG have largely been resolved. OISG’s charter also evolved and it

became an early preview forum for OMCI maintenance, an opportunity for sanity

checks and consensus building before new OMCI proposals were formally submitted

to the ITU-T process.

6 INTRODUCTION

Although OISG has not been formally disbanded, it is now dormant, both because

of its success at resolving interoperability issues and because of the shift of

responsibility from FSAN to BBF. Test plans are published as BBF technical reports,

specifically TR-247 and TR-255. Responsibility for plugfests also shifted from

FSAN to.

Broadband Forum entered the G-PON scene only recently, but in a major way.

Previously known as DSL Forum, BBF changed its name and expanded its scope to

include, among other things, the entire access network and everything attached to it.

In 2006, BBF published TR-101, which defined requirements for migration of the

access network from ATM to Ethernet, but still with a DSL mind-set. The ink had

scarcely dried on TR-101 when BBF began a project to define its applicability to the

special aspects of G-PON, resulting in TR-156 (2008).

Although there are some rough edges at the organizational boundaries, the scope

addressed by BBF is theoretically disjoint from the scope of the ITU-T recommenda-

tions. ITU-T specifies an interface between OLT and ONU and the fundamentals

of its operation. ITU-T includes tools for maintenance and tools from which

applications can be constructed and extends the tool set as necessary when new

applications arise.

In contrast, BBF views the overall network architecture as its scope. BBF takes the

ITU-T tools as a given, identifies preferred configuration options, and writes

network- and service-level requirements to serve these options. Vendors and

operators are, of course, free to develop other applications on the same base, but

if BBF does its job well, its model architectures prove to be satisfactory for most real-

world needs and are suitable as reference models even for applications that lie

beyond the strict bounds of the BBF architecture.

If BBF does its job well? In fact, TR-156 has largely been accepted by operators

worldwide as a satisfactory model for simple ONUs that deliver Ethernet service

to end users. Additional BBF technical reports (TR-142, TR-167) define how the

G-PON toolkit can be used to control only the ONU’s PON interface, with the

remainder of the ONU managed through other means. Chapter 2 expands this topic.

Once the G-PON standards began to mature and vendors busied themselves

bringing product to market, FSAN turned its attention to the question of the next

logical step after G-PON. Starting in about 2007, as the outline became clearer, the

operators launched a white paper project to define the requirements for the next

generation. FSAN completed the next-gen white paper in mid-2009. Parallel work

had begun in late 2008 to develop the details of the necessary recommendations.

FSAN structured its view of the future into two domains: Next-gen 1 (NG-1) was

the set of PON architectures that was required to coexist on the same fiber

distribution network with G-PON, while next-gen 2 designated the realm of

possibilities freed from that constraint, an invitation to take a long view of technology

to see what might make sense at some unspecified time in the future.

As it turned out, NG-1 PON bifurcated further, into versions known as XG-PON1

and XG-PON2, or often just XG-1 and XG-2, where X is the Roman numeral 10,

designating the nominal 10 Gb/s downstream rate. Both versions run downstream

data at 9.953Gb/s, another SDH bit rate. The upstream capacity of XG-PON1 is

EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 7

2.488Gb/s, while XG-PON2 runs at 9.953Gb/s upstream. The reason for the

distinction was the substantial difference in technological challenge, coupled with

the perception that market need was insufficient to justify the significantly higher

cost of high-speed upstream links.

XG-PON1 is standardized in the ITU-T G.987 series of recommendations. At the

time of writing, XG-PON2 is being held in abeyance, pending the convergence of

market demand and technological feasibility. Because IEEE has already standard-

ized a symmetric 10G form of EPON (10G-EPON, Chapter 7), it is possible that

XG-PON2 will not be pursued further. If that comes to pass, operators who need

symmetric 10G will deploy 10G-EPON, and the G-PON community will work

toward next-next-generation access, probably WDM PON.

8 INTRODUCTION

2

SYSTEM REQUIREMENTS

In this chapter:

. Overview of power-splitting PONs

. Optical network considerations

T Split ratio

T Maximum reach, differential reach

T PON protection

. Reach extenders

. Coexistence with future generations

. Types of ONU

. ONU powering

. Introduction to

T Dynamic bandwidth assignment; details in Chapter 6

T Quality of service; details in Chapter 6

T Security; details in Chapter 4

T Management; details in Chapter 5

Before diving into the details of how a G-PON works, we need to understand

something about the business case. We return repeatedly to business case questions

over the course of this book because, ultimately, everything we do must add value to

someone for something.

Gigabit-capable Passive Optical Networks, First Edition. Dave Hood and Elmar Trojer.� 2012 John Wiley & Sons, Inc. Published 2012 by John Wiley & Sons, Inc.

9

As with most companies, telecommunications operators are driven forward by

market opportunity, cost reduction, and competitive pressure, and they are held back

by existing investment and existing practices. New technology is comparatively easy

to justify in a greenfield development—we have to do something, so let us go for the

latest and greatest!—but most of the potential market is already served in one way or

another, even if it’s no more than ADSL (asymmetric digital subscriber line) from

a central office. The difficulty arises in making a business case for the deployment

of a new technology that may be of immediate interest to only a small number of

existing subscribers, in what is called, for contrast, a brownfield.

Civil works—right of way acquisition, permits, trenching for underground cable,

poles for aerial cable—are a very large part of the up-front cost of a change in

technology, for example, from copper to fiber. Estimates range from 65 to as much as

80% of the total cost. No matter how economical the equipment itself may be, this

cost must be paid. Once the business case has been made to install new fiber in the

outside plant infrastructure, it makes sense to place large fiber-count cables, or at

least a lot of empty ducts, through which fiber can easily be blown at a later date. The

cost of additional ducted fibers is comparatively small, even vanishingly small, in

fiber trunks. With the optical infrastructure in place and spare fibers available, it

becomes much easier to take subsequent evolutionary steps.

It is easier to develop a business case if all telecommunications services can be

provided by a single network. This is the idea behind the oft-heard term convergence,

a concerted effort to eliminate parallel networks, each of which serves only a subset

of the service mix. Software-defined features, Ethernet, and IP are major steps along

the road to convergence. The contribution of standards and of the network equipment

is to ensure that the investment, once made, can be used for a complete range of

services for decades to come. In keeping with the full-service focus of its FSAN

parent, G-PON is designed to deliver any telecommunications service that may be

needed.

2.1 G-PON OPERATION

2.1.1 Physical Layer

To recapitulate the brief overview in Chapter 1, a PON in general, and a G-PON in

particular, is built on a single-fiber optical network whose topology is a tree, as shown

in Figure 2.1. The OLT is at the root, and some number of ONUs connect at the

leaves. Downstream optical power from the OLT is split at the branching points of

the tree. Each split allocates an equal fraction of the power to each branch. The

achievable reach of a PON is a tradeoff of fiber loss against the division of power at

the splitter. Chapter 3 describes splitters and fiber loss in detail.

A power splitter is symmetric: Its loss upstream is the same as down, 3 dB for each

power of 2 in the split ratio.

The OLT transmits a continuous downstream signal that conveys timing, control,

management, and payload to the ONUs. The OLT is master of the PON. Based on

10 SYSTEM REQUIREMENTS

service-level commitments and traffic offered by the ONUs, the OLT continuously

develops an upstream capacity allocation plan for the near future—typically 1 or

2ms—and transmits this so-called bandwidth map to the ONUs. The ONU is

permitted to transmit only when explicitly given permission by a grant contained

in a bandwidth map. During its allocated time, the ONU sends a burst of data

upstream, data that includes control, management, and payload.

For the bursts to arrive at the OLT at precisely the proper interleaved times, each

ONU must offset its notion of a zero reference transmission time by a value

determined by its round-trip delay,* the time it takes for the signal from the OLT

to reach the ONU, plus ONU processing delay and the time it takes for the signal

from the ONU to reach the OLT. The OLT measures the round-trip delay of each

ONU during activation and programs the ONU with the compensating equalization

delay value.

Another low-level requirement on a PON is the discovery of new ONUs, be they

either newly installed devices or existing devices that have been offline for reasons

such as fiber failure or absence of power, whether intentional or not. The OLT

periodically broadcasts a discovery grant, which authorizes any ONU that is not yet

activated on the PON to transmit its identity. Since the round-trip time of a new ONU

is unknown, the OLT opens a quiet window, a discovery window, also called a

ranging window, a time interval during which only unactivated ONUs are permitted

to transmit.

It is possible that more than one unactivated ONU could attempt to activate at

the same time; if their transmissions overlapped in time, neither would succeed.

Worse, they could deadlock, repeatedly colliding on every discovery grant forever.

The ranging protocol therefore specifies that the ONU introduce a random delay in its

response to the OLT’s invitation. Even though the transmissions from two ONUs

may collide during a given discovery cycle, they will sooner or later appear as distinct

activation requests in some subsequent interval.

The size of the discovery window depends on the expected fiber distance between

the farthest possible ONU and the nearest possible ONU. This is called maximum

Downstream

SplitterOLTONU

Bidirectional single mode optical fiber

Optical power splitter

Aggregation and core network

Trunk fiber,Optical trunk link OTL

Optical drop

Home network

Access network

PONUpstream

Figure 2.1 Tree structure of a PON.

*True, the correction could instead be applied locally by the OLTas an offset in the bandwidth map, but it is

not.

G-PON OPERATION 11

differential reach, standardized with 10-, 20-, and 40-km options in G.984 G-PON.

In G.987 XG-PON, the maximum differential reach options are 20 and 40 km.

Chapter 4 goes into detail on all of these aspects of G-PON operation.

2.1.2 Layer 2

In terms of the OSI seven-layer communications model,* the access network largely

exists at layer 2. Perhaps the single most important concept underlying an Ethernet-

based access network—which G-PON is—is that of the virtual local area network

(VLAN), specified in IEEE 802.1Q. A very substantial part of the hardware,

software, and management of a G-PON is dedicated to classifying traffic into

VLANs, then forwarding the traffic according to VLAN to the right place with the

right quality of service (QoS). Although an ONU is modeled as an IEEE 802 MAC

bridge, MAC addresses are usually less important at the ONU than are VLAN tags.

The access network operates at layer 2, but it judiciously includes some layer 3

functions as well, particularly for multicast management. For practical purposes,

multicast means IPTV (Internet protocol television) service; it is expected to

represent a large fraction of the traffic and to yield a large part of the revenue

derived from a G-PON. The PON architecture is ideally suited for multicast

applications because a single copy of a multicast signal on the fiber can be

intercepted by as many ONUs as need it. Each ONU extracts only the multicast

groups (video channels) that are requested by its subscribers.

To determine which groups are requested at any given time, the ONU includes at

least an IGMP/MLD{ snoop function, about which we shall learn more in Chapter 6.

Snooping involves monitoring transmissions from the subscriber’s set-top box

(STB), based on which the ONU compares the requested channels with a local

access control list (ACL). If the requested content is authorized and is already

available on the PON, the ONU delivers it immediately without further ado. Also

acting as an IGMP/MLD snoop, or more likely as a proxy, the OLT likewise

determines whether a given multicast group is already available, or whether it needs

to be requested from yet a higher authority. As seen by a multicast router further up

the hierarchy, a proxy aggregates a number of physical STBs into a single virtual

STB, thereby avoiding unnecessary messages to the router and improving network

scalability.

It is an open question what statistical capacity gain should be expected from

multicast, now and in the future. Even if 80% of subscribers are watching the same

10 channels, a long statistical tail would require substantial capacity to carry

content of interest only to the remaining few. Will a PON with 50 subscribers,

each with 2 or more television sets and a recording device, need 50 multicast groups?

Thirty? Twenty?

* See ITU-T X.200.{ IGMP: Internet group management protocol (IPv4), MLD: multicast listener discovery (IPv6). We

usually spell out acronyms the first time they appear; acronyms are also listed in a separate section at the

end of the book.

12 SYSTEM REQUIREMENTS

There is a general expectation that video will move toward unicast, but no one is

prepared to say how soon. At the end of the day, it may not matter. The considerations

described above suggest that we should expect a busy hour load of two or three

multicast groups per subscriber. At bit rates on the order of 5Mb/s per multicast

group, G-PON has enough downstream capacity for that level of loading. If it were to

materialize, mass market demand for ultrahigh bandwidth unicast video, up to

65Mb/s per channel, could motivate further access network upgrade.

2.2 ONU TYPES

A variety of product configurations seeks to fit the range of operators’ needs

completely and optimally. Here we outline a few of the possibilities.

2.2.1 Single-Family ONU

At least in some markets, the single-family unit (SFU) is the most common form of

ONU. Predictably, there are many variations on the SFU theme. The SFU may be

located indoors or out. Power is always supplied by the subscriber, but the SFU may

or may not include battery backup. The SFU may be regarded as a part of the

telecommunications network, owned and managed by the operator, or it may be

considered to be customer premises equipment (CPE), owned by the subscriber.

The simplest SFU, such as the one illustrated in Figure 2.2 with its cover off,

delivers one Ethernet drop; it is essentially a G-PON-to-Ethernet conversion device.

This one is intended to be mounted on a wall, at the demarcation point between the

drop fiber and the subscriber’s home network. The single Ethernet feed would then

Figure 2.2 Single-family ONU with power brick.

ONU TYPES 13

be connected to a residential gateway (RG) at some location convenient to the

subscriber’s device layout.

Many SFUs, such as the one in Figure 2.3, add value by including several bridged

Ethernet drops, suitable for direct connection to several subscriber devices, for

example, two or three PCs and a set-top box. Some may also include built-in

terminations for one or two POTS lines. Other applications for home use might

include low-rate telemetry, for example, to read utility meters or to monitor intrusion

detectors. M2M (machine to machine) communications are expected to mushroom

over the next few years, and the SFU will surely play a part in backhauling

information to centralized servers.

The SFUmay also include a full residential gateway, with firewall, NAT (network

address translation) router, DHCP (dynamic host configuration protocol) server,

802.11 wireless access, USB ports, storage or print server, and more. This form of

SFU is typically managed jointly by the subscriber, by the ONU management and

control interface (OMCI, G.988) model of G-PON, and by an access control server

(ACS), the latter as defined in various Broadband Forum technical reports and

frequently short-handed as TR-69.*

2.2.2 Multi-Dwelling Unit ONU

The multiple dwelling unit (MDU) is an ONU that serves a number of residential

subscribers. It may be deployed in an apartment building, a condominium complex,

or at the curbside. The MDU is always considered to be part of the telecommunica-

tions network; that is, its power, management, and maintenance are the responsibility

of the operator. Depending on their target markets, MDUs typically serve from 8 to

24 subscribers. Very similar to the MDU, a G-PON-fed digital subscriber line access

multiplexer (DSLAM) may serve as many as 48 or even 96 subscribers.

Subscriber drops from an MDU may be Ethernet, but the IEEE 802.3 physical

layer is not specified to tolerate the stress of a full outdoor environment, specifically

lightning transients. Even if the MDU is housed indoors in the same building as the

Figure 2.3 SFU with enhanced functionality.

*BBF designates it TR-069. It is always pronounced without the zero, and we like to write it in the same

way we say it.

14 SYSTEM REQUIREMENTS

subscriber residences, it may be uneconomical to rewire the building with the cat-5

cable needed for Ethernet.

The alternative subscriber drop technology is DSL. When drops are short, the

preferred form of DSL is ITU-T G.993.2 VDSL2; such an MDUmay or may not also

offer POTS. Existing telephone-grade twisted pair runs from the MDU to the

subscriber premises, where there is a DSL modem and a splitter for POTS, if POTS

is included in the service. With the short drops implied by fiber to the curb, it is

feasible to deliver several tens of megabits per second—even 100Mb/s and more—

effectively overcoming the speed limitations of copper wiring. The rate-reach

maximum can be extended through bonding of services across two or more pairs,

while G.993.5 vectoring potentially increases attainable speed through crosstalk

cancellation.

2.2.3 Small-Business-Unit ONU

As well as the ubiquitous Ethernet service, a small business unit (SBU) is likely to

offer several POTS lines to a small-office customer. It may also support a few TDM

(time-division multiplex) services such as DS1 or E1 via pseudowire emulation

(Chapter 6 explains this). The SBU of Figure 2.4 has eight POTS lines, four Ethernet

drops, and four 2.048-Mb/s E1 TDM services.

The cellular backhaul unit (CBU) is a variation of the SBU—perhaps a new

category in its own right. In the cellular backhaul application, the ONU carries traffic

between the core network and a radio base station. Legacy mobile backhaul requires

interfaces such as DS1 or E1. As the cell network migrates from third to fourth

generation, Ethernet backhaul is displacing DS1 and E1. As well as the tightly

controlled frequency stability required of all TDM services, some wireless protocols

require a precise time of day reference, a function described in Chapter 4.

Another variation of the SBU is the multitenant unit (MTU), intended to be shared

by several small businesses. The target market is the small islands of commercial

activity common along major streets. The important distinction of the MTU from the

SBU is its need to isolate services one from another, both in terms of traffic—no

bridging between Ethernet ports—and in terms of service-level agreements (SLAs).

Figure 2.4 An SBU.

ONU TYPES 15

2.3 NETWORK CONSIDERATIONS

While some operators favor the CPE model, in which the ONU is indoors, located on

the subscriber’s desktop or perhaps mounted on an indoor wall, other operators wish

to deploy ONUs outdoors. To a considerable extent, this reflects a difference in the

operator’s perspective: ONU as part of the telecommunications network or ONU as

subscriber-owned device. Figure 2.5 illustrates such an outdoor ONU, which differs

from the device of Figure 2.2 in that it provides two POTS lines, as well as an

Ethernet drop, and is accessible to operator personnel without the need to enter the

subscriber’s home.

ONUs such as MDUs may go into equipment rooms or telecommunications

closets in buildings. ONUs may also be designed for curbside pedestals (Fig. 2.6) or

other outdoor housings, in which case they need to be fully hardened for outside plant

conditions. ONU components must generally be rated for the full industrial temper-

ature range, and ONU enclosures may be required to tolerate extremes of tempera-

ture and water exposure, including immersion (Fig. 2.7) and salt fog. Other

considerations for outdoor ONUs include lightning protection for all metallic wiring,

and insect and fungus resistance. All ONUs must satisfy regulatory requirements for

electromagnetic interference (EMI) generation and operator requirements for EMI

tolerance.

2.3.1 Power

ONU powering is indisputably a network consideration, but it warrants a separate

discussion in its own right. We defer this topic to Section 2.5.

Figure 2.5 Outdoor ONU, outer access cover open.

16 SYSTEM REQUIREMENTS