Gigabit-capable Passive Optical Networks E-Book

D. Hood

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 712

Veröffentlichungsjahr: 2012

Beschreibung

Gigabit-capable passive optical networks (G-PON) have a large and increasing base of support among telecommunications operators around the world. Written by two of the experts in the field, this book explains G-PON in detail, both the original 2.5 Gb/s version and XG-PON, the 10 Gb/s second generation. The foundation established by this book is also invaluable in understanding NG2 (next-generation 2) G-PON, which is built upon a number of XG-PON systems on parallel wavelengths. As well as a history that clarifies the reasons for many of the existing features, the book looks at current and evolving technology and discusses some of the alternatives for future access networks.

Leseprobe

Contents

Cover

Title Page

Copyright

Dedication

Acknowledgments

Publisher's Brief Review

Chapter 1: Introduction

1.1 Target Audience

1.2 Evolution of G-PON Technology and Standards

Chapter 2: System Requirements

2.1 G-PON Operation

2.2 ONU Types

2.3 Network Considerations

2.4 OLT Variations and Reach Extenders

2.5 ONU Powering

2.6 Technology Requirements

2.7 Management Requirements

2.8 Maintenance

Chapter 3: Optical Layer

3.1 Introduction

3.2 Optical Fiber

3.3 Connectors and Splices

3.4 WDM Devices and Optical Filters

3.5 Passive Optical Splitters

3.6 Power Budget

3.7 Coexistence

3.8 Optical Transmitters

3.9 Optical Receivers

3.10 G-PON Transceiver Modules

3.11 Optical Amplifiers

3.12 Reach Extension

Chapter 4: Transmission Convergence Layer

4.1 Framing

4.2 ONU ACTIVATION

4.3 ONU Transmission Timing and Equalization Delay

4.4 ONU registration

4.5 ONU Energy Conservation

4.6 Security

4.7 Update

Chapter 5: Management

5.1 The Toolkit

5.2 Equipment Management

5.3 Reach Extender Management

5.4 PON Maintenance

5.5 Obsolete Fragments of Information Model

5.6 Update

Chapter 6: Services

6.1 Basic Ethernet Management

6.2 Multicast

6.3 Quality of Service

6.4 IP Services

6.5 POTS

6.6 Pseudowires

6.7 Digital Subscriber Line UNIs

6.8 RF Video

Chapter 7: Other Technologies

7.1 Ethernet PON, EPON

7.2 Wireless Broadband

7.3 Copper

7.4 Ethernet, Point to Point

7.5 WDM PON

7.6 Access Migration

Appendix I: FEC and HEC in G-PON

I.1 Redundancy and Error Correction

I.2 Forward Error Correction

I.3 Hybrid Error Correction

Appendix II: PLOAM Messages

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

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

Update

References

Acronyms

Index

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 [phone], 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 [phone], 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 [phone].

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

621.38′275–dc23

2011028223

To Kent McCammon

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, Kent McCammon, Don McCullough, Derek Nesset, Peter Öhlen, Dave Piehler, Albert Rafel, Björn Skubic, and Mara Williams for 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 10 Gb/s evolution into XG-PON, this book describes the alternatives and the reasons for the choices that were made, the history and the tradeoffs.

Chapter 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) technology, 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.

Figure 1.1 G-PON taxonomy.

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 Standardization 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.

Table 1.1 PON Family Values.

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 serves more than one PON.1 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.

Figure 1.2 G-PON terminology.

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 requirements 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 communications 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 nevertheless 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 upon reflection, and subsequent PONs have been based on optical trees. In the early days, several companies2 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 (broadband 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 recommendations. Several data rates are standardized. Early deployments delivered data services only, at aggregate bit rates of 155 Mb/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 10 Gb/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 155 Mb/s, the technology, cost, and market requirements had evolved to the point that 622 Mb/s—another SDH speed—became the expected downstream rate.

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 recommendations 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 recommendations 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—622 Mb/s distributed across 32 subscribers is only (!) 20 Mb/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 2.488 Gb/s downstream, with 1.244 Gb/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.3 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 installations. The B-PON wavelength 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:

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 03 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 one way 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-T G.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.

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 recommendations. 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 9.953 Gb/s, another SDH bit rate. The upstream capacity of XG-PON1 is 2.488 Gb/s, while XG-PON2 runs at 9.953 Gb/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 standardized 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.

Notes

1. 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.

2. One of the original companies—DSC-Optilink—can be tied to today's Alcatel-Lucent; another—Raynet—can be linked to Ericsson.

3. 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.

Chapter 2

System Requirements

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.

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.

Figure 2.1 Tree structure of a PON.

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 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 2 ms—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,1 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 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,2 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/MLD3 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?

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 5 Mb/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 65 Mb/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 be connected to a residential gateway (RG) at some location convenient to the subscriber's device layout.

Figure 2.2 Single-family ONU with power brick.

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.

Figure 2.3 SFU with enhanced functionality.

The SFU may 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-69.4

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 telecommunications 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 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 MDU may 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 100 Mb/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.

Figure 2.4 An SBU.

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).

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.

Figure 2.5 Outdoor ONU, outer access cover open.

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 temperature range, and ONU enclosures may be required to tolerate extremes of temperature 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.

Figure 2.6 Chassis ONU.

Figure 2.7 Underground ONUs.

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.

2.3.2 Energy Conservation

Reducing the demand for power is an important topic. Power, especially remote power, is difficult and expensive to provide, and heat dissipation is a problem, especially in outdoor deployments subject to high ambient temperatures or direct sunlight.

The natural progress of technology is toward less power consumption for a given function. This is true not only in the silicon of the G-PON ONU itself, but perhaps more importantly, in the efficiency of the AC (alternating current) power converter and the backup battery and its charger.

Not least because it is politically correct, the operator community is interested in saving additional power by shutting down functions when they are not in use. This follows the fine tradition of POTS telephony, in which an on-hook line consumes no power.

Inactive user network interfaces (UNIs) are comparatively easy to power down, but the PON interface presents difficulties: if its PON receiver is powered down, how does the ONU know when a terminating call arrives? And if the ONU's transmitter is shut down, how does the OLT know that the ONU has not failed? The answer is to take only very brief naps, a few tens or perhaps a few hundreds of milliseconds at a time.

ITU-T supplement 45 to the G-PON recommendations outlines the energy-saving options, but the topic came to full maturity only in XG-PON. XG-PON defines two energy conservation modes, dozing and sleeping. In doze mode, the ONU keeps its receiver alive at all times. This is especially appropriate for one particular use case, IPTV, in which almost all of the traffic flows downstream. In contrast, sleep mode allows the ONU to shut down both transmitter and receiver. Both modes require the ONU to respond periodically, so that the OLT can confirm that the ONU is still alive and healthy, and to serve whatever traffic that may arrive.

Section 4.5 explores the energy conservation feature in detail.

2.3.3 Plug and Play

MDU ONUs are installed on engineering work orders and are maintained as telecoms equipment. While it is, of course, important that installation and provisioning be no more complex than necessary, it would never be expected that an arbitrary hitherto unknown and unexpected MDU might suddenly appear on the PON, with features and capacities only to be discovered after the fact.

At the other extreme is the desktop ONU. Some operators would like such an ONU to be purchased by the subscriber at the local electronics store and installed by simply plugging it in. The business aspects of installation can be dealt with: the subscriber calls the provider or browses to an introductory web page, signs up for service, provides billing details, receives some kind of license or login credentials, preferably implicit, whereupon everything just comes up and works.

More of an issue for the do-it-yourself subscriber is the physical installation. Although PON optics are rated to be eye safe, it is not really a good idea to leave optical fiber terminations exposed, launching even their small amount of invisible light in whatever random direction they lie. Not only that, but a single speck of dust in an optical connector can render the ONU nonfunctional; cleaning connectors requires tools and training beyond the level of the average subscriber. Because of the optical concerns, it may be that, even when the ONU becomes a commodity item, the operator will roll a truck to install it.

That said, we mention that the ONU of Figure 2.2 is intended to be installed just above floor level, with the optical connector facing down, minimizing concern about dust in the connector. The wall-mount unit and the optoelectronics module, suitably equipped with dust caps, can be uncovered and plugged together within a matter of seconds. So there may indeed be cases in which an ONU can effectively be installed or replaced directly by the subscriber.

2.3.4 How Far?

G-PON parameters specify a maximum reach of 60 km of fiber, with a maximum differential reach that defaults to 20 km.

Figure 2.8 illustrates what we mean by reach and differential reach. The reach is the total fiber distance from the OLT to the farthest ONU, in this case 30 km. Differential reach is the difference in fiber distance between the farthest and the nearest ONU. If our PON included only subscribers A, B, and C, the differential reach would be zero because each is 30 km from the OLT, measured along the fiber run. Add subscriber D to the PON, and the differential reach becomes 10 km.

Figure 2.8 Reach and differential reach.

It would be perverse to run the trunk fiber 20 km to a splitter, and then run a drop fiber back 10 km to subscriber A. In geographically spread-out locations such as imagined here, it often makes sense to deploy a cascade of splitters, as illustrated in Figure 2.9.

Figure 2.9 Multistage split.

The first splitter usually has a lower split ratio, typically 1:4. The shape of the serving area can be tailored by the locations of the splitters.

Reach and differential reach are primarily issues of upstream burst timing, which can be addressed by varying the OLT's delays and quiet intervals. But greater reach, or a larger split ratio, also imply greater optical loss.

The need to go further with more splits made it natural to define G-PON reach extenders (REs). In its simplest form, an RE is simply an optical amplifier or an optical/electrical regenerator in each direction. More sophisticated REs may extend a number of PONs, with the OLT (trunk) side either using a separate fiber for each PON or separate wavelengths on a single fiber. A multi-PON RE is also a likely candidate for trunk-side protection. Sections 2.4, 3.12, and 5.3 discuss reach extenders in more detail.

2.3.5 PON Protection

G-PON protocols do not directly support protection of the nature defined in classical transmission protocols such as SDH—linear or ring protection, for example, with or without bidirectional signaling—but several forms of PON protection are possible. Recommendation G.983.5 describes protection scenarios and works out the details of message exchanges. As far as we know, it was never implemented, and the protection definitions of G.984.3 and G.987.3 omit such details.

Figure 2.10 illustrates the simplest, namely OLT port protection. The trunk fiber is connected to the OLT with a colocated 2:1 splitter, at the cost of an additional 3 dB of loss. Both OLT ports receive the upstream signal, but only one port transmits at any given time. The OLT triggers protection switching if one of its ports fails or is unplugged, or if it declares loss of signal from all ONUs. The ONUs themselves do not know about PON protection. Depending on the OLT's architecture, fast switching is possible, less than the classical target of 50 ms. Depending on the OLT's architecture, it may be necessary to reinitialize or rerange the ONUs after a switch.

Figure 2.10 OLT port protection.

OLT port protection covers failures at the OLT itself but does not address issues such as cable cuts in the outside plant. Lack of protection against cable cuts is not necessarily a show stopper because cables in the access network are usually not routed diversely anyway. If a backhoe cuts one cable, it probably also cuts whatever redundant fiber might have been present in an adjacent cable. This is one reason why some operators are considering stationary wireless links for PON protection.

As shown in Figure 2.11, we can readily protect against cable cuts of the trunk fiber by using a 2:N splitter at the remote site. This layout also recovers 3 dB of optical budget that was lost in Figure 2.10. The ONUs cannot tell the difference between the feeders of Figures 2.10 and 2.11. As to the OLT, because the trunk cables are presumably routed diversely (else why protect them?), this protection design requires redetermination of the ONUs' equalization delays after a protection switch, although only the trunk delay differs. Both G-PON and XG-PON include the ability for the OLT to minimize recovery time by retiming a single ONU, deriving a correction factor, and broadcasting it to all ONUs.

Figure 2.11 Trunk fiber protection.

PON protection may be generalized to use ports on separate OLTs, thereby protecting against complete OLT failure. Further, the separate OLTs may be located in separate central offices, providing at least some degree of protection from large-scale disasters. Dual homing, as this is called, raises additional issues in coordinating PON provisioning, the uplinks from the OLTs, and the real-time switch-over between working and protect PONs.

In Figures 2.10 and 2.11, the ONUs need not know anything about PON protection. Figure 2.12 illustrates an ONU designed for protection, with two optical interfaces. It is possible for such an ONU to have only a single PON MAC (medium access control) device, but if we are going to pay for two optics modules, it could make sense to include two MAC interfaces, with the ability to carry traffic on both PONs at the same time, either duplicate traffic or extra traffic of low priority that could be dropped in the event of a switch.

Figure 2.12 Complete redundancy.

Another possible merit of Figure 2.12 is that only some, but not all, ONUs need be protected, for example, those serving business customers, large MDUs, DSLAMs, or mobile base stations. Figure 2.12 follows the classical precedent of SDH, a core network technology that generally justifies higher costs. Because of high development cost for a low-volume product, the market for dual MAC ONUs has not yet developed.

There are other ways to do protection, specifically Ethernet link aggregation (originally in IEEE 802.3ad, now in 802.1AX). Figure 2.13 illustrates how individual ONUs may be protected on an end-to-end basis, end-to-end from the layer 2 viewpoint, at least. In this configuration, the ONUs, PONs, and OLTs need know nothing about protection. Standards and equipment already exist, avoiding the need for the PON subnetwork to reinvent the wheel.

Figure 2.13 Link aggregation protection.

2.3.6 How Many?

If we wish to dedicate a 50-Mb/s average downstream data rate to each subscriber, a 2.5-Gb/s G-PON can serve about 50 subscribers; a 10-Gb/s XG-PON about 200. Some operators would like to serve 500 subscribers per PON; others would like to be able to deliver 100 Mb/s to each subscriber. Of course, multicast and bursty traffic patterns mean that these numbers are fairly arbitrary, but they do provide some indication of the capacity available.

When a PON is equipped with MDUs, it may be cost-effective to connect only 16 ONUs, or even fewer. For single-family ONUs, common planning numbers are 32–64 ONUs per PON. Although there is clearly a point of diminishing returns, operators find it economical to pay for higher split ratios, rather than installing additional fibers and OLT blades. Some operators talk about 128-way splits and even more. In the discussions leading up to XG-PON, a PON with 256 ONUs was the largest number anyone could imagine—but understanding how imagination works, the community allocated 10 bits to the ONU-ID, so that in theory, 10235 ONUs could be connected to an XG-PON.

The optical loss budget ranges from 28 dB (G-PON class B+) or 29 dB (XG-PON1 class N1), right up to 35 dB (XG-PON1 extended class E2). The standards put the options into the OLT as much as possible. Limiting the number of ONU types recognizes the fact that the ONU is the point at which high-volume components matter, and where the operator's inventory and logistics costs make a big difference. The OLT is also likely to support plug-in optics, while for cost reasons, the ONU is more likely to have integral optics.

Keep in mind that each 1:2 split costs something over 3 dB, so a 10-deep splitter (210 = 1024 ONUs) would pretty well use up the most aggressive optical budget, all by itself. Having said that, nothing prevents the development of a reach extender that could indeed support, say eight 128-way splitters from a single PON. Nothing, that is, but the operators' understandable reluctance to deploy powered and managed equipment deep in the field.

2.3.7 Coexistence

Although G-PON will have a long service life, the nature of progress is such that someday, G-PON will be superseded by technologies that better satisfy evolution in demand, in services, in technology, and in revenue. How will we someday replace G-PON with the next generation? It is safe to assume that the next generation, whatever it may be, will be based on single-mode optical fiber, to or near the subscribers' premises.

The easiest answer would be to install a new optical distribution network in parallel with the existing one, and when all is said and done, this may well be the least-bad solution at some point. But particularly in residential areas—beyond the first of several possible splitters in tandem—this may be difficult. There is no guarantee that there will be spare fibers or ducts in existing distribution cable, and laying new cable is very expensive. Nor is it feasible to visit 32 or 64 or 128 subscriber premises simultaneously to replace their ONUs.

Indeed, the most complicated factor in the evolution story is that only a few subscribers will need to be upgraded anyway—G-PON ONUs are expected to satisfy the needs of most users for many years to come—and it is hard to justify a large new investment for only that first pioneering upgrade subscriber, especially when the take rate may be quite modest for many years to come.

It is therefore required that G.984 G-PON and next-generation G.987 XG-PON coexist on the same ODN indefinitely, and further, that upgrade not disrupt existing services more than momentarily—zero disruption is the target. Coexistence is achieved through compatible wavelength plans and optical budgets, as discussed further in Chapter 3.

Beyond G.987 XG-PON, the technology options are open. Further migration is sure to be required on existing ODNs, coexisting with at least one of G.984 G-PON or G.987 XG-PON, and possibly both. WDM (wavelength division multiplexing) PON is regarded as a prime candidate, but its parameters remain under discussion. Chapter 7 outlines some of the issues and options of WDM PON.

2.3.8 Unbundling

For business benefit or regulatory compliance, more than one company may be involved in delivering telecommunications services to the subscribers of a PON.

In the context of a G-PON, suppose that company A owns the local network of optical cables or fiber ducts. Physical layer unbundling occurs when company A leases duct space or dark fibers to company B. Generally, this means that the fiber terminates at a fiber distribution frame and is patched to some separate network element that is owned or controlled by company B. Repairs to ducts and cables are the responsibility of company A.

Wavelength unbundling occurs when company A or B6 leases one or more of the wavelengths on the fiber to company C. Generally, this means that company A is responsible to provide a filter and to break out the contracted wavelengths to a fiber distribution frame for patching to separate network elements. The contract also binds all parties not to cause harmful mutual interference, for example, by transmitting excessive power levels. Physical repairs are the responsibility of company A.

In physical and wavelength unbundling, the lessor is free to modulate the fiber with its choice of signal format, subject to contracted channel characteristics and interference constraints.

In layer 2 unbundling, company A lights up the fiber with its own protocol—G-PON, for example—and company D leases capacity within that protocol. Typical lease parameters would include VLAN IDs and service-level commitments. The fiber terminates in an OLT owned by company A, and the unbundled stream is switched at layer 2 into network elements owned by company D. Diagnosis and repair is largely the responsibility of company A.

All of these options are important in terms of the companies' operations and business practices, but duct and fiber unbundling do not affect G-PON. Wavelength unbundling only affects G-PON in the sense of assigning wavelengths. In terms of G-PON requirements, layer 2 unbundling may include requirements to groom traffic into separate bundles, even when the committed QoS of one bundle is identical to that of another, differing only by contractual relationship.

One particular higher layer unbundling feature is wholesale multicast service, in which company A may offer IPTV bundles from companies E, F, G, and so on. This option has implications in the complexity of multicast provisioning, inasmuch as a subscriber may mix and match from a menu of offerings, some of which may overlap. Multicast, and in particular multiprovider multicast, is discussed in Chapter 6.

2.3.9 Synchronization

It is rather taken for granted that a G-PON OLT is timed from a stratum-traceable source, with at least stratum 47 and usually stratum 3 or 3E holdover, and a frequency accuracy within four parts per billion. The G-PON itself is synchronous, so a PON-derived frequency reference at the ONU is also stratum traceable. A stratum-traceable frequency reference is important for services such as DS1/E1 circuit emulation. An OLT may derive its timing reference from a building integrated timing supply (BITS), but if it is located in a controlled environment vault or a remote cabinet, the OLT may alternatively be timed via synchronous Ethernet or IEEE 1588.

As well as precise frequency, some radio protocols also require a precise time of day, preferably to be supplied by the mobile backhaul ONU. The underlying reason is that these technologies share spectrum on a time-divided basis among several devices. If separately located transmitters are to know when they are allowed to use the spectrum, they need an accurate time reference. One-microsecond accuracy was provisionally specified for G-PON, in the absence of a better value. As the community works through the standardization issues of next-generation radio systems, it appears that a G-PON system will be asked to reduce its allocation quite considerably, perhaps to as little as 100 ns. This accuracy is a question of hardware design, not a standards issue.

Time of day is not available from a frequency reference. Time of day can be conveyed via IETF (Internet Engineering Task Force) network time protocol [NTP, RFC (request for comments) 5905] or simple NTP (SNTP, RFC 2030). Time of day is also available from GPS (global positioning system) receivers, which are regarded as too expensive to be desirable in every endpoint—nor can every endpoint rely on having a clear view of the sky. The favored candidate for time distribution is IEEE 1588.

The baseline assumption for packet timing is that delay through the network is (a) short, (b) symmetric, and (c) stable. None of these is necessarily true in a G-PON, where the upstream direction is delayed and subject to bandwidth allocation irregularities. Chapter 4 explains how the G-PON protocols include a way to transport time of day over the PON, using the PON ranging parameters for each given ONU. Transparent timing is also possible, in which the equipment merely records and forwards the transit delay of each given timing packet, a delay that can subsequently be used as a correction factor.

2.4 OLT Variations and Reach Extenders

The OLT is the interface between the PON and the telecommunications aggregation or core network. Conceptually, it is located in a central office, but in practice it may be located in a controlled environment vault (CEV) or an outdoor cabinet, as a way to extend the reach of the PON. Another way to extend reach is the so-called reach extender (RE; Fig. 2.14). Conceptually, a reach extender is just a repeater, either based on optical amplification or on electrical regeneration. The reach extender is usually located at the same site as the splitter; indeed the splitter may be integrated into the RE equipment itself.

Figure 2.14 Reach extender.

Because a reach extender requires power, management, and possibly facility protection, it makes economic sense to extend several PONs with a single equipment unit. In this case, the reach extender may have one trunk fiber per PON, or may multiplex several PONs onto a single trunk fiber through WDM, either coarse (CWDM) or dense (DWDM).

2.4.1 Why Reach Extenders? The Business Case

In Payne et al. (2006) British Telecom (BT) observes that the demand for bandwidth increases faster than can be supported by the combination of revenue growth—subscribers want more bandwidth but are not willing to pay very much for it—and the normal year-over-year erosion of equipment cost. This makes it difficult to develop a business case that justifies investment for broadband access, be it G-PON or anything else. Some other economic factor must be folded into the analysis.

In the absence of a clearly visible killer app that will completely redefine the economics of telecommunications, operators look to cost reduction. The BT chapter in Payne et al. (2006) summarizes a study in which a number of best-case assumptions were made as a way to understand the best possible cost savings.

The study concluded that, ignoring the real-world issues, a dual-homed access network with a reach of 100 km could allow as few as 100 well-chosen local exchanges (central offices) to cover the United Kingdom, replacing the 5000 that exist today. Exchange consolidation could represent a major cost savings.

In view of the real world, in particular the capabilities of G-PON:

Substantial exchange consolidation is possible, even with only 20 km of reach. Twenty kilometers far exceeds the range achievable with the current exchange-fed copper infrastructure.G-PON's reach could be extended with C+ optics (explained in Chapter 3). Under this assumption, a very high percentage of the United Kingdom's population could be served with dual-homed G-PON.

It will not come as a surprise to learn that BT is very interested in extending the reach of G-PON in any way possible, or that BT continues to push for dual-homed redundancy.

If the optical network cannot be completely passive, BT would like to see the simplest possible reach extenders, ideally nothing more than optical amplifiers in footway enclosures. BT views this as a better choice than remote OLTs. As much as anything, this preference is a consequence of the increased power demanded by a remote OLT, deployed in an environment where every watt is precious.

In Edmon et al. (2006), SBC (now part of AT&T) considers somewhat the same problem in light of U.S. geography. They conclude that fiber to the home (G-PON) is the right solution for greenfield deployment. There is no question that new cable must be installed to serve new subdivisions, and it might as well be optical fiber. Greenfield developments are likely to be well away from the central city, so reach is an issue. Like most operators, AT&T has consistently pushed for increased optical budgets, just a few decibels more. Each decibel expands the circle that a central-office-based OLT can serve, and like BT, AT&T is keenly aware of the disproportionately higher cost of remote siting. These discussions have led to higher loss budget classes in both G-PON (32 dB C+) and XG-PON (extended classes up to 35 dB).