Microprocessor Report (MPR) Subscribe

10Gbps PONs Follow Two Paths

XGS-PON and NG-PON2 Address Diverging Deployments

February 6, 2017

By Loring Wirbel


The International Telecommunication Union (ITU) has defined two distinct paths for passive optical networks that deliver 10Gbps or greater: XGS-PON and NG-PON2. They join early 10G-EPON networks based on IEEE standards. XGS-PON (G.9807.10) provides a straightforward evolution to symmetrical 10Gbps connections. The G.989.2 Amendment 1 standard enhances an existing proposal, NG-PON2, which relies on a combination of time-division multiplexing (TDM) and wave-division multiplexing (WDM) to support four or eight symmetrical 10Gbps links on one fiber.

Easing this apparent bifurcation for faster PONs, both standards rely on a common MAC and PHY base. Broadcom’s dual-mode BCM68620, which sampled in 4Q16, demonstrates the likely response by the dwindling pool of PON-semiconductor vendors. The company offers a single optical-line-termination (OLT) device for both standard symmetrical and “TWDM” 10Gbps PONs.

Despite a handful of 10Gbps trials at U.S.-based service providers, the willingness of operators in North America and Europe to deploy fiber to the home and fiber to the curb (FTTx) has declined in recent years. Chinese PON deployment continued at breakneck pace through 2015, although most of the build-outs from China Telecom and China Unicom are DSL replacements. China Mobile, however, has focused on gaining new broadband subscribers. That nation remains in the early stages of 10G-EPON upgrades, and the ITU specifications for XGS-PON and NG-PON2 apparently complicate those upgrade plans.

Proponents believe XGS-PON is the simplest transition path for business and multi-dwelling-unit (MDU) PONs, which may employ a dedicated fiber for the fastest symmetrical speeds. NG-PON2 deployments can use a single wavelength per fiber, offering a simple upgrade from GPON. Their advantage over XGS-PON is that they also can add per-fiber WDM for fast scaling in some residential and MDU cases.

Even the strongest XGS-PON defenders view that standard as transitional to NG-PON2, which is more flexible. But the use of tunable lasers and filters in both OLTs and client ONTs makes NG-PON2 more complex than most passive networks and can increase overall component costs. Although combined use of GPON, XGS-PON, and multi-wavelength NG-PON2 gives a service provider the most flexibility, it requires the addition of a special multiplexer node, called a coexistence element, near the OLT.

Herding GPON and EPON Cats

The desire to provide optimal upgrades to all PON topologies has yielded the current two-fork roadmap. This roadmap is partially the result of an original ITU plan that sought to include Asynchronous Transfer Mode (ATM) cells in Broadband PON (BPON), which became Gigabit PON (GPON) in 2008. BPON’s downstream throughput topped out at 75Mbps, whereas GPON offered only 2.5Gbps, minus overhead; data rates at the ONT in the field were typically much less than 2.5Gbps, however. BPON deployments crested in 2009 and 2010, and GPON likely peaked in 2016.

Meanwhile, the IEEE offered a separate suite of standards based on Layer 2 Ethernet frames. IEEE 802.3ah specifies 1Gbps EPON with native Ethernet frames, and 802.3av extended this feature to 10GbE speeds. The latter task force laid out two 10G-EPON modes: a symmetrical mode operating at 10Gbps in both directions and an asymmetrical mode operating at 10Gbps downstream and 1Gbps upstream. A newer task force, 802.3ca, is nominally targeting 100G-EPON, though it’s exploring nearer-term 25GbE and 50GbE options. It expects to deliver a draft standard by June 2017.

Even though Ethernet began as a LAN with distinct U.S. and European roots, service providers in the U.S. have tended to favor GPON because of their familiarity with ATM and BPON. Service providers in most regions display a natural preference for carrier-oriented standards, such as those from the ITU, rather than IEEE standards with enterprise roots. Asian service providers, which have more green-field FTTx opportunities, began EPON rollouts because of the common framing interface, which can decrease component costs in some cases. In early Japanese and Korean deployments, EPON could use cheaper components than BPON/GPON. More recently, Chinese deployments have favored GPON over EPON. The ITU developed the two new standards to give all service providers optimum evolutionary paths to 10Gbps and beyond.

Standards Emphasize Continuity

The new ITU standards build on GPON and the asymmetric XG-PON1 (10Gbps/2.5Gbps) using similar PHY and MAC (transmission convergence, or TC) layers, though NG-PON2 adds TC support for tunability at the client and central office. Both standards enhance GPON’s TC/MAC framing by aligning the frame and field design to word boundaries and by adding dynamic-bandwidth-assignment logic. TC specifications for both XGS-PON and NG-PON2 mandate three authentication methods, compared with GPON’s optional security methods. One reason why the ITU chose TWDM PON over alternatives such as OFDM PON was a desire to ensure backward compatibility with GPON and XG-PON1. Table 1 compares allocated wavelengths and multiplexing options.

Table 1. Comparison of the ITU’s 10Gbps and higher PON standards. The symmetric XGS-PON, though last to be approved, was conceived first as a symmetric upgrade to GPON. NG-PON2 is a time- and wave-division-multiplexed standard that gives service providers maximum wavelength-allocation flexibility. (Source: ITU)

NG-PON2 supporters wanted dual multiplexing of the time and wavelength domains to show the technology could reach 40Gbps, and eventually 80Gbps, through channel bonding. Service providers, however, saw a need for simpler and more-opportunistic ways of providing symmetrical 10Gbps service. Often, the target was not the residential user. In some cases, FTTx networks pass business parks that have yet to be served by other optical transport networks, and business customers can be served more quickly by XGS-PON. Verizon and other operators also are preparing for backhaul to LTE and 5G small cells, as well as C-RAN fronthaul. Backhaul could be based temporarily on XGS-PON, but combined mobile fronthaul/backhaul would require NG-PON2. Both 10Gbps PONs also are seen as ideal feeders for fiber-to-the-curb (FTTC) and fiber-to-the-distribution-point (FTTDP) topologies, which use fast copper technologies such as G.fast in the final 200 meters.

Service providers are well aware that moving to NG-PON2 requires new lasers and filters at the ONT, new active and passive components at the OLT, and a special system-level passive multiplexer, the coexistence element. But they view these requirements as an acceptable price to pay for more-effective bandwidth management.

XGS-PON was proposed in 2014 as a fast way to roll out symmetrical business services, and it garnered approval in June 2016. It offers a maximum symmetric capacity of just 8.56Gbps, however, because the standard mandates forward error correction. It was designed to employ existing optics for EPON/GPON and to support asymmetric XG-PON1 using a dual burst-mode transceiver. XGS-PON is limited to one upstream link and one downstream serial link, using 1,260–1,280nm wavelengths up and 1,575nm down. Unlike NG-PON2, it’s unable to independently offer point-to-point links that are separate from the main content service.

NG-PON2 received approval in 2015, but it’s more complex than the symmetric XGS-PON. The concept of using TWDM in subscriber management stemmed from an early proposal by Huawei and others called opportunistic and dynamic spectrum-management PON (ODSM PON). NG-PON2 offers up to eight upstream and downstream links, as well as a similar number of point-to-point links. Subscribers could simultaneously receive 10Gbps/2.5Gbps or 2.5Gbps/2.5Gbps, as well as reserved point-to-point services. The OLT creates a WDM channel by multiplexing light from four fixed-wavelength lasers, and the ONT filters the stream using a tunable filter. On the upstream path, the ONT dynamically selects the laser wavelengths for the return WDM. This capability implies the existence of tunable components at the ONT.

Offsetting NG-PON2’s greater component cost is the ability to manage load balancing, capacity growth, and power efficiency. The standard can also deliver true 10GbE by deactivating FEC. The ITU defined four typical NG-PON2 modes: basic, with 40Gbps on the downlink and 10Gbps on the uplink using four wavelengths; extended, with 80Gbps down and 20Gbps up using eight wavelengths; business symmetrical, available in 40Gbps and 80Gbps versions; and mobile fronthaul, a point-to-point WDM overlay. Research firm Ovum believes the fronthaul and backhaul needs of mobile operators are driving NG-PON2 more than FTTH is.

Waiting for the Fiber Exhaust

Asian service providers are still trying to complete first-generation EPON build-outs and initial 10G-EPON upgrades. China is expected to become the largest 10Gbps market. We estimate it had more than 200 million FTTH subscribers at the end of 2016, representing three-fourths of the global total. China Mobile has been the most aggressive in adding new GPON subscribers. China Telecom and China Unicom are primarily upgrading their DSL customer base to FTTH, though the former added 10 million new subscribers in 2016 while the latter added 2 million. Chinese carriers are in the early stages of 10G-EPON deployment and are likely to delay consideration of newer ITU networks for several years.

Today, Verizon and AT&T are flag-bearers for future North American PONs, whereas Vodafone and Alcatel-Lucent are launching NG-PON2 trials in Europe. Verizon admits that early U.S. FiOS customers won’t begin to run out of fiber transmission capacity (“fiber exhaust”) until the end of this year, though broadband proponents continue to expect 4K HDTV to suddenly create demand for multigigabit speeds in the home. The FTTH Council Europe expects global data demand to top 11 exabytes per month in 2017, with data usage growing 66% per year. Because FiOS demand has exceeded Nielsen’s Law of 50% annualized growth, Verizon expects residential customers will demand greater than 1Gbps in 2017, as Figure 1 shows.

Figure 1. Verizon’s anticipated bandwidth growth per subscriber. Expected increases in 2017–2019 leave GPON lacking, according to the company. (Source: Verizon)

Verizon agrees with market-research firm IHS Infonetics that XGS-PON build-outs will be short term and tactical in nature, intended primary for business-park bypasses and occasional mobile-backhaul duties. Nevertheless, the Broadband Forum and FSAN Group in 4Q16 cosponsored an XGS-PON plugfest in Tauxigny, France, under the assumption that service providers will still need help deploying this simpler symmetric GPON upgrade.

Verizon has dedicated the bulk of its development effort to NG-PON2 and has a goal of reducing component and passive-system-platform costs to make subscriber equipment less expensive. AT&T sees XGS-PON as a more feasible upgrade for the next two to three years, but it emphasizes that most of the R&D effort must quickly shift to practical NG-PON2 trials. AT&T distinguished fellow Tom Anschutz told the Broadband World Forum in 2016 that a “disaggregated architecture” combining mix-and-match filters with Open Compute Project infrastructure is the future of FTTH.

Verizon and AT&T are prodding other service providers and OEMs to adopt the NG-PON2 model. Along with Adtran and Calix, they formed the NG-PON2 Forum in July 2016 to promote TWDM-based technologies. In January 2017, Verizon pledged to publish open specs for its ONU Management and Control Interface (OMCI) to simplify product development for OEMs and even competing FTTH providers.

Affordable NG-PON2 Silicon and Optics

FTTH providers realize they must work closely with OEMs and component manufacturers to reduce costs at both the central office and the client ONT. Doing so is less a concern for XGS-PON, which allows bulk optics developed for GPON and EPON to be adapted for symmetric 10Gbps upgrades. Both XGS-PON and NG-PON2 can in early generations employ pluggable optics; service providers, however, assume that dedicated lasers and filters designed for the new standards will emerge within a year or two, many based on photonic integrated circuits (PICs).

Adtran and Calix have both claimed optical-interface breakthroughs in equipment trials with Verizon, but neither has disclosed details. Adtran claims it has made tunable lasers more affordable for the central office. Despite the lack of details, Dave Burstein of DSL Prime and Fast Net News said the company has a good record in many technologies and may well have made breakthroughs in optical-subassembly design. Burstein added that Adtran and Calix also have an inside track with Verizon because of their pledge to seek compatibility with SDN/NFV. Nokia’s (Alcatel-Lucent’s) slow move to open SDN/NFV could prevent the company from participating in 10Gbps PON trials, he said.

Dicon is one of many manufacturers designing filters for OLT, ONT, and the new coexistence-element mux. Verizon sees multiple opportunities for combiners, splitters, and filters in these network elements, as Figure 2 shows.

Figure 2. NG-PON2 network architecture. CO=central office. The various multiplexing points in the PON allow GPON and NG-PON2 to coexist, while providing new opportunities for silicon, filters, and lasers.

On the silicon front, Broadcom is enabling both standards in its OLT chip, potentially representing the typical implementation for FTTH networks. The departure of many players from the GPON/EPON silicon market over the last five years may limit the pool of ASSP providers for new 10Gbps PONs to Broadcom and HiSilicon. On the OLT end, those two companies will face competition from FPGA vendors and possibly OEM ASICs.

Broadcom’s BCM68620, which sampled in 4Q16, is the first merchant OLT chip to support both XGS-PON and NG-PON2. The company touts it as the most adaptable OLT SoC. Using common APIs, the device offers backward compatibility with the ITU’s G.984.2 GPON as well as the IEEE’s 802.ah 1Gbps EPON and 802.3av 10Gbps EPON. It also integrates a burst-mode serdes.

Although just a third of OLT chips are merchant ASSPs (the rest are based on ASICs or FPGAs), Broadcom’s acquisitions of Teknovus for EPON and BroadLight for GPON have increased its market share to 45% of merchant PON chips for both OLTs and ONTs. The company dominates in GPON OLTs, whereas Microsemi supplies most EPON OLT chips thanks to its PMC-Sierra acquisition. PMC, however, had halted new PON-device development for both the OLT and ONT long before it was acquired.

HiSilicon is the only other significant merchant OLT supplier, owing to its captive alliance with Huawei as well as merchant sales to Nokia (Alcatel-Lucent) and other OEMs. Realtek acquired Cortina Access, which is expected to pursue some 10Gbps PON opportunities, but it has yet to indicate whether it will participate in OLT, ONT, or both markets; a sole ONT focus appears likely, however. Intel (Lantiq) and MediaTek are only pursuing ONT silicon for customer premises equipment. To date, none of the ONT vendors has announced plans to support either 10Gbps PON standard. Marvell and Qualcomm have discontinued PON development.

Xilinx and Intel’s Programmable Solutions Group (formerly Altera) have worked on direct interfaces between high-speed serdes and active optical transceivers. Such efforts could benefit integration with future PON PICs, giving FPGA vendors a leg up on NG-PON2 implementations. FPGAs may have an important role in OLT designs, though FPGAs would be unlikely to appear in ONTs after initial trials.

FTTH’s Delayed Adolescence

One reason why so many silicon vendors have left the PON space is that Verizon and AT&T are the sole proponents of FTTH build-outs in North America. In Japan and South Korea, FTTH build-outs have reached a plateau. China is at the end of its DSL-replacement wave, and costs will likely make its service providers pause before implementing XGS-PON or NG-PON2 upgrades, particularly for the latter standard. Huawei and other Chinese OEMs have locked in their service-provider relationships, giving HiSilicon in particular a design edge.

We see XGS-PON as a useful but tactical play to give service providers a short-term GPON upgrade. Operators that have already implemented 10G-EPON may bypass XGS-PON. The opportunity for XGS-PON is likely to remain for three to five years, after which time FTTH providers will be well on their way to NG-PON2. Nevertheless, it makes sense for Broadcom to add enhanced TC support in its combined BCM68620 device now so that service providers can avoid additional upgrades later.

FTTH deployments exceeded xDSL in total CPE subscriber base three years ago, and most central-office DSL services were in decline (VDSL2 and G.fast are local services that can complement fiber to the curb). Total FTTx subscribers worldwide are on track to reach 300 million in 2018, with more than 90% of those networks being based on PONs. Recent Chinese EPON and 10G-EPON upgrades, however, may suffer delays in moving to alternative 10Gbps technologies, whether symmetrical XGS-PON or the more ambitious NG-PON2.

Silicon PON players dropped out because of fits and starts in U.S. and European deployments, as well as the limited opportunities for winning Asian designs. Currently, the developing market for OLT devices that support NG-PON2 will likely continue to be served by Broadcom, HiSilicon, and FPGA vendors through carrier trial phases.

The role of Verizon and AT&T in encouraging an open NG-PON2 development environment will be critical to expanding the opportunities for silicon and optical-component vendors. Long-term growth will favor NG-PON2, but the success of that standard will depend on reducing component costs and node complexity while preserving its flexibility.

For More Information

The ITU’s PON specifications are available at www.itu.int/rec/T-REC-G/en. The NG-PON2 Forum maintains a marketing web site at www.ngpon2forum.com. For the IEEE’s 10G-EPON standard, access www.ieee802.org/3/av/; for the emerging 100G-EPON standard, access www.ieee802.org/3/ca/. Broadcom provides limited information on its new dual-mode OLT chip at www.broadcom.com/products/broadband/xpon/bcm68620#overview. A Calix article on 10Gbps PON services from a carrier’s perspective is at www.telecompetitor.com/from-gigabit-to-10g-services-defining-10g-pon-standards/.

Events

Linley Processor Conference 2017
Covers processors and IP cores used in embedded, communications, automotive, IoT, and server designs.
October 4 - 5, 2017
Hyatt Regency, Santa Clara, CA
More Events »

Newsletter

Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »