I. Introduction

Technology roadmaps are a useful tool to focus the access network ecosystem on common objectives and timescales. As fiber access is such a cost-sensitive but, potentially, high-volume business, it is essential that systems are developed to meet the requirements of network operators worldwide. Hence, there is a major role for standards in defining such systems to drive volumes of common components and encourage diversity in the supply chain to lower costs through competition.

With such a large capacity already within the capability of NG-PON2, it seems that purely increasing the capacity may not be the only consideration for the next steps on the PON roadmap. In this paper, we examine where the PON systems market is today, the applications for existing next-generation PON systems, and the future drivers for a new PON roadmap.

II. FSAN Roadmap and ITU-T PON Standards

The full service access network (FSAN) group [1,2] has been using the roadmap shown in Fig. 1 to guide the development of passive optical networking (PON) standards beyond GPON (gigabit PON). The FSAN roadmap shows two generations of PON beyond GPON: so-called next-generation PON (NG-PON) 1 and 2, respectively. Although this roadmap figure is fairly simple, it conveys several key guiding principles that have served to steer the development of ITU-T PON standards and systems for the past 10 years or so.

 

Fig. 1. FSAN roadmap.

Download Full Size | PDF

The first key message conveyed by the roadmap is that the power splitter based fiber infrastructure deployed for GPON will need to be reusable for some considerable time. Second, the enduring principles of coexistence of generations of PON, and the gradual migration of customers to new PON systems, are captured. Third, clear market windows are provided for each technology generation to avoid market confusion and competition among systems that could damage the objective of high global volumes of common components.

The FSAN roadmap has largely been delivered as originally conceived, with XG-PON1 standards published in 2010 [3] and the NG-PON2 standards [4] in place by the end of 2015. Now, the network operators in FSAN are very actively looking to consolidate their ideas on the next roadmap for PON.

There are many directions that PON technology could develop, and these need to be narrowed down to those that make most sense for a global market. With NG-PON2 systems just being released by vendors [5], and deployments planned [6], FSAN needs to ensure that there is a clear window of opportunity for that technology to establish a market without adding confusion to the technology landscape.

NG-PON2 is already a very high-capacity fiber access system with many options available for different use cases. Just considering the TWDM-PON (time and wavelength division multiplexing) aspect, recent updates to the physical media dependent (PMD) layer standard (G.989.2) enabling up to 80 Gbps aggregate capacity have been consented in the ITU-T. Furthermore, bit-rate increases from 10 Gbps per channel to 25 Gbps have been discussed. Furthermore, PtP WDM (point-to-point wavelength division multiplexing) channels have been defined down to 50 GHz channel spacing, with the flexibility to exploit $\sim 100\mathrm{nm}$ of spectrum when using the Expanded Spectrum wavelength plan.

III. PON Systems Market

PONs are the basis of the majority of FTTH deployments, and the estimated total worldwide revenue for PON equipment was around $5B in 2015 [7].

While several next-generation PON technologies have been standardized in the ITU-T, it is indicated in market research reports [7] that GPON will dominate the number of optical line termination (OLT) port shipments for some considerable time. In the Ovum report [7], it is estimated that the number of GPON OLT ports shipped in 2021 will be 10 times the volume of all other ITU-T type next-generation PONs combined.

While market predictions are clearly subject to error, the expectations regarding the general trends are clear and next-generation PON technology volumes will remain relatively low. This does not mean there are no applications for the more advanced PON technologies, but just that mass market FTTH is well served by GPON. Other applications are expected to be lower volume, and this, in addition to the lower degree of technology maturity, will keep the price of next-generation PON transceivers higher than those for GPON. For example, in the report referenced above [7], the average selling price of an NG-PON2 optical network unit (ONU) is predicted to be more than 20 times that of a GPON ONU. Such a price premium would not be sustainable in the residential FTTH market where subscriber expectations are that service bandwidths are continually increasing with minimal or no increases to the tariff for basic connectivity. Service providers could only justify a large price premium if they gain cost advantages elsewhere in their networks and/or they may increase revenues through offering additional premium services, e.g., pay TV.

Despite the expected dominance of GPON in future port shipment volumes, there are applications that may sustain higher equipment costs. Despite the lower number of ports, the higher equipment costs can result in a significant proportion of revenue for next-generation PON systems. In the next section, we will look at these applications.

GPON has seen mass deployments since around 2007 and has a mature ecosystem with all the major technology challenges addressed. There is a GPON ONU certification program [8] in the Broadband Forum (BBF) that provides a high level of confidence that ONUs will interoperate with standards compliant OLTs to offer basic Ethernet services as defined in TR-156 [9]. Such certification programs are not yet in place for next-generation PONs. To date there have been discussions in FSAN about early stage interoperability events for NG-PON2, but none have yet taken place. For XG-PON, there have been some early stage interoperability events that verify the basic connectivity and service configuration. However, the early stage interoperability events have not been followed up by more events that probe a wider range of configurable features.

There is a need to mature the single wavelength channel 10-gigabit-class PON technologies (XG-PON and XGS-PON) to drive down the equipment costs (CAPEX) and lower operational costs (OPEX) in deployment, e.g., arising from a lack of interoperability. This can stimulate the market and tip the market balance in favor of these systems so they may become the new mass market solution of choice. Similarly, the multi-wavelength NG-PON2 systems need to begin these steps so they may follow this same progression. However, there are many physical layer options for NG-PON2 today, and some convergence in the market and some fine-tuning of the standards may be needed to realize the full potential and accelerate this process.

IV. Service Drivers

As has been mentioned, the expected dominance of GPON in future OLT port shipments does not preclude the creation of a meaningful market for next-generation PON systems. Here we will discuss potential services and applications that might be addressed with such systems.

In the main, fiber-based business connectivity services are delivered using a dedicated fiber (or fiber pair) for each connection and a single wavelength channel per direction. This has many advantages in terms of simplicity, link budget (reach), and security (through physical isolation of traffic from other users). However, such an approach can be problematic for network operators where fiber or duct resources are scarce. New service activation can be time consuming and costly if a new fiber and duct build is necessary, and, furthermore, additional capacity can only be provisioned by adding an extra fiber. Therefore, more fiber efficient service delivery can be attractive to both subscribers and network service providers alike. Nevertheless, there will still be subscribers for whom a dedicated fiber is the preference for their needs.

Until recently, 100 megabit Ethernet services were very common for businesses, but gigabit Ethernet services are becoming increasingly popular with large businesses and small to medium enterprises alike. Such services can be effectively delivered over a PON using next-generation PON technologies. The tree architecture is very fiber efficient as only a single feeder fiber is used to connect a central office to the local service area where a passive optical splitter provides onward connectivity to each subscriber.

Typically, business users expect their services to provide symmetric bandwidth. This has been a strong motivation behind the recent standardization in the ITU-T of the XGS-PON system [10]. The highly asymmetric service profile of XG-PON (10 Gbps/2.5 Gbps) was largely motivated by delivering a cost effective residential access service where the ONU needs to be very low cost. In 2010, when XG-PON was standardized, 10 Gbps transmitters were not foreseen to meet the expected cost targets. Furthermore, as the market for residential services delivered through XG-PON has yet to emerge, and demands are growing for lower-cost delivery of business services, this has driven the industry toward the development of a symmetric 10-gigabit-class PON.

Business leased line services command a price premium as they come with the stringent service level agreements (e.g., fast repair time) that such mission critical connectivity demands. Furthermore, additional features, not needed for residential broadband, are often necessary, e.g., resilience. With the fiber savings enabled by PON, as highlighted above, coupled with the higher service revenues, this provides an opportunity for next-generation PON systems such as XGS-PON and NG-PON2. Revenues from such premium services may be sufficient to justify higher CAPEX for next-generation systems. As to which of the next-generation PON systems will be best for these services, this will be very dependent on the service provider and their market conditions. For the foreseeable future, there will be a cost advantage for the fixed wavelength transceivers of XGS-PON compared to the tunable transceivers in NG-PON2. Even though NG-PON2 can be deployed in a pay-as-you-grow manner using one wavelength channel at a time, the need for a wavelength multiplexer/demultiplexer and (probably) optical amplifier(s) at the OLT make the cost of the first NG-PON2 channel higher than XGS-PON. Nevertheless, some service providers may see a longer-term commercial advantage in being able to then grow their PON capacity with NG-PON2 by adding further NG-PON2 wavelength channels. Others may see this as an upfront cost on which it is hard to get a financial return, and so they will prefer to deploy XGS-PON until NG-PON2 equipment prices reduce sufficiently. NG-PON2 can coexist with XGS-PON, and, in principle, common line cards can be used at the OLT and common host equipment at the ONT, so only the optical transceivers should need to change.

The exponential rise in mobile data consumption, enabled by the proliferation of smart phones and the deployment of fourth-generation (4G) radio access networks (RANs), is driving demand for more cell site backhaul capacity. Furthermore, to increase the overall RAN capacity, mobile network operators are looking to increase the density of cell sites and the use of small cells. This explosion in cell sites and mobile backhaul demand is another driver for the next-generation PON market. Although there are other options (e.g., microwave links) for mobile backhaul, there is an expectation that fiber is going to be a significant backhaul delivery mechanism that will grow in importance [11]. Additional system features that are important in the mobile backhaul application include frequency and time-of-day synchronization. These features are well described in, and supported by, standards for PON.

V. Infrastructure Reuse

Around the world, network operators have been pushing fiber deeper into their access networks. Depending on the local circumstances, this may involve terminating the fiber at an active cabinet and shortening the existing copper loop, or deploying fiber all the way to a building (FTTB) or living unit (FTTH). The installation of this fiber infrastructure (duct, cables, and joints) is the major cost and, for FTTH, reaching up to 80% of the total cost (according to some estimates [12]), so it is highly desirable to reuse this infrastructure for more than just a single service such as residential broadband. For example, the mixing of residential and business users or mobile cell sites on a common PON ODN is an attractive proposition to increase revenues and PON utilization.

For the generations of ITU-T PON beyond GPON, it has been a key requirement from network operators that these systems can coexist on the same fiber infrastructure as GPON. As well as facilitating the wholesale reuse of the deployed fiber infrastructure, this also enables a smooth evolution of individual subscribers from one system to the next without forced migration of all subscribers. Furthermore, different subscriber types can be segmented by PON technology, e.g., residential subscribers connected using GPON and business subscribers using XGS-PON.

In ITU-T PON standards, this coexistence is facilitated by the wavelength plans for each system. This is illustrated in Fig. 2 for GPON and XG(S)-PON. All that is needed, to ensure a hitless addition of a next-generation PON system to a deployed GPON, is for appropriate blocking filters to be included in each ONT and for a wavelength band multiplexer (coexistence element) to be installed at the OLT. The coexistence of such PON systems is described in the ITU-T recommendation G.984.5 [13].

 

Fig. 2. GPON and XG(S)-PON wavelength plans. U/S, upstream; D/S, downstream.

Download Full Size | PDF

With the advent of XGS-PON and NG-PON2 there are extra options available for reuse of an existing PON ODN. With XGS-PON, the standard supports TDM/TDMA (time-division multiplexed/time-division multiple access) coexistence with XG-PON by sharing the same wavelength plan and/or wavelength overlay coexistence with XG-PON through XGS-PON occupying the GPON wavelength bands. In the latter case, coexistence with GPON is not supported and residential FTTH may be served with an XG-PON that supports 2.5 Gbps in the upstream. Alternatively, 20 Gbps of symmetric PON capacity may be offered by simultaneously using XGS-PON in both the GPON and XG-PON wavelength bands. This gives a lot of flexibility to the network operator in tailoring their deployments to meet market needs.

With NG-PON2, the network operator has the flexibility of adding PON capacity in increments of NG-PON2 wavelength channels. In the case of TWDM, these wavelength channels can use a variety of bit-rate options, i.e., the basic option of 10 Gbps DS and 2.5 Gbps US or one of two symmetric line rate options at 2.5 or 10 Gbps. Thus, the network operator can choose to add capacity when needed (pay-as-you-grow) and serve different user demands on different wavelengths depending on the service offering. Once again, there is a considerable flexibility built into the NG-PON2 standards.

VI. When Do We Need NG-PON3?

With the proliferation of next-generation PON standards having been developed in recent years at line rates up to 10 Gbps, there is a question now as to whether the industry needs to plan for another evolution in the near term (say, in the next five years). A clear objective of the standards developed so far is that these need to, ultimately, be suitable for use in residential FTTH type applications; i.e., the technology needs to be able to be cost reduced when shipping in the massive volumes observed for that application.

Given that there is an expectation that GPON technology has a very long life left in it, and that the market in the next five years for next-generation PON is likely to be in lower volume but higher revenue applications, it is hard to see that another PON system generation is needed in five years time with a future FTTH market in mind. The IEEE 100G-EPON Task Force [14] is targeting a 100 Gbps capacity PON ($4\times 25$ Gbps) to be required in 2025. With NG-PON2 already offering 80 Gbps capacity for TWDM channels alone [15], and 240 Gbps if 16 PtP WDM channels are added, this seems like enough capacity for the foreseeable future. NG-PON2 was conceived as a long-term evolution for network operators well into the next decade, with the pay-as-you-grow capability for increasing capacity through adding wavelength channels. This does not mean that there will not be any evolution of the existing PON standards, but rather that there is no obvious requirement for another technology leap in the foreseeable future to meet the needs of FTTH.

Enhancements being discussed for the next-generation PON standards already in place (i.e., XG(S)-PON and NG-PON2) include the increase in per-channel line rates from 10 to 25 Gbps. This could leverage the industry ecosystem developing around 25 Gbps components for other applications and offer higher peak rate services. Another enhancement option being discussed, to enable higher peak rate, is the use of wavelength channel bonding in NG-PON2. In this scheme, a single ONU would have access to more than one wavelength channel on the PON and could use free capacity on any channel. In the limit, the full PON capacity could be accessible by a single ONU for higher rate service tiers. Bandwidth resources on the PON would need to be coordinated to make efficient use of capacity, i.e., by using a dynamic bandwidth allocation scheme in both the time and wavelength domains.

Whatever enhancements to existing next-generation PON are pursued, they will have to meet the existing requirements for compatibility with legacy, power splitter based ODNs and coexistence with legacy PON systems. The definition of a new generation of PON system beyond these enhancements may be motivated by other applications than FTTH, and it could be that new ODNs (e.g., purely wavelength splitter based) are adopted. We are running out of technical levers that enable higher PON capacity while needing to support ODN losses up to 35 dB, and lower loss ODN types could be attractive. Furthermore, full coexistence with all legacy PONs may also be unnecessary and limited to certain systems only. The scope to add new systems through wavelength overlay is severely limited by the availability of unallocated optical spectrum.

VII. Mobile Networks as a PON Standards Driver

Mobile backhaul has already been mentioned in Section IV as a service driver for next-generation PON system deployments. There are other ways in which PON could be utilized in future mobile networks.

First, transport of mobile fronthaul, whereby in-phase and quadrature (IQ) samples from the air interface are carried between a remote radio head (RRH) and centralized base-band unit (BBU), has received considerable attention [16]. Mobile fronthaul transport is the key to so-called centralized RAN (CRAN) whereby the BBU resources are pulled back from the cell site and into a centralized network location. Considerable OPEX savings are expected from the simplification of the cell site and pooling of BBU equipment into a central office [17]. BBU centralization also facilitates low-latency inter-BBU communications that enable improved radio performance at cell boundaries by mitigating against inter-cell interference effects. Low-latency, continuous-mode, and high-bit-rate (up to 10 Gbps) fiber transmission is essential to the CRAN approach, but it can be very fiber resource hungry as one fronthaul link is required per RRH.

In the NG-PON2 standards the transport of mobile fronthaul was a key driver behind the specification of PtP WDM overlay channels in addition to TWDM. These channels offer dedicated wavelength connectivity to each end point on a PON in three line rate classes covering the main CPRI/OBSAI [18,19] line rates from 1 to 10 Gbps. Forward error correction (FEC) is not used for these channels in order to avoid additional latency.

On the near horizon is the upcoming fifth-generation (5G) mobile system, the standards for which are expected to be complete in 2020 [20]. Some of the key features of 5G are expected to be up to 10 Gbps capacity on the air interface, massive MIMO (multiple-input multiple-output), dense small cells, and a new functional split for fronthaul. With such features in scope, fiber is expected to be crucial to enable 5G RANs. However, with a new functional split, future fronthaul is expected to be packet based [21] and, possibly, could be transported over a TDM/TDMA PON such as NG-PON2 [16].

VIII. Possible Evolution Path for PON

The convergence of residential, business, and mobile backhaul/fronthaul is already enabled by NG-PON2. For example, using TWDM and PtP WDM (plus coexistence with GPON, XG(S)-PON) one can serve a mix of residential, business, and mobile clients. Any future roadmap for PON systems should take into account wider market developments beyond FTTH, e.g., 5G backhaul and fiber business services. However, some preconceived requirements from FTTH may be discarded, and a future optical access system might be something optimized for a Greenfield application that is free from legacy constraints. For example, the emphasis of a future PON may shift to be a mobile optimized system with FTTH as a “nice to have” feature. The requirements for such a system would be driven by RAN specific requirements and coupled to the RAN technology evolution.

Possible pathways forward for ITU-T PON may be to evolve and specify in detail the NG-PON2 PtP WDM, Expanded Spectrum option (1524–1625 nm). This would make full use of the available C+L band spectrum to offer many more channels but at the expense of coexistence with TWDM. It is possible that the PMD used could be that from the G.metro standard, being developed in the ITU-T SG15/Q6 group, which aims to lower the cost of DWDM PON based on wavelength agnostic tunable ONUs. This could be coupled with an NG-PON2 PtP WDM TC-layer (Transmission Convergence) and any necessary G.988 OMCI (ONU Management and Control Interface) extensions to provide a fully managed PON system.

IX. Beyond New PON Systems Definition

With 5G mobile being the most disruptive technology that might drive a future PON system standard, there is a role for industry groups such as FSAN and the BBF in maturing the already standardized NG-PON1/2 technologies. Current NG-PON technologies must mature and drive out costs in the supply chain and continue to make PON even easier to deploy and operate.

Currently, GPON (OLT-ONU) interoperability is mature, with an established ONU certification process in BBF [8], and the XG-PON interoperability assurance has started with a couple of initial test events. On the contrary, NG-PON2 and XGS-PON interoperability is still in the planning phase and needs to progress rapidly to catch up with GPON.

There are currently many options in the NG-PON2 standards that may be a barrier to cost reduction through the manufacture of high volumes of common components. The inclusion of options is commonplace in the standardization process, and the perceived usefulness of some of these may diminish over time and some options may just be low priority. Where possible, options can be deprecated from the base standards themselves during routine maintenance activities. However, another useful approach is for network operators to identify the options of most interest. In FSAN, this can be done using a Common Technical Specifications document. Such a process is underway for NG-PON2 to focus early system implementations. Some options may be naturally deprecated through a market lead approach that is driven by commercial or technical realities.

Given the high value services that are to be addressed by next-generation PON, it is expected that protection and resilience are going to grow in importance. Even residential fiber services that carry premium TV and video content are candidates for increases in service availability and enhanced customer experience. Such features have been included in PON standards for some considerable time. However, there is now much more study in FSAN/ITU-T on requirements, detailed schemes, and their specifications.

Another mechanism to offer higher service availability and improve customer experience is to automate accurate fault diagnostics. This may be through automatic ODN monitoring to allow more stringent service level guarantees for PON-based services, e.g., by embedding OTDR (optical time-domain reflectometry). Such an approach can enable rapid fiber fault location and hence shorter repair times in the event of faults. Such schemes have been described in BBF TR-287 [22], but the underlying technology needs development and cost reduction to enable widespread deployments.

Power-saving mechanisms have been included in amendments to GPON and XG-PON standards and in XG(S)-PON and NG-PON2 standards from inception. As yet, these mechanisms have not been widely implemented in PON equipment and need further development to drive more widespread adoption by network operators. The impact of these mechanisms on services and the OLT-ONU interoperability of these schemes require further study to give network operators greater confidence in deploying them.

Network operators are continually looking for ways to cost reduce their networks and meet the expectations that subscribers have that the price per unit of bandwidth will continue to decline. Furthermore, convergence is a key industry trend that enables more services to be delivered from common platforms where they used to be separate. New, and potentially disruptive, approaches to meet these needs are emerging [23] based on network function virtualization (NFV) and software defined networking (SDN) alongside other ideas from the data center industry such as White Box Switching. As well as promising both CAPEX and OPEX savings, it is hoped that such approaches will also enable rapid service activation through the use of software components running on programmable, commodity hardware. It remains to be seen whether, and when, the expected benefits can be fully realized, but this is certainly an approach that has a lot of industry momentum and resources behind it.

X. Conclusion

GPON has been an extremely successful technology that is widely deployed around the world to deliver FTTH services. GPON is expected to continue dominating the volume of PON ports shipped for many years to come. Technologies beyond GPON (i.e., next-generation PON) have been standardized but not yet widely deployed. It is expected that these will start to see deployments picking up in the next few years. This paper has reviewed the services that will drive the deployment of next-generation PON. The future evolution of PON systems and standards has also been reviewed. It may be concluded that FTTH will probably not be the main driver for a new PON technology generation as there are enough PON systems to meet the expected capacity requirements for this market. Any new PON technology may be more mobile-centric and target 5G RAN (and beyond) and/or future business services. In the meantime, there is a need for the industry to focus on maturing the existing NG-PON technologies to match the maturity of GPON and replicate the success of that technology in the market.