As 10 Gigabit Passive Optical Networks (PONs) continue in volume deployment, the industry is looking to the next step. I have argued that this next step is about 50 Gbps (see Is 25 G or 50 G the Next Step for Passive Optical Networks? and Technology Roadmap for Passive Optical Networks: The Next Step is 50G-PON). There is an inevitability to periodic increases in ‘speeds and feeds’, but other than technological determinism and vendor competition, what is the rationale for a 50G-PON? After all, there is widespread agreement that consumers aren’t able to use the capacity of even 10 GPON and there is little prospect of that changing in the foreseeable future. That eliminates Fiber-to-the-Home, historically the main PON market driver, as a use case for 50G-PON.
However, new use cases are emerging for higher speed PONs in high-value, less cost-sensitive applications. They have high economic value, and operators can sell their connectivity at a large premium over residential and small business broadband services. Together, these use cases will not support anything like the hardware production volumes enjoyed by GPON, GE-PON, XG(S) PON and 10GE-PON, meaning that 50G-PON will not ride the same learning curve, or experience its cost reductions
Since TDM PON is a packet-based technology and data, voice and video traffic are bursty in nature, networks can offer total service rates to users at a multiple of the aggregate rate of the network. This multiple is called ‘statistical multiplexing gain’. Network designers use statistical assumptions about user traffic and performance objectives to determine the maximum acceptable statistical multiplexing gain. For PONs, this implicates numbers of customers of each type that can be provisioned on a PON, and, indirectly, split ratios. In the examples below, I use conservative assumptions for statistical multiplexing gain. Operators can engineer more aggressive statistical multiplexing gains, at the expense of latency, jitter and packet loss objectives.
Note that 50G-PON has a line rate of just over 50 Gbps. That includes overhead, especially for Forward Error Correction (FEC). Less the overhead, its downstream payload capacity is just over 40 Gbps.
Enterprise and Business Services
50G-PON will be a valuable tool in operators’ toolkits for business services. It is well suited to serving buildings in a smart campus, medical clinics, video production studios, smart factories, edge data centers, and cloud-centric mid-sized businesses and branch offices – almost anywhere XG(S)-PON does not have adequate shared capacity. It is likely that 50G-PON ONUs will become available in the SFP-28 (“stick”) form factor, allowing them to directly plug into an enterprise-class router or Ethernet switch.
XGS-PON supports MEF Carrier Ethernet E-Line, E-LAN and E-Tree service at about 8 Gbps. Occasionally, a true 10 Gbps service is necessary. In addition, operators may wish to offer yet higher rate services, such as at 25 Gbps. These services include Service Level Agreements (SLAs), so these networks must be engineered with lower statistical gains than residential networks. A 50G-PON can accommodate four 10 Gbps carrier Ethernet services, or one 25 Gbps service plus one 10 Gbps service with no statistical gain. With slightly relaxed SLAs (e.g., 95th percentile rather than 99.9th percentile throughput objectives), operators can use traffic models to engineer some statistical gain into the PON while still meeting the SLAs. This would accommodate split ratios greater than 4:1.
Another way to assure SLAs for mission-critical, jitter and latency sensitive applications is network slicing. ITU-T Study Group 15 recently approved a new supplement to the PON Recommendations, G.sup 74. It describes how to use existing mechanisms to isolate slices from each other in the User plane. Slices in the PON can complement slices in 5G in backhaul applications, allowing a slice to be end-to-end.
50G-PON also complements other enterprise-focused technologies, such as SD-WAN, edge computing, private wireless LAN, massive machine vision, augmented reality and virtualized CPE (uCPE). In addition to being a service provider access platform, 50G-PON can serve in passive optical LAN (PO-LAN) applications. PON topology brings enterprises the same benefits as it offers to service providers: higher logical port density, lower power, less cooling, less rack space, less congested fiber distribution frames, ducts and plenums and lower total cost of ownership (TCO).
WiFi 6, WiFi 6e and WiFi 7
50G-PON is well matched to backhaul applications for emerging and future WiFi, including public hotspots and private wireless LANs. WiFi 6, which is now being widely adopted, has a maximum downstream rate of 9.6 Gbps, versus 1.3 Gbps for WiFi 5. In practice, this is rarely achieved due to interference in the unlicensed 5 GHz band. 1200 MHz of unlicensed spectrum in the 6 GHz band has been opened in some countries (with others to follow) and IEEE 802.11 has issued an addendum to the WiFi standard that extends WiFI 6 into that spectrum. The WiFi Alliance dubs this WiFI 6e. With little interference in the newly released spectrum and leveraging numerous improvements in WiFi 6 over WiFi 5, users will be able to burst at or near 9.6 Gbps more often.
For example, consider a WiFI 6e deployment with PON backhaul. It will likely be dual-band, with both 5 and 6 GHz spectrum. The objectives are that WiFi APs and Stations be able to burst at speeds limited only by the radio environment (and not the PON), and that the split ratio be at least 8:1. Assuming for the sake of discussion that the aggregate peak data rate (both bands) is 10 Gbps. This means that the PON will run at a statistical gain of 2. In less critical applications, higher statistical gain, and thus a higher split ratio, may be acceptable.
IEEE 802.11 is working on a new amendment, IEEE 802.11be, which the WiFi Alliance will market as WiFi 7. It will have rates up to 30 Gbps and will support ultra-low latency and jitter applications. The standard is expected to be approved in May 2024 and implementations will pre-date it. This means that its adoption curve will coincide with the 50G-PON adoption curve.
Multiple Dwelling Units (MDU) and, Multi-Tenant Units (MTU)
The evolution of broadband services in multi-tenant residential and office buildings are another use case for 50G-PON.
Fiber-to-the-Building (FTTB) is common in older MDUs and MTUs. It uses existing copper or coax for in-building distribution, using G.fast, G.hn or MoCA Access. All of these technologies have extended roadmaps:
- G.fast has a data rate of 1 Gbps. The recently standardized MGfast supports data rates of either 4 or 8 Gbps. In addition, researchers are working on using copper as a waveguide, possibly at 10 Gbps over 500 meters.
- G.hn (or Gigawire), depending on physical media, operates at up to 4 Gbps half duplex or 2 Gbps symmetrical. Higher speeds are planned.
- MoCA Access 2.5 has a data rate of 2.5 Gbps downstream, and MoCA Access 3.0, to be published in Q4 2022, will have a 10 Gbps data rate.
For all these technologies, the roadmap for the next few years has a rate increase of 4x or more. Take the example of a high-rise MDUs with 16 units per floor. Each unit is served over coax with G.fast at a 500 Mbps service rate. An XGS-PON could serve four floors with a 4:1 split ratio and statistical gain factor of 4. If the operator wanted to upgrade to 2 Gbps service using MGfast, it would also have to upgrade the PON by a factor of 4 to maintain the same statistical gain factor; thus 50G-PON.
In some buildings, fiber retrofit is feasible. This would be a good application of Fiber-to-the-Room (FTTR) technology. The same logic applies: the 10 Gbps in-building distribution network allows multi-Gigabit service rates, and to allow reasonable split ratios and statistical gain, the FTTB service must be a significant multiple (such as 4) of 10 Gbps.
5G X-haul
Operators are converging their residential, enterprise, mobile and wholesale fiber networks onto a common fiber infrastructure to reduce duplication, leverage economies of scale and provide flexibility to respond to new customer needs. For many operators, this converged infrastructure is at least partly PON-based. For example, China Mobile recently conducted a field trial of a small cell backhaul use case for 50G-PON. In addition, in-building small cells, such as in factories, shopping malls and airports, can be backhauled with PON. .
There are a lot of factors that determine the throughput of the upstream and downstream fronthaul at a given instant, too many to enumerate here. There is no easy answer to the question of how much bandwidth is necessary for fronthaul and mid-haul applications. Following are some examples of how 50G-PON would apply to these use cases.
The first example is a 4X4 SU-MIMO mid-band urban macro-cell with 100 MHz carrier bandwidth per sector and 30 kHz subcarrier spacing. In this configuration, the maximum possible data rate per sector for Split 7.2 is 6.73 Gbps, a peak that would rarely be reached, and thus is susceptible to statistical multiplexing by PON. According to simulations by Bidkar et al , with that radio configuration and Split Option 7.2, a 50G-PON could support an aggregate of up to 12 sectors with a statistical gain of 2.
Next, consider a possible small cell deployment. Each site has three sectors and 200 MHz per sector is allocated. Split Option 2 (F1 Interface) is employed. According to calculations by Huawei, for the downlink backhaul, peak data rate is 10 Gbps and average is 3.86 Gbps. A 50G-PON can support traffic from four of these small cells, at peak rate. With Split 2, statistical multiplexing gain is feasible, perhaps by a factor of 2, for 8 of these cells.
Finally, a 5G fixed wireless access (FWA) deployment. The operator offers two broadband services, one FTTH, the other FWA, served by a common PON. The combined RU/CU/DUs operate in licensed mmWave spectrum, TDD, 4x4 MU MIMO with 2 streams, 256 QAM modulation, 400 MHz carriers, 120 kHz subcarrier spacing. Based on the formula provided in 3GPP TR-38.306 via the O-RAN Alliance X-Haul Transport Requirements document, using the 5G NR Throughput Calculator, peak backhaul data rate for each cell is 14.8 Gbps. A 50G-PON could accommodate four such cells with a statistical multiplexing gain of 1.5.
In addition to meeting data rate requirements for a variety of configurations, the High Speed Common TC, Recommendation G.9804.2, which is used by 50G-PON, includes some necessary accommodations for fronthaul applications, such as Quiet Window elimination and allocation period reduction.
Commentary
These are a few examples of realistic deployment scenarios. There are many variations that would also implicate 50 G PON.
Optical power splitting is a central tenet of PON. The benefits of PON scale with the split ratio: less fiber, smaller fiber frames, N+1 transceivers rather than N*2, with resulting savings in capex, power, rack space, and inventory, and statistical multiplexing gain. As split ratio declines, these advantages over point-to-point similarly diminish and the higher cost of PON transceivers becomes an economic factor. Many operators use 1:8 split ratios for non-residential PONs, and 1:32 or 1:64 split ratios – typically distributed -- for residential. This is on the assumption that residential customers have lower traffic intensity and more tolerance for queuing delay and dropped packets than non-residential customers. The examples above illustrate cases where provisioned bandwidth needed to satisfy the use case is 5 – 10 Gbps, or equivalent to the capacity of a 50G-PON at 4:1 or 8:1 split.
The key dependent variable is statistical multiplexing gain. Higher line rates reduce the effects of statistical multiplexing gain by shortening the packet transmission time. Thus, engineers can target higher split ratios and provisioned service rates on higher-speed PONs, with minimum effect on QoS.
Why not 25 Gbps? These particular use cases would work only if the split ratio were halved, or the statistical gain factor doubled. Neither would be desirable. That is not to say that there are no use cases for 25 Gbps, but rather that there are many more for 50 Gbps.
Why not 100 or 200 Gbps? These are still in the research phase. Someday it will be a practical option. Not for the immediate future.