Open Access
Add to BibSonomyAbstract
Passive optical networks (PONs) have been widely used in access networks and are today the access technology of choice for operators, especially when they need to build new infrastructure. Significant effort has been put in place across the years to scale the speed of passive optical networks. The super-PON concept has also been proposed in order to scale reach and customer aggregation of PONs. However, it did not produce any standardization efforts nor significant commercial deployments. In November 2018, the IEEE 802.3 Working Group formed the P802.3cs “Super-PON” Task Force to standardize a flavor of super-PON able to support an increased optical reach of up to 50 km and an expanded customer coverage of up to 1024 customers per fiber over a passive optical distribution network. A project with near identical objectives was later started in ITU-T Q2/SG15. This paper presents the latest updates from these industry standardization efforts, with special attention to the activity of the IEEE P802.3cs Task Force, and the technical reasoning behind some of the key decisions that have been made.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. INTRODUCTION
Passive optical networks (PONs) have been widely used in access networks and are today’s access technology of choice because:
- • PONs are more cost effective to build and operate because their multi-access nature enables them to serve many customers with a single fiber coming from the central office (usually between 16 and 64 subscribers per fiber);
- • PONs provide high burst bandwidth to customers, enabling high perceived speeds (common link speeds range from 1 to 10 Gb/s);
- • optical fibers are a future proof way to build infrastructure, given their ability to carry higher and higher rates as technology evolves.
Figure 1 shows the structure of a typical PON deployment, with one strand of fiber from the central office (CO) serving up to 64 customers, typically spread over a maximum radius of 20 km. For example, both class PR40 in 10G-EPON [1] and class E1 in XGS-PON [2] support 64 subscribers over 20 km. The 20 km distance limitation is due to optical power budget constraints, and thus it is related to the split ratio of the optical distribution network (ODN). PON upper layer protocols are able to support longer distances (e.g., up to 60 km for the G-PON transmission convergence (TC) layer [3]) and larger split ratios (e.g., up to 1:1024 for the XGS-PON TC layer [2]).
The speed of PONs has evolved significantly during the years. The basic gigabit speeds of EPON (1 Gb/s symmetric service; see [1]) and G-PON (2.5/1.25 Gb/s asymmetric service; see [4]) account for the majority of PONs installed to date. However, today’s PON speed of choice, especially for new deployments, is 10 Gb/s. 10G-PON is available in two variants: 10G-EPON, specified in [1]; and XGS-PON, specified in [2]. Both specifications are capable of providing a 10 Gb/s symmetric service (the actual throughput is around 8.5 Gb/s after taking out the overhead).
Looking into future speeds, after completing the definition of the parallel 40 Gb capable NG-PON2 system [5], Question 2 in ITU-T Study Group 15 (Q2/SG15 [6]) is now working on the 50 Gb/s PON serial speed in the G.HSP project. At the same time, the IEEE 802.3 working group is completing the standardization of 25 Gb/s and 50 Gb/s PONs based on 25 Gb/s serial streams in the IEEE P802.3ca Task Force [7].
This evolution shows that significant effort has been put in place across the years to scale the speed of PONs. The PON upper layer protocols’ scalability has also been increased [e.g., the G-PON TC layer is able to address up to 256 optical network unit identifiers (ONU-IDs), and the XGS-PON one can address up to 1024]. However, not much has been done at the optical layer to scale the spatial dimensions (e.g., reach and aggregation) of PONs to enable more efficient outside plant (OSP) designs. To serve areas with a radius greater than 20 km, current PON technologies require building multiple active sites, such as COs or active cabinets, or using a low optical split ratio, which drives up the number of fiber strands required.
Building infrastructure is the most expensive component of building an access network. Figure 2 shows a typical deployment layout using current PON solutions for metropolitan areas with a ${\gt}{20}\;{\rm km}$ radius. Building multiple COs can be difficult because of permitting processes, power availability, or other concerns. Additionally, each CO is associated with a fixed maintenance cost, so more COs result in higher operational costs. Moreover, the customer coverage (up to 64 per fiber) of current PON solutions requires many fiber strands in the ground to cover larger areas such as the one shown in Fig. 2. A commonly used work-around for unavailability of COs is to use remote active cabinets. Cabinets reduce but do not eliminate the deployment challenges of permits and power as well as the increased operational complexities associated with distributed active sites, such as remote monitoring, power backup, and generational upgrades. A more spatially scalable optical solution, providing increased reach and more customers per feeder fiber, would be helpful for these cases. This is what Super-PON aims to provide.
Reach extension for time division multiplexing (TDM) PONs by using optical amplification at the optical line terminal (OLT) has been investigated [8] but has not resulted in widely available commercial products. Mid-span reach extenders have been standardized for G-PON [9] and XG(S)-PON [10]. However, they are designed for a remote location and need an additional active site between the CO and the end users. This is perhaps one reason why they have seen limited commercial application.
The term “super-PON” was coined to identify architectures aimed at centralizing the packet-switching functions of an extended metropolitan area into one location. Some methods proposed to use optical amplification at an intermediate site with only optical regeneration [11–14]. Other architectures used optical amplification at the OLT location [15,16], reducing the number of powered sites needed for a deployment. However, despite a large body of quality research [17–19] and some experimental field demonstrations [20], these architectures have never been widely deployed commercially nor standardized. The shortfall in industry activity on these architectures was identified first in IEEE 802.3, which initiated the P802.3cs Task Force. This article presents the super-PON architecture that is being developed and standardized in the IEEE P802.3cs Task Force. This architecture is referred to as “Super-PON,” with capitalization.
2. SUPER-PON IDEA
Super-PON is a PON optical layer able to support an increased optical reach of up to 50 km and an increased customer coverage of up to 1024 customers per fiber over a passive ODN as shown in Fig. 3. As such, the focus of Super-PON is more on scaling the PON spatial properties rather than on scaling the speed. In terms of speed, Super-PON supports the 10 Gb/s speed.
With the Super-PON optical layer, each CO is able to serve a much larger region and a higher number of customers, which thus significantly simplifies the underlying physical infrastructure. Extending the reach from 20 to 50 km also enables CO consolidation, reducing the number of powered sites needed to serve a large area. As an example, a single consolidated CO would be enough to serve a 50 km radius as shown in Fig. 4.
Super-PON achieves these results by leveraging three mature optical technologies not commonly used in access networks: wavelength-division multiplexing (WDM), wavelength routing, and optical amplification.
WDM enables the transport of multiple independent signals over one fiber strand by using different wavelengths. In Super-PON, multiple PON instances are mapped to different wavelengths and transmitted over the same feeder fiber. The number of subscribers is scaled by:
- • associating multiple channels to different wavelength pairs;
- • multiplexing/demultiplexing these wavelengths in the fiber exiting the CO; and
- • routing the wavelengths in the field through a passive wavelength router in the ODN.
By using 16 different wavelength pairs, it is possible to transport 16 PON instances, each supporting 64 customers, for a total of ${16}{*}{64} = {1024}$ customers per feeder fiber exiting the CO.
An ODN incorporating a passive wavelength router is channelized: it transmits only the wavelengths belonging to the router wavelength plan and filters out all other wavelengths. By comparison, a power-splitting-only ODN transmits any wavelength. This makes a channelized ODN much more constrained than a power-splitting-only ODN and brings specific wavelength plan requirements to make it more future proof (see Section 5.D). On the other side, in a channelized ODN the wavelength pair over which an ONU operates depends only on the point of attachment to the ODN topology. This enables ONUs with a broadband receiver (because the wavelength router delivers only one wavelength to the ONU) and flexibility in the transmitter implementation as follows:
- • a fixed transmitter, operating on the channel required by the point of attachment to the ODN;
- • a partially tunable transmitter, able to operate over a contiguous subset of the defined channels; or
- • a fully tunable transmitter, able to operate over the full set of the defined channels.
A potential solution for a fully tunable ONU transmitter is a distributed Bragg reflector (DBR) laser [21]. The laser driver is critical, as it can enhance tuning range and decrease wavelength drifts from burst mode operation [22]. Another solution is a distributed feedback (DFB) laser, which is currently the most mature and lowest cost technology. Since temperature stabilization is required for dense wavelength division multiplexing (DWDM) operation, this solution will be inherently a partially tunable solution capable of tuning over around four channels.
A major difference between a Super-PON ONU and an NG-PON2 ONU is the absence of the tunable optical filter from the ONU receiver. This removal simplifies the optical components, which have been the largest challenge in the realization of NG-PON2 [23]. As for NG-PON2, Super-PON ONU transmitters need to be wavelength stabilized for burst mode operation on narrow DWDM channels [24,25].
The reach is scaled by using optical amplification in the CO. Super-PON mitigates the cost of optical amplification by applying it to all multiplexed wavelengths at once. In this way, the cost is distributed over a large number of customers.
The Super-PON architecture is depicted in Fig. 5. The wavelengths’ multiplexing/demultiplexing functions and the amplification are put together in a single unit hosted in the CO, which is called MUX/Amp in Fig. 5. This unit can also provide dispersion compensation functions, if needed.
In the downstream direction, the downstream wavelengths belonging to the channels used by the $m$ OLTs are multiplexed, amplified, and finally merged with the upstream wavelengths on the ODN fiber. In the ODN, these wavelengths are routed by the remote $\lambda$ router over its physical ports. Each wavelength passes a power splitter and then reaches the destination ONUs. The point of attachment to the ODN determines the wavelength pair an ONU uses to operate.
In the upstream direction, an ONU transmits using the upstream wavelength belonging to the channel associated with the OLT to which it is connected. The upstream wavelengths are multiplexed by the $\lambda$ router toward the CO. In the CO, the upstream wavelengths are separated from the downstream wavelengths (e.g., using a bandpass thin-film filter), pre-amplified, processed by a dispersion compensation module (DCM), demultiplexed, and delivered to the appropriate OLT port.
The differential length beyond the $\lambda$ router is assumed to be ${\lt}{20}\;{\rm km}$. The main limitation is from power imbalance, both between the wavelength channels, which could cause excess crosstalk, and within the wavelength channel between splitter outputs, which increases the dynamic range of the OLT receiver. This is expected to not be a problem for most deployments, as operators typically want to maximize the fiber length in the feeder fiber in order to minimize fiber counts.
Super-PON decreases capital expenditures (CAPEX) in two ways. First, by carrying multiple channels per fiber, Super-PON requires fewer fibers to serve a certain area, which translates into smaller cables. Smaller cables are significantly easier to handle, repair, and install. In particular, cable installation can be done with microtrenching techniques, which are much less expensive than traditional trenching. This benefit is most relevant to developed urban and suburban areas, where household densities result in tens or hundreds of thousands of households per CO. Second, the reduction of the number of COs needed to serve a geographical area decreases construction costs and lowers the risk of delays from CO permitting. In addition, by consolidating the active equipment that requires power and maintenance in fewer COs, Super–PON enables a significant reduction in infrastructure operating costs [i.e., lower operation expenditures (OPEX)].
Super-PON has a drawback: the customer equipment (i.e., the ONU) is more expensive than a conventional XGS-PON or 10G-EPON ONU because its laser transmitter needs the ability to operate on a narrow wavelength-stabilized WDM channel and may need to support multiple channels if the operator wishes to reduce the quantity of part numbers in inventory. This higher cost is due in part to the fact that Super-PON is a new technology with no current significant volumes and the ONU cost is success-based (i.e., you only incur this cost when you have paying customers). With the standardization and adoption of the technology, volumes are expected to grow, and the associated costs per unit are expected to become lower. In any case, the Super-PON saving on the infrastructure side of new deployments can easily make up for the more expensive ONUs.
3. SUPER-PON MOTIVATION
Reducing the access infrastructure building and operating cost is the main goal of Super-PON. An example can help to understand how Super-PON can achieve this result. Figure 6 shows how different the optical access infrastructure needed to serve a medium sized US city is by using the conventional XGS-PON optical layer versus the Super-PON optical layer. This area has around 600,000 households, consisting mostly of residential customers.
As shown in Fig. 6, 16 COs are needed to serve this city area with a conventional PON optical layer. This number is reduced to 3 COs when Super-PON is used. Moreover, the smaller fiber cables required by Super-PON allow the use of cheaper microtrenching techniques in place of traditional trenching or directional boring.
A quantitative analysis of the infrastructure building cost reduction for this example has been performed as part of the Super-PON standardization process [26]. Performing cost analysis is always tricky, because cost is a parameter dependent on many factors, including non-technical ones, and operators do not share their cost structure. This analysis leveraged public cost data available on the US Department of Transportation (DoT) website [27]. Figure 7 shows the results of the cost analysis. As shown, using Super-PON enables a significant reduction in the total cost to build the OSP, in terms of both labor and materials, as well as a reduction in CO construction cost. On the other side, the cost of the ONUs and the cost of optical components in the CO are significantly higher with Super-PON. However, Fig. 7 shows that, putting together all the cost components, Super-PON enables a significantly lower overall infrastructure building cost.
The benefits of Super-PON are most impactful in countries that are building infrastructures right now; countries such as India, Brazil, Indonesia, Thailand, Vietnam, South Africa, Morocco, Kenya, the Philippines, and many more. A technology that allows cheaper infrastructure building makes a lot of sense for these countries. Moreover, reliable power is not available everywhere. By reducing the locations that require reliable power (i.e., the COs), Super-PON is able to significantly help infrastructure builds in those areas.
In addition, in countries where broadband is widely available, it is usually more readily available in urbanized areas, while suburban or rural areas are often underserved. It is relatively easy to provide service in the more densely populated urban areas. However, serving the less densely populated suburban and rural areas requires an economic investment by an operator that can be more difficult to absorb with current technologies. With its increased reach and high customer aggregation, Super-PON is especially attractive as a solution to provide service to more sparsely populated areas. As an example, an operator could use the COs at the periphery of its urban area of service as the “launch point” to deploy Super-PON to serve the suburban areas.
Just considering the United States, there are still millions of households underserved in the rural parts of the country, regions that the Federal Communications Commission (FCC) continues to help through its Connect America Fund, which was created in 2011. There are thousands of small operators in the United States that work to provide broadband connectivity to suburban and rural areas. Super-PON is a technology that can help smaller operators serve these customers.
Finally, some operators may find it advantageous to retrofit their current infrastructure to Super-PON in order to reduce the number of powered sites needed to serve a certain area (CO consolidation). Each powered site requires monitoring and maintenance. Reducing the number of powered COs reduces operational costs and increases network reliability, because the infrastructure becomes simpler and with fewer points of control and is therefore also more reliable.
4. SUPER-PON STANDARDIZATION
It is difficult for a network technology to be successful without being an industry standard. PON technologies are advanced and standardized by three organizations:
- • The IEEE 802.3 working group, for the EPON suite of protocols (including EPON, 10G-EPON, 25G-EPON, 50G-EPON);
- • ITU-T Study Group 15 Question 2 (Q2/SG15), for the G-PON suite of protocols (including G-PON, XG-PON, XGS-PON, NG–PON2); and
- • The Full Service Access Network (FSAN), an industry consortium that brings together network operators to create requirements for ITU-T Q2/SG15.
The Super-PON idea was first presented at an interim meeting of IEEE 802.3 on January 26, 2018, at a New Ethernet Application (NEA) meeting [28]. A call for interest on Super-PON happened at the July 2018 IEEE 802.3 plenary meeting, and the formation of the Physical Layers for increased-reach Ethernet optical subscriber access (Super-PON) Study Group was approved on July 12, 2018.
Things moved fast, and at the November 15, 2018 plenary, the request to form the Increased-reach Ethernet optical subscriber access (Super-PON) Task Force was approved as well. The final IEEE Standards Association board approval arrived on December 5, 2018. The IEEE P802.3cs Task Force [29] held its first meeting at the January 2019 IEEE 802.3 interim meeting. Since then, the Task Force has been working to define the Super-PON technology.
In addition to the activity in IEEE 802.3, ITU-T Q2/SG15 approved standardization of Super-PON at its July 2019 plenary meeting. Super-PON will be defined in ITU-T recommendation G.9807.3, “Wavelength multiplexed point-to-multipoint 10-gigabit-capable passive optical network” [30]. This is extremely important, because in this way the same optical devices can be leveraged by both protocol suites, resulting in larger volumes and lower cost per component.
5. IEEE P802.3CS DEVELOPMENTS
A. Task Force Objectives
The objectives of the IEEE P802.3cs Task Force are to provide Physical Layer specifications that:
- • preserve the Ethernet frame format utilizing the Ethernet media access control (MAC);
- • support a bit error rate (BER) of better than or equal to ${{10}^{- 12}}$ at the MAC/PLS [1] service interface (or the frame loss ratio equivalent);
- • support a passive point-to-multipoint ODN with a maximum reach of at least 50 km with at least a 1:64 split ratio per wavelength pair;
- • support at least 16 wavelength pairs for point-to-multipoint PON operation;
- • support the MAC data rate of 10 Gb/s downstream;
- • support the MAC data rates of 2.5 Gb/s and 10 Gb/s upstream;
- • leverage existing EPON physical coding sublayer (PCS) and physical medium attachment (PMA) to support the above MAC data rates; and
- • support tunable transmitters.
B. Specification Method
IEEE 802.3 specifies physical medium-dependent modules (PMDs) at the medium-dependent interface (MDI). However, in Super-PON there are multiple devices between two MDIs, as shown in Fig. 5: MUX/Amp, $\lambda$ router, and power splitter. To reconcile Super-PON with the IEEE 802.3 specification conventions, the P802.3cs Task Force decided to use the so-called “black link” approach [31].
With this approach, the normative standard specification defines the PMD parameters at the MDI interfaces (so everything that is between them is a “black link”) as shown in Fig. 8.
An informative annex of IEEE 802.3cs will specify a possible implementation of the black link according to the Super-PON architecture as shown in Fig. 9.
This way of specifying Super-PON is actually very flexible from an implementation standpoint. It enables the standard to define the OLT and ONU PMDs independently from the implementation of the black link. This allows for easy extensibility of the system in the future (e.g., future better amplification systems or DCMs may enable further reach or higher aggregation without changing the PMDs at the end points but just changing the way in which the black link is implemented).
C. Amplification/Wavelength Region
The first problem addressed by the Task Force was identifying the optical amplification technology and the wavelength region best suited for Super-PON [32,33]. By placing both downstream and upstream amplification in the CO, Super-PON is power limited in the downstream direction and noise limited in the upstream direction. Previous studies have shown that erbium-doped fiber amplifiers (EDFAs) [34] are most effective for amplifiers co-located with OLTs [35]. Moreover, the analysis showed that EDFAs are preferable to semiconductor optical amplifiers (SOAs) because of their lower noise figure and minimal interwavelength interactions.
The burst mode nature of the upstream traffic causes gain transients in a conventional EDFA. However, many different gain-clamping methods have been demonstrated to greatly reduce these transient effects to allow EDFA amplification of burst mode signals. For example, the use of an optical feedback loop [36], a saturating signal [37], or a feedback loop controlled optical stabilizing laser [38] can all greatly reduce the transient effects caused by TDM bursts.
Using EDFAs limits the spectral region where Super-PON has to operate to the C-band and L-band. This is also the region of the spectrum where fiber attenuation is lowest, which is very important for a long-reach access system [39].
D. Wavelength Router/Wavelength Plan
Super-PON operates over a channelized ODN, incorporating a passive wavelength router that channelizes according to its wavelength plan. The presence of a wavelength router in the ODN makes the ODN very specialized: only the wavelengths belonging to the router wavelength plan are allowed to pass and are routed to specific physical ports of the wavelength router. All other wavelengths are filtered out by the router. This poses a requirement to make the channelized ODN more future proof: the router wavelength plan needs to support multiple bands, with a wavelength in each band allocated to each router output. This allows support of multiple generations of TDM systems for each wavelength router output. Super-PON uses a 100 GHz nominal channel spacing, four wavelength bands, and 16 channels per band over the combination of the C- and L-bands. We name these bands as follows: C-band 1, C-band 2, L-band 1, and L-band 2.
Having four wavelengths associated with each wavelength router output port allows Super-PON to support seamless generation upgrades. While a first-generation ${ X}$ system operates by using wavelengths belonging to the first pair of wavelength bands (e.g., C-band 1 and L-band 1), a second-generation ${Y}$ system can be supported over the same ODN by using wavelengths belonging to the second pair of wavelength bands (e.g., C-band 2 and L-band 2) as shown in Fig. 10.
Once the upgrade cycle from generation ${X}$ to generation ${Y}$ is completed, it is then possible to leverage the pair of wavelength bands used for generation ${X}$ to support a subsequent upgrade cycle.
A natural choice to implement the Super-PON wavelength router is a cyclic arrayed waveguide grating (CAWG) [40]. It is possible to design CAWGs satisfying the Super-PON requirements of four wavelength bands over the C- and L-bands by using a free spectral range (FSR) wider than the number of channels but narrower than one-quarter of the combined C- and L-bands. Super-PON uses a 22-channel wide FSR: the 16 central channels of each spectral band are used for the 16 Super-PON channels; the two edge channels, which have substantially higher loss, are used as a guard band; and the four shoulder channels are not used in the standard, but are available for other applications at the discretion of an operator. Athermalization of the CAWGs is fundamental for Super-PON because they need to be placed outside, subject to significant temperature variations [41,42]. This can be achieved either with silicone-filled triangular grooves [43] or using mechanical compensation techniques [44]. Low-loss CAWG designs with optimized athermalization are also possible [45–47]. Table 1 shows the cyclic wavelength plan adopted by the IEEE P802.3cs Task Force for Super-PON.
The Super-PON wavelength plan defines two wavelength-band pairs, called FSR sets. Each FSR set includes two spectral bands, one for upstream and the other one for downstream. Sixteen channels are defined in each spectral band. Within each wavelength channel, the ripple is specified to be lower than 2 dB between ${-}{15}\;{\rm GHz}$ and ${+}{15}\;{\rm GHz}$ from the nominal frequency. FSR Set 1 is the one used by the current Super-PON definition; FSR Set 2 is reserved for migration to a subsequent generation and not currently used. The speed and nature of future generation systems is beyond the scope of this paper.
Figure 11 shows how the wavelength region selected for Super-PON compares with other PON technologies. The partial overlap between Super-PON and XGS-PON/NG-PON2 wavelengths may enable component sharing among these technologies and is not a concern because Super-PON operates over a different ODN than that of XGS–PON and NG-PON2.
E. Dispersion Compensation
Operating in the C-band and the L-band requires careful consideration of the effects of chromatic dispersion, especially at the 10 Gb/s speeds (chromatic dispersion is not expected to be a significant issue for the 2.5 Gb/s upstream speed). OLT transmitters need to have a high extinction ratio (ER) and relatively low output power, since they are immediately followed by the multiplexer and booster amplifier. These requirements likely result in the use of external modulation, either a deeply biased electron absorption modulator (EAM) [48] or a Mach–Zehnder modulator [49]. These external modulators are able to create signals with zero or negative chirp to combat chromatic dispersion and are commonly used for ${\gt}{70}\;{\rm km}$ transceivers. Therefore, it is possible to achieve 50 km of transmission with little penalty, avoiding the need for a DCM in the downstream direction.
Table 1. IEEE P802.3cs Wavelength Plan
The ONU transmitter is required to output a higher power and is a component extremely sensitive to cost. It is important to allow multiple technologies to be used in the ONU transmitter, including directly modulated lasers (DMLs), to enable cost-effective solutions. DMLs and low-bias EAMs tend to generate signals with significant positive chirp. Positively chirped signals have optimal sensitivity at a net negative residual dispersion and are very intolerant to net positive dispersion.
We experimentally analyzed a commercial low-bias EAM transmitter and a DML [50]. Figure 12 shows the sensitivity penalty of the two 10 Gb/s transmitters after 50 km of transmission and a tunable DCM. The reference sensitivity was back-to-back without the DCM. This shows that both the DML and the externally modulated laser (EML) have improved sensitivity in the negative residual dispersion region. Therefore, it is better to overcompensate for link dispersion than to undercompensate. These results suggest that both DML- and EML-based solutions can be used in ONU transmitters if the residual dispersion of all links can be contained between ${-}{500}$ and 0 ps/nm.
The Super-PON architecture incorporates into the MUX/Amp module a DCM for upstream, but not for downstream (see Fig. 5). Chromatic dispersion values typically range from 17 to 20 ps/nm/km in the L-band. To enable strongly chirped DMLs, two different DCMs are needed: a 50 km DCM to compensate for 25 to 50 km links, and a 30 km DCM to compensate for 0 to 30 km links [50]. The 5 km overlap allows ONUs located between 25 and 30 km away from the OLT to be serviced by either DCM. In this case, links ${\lt}{5}\;{\rm km}$ will suffer from an elevated sensitivity penalty. This is, however, acceptable, as short links are likely to have a significantly higher optical signal to noise ratio (OSNR) and power at the OLT receiver than longer links to overcome the penalty.
Two types of DCMs have been considered for Super-PON:
FBG DCMs are channelized and able to cover the full C- and L-bands [51]. They have a fixed insertion loss and a fixed latency. The latency of FBG DCMs is very small (i.e., ${\lt}{0.025}\;\unicode{x00B5} {\rm s}$), and therefore their usage does not alter the latency properties of PON links between downstream and upstream.
Table 2. Typical Loss of a 50 km Super-PON Access Link
Negative dispersion fiber DCMs are just spools of negative dispersion fiber. Their insertion loss and latency are dependent on the length of the fiber used to build them and they are not channelized [52]. To compensate for the chromatic dispersion of a 50 km optical fiber link, a negative dispersion fiber DCM introduces a latency of around 25 µs, which is significant. Since a DCM is only proposed for the upstream direction, a fiber-based DCM results in more upstream latency than downstream latency. However, PON protocols are tolerant of latency asymmetry, and therefore this type of DCM is also acceptable for Super-PON.
Both being acceptable, the IEEE P802.3cs Task Force decided to allow both types of DCMs to be used for Super-PON.
F. Optical Power Budget
One of the activities carried forward by the Task Force in order to compute the power loss of a typical 50 km access link has been determining an updated value for the link fiber attenuation, because the values specified in G.652 [53] appeared to be too conservative in respect to current fiber deployments. Appendix 1 of G.652 recommends the values 0.275 dB/km for the C-band and 0.35 dB/km for the L-band; based on the analysis in [54], the Task Force decided to adopt the value 0.24 dB/km for both the C-band and L-band. An informative Annex has been included in the draft P802.3cs standard to document this decision. This is consistent with the specification method, where the black link implementation is informative as well.
Table 2 shows the typical loss of a 50 km Super-PON access link, from the MUX/Amp module to an ONU, as approved by the IEEE P802.3cs Task Force and based on the updated link fiber attenuation value. Table 3 shows the assumed transmission penalties for all transmission effects other than Raman, which is accounted for separately.
Table 3. Link and Receiver Assumptions
It is common practice in the industry to use L-band transmitters for downstream and C-band transmitters for upstream (e.g., this is what has been chosen for NG-PON2). However, from a physics point of view, this arrangement is suboptimal due to the Raman effect [55]. The Raman effect causes a transfer of energy from the shorter wavelengths to the longer wavelengths when multiple wavelengths copropagate or counterpropagate in a single fiber strand [56]. This transfer of energy is given by Eqs. (1) and (2) from [57]. For Super-PON, this means a transfer of energy from wavelengths in the C-band to wavelengths in the L-band. When using C-band wavelengths for upstream, the Raman effect results in a significant power transfer from the low-power upstream signals to the high-power downstream signals. The power reduction for upstream can be calculated by considering each L-band downstream wavelength individually as given by
- • ${A_{\rm on - off,dB}}$ is the on–off Raman-induced power loss in decibels (dB);
- • ${N_{\rm ch}}$ is the total number of Stokes channels (power receiving);
- • ${\lambda _{\rm pump}}$ is the wavelength of the pump channel (power donor);
- • ${\lambda _{{\rm Stokes},i}}$ is the power of the $i$th Stokes channel;
- • ${C_{R,i}}$ is the Raman efficiency coefficient for the $i$th Stokes channel;
- • ${P_{{\rm Stokes},i}}$ is the power of the $i$th Stokes channel; and
- • ${\alpha _{{\rm dB,Stokes},i}}$ is the attenuation of the $i$th Stokes channel in decibels per kilometer (dB/km).
The ${{ C}_{R,i}}$ increases with wavelength separation, so the greatest Raman penalty is experienced by the shortest-wavelength upstream channel.
The downstream channels have ${\gt}{20}\;{\rm dB}$ more power than the upstream channels in the feeder fiber, because the upstream channels have already passed the power splitter and CAWG before arriving in the feeder fiber. Therefore, the Raman effect between upstream channels can be ignored. In contrast, Raman interactions between the downstream channels are significant, with power transferring from the shorter downstream wavelengths to the longer downstream wavelengths. The same equation can be used to calculate the Raman penalty for the shortest downstream channel by using its wavelength as the pump wavelength and all other downstream wavelengths as Stokes wavelengths. The downstream Raman penalty is much smaller because of the close pump-Stokes separation (i.e., lower ${{C}_{R,i}}$ values). The resulting penalty can be handled by increasing the gain and maximum power of the downstream EDFA in the MUX/Amp module.
The power reduction was calculated using two methods: a numerical ordinary differential equation (ODE) solver of Eqs. (1) and (2) from [57] and algebraically using the equation above. Input powers across all 16 wavelengths were uniform, and scrambled polarizations were assumed. The two methods produced results within 0.2 dB of each other. Figure 13 shows the Raman penalty of the worst channel for the upstream and downstream wavelengths using the numerical ODE solver.
The upstream C-band transmitters need to be more powerful to counterbalance the Raman penalty. Figure 14 shows the required power levels as a function of the link budget. The transmission penalties and receiver sensitivities are shown in Table 3. The Raman power penalty is determined by the downstream power. The calculations assumed the downstream powers were 2 dB higher than the minimum required to close the link. This allows for some channel non-uniformity in the downstream direction.
With a 41 dB link budget, this results in an upstream power level of 7.5 dBm at 10 Gb/s. Lasers at this high power level are very expensive because they are very difficult to manufacture, resulting in very low yields. Requiring these lasers for the ONUs, the most cost-sensitive component of any PON, is not economically feasible. For this reason, the Task Force decided to diverge from industry practice and use L-band wavelengths for upstream and C-band wavelengths for downstream [58,59].
When using L-band wavelengths for upstream, the Raman effect transfers power to the upstream signals. This is illustrated by the negative Raman penalty in Fig. 15. This Raman gain cannot be relied upon for two reasons:
- • the gain in power will only be a fraction if fewer wavelengths are used, and
- • gain from a modulated pump will also result in additional noise.
Therefore, the Raman gain/penalty is assumed to be zero when calculating the required powers for upstream operation in the L-band. The Raman penalty for the downstream wavelengths is almost unchanged because the downstream channels have a much higher power than the upstream ones.
Figure 16 shows the required power levels as a function of the link budget when L-band wavelengths are used for upstream. A 41 dB link budget requires an upstream launch power level of 4 dBm at 10 Gb/s, the same power level of the NG-PON2 standard, something the industry is comfortable with.
Table 4 summarizes the required launch power levels for upstream and downstream transmitters at an extinction ratio of 6 dB as a function of which wavelength band of a Super-PON FSR Set is used for what. As shown, using C-band for downstream and L-band for upstream results in reasonable power levels at the 10 Gb/s upstream speed and in comfortable power levels at the 2.5 Gb/s upstream speed.
Table 4. Transmitter Launch Power Levels
6. DEPLOYMENT EXPERIENCE
Google Fiber is using in production a preliminary version of Super-PON [60] in the city of San Antonio, TX, USA. This version operates at the G-PON speed (i.e., 2.5 Gb/s downstream, 1.25 Gb/s upstream) and leverages the G-PON TC layer. Using Super-PON sped up the initial deployment by 8 to 10 weeks and increased the first year revenues by 36%. This is a real-life demonstration of the viability and usefulness of Super-PON.
One of the issues Google Fiber initially encountered in building the San Antonio infrastructure was finding suitable locations to build COs. The long reach of Super-PON was fundamental to enable a good coverage of the city with few COs. This avoided the need to negotiate with the city for huts or cabinet locations, preventing unexpected delays. Additionally, the high subscriber aggregation of Super-PON enabled the use of thin-profile microtrenching, which enabled a rapid network buildout.
7. CONCLUSION
Super-PON is an upcoming PON optical layer able to support an increased optical reach of up to 50 km with up to 1024 customers per feeder fiber over a passive ODN. This is achieved by leveraging WDM, optical amplification, wavelength routing, and existing TDM PON protocols. There is significant interest in the industry as demonstrated by the two standardization efforts concurrently happening in IEEE 802.3 and ITU-T Q2/SG15. The viability of this technology has been proven by the production deployment of Google Fiber.
Acknowledgment
The authors acknowledge all contributors to the work of the IEEE P802.3cs Task Force, in particular Duane Remein, Mark Laubach, Xiangjun Zhao, Frank Effenberger, Marek Hajduczenia, Vincent Ferretti, Glen Kramer, Weyl Wang, Weiqing Zhang, Qing Wei, Junming An, Daisuke Ogawa, Henk Bulthuis, Patrick Lebeau, Robert Lingle, Jesper Bevensee Jensen, and Earl Parsons.
REFERENCES
1. “IEEE Std 802.3-2018—IEEE Standard for Ethernet,” Aug. 2018.
2. “10-Gigabit-capable symmetric passive optical network (XGS-PON),” ITU-T Recommendation G.9807.1, June 2016.
3. “Gigabit-capable passive optical networks (G-PON): transmission convergence layer specification,” ITU-T Recommendation G.984.3, Jan. 2014.
4. “Gigabit-capable passive optical networks (G-PON): physical media dependent (PMD) layer specification,” ITU-T Recommendation G.984.2, Aug. 2019.
5. “40-gigabit-capable passive optical networks 2 (NG-PON2): physical media dependent (PMD) layer specification,” ITU-T Recommendation G.989.2, Feb. 2019.
6. https://www.itu.int/en/ITU-T/studygroups/2017-2020/15/Pages/default.aspx.
7. http://www.ieee802.org/3/ca/index.shtml.
8. J. Sugawa and H. Ikeda, “Development of OLT using semiconductor optical amplifiers as booster and preamplifier for loss-budget extension in 10.3-Gb/s PON system,” in Optical Fiber Communication Conference and Exhibition and the National Fiber Optic Engineers Conference (OFC/NFOEC), Los Angeles, California, USA, 2012, paper OTh4G.4.
9. “Gigabit-capable passive optical networks (GPON): reach extension,” ITU-T Recommendation G.984.6, March 2008.
10. “10 gigabit-capable passive optical networks (XG(S)-PON): reach extension,” ITU-T Recommendation G.9807.2, Aug. 2017.
11. C. Antony, P. Ossieur, A. M. Clarke, A. Naughton, H. G. Krimmel, Y. Chang, A. Borghesani, D. Moodie, A. Poustie, R. Wyatt, B. Harmon, I. Lealman, G. Maxwell, D. Rogers, D. W. Smith, D. Nesset, R. P. Davey, and P. D. Townsend, “Demonstration of a carrier distributed, 8192-split hybrid DWDM-TDMA PON over 124km field-installed fibers,” in Optical Fiber Communication Conference (OFC), March 2010.
12. I. Van de Voorde, C. M. Martin, I. Vandewege, and X. Z. Oiu, “The superPON demonstrator: an exploration of possible evolution paths for optical access networks,” IEEE Commun. Mag.38(2), 74–82 (2000). [CrossRef]
13. G. Talli and P. D. Townsend, “Hybrid DWDM-TDM long-reach PON for next-generation optical access,” J. Lightwave Technol.24, 2827–2834 (2006). [CrossRef]
14. D. P. Shea and J. E. Mitchell, “Long-reach optical access technologies,” IEEE Netw.21, 5–11 (2007). [CrossRef]
15. E. Wong, M. Muller, and M. C. Amann, “Colourless operation of short-cavity VCSELs in C-minus band for TWDM-PONs,” Electron. Lett.49, 282–284 (2013). [CrossRef]
16. Z. Li, L. Yi, W. Wei, M. Bi, H. He, S. Xiao, and W. Hu, “Symmetric 40-Gb/s, 100-km passive reach TWDM-PON with 53-dB loss budget,” J. Lightwave Technol.32, 3991–3998 (2014). [CrossRef]
17. D. B. Payne and R. P. Davey, “The future of fibre access systems?” BT Technol. J.20, 104–114 (2002). [CrossRef]
18. F. Effenberger, D. Cleary, O. Haran, G. Kramer, R. D. Li, M. Oron, and T. Pfeiffer, “An introduction to PON technologies [Topics in Optical Communications],” IEEE Commun. Mag.45(3), S17–S25 (2007). [CrossRef]
19. P. P. Iannone and K. C. Reichmann, “Optical access beyond 10 Gb/s PON,” in 36th European Conference and Exhibition on Optical Communication (2010), pp. 1–5.
20. D. Carey, N. Brandonisio, S. Porto, A. Naughton, P. Ossieur, N. Parsons, G. Talli, and P. Townsend, “Dynamically reconfigurable TDM-DWDM PON ring architecture for efficient rural deployment,” in European Conference and Exhibition on Optical Communication (ECOC), September 2016.
21. L. A. Coldren, “Monolithic tunable diode lasers,” IEEE J. Sel. Top. Quantum Electron.6, 988–999 (2000). [CrossRef]
22. T. Zhang, “Tunable laser drivers for next generation WDM-based PON networks,” in Optical Fiber Communication Conference (OFC), March 2019.
23. R. Bonk, W. Poehlmann, D. van Veen, J. Galaro, R. Farah, H. Schmuck, and T. Pfeiffer, “The underestimated challenges of burst-mode WDM transmission in TWDM-PON,” Opt. Fiber Technol.26, 59–70 (2015). [CrossRef]
24. D. van Veen, W. Poehlmann, B. Farah, T. Pfeiffer, and P. Vetter, “Measurement and mitigation of wavelength drift due to self-heating of tunable burst-mode DML for TWDM-PON,” in Optical Fiber Communication Conference (OFC), Optical Society of America, San Francisco, California, USA, 2014, paper W1D.6.
25. X. Wu, D. Liu, H. Lin, X. Liu, and F. Effenberger, “Burst mode wavelength stabilization,” presented at the IEEE 802.3 meeting, Jan. 2019, http://www.ieee802.org/3/cs/public/201901/201901_02-Burst-Mode_Wavelength_Stabilization.pdf.
26. L. Du and C. DeSanti, “Super-PON economic feasibility,” presented at the IEEE 802.3 meeting, Nov. 2018, http://www.ieee802.org/3/SUPER_PON/public/201811/Super-PON_Economic_Feasibility_v1b.pdf.
28. C. DeSanti, L. Du, C. Lam, and J. Jiang, “Super-PON,” presented at the IEEE 802.3 meeting, Jan. 2018, http://www.ieee802.org/3/ad_hoc/ngrates/public/18_01/desanti_nea_01a_0118.pdf.
29. http://www.ieee802.org/3/cs/index.html.
30. https://www.itu.int/ITU-T/workprog/wp_item.aspx?isn=15208.
31. C. DeSanti, “Specifying Super-PON,” presented at the IEEE 802.3 meeting, March 2019, http://www.ieee802.org/3/cs/public/201903/20190312-Specifying_Super-PON.pdf.
32. W. Wang and W. Zhang, “Optical amplification for a Super-PON system,” presented at the IEEE 802.3 meeting, Sept. 2018, http://www.ieee802.org/3/SUPER_PON/public/201809/Optical_Amplification_v1.pdf.
33. Q. Wei, “Burst-mode capable EDFAs,” presented at the IEEE 802.3 meeting, Jan. 2019, http://www.ieee802.org/3/cs/public/201901/201901_07-Burst-Mode_capable_EDFAs.pdf.
34. E. Desurvire, J. R. Simpson, and P. C. Becker, “High-gain erbium-doped traveling-wave fiber amplifier,” Opt. Lett.12, 888–890 (1987). [CrossRef]
35. N. Genay, P. Chanclou, F. Saliou, Q. Liu, T. Soret, and L. Guillo, “Solutions for budget increase for the next generation optical access network,” in International Conference on Transparent Optical Networks (ICTON), IEEE, Rome, Italy, 2007, paper Tu.A4.7.
36. A. Joon Tae and K. Kyong Hon, “All-optical gain-clamped erbium-doped fiber amplifier with improved noise figure and freedom from relaxation oscillation,” IEEE Photon. Technol. Lett.16, 84–86 (2004). [CrossRef]
37. A. Kaszubowska-Anandarajah, R. Oberland, E. Bravi, A. Surpin, O. Aharoni, U. Ghera, R. Giller, E. Connolly, E. K. MacHale, M. Todd, G. Talli, and D. McDonald, “EDFA transient suppression in optical burst switching systems,” in 14th International Conference on Transparent Optical Networks (ICTON), July 2012.
38. H. G. Krimmel, T. Pfeiffer, B. Deppisch, and L. Jentsch, “Hybrid electro-optical feedback gain-stabilized EDFAs for long-reach wavelength-multiplexed passive optical networks,” in 35th European Conference on Optical Communication, September 2009.
39. L. Du, “Super-PON wavelengths considerations,” presented at theIEEE 802.3 meeting, March 2019, http://www.ieee802.org/3/cs/public/201903/20190312-Super-PON_Wavelength_Considerations.pdf.
40. A. Kaneko, T. Goh, H. Yamada, T. Tanaka, and L. Ogawa, “Design and applications of silica-based planar lightwave circuits,” IEEE J. Sel. Top. Quantum Electron.5, 1227–1236 (1999). [CrossRef]
41. J. An, “Cyclical AWG for Super-PON system,” presented at theIEEE 802.3 meeting, March 2019, http://www.ieee802.org/3/cs/public/201903/20190312-CAWG_for_Super-PON.pdf.
42. D. Ogawa, “Cyclic Athermal AWG λ Router for Super-PON,” presented at the IEEE 802.3 meeting, March 2019, http://www.ieee802.org/3/cs/public/201903/20190312-CAWG_Router_for_Super-PON.pdf.
43. Y. Inoue, A. Kaneko, F. Hanawa, H. Takahashi, K. Hattori, and S. Sumida, “Athermal silica-based arrayed-waveguide grating multiplexer,” Electron. Lett.33, 1945–1947 (1997). [CrossRef]
44. L. Leick, M. Boulanger, J. G. Nielsen, H. Imam, and J. Ingenhoff, “Athermal AWGs for colourless WDM-PON with −40°C to +70°C and underwater operation,” in Optical Fiber Communication Conference (OFC) (2016).
45. H. Bulthuis, “Low loss CAWG design for Super-PON systems,” presented at the IEEE 802.3 meeting, July 2019,http://www.ieee802.org/3/cs/public/201907/20190717-Low_Loss_CAWG.pdf.
46. S. Kamei, M. Kohtoku, T. Shibata, and T. Kitoh, “Athermal Mach-Zehnder interferometer-synchronised arrayed waveguide grating multi/demultiplexer with low loss and wide passband,” Electron. Lett.44, 201–202 (2008). [CrossRef]
47. K. Hirabayashi, N. Koshobu, J. Kobayashi, M. Itoh, and S. Kamei, “Reduction of second-order temperature dependence of silica-based athermal AWG by using two resin-filled grooves,” IEEE Photon. Technol. Lett.23, 676–678 (2011). [CrossRef]
48. M. Ishizaka, M. Yamaguchi, J. Shimizu, and K. Komatsu, “The transmission capability of a 10-Gb/s electroabsorption modulator integrated DFB laser using the offset bias chirp reduction technique,” IEEE Photon. Technol. Lett.9, 1628–1630 (1997). [CrossRef]
49. J. C. Cartledge, “Performance of 10 Gb/s lightwave systems based on lithium niobate Mach-Zehnder modulators with asymmetric Y-branch waveguides,” IEEE Photon. Technol. Lett.7, 1090–1092 (1995). [CrossRef]
50. L. Du, “Super-PON chromatic dispersion and DCM for black link,” presented in the IEEE P802.3cs concall, April 2020, http://www.ieee802.org/3/cs/public/AdHoc/20200430-Du_3cs_01.pdf.
51. P. Lebeau, “Chromatic dispersion compensation in Super-PON networks with FBG-based, multi-channel chromatic dispersion compensators,” presented at the IEEE 802.3 meeting, Jan. 2019, http://www.ieee802.org/3/cs/public/201901/201901_05-Chromatic_Dispersion%20Compensation_in_Super-PON.pdf.
52. R. Lingle, “Chromatic dispersion compensation with negative dispersion fiber,” presented at the IEEE 802.3 meeting, May 2019, http://www.ieee802.org/3/cs/public/201905/20190521-Fiber_Based_DSCMa.pdf.
53. “Characteristics of a single-mode optical fibre and cable,” ITU-T Recommendation G.652, Nov. 2016.
54. V. Ferretti, “Considerations for Link Loss–2,” presented at theIEEE 802.3 meeting, Nov. 2019, http://www.ieee802.org/3/cs/public/201911/20191112-Ferretti_3cs_01.pdf.
55. R. Gaudino, V. Curri, and S. Capriata, “Propagation impairments due to Raman effect on the coexistence of GPON, XG-PON, RF-video and TWDM-PON,” in European Conference and Exhibition on Optical Communication (ECOC) (2013), pp. 6–19.
56. R. W. Hellwarth, “Theory of stimulated Raman scattering,” Phys. Rev.182, 482–494 (1963). [CrossRef]
57. J. Bromage, “Raman amplification for fiber communications systems,” J. Lightwave Technol.22, 79–93 (2004). [CrossRef]
58. L. Du, “Super-PON link budget analysis—effect of Raman,” presented at the IEEE 802.3 meeting, Jan. 2020, http://www.ieee802.org/3/cs/public/202001/20200121-Du_3cs_02a.pdf.
59. L. Du, “Super-PON link budget analysis,” presented at theIEEE 802.3 meeting, March 2020, http://www.ieee802.org/3/cs/public/202003/20200402-Du_3cs_01.pdf.
60. L. Du, X. Zhao, S. Yin, T. Zhang, A. E. T. Barratt, J. Jiang, D. Wang, J. Geng, C. DeSanti, and C. F. Lam, “Long-reach wavelength-routed TWDM PON: technology and deployment,” J. Lightwave Technol.37, 688–697 (2019). [CrossRef]
Claudio DeSanti (M’09) received his Master cum laude in Computer Engineering from University of Pisa, Italy, in 1995 and his Ph.D. in Computer Engineering from Scuola Superiore di Studi Universitari e di Perfezionamento S. Anna, Pisa, Italy, in 2000. In March 2020, he joined Dell Technologies in Santa Clara, CA, USA, as an Engineering Technologist. From 2000 to 2017, he worked in Cisco Systems up to the Fellow level. He contributed to the development and success of many Cisco products, including Cisco’s UCS, Cisco’s Nexus 7K and 5K, Cisco’s MDS, and Cisco’s Catalyst switches. From 2017 to February 2020, he worked in Google Fiber in Mountain View, CA, USA, as a Senior Network Architect Lead. Dr. DeSanti is highly recognized in standards bodies and industry associations, such as INCITS T11, IEEE 802.3/802.1, IETF, and ITU-T Q2/SG15. He has been chairman of conferences and standard committees as well as the technical editor of multiple standards. He is the author of several patents and recipient of four INCITS technical awards. He is chairman of the IEEE P802.3cs “Super-PON” Task Force and editor of ITU-T recommendation G.9807.3.
Liang B. Du (M’08) received his B.E. and B.Com degrees majoring in electrical engineering and finance from Monash University, Victoria, Australia, in 2007, and his Ph.D. in electrical engineering also from Monash University in 2011. He was a research fellow at Monash University from 2011 to 2013. In 2013, he joined Google Fiber, Mountain View, CA, USA as a Photonics engineer designing next-generation access networks.
Jhon Guarin received his B.S. in Civil Engineering from the University of Medellin, Colombia in 2000. He joined Google Fiber San Antonio in 2015, where he is currently the Technical Operations Lead responsible for the FTTH network planning, implementation, installation, and maintenance. Jhon has been managing the network implementation of the Super-PON system in San Antonio since its inception. Prior to Google Fiber, Jhon worked for seven years with Rogers Communication, where he contributed to the deployments of FTTx, FTTH—GPON, and RFOG.
Jason Bone received his B.S. in Computer Engineering Technology from the University of Houston in 2001. He joined Google Fiber San Antonio in 2016, where he is currently the Engineering and Operations Lead responsible for FTTH network planning, engineering, operations, and maintenance of the San Antonio Super-PON network. Prior to Google Fiber, Jason worked at En-Touch Systems in Houston, where he managed both the inside and outside engineering of a network consisting of several types of deployments, including telephony (twisted pair), RF, and FTTH.
Cedric F. Lam (SM’07) received his B.Eng. with first-class honors in electrical and electronic engineering from the University of Hong Kong in 1993 and his Ph.D. in electrical engineering from UCLA in 1999. From 1999 to 2002, he worked at AT&T Labs–Research as a Senior Technical Staff Member. From 2002 to 2009, he was Chief System Architect at OpVista, leading the development of an ultradense DWDM transport system. He joined Google in Mountain View, CA, USA, in 2009 as a Datacenter Network Architect. In 2010, he cofounded the Google Fiber program and is now an Engineering Director at Google Fiber, responsible for network architecture and network technology development. Dr. Lam is a Fellow of The Optical Society (OSA).
