Skip to main content
Publications lead hero image abstract pattern


Written By:

Mark Laubach, Ed Boyd, James Harley, and Fernando Villarruel, Ciena Corporation

Published: 26 Apr 2024


CTN Issue: April 2024

A note from the editor:

Coherent technologies offer promising solutions for high-speed optical access beyond the rate of 100 Gb/s. By embracing coherent detection, optical access system capacity can be enhanced by encoding information in amplitude, phase, as well as polarization components of a carrier.  With the use of digital signal processing (DSP), coherent receivers can mitigate signal impairments through the use of distortion equalization and fiber dispersion compensation. Compared to Intensity Modulation / Direct Detection (IMDD), coherent detection excels in chromatic dispersion (CD) equalization, enabling the utilization of C-band wavelengths, achieving a higher access link budget. On the other hand, system complexity is one of the challenges of integrating coherence technologies into optical access systems. There are ongoing efforts focused on streamlining coherent systems to meet the optical access requirements of cost and complexity reduction.

This month’s issue of CTN offers a comprehensive review of the industrial landscape on coherence for optical access. The authors introduce the principles of coherent signal transmission and reception while delving into standards development within IEEE, ITU, and other groups. Candidate solutions are analyzed to underscore future directions in this area. We hope you enjoy this tour of an up-and-coming technology that can enhance how information is delivered to homes and businesses.

Yuanqiu Luo, CTN Guest Editor

Transitioning PON to Coherent Technology for 100 Gbps and Beyond: Expanding PON Service Rates from Residential and Small Business to Enterprise

Mark Laubach

Mark Laubach

Ciena Corporation

Ed Boyd

Ed Boyd

Ciena Corporation

James Harley

James Harley

Ciena Corporation

Fernando Villarruel

Fernando Villarruel

Ciena Corporation


The market growth of Passive Optical Networks (PON) and data service rates suggest 100 Gbps as a viable next step for access network service technology.  In parallel, the point-to-point market for coherent single-carrier (SC) optical transport technologies for data center and extended campus applications are indicating increasing volumes with established ecosystems leading to expected cost reductions for single-carrier coherent optical modules.  This alignment in time with the demand for higher-speed PON access network services enable the cost-effective reuse of coherent modulation technology into the next generation of PON, known as Coherent PON (CPON).  The use cases for CPON derive from lowest relative-cost ONU for 100+ Gbps PON, overlay of CPON onto existing brownfield optical data networks (ODNs) preserving existing revenue-bearing legacy PON services, true Ethernet service rates, and extensibility to support enterprise Ethernet service rates and service level agreements (SLAs) competitive to what is offered on point-to-point transport fiber optics and with lower operator OPEX.

ITU-T Study Group 15 Question 2 [1] on optical access networks is drafting “beyond 50G” requirements and candidate technologies for their G.Suppl.VHSP supplement development.  In parallel, CableLabs CPON Working Group [2] efforts are developing requirements and design specifications for 100+ Gbps CPON systems.  It is expected that vendors together with CableLabs will bring proposals into the Q2 process for consideration. Both ITU-T and CableLabs activities are work-in-progress. This article represents our technical contributions to date on single-carrier coherent for next-generation CPON.

What is PON?

Over two decades ago, standards were published separately in both the ITU-T and IEEE for providing broadband subscriber access services over a passive ODN (see Table 1 and Table 2).   The term PON has been socialized over time to mean the suite of these standards and the equipment built to those standards. A typical PON system consists of an Optical Line Terminal (OLT) connected to a network of fiber and optical splitters providing a point-to-multipoint direct fiber connections to an Optical Network Terminals/Units (ONTs or ONUs) in subscriber homes.  An OLT will typically service up to 64 subscribers or 128 subscribers with distances typically up to 20km or longer.   OLTs today can be located in a central office, headend, hub, residential pedestal, or a utility pole “strand mounted” enclosure referred to as a Remote OLT. Figure 1 is an illustration of a general PON network comprised of backend operator services, the OLT, the ODN consisting of passive optical fiber links and splitters, and subscriber/small-business ONUs.

Figure 1: General PON access network architecture
Figure 1: General PON access network architecture

The perceived capacity of a PON standard relates directly to the line rate of the optical signal used in the downstream and upstream directions. For example, at 10 Gbps symmetric PON in either ITU-T or IEEE EPON standard has roughly a 10 Gbps optical line rate. The actual service rate capacity is reduced proportionally related to the framing and forward error correction (FEC) overheads. This is typically about 13% to 15%.   Therefore, if a service provider today advertises a “10 Gbps” PON fiber service, the maximum datarate any subscriber can experience is approximately ~8.7 Gbps. With legacy PON deployments the advertised line speed of the technology is not what the customer experiences.

Each generation of PON increases the downstream and upstream speeds.  Table 1 and Table 2 separately summarize ITU-T PON and IEEE EPON standards and nominal downstream and upstream line rates from 2003 to 2023:

Table 1: ITU-PON generations and nominal line rates
ITU-T PON Standards
YearNameSpecificationDown Line Rate GbpsUp Line Rate Gbps
2003Gigabit PON (GPON)G.984.1 [3]2.51.25
2010XG-PONG.987.1 [4]102.5
2013NG-PON2 1G.989.1 [5]4 x 104 x 10
2016XG(S)-PONG.9807.1 [6]1010
2019/23HSP PON 1G.9804.1 [7]5050, 25, or 12.5
202125GS-PON 2V3.0 [8]2510 or 25


NG-PON2 provides up to four symmetric channel-pairs of PON services at 10 Gbps each, with an aggregate OLT capacity of 40 Gbps using a time and wavelength division multiplexing (TWDM) method.  HSP TWDM-PON is four channel-pairs of 50 Gbps each, with an aggregate of 200 Gbps. 

2 25GS-PON was developed by the 25G-PON MSA to answer a market need for 25 Gbps symmetric services. A “delta” specification was produced based on the XG(S)-PON Transmission Convergence (TC) layer from the ITU-T combined with optical transceiver and FEC aspects of the 25 Gbps physical layer developed for IEEE 802.3 Ethernet 25G + 50G EPON.

Table 2: IEEE EPON generations and nominal line rates
IEEE 802.3 Ethernet PON Standards [9]
YearName802.3 ClauseDown Line Rate GbpsUp Line   Rate Gbps
2003EPON64, 6511
200910G-EPON74, 75, 76, 771010
202025G + 50G EPON141, 142, 143, 14425+25, 2525+25, 25, 10


Supporting multiple generations of PON on the same ODN has been a consistent challenge.  Initially in 2003, the PON standards focused on re-using “cheap” uncooled upstream lasers (for lowest cost ONUs).  The first generation of GPON and EPON each had wide upstream channels located in the O-band.  As speeds increased the lowest-cost ONU laser both narrowed and stayed in the O-band.  25GS-PON and 50G HSP PON also use the O-band for downstream and upstream.  As illustrated in Figure 2, there are some uses of the S- and L-bands for GPON and 10G-EPON/XG(S)-PON downstream, and NG-PON2 and HSP TWDM PON use the C/L-band.  Notably, the legacy “RF Video” channel in the C-band does not appear as a coexistence requirement from the operators at this time in the ITU-T Beyond 50G supplement work. The C/L-band can accommodate many coherent channels.  In both ITU-T and IEEE care has been taken with standardized wavelength choices in the O-band to permit dual-generation and triple-generation coexistence options. The intent being that existing deployed revenue generating PON services will continue to operate, and individual operators will choose their own business paths.  Also note that NG-PON2 and HSP TWDM PON deploy in the C/L-band and fortunately have stayed out the accumulating and crowded use of the O-band.

Figure 2: Generational wavelength use of legacy PON systems
Figure 2: Generational wavelength use of legacy PON systems


  1. Original GPON G.984.2 [14] upstream “cheap” lasers were wide ±50. G.984.5 [15] provided US options: regular: ±50, reduced ±20 (shown), and narrow ±10 and Enhancement Bands 1 (not shown, in water peak area) and 2 (shown) for DS and US. 1G EPON IEEE 802.3ah (2004) aligned with the 1490 DS and 1310 US wavelengths as GPON with upstream lasers remaining wide ±50.
  2. NG-PON2 (TWDM-PON, Point-to-point WDM PON) G.989 series. The TWDM PON DS and US bands each have 8 channels arranged in DS+US channel pairs in the C/L bands and tunable ONT/ONU.   

Dynamic Bandwidth Allocation Technology

In PON, the downstream transmission signal is a broadcast from the single OLT transmitter to many ONU receivers and the upstream is many ONU transmitters to a single OLT receiver.  The downstream and upstream sharing method for single-carrier channels (one downstream channel paired with one upstream channel) are known as Time Domain Multiplexing (Access) or TDM/TMDA.  In the upstream direction individual ONUs are given permission (granted) to transmit by the Dynamic Bandwidth Assignment (DBA) function within the OLT.  This is to control and share access and to prevent signal-corrupting collisions from multiple ONUs transmitting at the same time.  The DBA’s granting of transmission opportunities is flexible in time and accounts for ONU data service needs (requests) as well as other quality of service (QoS) scheduling needs per SLA that include a mixture of fixed, guaranteed, and best effort upstream resource allocations.  Grants are allocated separately per ONU “service flow” as configured by the service provider (e.g., separate service flows for high-speed data, voice, video, network slices, etc.).

DBA scheduling for access networks combines both reactive and predictive methods. These techniques and the statistically bursty and time of day nature of residential and small business traffic, directly support oversubscription models. That is, the operator can offer service to more customers (ONUs) than could be supported in a static enterprise fixed-service rate allocation model.  Note that DBA operation and QoS controls are directly observable by customers measuring data rates, response times, and jitter of their upstream traffic (i.e., speed tests). 

Flexible Enterprise Service Rates, Night-time Burst Services, and OPEX Upgrade Savings

In addition to the fundamental service rate and QoS management, the DBA functions allow the operator to provision specific response times per customer.    As an example in a 100 Gbps CPON, one or more enterprise customers may desire a 10 Gbps symmetric service with low response time delay equivalent to a dedicated point-to-point 10 Gbps optical service. Other customers may desire a 5 Gbps service, with a longer predictable response time, with some amount of guaranteed burst rate over the base 5 Gbps dedicated rate.  Other customers might contract for simple best effort services where delay and delay-variation are unspecified, but still a very usable service.  A key point is that when the operator installs a 100 Gbps CPON ONU at the customer’s premises the DBA is responsible for maintaining the customer’s contracted service rate as configured by the operator.  Example 1: no hit to OPEX for SLA changes: at any given time, a customer may wish to make changes to their SLA. For example, a customer contracts for a 5 Gbps service and at a later time needs to upgrade to a higher service rate, e.g. 8 or 12 Gbps.  Aside from billing system changes, the operator simply re-configures the OLT system, and the DBA now provides the new service rate without changing the ONU hardware.  Example 2: customers may also contract for a guaranteed/best effort higher rate bursty service at night to accommodate cloud-based backup applications.  If the customer subscribes to a fixed rate 24-hour transport rate (e.g. 5 Gbps), nightly bursty allocation can easily be 25 Gbps or higher, all on the same 100 Gbps CPON ONU hardware.  The customer’s peak night-time service rate will be limited by the speed of the User-Network Interface (UNI), e.g. Ethernet speeds of 10 Gbps, 25 Gbps, and higher. So long as the customer’s bursty night-time requirements are satisfied by the UNI service rate speed, no truck-roll or other OPEX is required.

CPON Use Cases

There are two main use cases that are enabled by the shift from legacy direct detection modulation to single-carrier coherent modulation at 100 Gbps.

Use Case 1: Coherent 100Gbps CPON Entry

This use case supports the following important aspects of entry into the CPON market:

  • Lowest relative-cost ONU for 100 Gbps single-carrier coherent
  • Brownfield overlay of legacy PON deployments on existing ODNs
  • True Ethernet rates
Lowest Relative-cost ONU for 100Gbps

Within several years single carrier (SC) coherent point-to-point technology will have both an established ecosystem and multiple-vendor mass international deployment. These are pre-requisite conditions that lead to consideration for lower relative cost for optical components and re-use for CPON.  The growing Metro DCI, extended campus, and edge aggregation markets provide examples of an existing coherent technology such as the OpenZR+ MSA work [10] that target small form factor pluggable modules and speeds of 100Gbps, 200Gbps, 300Gbps, and 400Gbps as well as single carrier 400G ZR to 800G ZR from OIF [11]. While 400Gbps is likely not the immediate next speed increase for CPON, it is worth noting that the SC coherent point-to-point ecosystem has a 100Gbps entry point with steps up to 400Gbps at distances that are interesting to both brownfield overlay and new greenfield deployments of CPON.

Figure 3: Relative ONU cost estimates for moving from 10G to 100G CPON
Figure 3: Relative ONU cost estimates for moving from 10G to 100G CPON

As shown in Figure 3, 25GS-PON is a simple line-rate speed up of 10G-PON using the same optical module technology.  No silicon optical amplifier (SOA) or digital signal processor (DSP) is required and less expensive DML lasers are supported.  However, moving to 50G HSP PON is based on legacy intensity-modulated direct-direct (IMDD) modulation and is a moderate increase in both relative cost and power as both an SOA and DSP are needed as well as the shift to a higher cost EML laser and cooling.  There is a much smaller relative cost differential between 50G speeds and single-carrier coherent 100G CPON as coherent modulation has inherently more signal coding gain than IMDD requiring only a DSP and no SOA to meet brownfield ODN optical link budgets.   The predicted relative cost difference between 50G IMDD and 100G coherent PON is much less than the relative difference of 25G and 50G, which might be very attractive to operators wishing to go directly from 25G line rate with 21G service rate to 100G CPON with a true Ethernet service rate of 100G.

Brownfield Overlay of Legacy PON Deployments on Existing ODNs

Figure 2 shows the generational use of the O-, S-, and L-bands for legacy PON systems with previously noted exceptions.  History is showing that if a given generation of PON continues to meet an operator’s business plan for generating revenue, it does not go away.  Twenty plus years since its first publication in 2003, GPON is still being deployed today!  Two to three years from now, the market will see 25GS PON and 50G HSP PON systems also deployed in addition to legacy XG(S)-PON and GPON (i.e., triple-generational support) [19,20].   As a priority it is up to the operator’s business plan and needs for moving forward for new PON services independent on when to deprecate a prior generation PON service.  Our observation: O-band will be a bit challenged to add any new systems in the future.

Coherent modulation operates best in the C/L-bands.  This fortunate characteristic permit 100+ Gbps coherent channel pairs to be added as an overlay to existing brownfield PON ODNs as previously discussed. This overlay permits an operator to deploy CPON in ODNs where there are new service opportunities, without having to cannibalize existing revenue-generating services.

True Ethernet Rates – 100 Gbps PON Should Mean True 100 Gbps Ethernet Service Rate

Today’s market is competing on advertisable Ethernet rates for PON services.  The feedback received from some operators and customers is that a 10Gbps service should be a 10Gbps Ethernet rate. XG(S)-PON is ~8.7Gbps not 10Gbps, 25GS-PON is ~21Gbps not 25Gbps, etc. The rates do not match marketing competition expectations.  Starting at 100 Gbps PON, both the operators participating in the ITU-T PON standards as well as the operators participating in the CableLabs CPON development separately and mutually agree that the advertised PON rate should be the service rate and not the nominal line rate.  Increasing the nominal line rate to accommodate a post-overhead true Ethernet service rate requires some small changes to the ITU-T transport convergence and physical layers. The proposed functional layer architecture for moving forward is shown in Figure 4.

Figure 4: Example layer architecture for ITU-T PON supporting true Ethernet rates
Figure 4: Example layer architecture for ITU-T PON supporting true Ethernet rates

Figure 4 shows two important changes from legacy PON architectures: the nominal line rate will raise to ~117 Gbps to offset the overheads for delivery 100G Ethernet service rates and the FEC function should be moved to the coherent optical module adjacent to the DSP function.  This adjustment has both beneficial coherent optical module re-use relative cost as well as internal noise-reduction implementation aspects that are beyond the scope of this article.  The above example also preserves both the 125 µsec framing rate and the fundamental line rate which are essential for re-using the ITU-T Transmission Convergence (TC) layer with just a “simple” speed increase: the medium access control (MAC) layer is already defined and usable for 100 Gbps single-carrier coherent for 100Gbps and higher services rates multiples, e.g. 200 Gbps, 400 Gbps, etc.

Use Case 2: Low Cost Coherent TWDM-PON Overlay for 4 x 100G

With single-carrier coherent modulation the OLT and ONU can tune to any one or four adjacent CPON wavelengths at 100GHz spacing for no added cost. Essentially, the operator can deploy up to four adjacent 100G CPONs on the same ODN where the OLT and ONU skew (part) numbers are the same.  ITU-T NG-PON2 and HSP TWDM-PON standards have already incorporated the necessary protocols to tune an ONU from one channel pair to another.   Figure 5 illustrates four CPON channel-pairs as an overlay on the same ODN as legacy PON.

Figure 5: CPON deployment leveraging and NG-PON2 / HSP TWDM-PON architecture.
Figure 5: CPON deployment leveraging and NG-PON2 / HSP TWDM-PON architecture.

Combined, and as needed, four 100G CPONs allow for large, guaranteed throughput to address a typical number of ONU endpoints in existing ODN’s, e.g., 32, 64, 128, etc. Note that each channel-pair is a separate TDM/TDMA PON domain and managed by the OLT’s DBA for that pair.  Keeping costs low, each ONU has a single channel-pair transceiver.  Tuning between channels is a much slower non-real time control loop available to the operator, e.g., for load balancing. 

Other Aspects of Evolving to Single-carrier Coherent PON

Due to the nature of single carrier 100G CPON signaling, and the independence of coexisting CPONs, the re-use of the existing ITU-T TDM/TDMA PON architecture by accommodating the higher speed has the direct benefit of leveraging an operator’s existing back-office management system without requiring replacement, minimizing OPEX for deploying the new technology. Likewise, impact to the Physical Layer Operation, Administration, and Maintenance (PLOAM) and the ONU Management and Control Interface (OMCI) protocol specifications is predictable and accommodates the higher speed options and any additional ONU capabilities, etc.  The amount of added OPEX needed to deploy and manage CPON systems is minimized by single-carrier CPON.

Maintenance and troubleshooting of legacy PON systems extend directly to single-carrier CPON systems. Example: an operator’s procedures for rogue detection and mitigation apply directly to CPON deployments without modification [21].

An additional beneficial aspect of coherent receiver processing is the examination of phase and state of polarization for fiber sensing using coherent transceiver (DSP equalizer processing output).  For example, geophysical sensing of earthquakes and water waves is in laboratory demonstrations.  This is a promising area of development that may permit every ONU receiver to provide sensor data that would help correlate physical disturbances to the ODN [13]. 

Working Towards a CPON Standard

As with each prior generation of PON, moving to the next increase in speed always has faced engineering challenges that are met and overcome by the consensus of the standards development organization.  Bringing single-carrier coherent optical modulation to PON is a new application space for what is well-known and applied in the point-to-point transport market.   For PON, the known areas of work include:

  • Coherent upstream burst reception and channel characteristics [12]
  • FEC selection: balancing power (watts) with optical coding gain 
  • Link budgets, balancing transmitter optical launch power with receiver sensitivity and relative cost impact to both OLT and ONU for brownfield overlay. Defining other channel models, etc.
  • Band plan: DWDM grid, channel spacing, tuning ranges, etc.

Good news as a pair: 1) the ITU-T Transmission Convergence (TC/MAC) architecture can be used largely “as is” with the speed increase for true Ethernet rates and the architectural changes to move the FEC into the optical model, and 2) there is a growing ecosystem for mass-market 100G/400G OpenZR+ that will enable direct and cost-effective re-use into CPON.


The relative cost of the PON ONU makes or breaks the market.  Achieving the lowest relative-cost ONU for 100 Gbps single-channel coherent PON is a primary focus.  The demand for higher services rates comes from market and competition growth as well as the opportunity for extending PON into enterprise service rates and SLAs. This article has provided an overview on system and technology requirements for single-carrier CPON as well as a viable brownfield overlay model for deploying up to 4 x 100Gbps CPONs on the same ODN while maintaining any legacy PON systems that the operator has previously deployed.  Single carrier coherent PON is work-in-progress in both the ITU-T Study Group 15 Question 2 and CableLabs CPON efforts.


  1. ITU-T Study Group 15 - Networks, technologies and infrastructures for transport, access and home​,  Note: work-in-progress documents are available to SG15 members only.
  2. CableLabs, “CPON Architecture Specification”, CPON-SP-ARCH-I01-230503,, May 2023.
  3. ITU-T, G.984.1 Gigabit-capable Passive Optical Networks (G-PON): General characteristics ,!!PDF-E&type=items, March 2003.
  4. ITU-T, G.987.1 10-Gigabit-capable passive optical networks (XG-PON): General requirements,, March 2016.
  5. ITU-T. G.989.1 40-Gigabit-capable passive optical networks (NG-PON2): General requirements Amendment 1,, August 2018.
  6. ITU-T, G.9807.1 10-Gigabit-capable symmetric passive optical network (XGS-PON) Amendment 2,!!PDF-E&type=items, October 2020.
  7. ITU-T. G.9804.1 Higher speed passive optical networks – Requirements Amendment 1,!Amd1!PDF-E&type=items, August 2021.
  8. 25GS-PON MSA, Specification 25 Gigabit Symmetric Passive Optical Network Version 3.0,, November 2023.
  9. IEEE, IEEE Std 802.3™‐2022 Standard for Ethernet,, July 2022.
  10. OpenZR+ MSA, “OpenZR+ Specifications, version 3.0”, , September 2023.
  11. OIF, “Implementation Agreement 400ZR OIF-400ZR-02.0”,, November 2022.
  12. Junwen Zhang, Zhensheng Jia, Mu Xu, Haipeng Zhang, and Luis Alberto Campos, “Efficient preamble design and digital signal processing in upstream burst-mode detection of 100G TDM coherent-PON”, Journal of Optical Communications and Networking, A135, Vol 13 No 2,, February 2021.
  13. Christian Dorize, Sterenn Guerrier, Elie Awwad, Haïk Mardoyan, and Jérémie Renaudier , “From Coherent Systems Technology to Advanced Fiber Sensing for Smart Network Monitoring “,, Journal of Lightwave Technology Vol. 41, Issue 4, pp. 1054-1063, 2023.
  14. ITU-T, G.984.2 Gigabit-capable Passive Optical Networks (G-PON): Physical Media Dependent (PMD) layer specification,!!PDF-E&type=items, March 2023.
  15. ITU-T. G.984.5 Gigabit-capable Passive Optical Networks (G-PON): Enhancement band,!!PDF-E&type=items, September 2007. 
  16. Dell’Oro Group, “Growing Interest in 25GS-PON Drives Forecast Increases”,, August 2023.
  17. CCS Insight, “Growing Adoption of 50G-PON in Enterprise and Consumer Segments”,, October 2023.
  18. ITU-T, G-series Supplement 49 Rogue optical network unit (ONU) considerations,!!PDF-E, September 2020.

Statements and opinions given in a work published by the IEEE or the IEEE Communications Society are the expressions of the author(s). Responsibility for the content of published articles rests upon the authors(s), not IEEE nor the IEEE Communications Society.

Sign In to Comment