Coexistence Options and Performance Analysis of 100 Gbit/s Coherent PON in Brownfield DWDM Networks

Gabriele Di Rosa; Jim Zou; J\"org-Peter Elbers; Martin Kuipers; Ognjen Jovanovic

REVIEW 3 major objections 4 minor 11 references

Coexistence Options and Performance Analysis of 100 Gbit/s Coherent PON in Brownfield DWDM Networks

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A dual-channel DWDM layout lets 100 Gbit/s coherent PON tolerate 12 GHz laser drift with less than 1 dB penalty.

desk verdict Useful quantitative comparison of CPON-over-DWDM architectures on measured filter profiles; the 12 GHz laser-accuracy number is real within the model but omits filter centering error, so treat it as a component budget, not a deployment guarantee. read the letter →

arxiv 2412.15743 v1 pith:GOT7NYLH submitted 2024-12-20 eess.SP cs.NI

Gabriele Di Rosa , Martin Kuipers , Jim Zou , Ognjen Jovanovic , Jörg-Peter Elbers This is my paper · ORCID

classification eess.SPcs.NI

keywords coherentPONDWDMcoexistencebrownfieldnetworkslaserfrequencyaccuracyfilteringpenaltydual-channelarchitecturesingle-channel100Gbit/sQPSK

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper shows that a future 100 Gbit/s coherent PON can share an existing DWDM backhaul without forcing operators to upgrade the passive multiplexers or to install expensive, tightly locked lasers. Using measured passband profiles of three deployed 100 GHz multiplexer/demultiplexer filters, the authors simulate the upstream and downstream signals in two separate DWDM channels and find a filtering penalty below 1 dB over a 38 dB link budget, even when the laser frequency is off by up to 12 GHz. This matters because the relaxed laser accuracy removes the main transceiver cost driver that would otherwise make coherent ONUs too expensive. The paper contrasts this dual-channel architecture with a single-channel design using circulators, which needs flatter filters and tighter laser control to stay below 1 dB of penalty.

What carries the argument

The argument is carried by a simulation model in a commercial photonic design suite, anchored to three measured multiplexer/demultiplexer amplitude profiles on a 100 GHz grid: a Gaussian-shaped filter with about 60 GHz 3 dB bandwidth and two flat-top designs with about 80 GHz 3 dB bandwidth. The central comparison variable is the laser frequency accuracy, defined as the joint offset of the OLT and ONU lasers from the nominal channel center, and the performance metric is the reduction of the PON loss budget at the 2% pre-FEC bit-error ratio threshold. The dual-channel architecture separates upstream and downstream with inexpensive diplexers; the single-channel architecture uses circulators and either one or two lasers per module. The load-bearing step is sweeping the laser offset through the measured filter passbands and recording how much of the 38 dB budget is consumed by filtering and reflection.

What would settle it

A hardware testbed would settle the claim: connect a 100 Gbit/s QPSK transmitter and coherent receiver through a real 100 GHz Gaussian multiplexer/demultiplexer with about 60 GHz 3 dB bandwidth, 30 km of standard single-mode fiber, and 38 dB of attenuation, then measure the pre-FEC bit-error ratio while stepping the laser offset from 0 to 12 GHz. A penalty above 1 dB at 12 GHz, or a deployed filter whose passband differs materially from the three measured shapes, would mean the conclusion does not transfer to that plant.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central result is that the dual-channel architecture—placing upstream and downstream in two different DWDM channels separated by a full 100 GHz grid spacing—preserves the full 38 dB PON budget under the worst measured filter shape. For the 60 GHz Gaussian passband, the penalty remains below 1 dB even for a joint laser frequency offset of 12 GHz, while the single-channel alternatives on that same filter lose almost 2 dB of budget even with perfect laser locking. For the two flat-top 80 GHz filters, the dual-channel penalty stays below 0.5 dB across the studied range, and the single-channel options become usable only with smoother filter transitions and tighter accuracy limits (below roughly 3.5 to 7 GHz). A residual penalty of about 0.3 dB in the single-channel case comes from circulator reflections that saturate the receiver.

Load-bearing premise

The load-bearing premise is that the three measured multiplexer/demultiplexer filter profiles used in the simulation represent what a 100 Gbit/s coherent PON will actually encounter in existing DWDM plants, and that the simulation reproduces those filters and the rest of the physical layer faithfully.

Editorial extensions

If this is right

A dual-channel 100 Gbit/s coherent PON can be deployed over existing 100 GHz DWDM filters with less than 1 dB of the 38 dB PON budget lost to filtering, leaving the rest for fiber, splitter, and connector losses.
Coherent ONUs for this architecture do not need full-band tunable, tightly locked lasers; a frequency accuracy of up to 12 GHz is sufficient on the worst measured filter.
The single-channel architecture is not workable on narrow Gaussian filters, but on flat-top filters it can meet the 1 dB target if laser accuracy is held below about 3.5 to 7 GHz depending on the laser configuration.
Choosing between the architectures is a trade of spectrum versus cost: dual-channel uses two DWDM slots per bidirectional service but keeps transceivers simpler, while single-channel halves the slot count at the price of stricter laser control and either wider-band electronics or two integrated lasers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper does not quantify the cost saving, but relaxing laser accuracy from the full-band tunable DWDM class to a 12 GHz-tolerant class would most directly reduce the ONU bill of materials, which is the critical cost constraint for access networks.
Because the penalty is dominated by the measured filter shape, operators with flat-top filters can expect even better margins than the Gaussian worst case, possibly allowing lower-grade lasers or higher-order modulation in future upgrades.
A natural hardware extension is a field trial that places a 100 Gbit/s coherent ONU with a loosely calibrated laser behind a real 100 GHz Gaussian multiplexer; reproducing the <1 dB penalty at 12 GHz offset would confirm the simulation's central claim.
The same simulation method could be rerun for 200 Gbit/s by increasing baud rate or modulation order; the narrower effective passband would tighten the allowed laser accuracy, but the dual-channel architecture is likely to remain the safer choice.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper gives the CPON standardization community a concrete result: a dual-channel architecture keeps filtering penalty under 1 dB at 12 GHz laser frequency error even on the worst 60 GHz Gaussian filter, while the single-channel alternatives take about 2 dB hits even with perfect locking. That comparison, using measured field-deployed MUX/DEMUX profiles, is the genuinely useful new contribution. The architectures themselves are known from prior CPON and hybrid PON/WDM work, but nobody had put actual brownfield filter shapes into a simulator and pulled out laser-accuracy thresholds for the different options.

The simulation is clearly described: VPIphotonics model, 30 Gbaud QPSK, 38 dB budget, standard DSP chain, and three measured filter profiles. Parameters are stated, including insertion loss, circulator isolation, and the single-laser sideband setup. The analysis distinguishes the trade-offs between dual-channel and single-channel cleanly, and the qualitative ranking is robust: on the Gaussian filter, single-channel is dead on arrival regardless of laser accuracy. That is worth knowing.

The soft spots are real but not disqualifying. The measured filter profiles are provided by CableLabs but not published, so the central penalty curves cannot be independently replayed. There is also no hardware validation of the full setup; receiver sensitivity and FEC assumptions come from prior experiments and OpenZR+ specifications, which is reasonable but not an end-to-end demonstration. The more material issue, flagged by a stress test, is that the 12 GHz accuracy is computed relative to the filter passband center. The paper assumes the nominal laser frequency coincides with that center. In a brownfield network, a transceiver locks to the ITU grid, and the deployed filter has its own manufacturing offset and thermal drift. The combined detuning is therefore laser error plus filter error, so the <1 dB penalty at 12 GHz is not a guarantee for a real filter whose center is off by several GHz. This does not change the architectural conclusion—dual-channel wins regardless—but it does mean the absolute threshold should be quoted as a component budget, not a deployment specification.

There is also a mild overclaim on cost-effectiveness: the paper asserts cost savings from relaxed laser accuracy but gives no cost model. That is fine as a directional statement, but it is not quantified.

Overall this is a well-scoped engineering simulation, honest about its assumptions, and useful for the ongoing CPON standardization discussions. It deserves serious peer review, ideally with the filter data published as an artifact and a short paragraph addressing filter centering error. I would cite it for the architecture comparison, with that caveat.

Referee Report

3 major / 4 minor

Summary. The manuscript presents a simulation study of two architectures for carrying 100 Gbit/s coherent PON signals over existing brownfield DWDM MUX/DEMUX filters: a dual-channel architecture with diplexers and a single-channel architecture with circulators, the latter in one-laser and two-laser variants. Using three measured MUX/DEMUX filter profiles (one Gaussian with ~60 GHz -3 dB bandwidth and two flat-top with ~80 GHz bandwidth), the authors simulate downstream transmission over a 30 km DWDM trunk plus a 20 km PON reach and evaluate the loss-budget penalty as a function of laser frequency accuracy. The main conclusion is that the dual-channel architecture incurs less than 1 dB penalty up to 12 GHz laser frequency accuracy even for the worst-case Gaussian filter, enabling relaxed wavelength control in cost-constrained PON transceivers.

Significance. If the central claim holds, the paper provides a concrete, cost-relevant specification for future coherent PON transceivers in brownfield DWDM deployments. The use of measured filter profiles from the field, a realistic DSP chain (pilot-assisted carrier phase recovery, 6-bit ADC, matched filtering), and a clear comparison of architecture options are strengths. The paper is also useful in identifying the single-channel architecture's sensitivity to filter shape. However, the headline accuracy claim is defined relative to the measured filter passband center rather than to the ITU grid, and the paper does not account for the deployed filter's own center-frequency error and drift. This is a load-bearing gap that limits the strength of the main conclusion as stated.

major comments (3)

[Results and Conclusions] The '<1 dB penalty at f_acc < 12 GHz' claim is computed with the nominal laser frequency set to the center of the MUX/DEMUX optical passband (stated in Results: 'The nominal frequency coincides with the center of the MUX/DEMUX optical passband'). In a brownfield deployment, however, the transceiver is tuned to the ITU grid, not to the unknown center of a deployed filter, and the filter itself has manufacturing center-frequency tolerance and thermal drift. The effective detuning is therefore f_acc plus the filter-center error. The paper does not quantify this filter error or argue that the architecture calibrates the laser to the actual passband center. For the 60 GHz Gaussian profile, the attenuation at 12 GHz offset is already near 1 dB at the carrier; an additional 5-7 GHz filter offset would push the effective detuning to 17-19 GHz, where the penalty can reasonably be expected to exceed 1 dB. Please include a filter-center error in the budget and report worst-case penalties, or revise the claim to state that the result holds when the laser is referenced to the passband center rather than to the ITU grid.
[Simulation Model and Results] The simulation considers only downstream transmission (stated in Results: 'downstream transmission is considered') and assumes three adjacent signals with equal spectral characteristics. Since the paper's goal is coexistence in brownfield DWDM networks, the upstream direction (where the cost-critical ONU laser is the transmitter) and the presence of legacy DWDM services with different modulation formats in neighboring slots are both relevant. The authors should state whether the symmetric upstream case gives the same penalty and whether adjacent-channel crosstalk from non-CPON services was included in the 'worst-case' assessment, or should they explicitly list these as limitations.
[Simulation Model] The three measured MUX/DEMUX filter profiles are central inputs, but only a qualitative plot is shown in Fig. 1c; the numerical profiles are not provided. Because the quantitative conclusions (e.g., the 12 GHz threshold) are specific to these measured profiles, the manuscript should include the numerical filter data as supplementary material or a data repository to allow independent reproduction and checking of the reported penalties.

minor comments (4)

[Simulation Model] The equal 3.5 dB insertion loss for the Gaussian and flat-top filters is an idealization. Since the penalty is computed relative to a no-filtering baseline, this does not affect the relative penalty, but the absolute 38 dB link budget in a real deployment would depend on the actual insertion losses; please state this explicitly.
[Results] The description of Fig. 2 does not define the horizontal axis or the plotted quantity beyond 'loss budget penalty'; adding a clear axis label and a brief explanation of how the penalty is read from the curves would improve readability.
[Introduction] The claim that a relaxed laser frequency accuracy may be a major market enabler for CPON products is plausible, but the paper does not quantify the cost impact; a reference or a short argument linking accuracy to transceiver cost would strengthen this motivation.
[Simulation Model] The receiver sensitivity of -38 dBm at 2% BER is based on previous experiments and OpenZR+ FEC assumptions; the paper should state explicitly that this is an assumed operating point rather than a measurement of the proposed architecture.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the <1 dB penalty result is a simulation outcome computed from measured filter profiles and standard DSP, with no fitted parameter or self-citation chain doing the work.

full rationale

The paper's central claim is that a dual-channel architecture keeps the filtering penalty below 1 dB for laser frequency errors up to 12 GHz when propagating a 100G coherent PON signal over deployed DWDM filters. This claim is produced by a VPIphotonics simulation whose inputs are measured MUX/DEMUX profiles supplied by CableLabs (Fig. 1c), a receiver sensitivity of -38 dBm based on prior external experiments [9], FEC assumptions from OpenZR+ [10], and a standard DSP chain. No parameter is fitted to the output penalty, and the laser frequency accuracy facc is an independent sweep variable, not a quantity defined in terms of the simulated penalty. The only self-citation, [7], supports a motivational cost statement about tunable lasers and is not load-bearing for the simulation result. The skeptic's concern that the 12 GHz accuracy is referenced to the measured passband center rather than the ITU grid plus deployed filter center error is a modeling-validity limitation, not circularity: it questions whether the inputs represent field conditions, not whether the derivation reduces to its own assumptions. Therefore no circular step is present, and the honest finding is a non-circular score of 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper's central claim rests on a simulation with external inputs: measured filter profiles from CableLabs and an operating point from prior experiments. It introduces no new physical entities and fits no parameters to the target result. The main unverified premises are the representativeness of the filter profiles and the accuracy of the VPIphotonics model. These are domain assumptions rather than circular dependencies.

free parameters (4)

Receiver sensitivity at 2% BER = -38 dBm
Assumed from prior experimental work and OpenZR+ FEC; anchors the 38 dB link budget and all penalty comparisons.
Circulator isolation between Tx and Rx ports = -30 dB
Hand-picked assumption used to include reflection effects in single-channel architectures; affects the roughly 0.3 dB penalty observed.
Digital carrier offset for single-laser single-channel operation = 16.5 GHz
Chosen to prevent spectral overlap between upstream and downstream signals; affects single-channel single-laser performance.
Equal insertion loss for all three filter types = 3.5 dB
Set equal to isolate filtering effects; this ignores additional loss that flat-top designs may incur in practice, as the paper notes.

assumptions (4)

domain assumption The three measured MUX/DEMUX filter profiles from CableLabs (Gaussian 60 GHz, flat-top 80 GHz sharp and smooth) are representative of deployed brownfield DWDM networks.
All filter penalty results depend on these profiles, which are inputs from field deployment rather than from a statistical survey. Invoked in Simulation Model and Fig. 1(c).
domain assumption The VPIphotonics Design Suite simulation accurately models the physical layer, including filtering, fiber dispersion, noise, and DSP.
No experimental validation of the full setup is presented; the simulation tool is treated as a faithful model. Invoked throughout Simulation Model.
domain assumption The receiver operating point of -38 dBm sensitivity at 2% BER before FEC, taken from prior experiments [9] and OpenZR+ FEC [10], is applicable to the simulated transceivers.
The 38 dB link budget and the penalty thresholds are anchored to this assumption. Invoked in Simulation Model.
domain assumption The DSP chain (orthonormalization, CD compensation, matched filtering, MIMO equalization, pilot-assisted CPR) works ideally as configured, with no additional implementation penalties.
The simulation assumes 6-bit ADC at 2 samples per symbol and pilot symbols, but does not model all implementation impairments. Invoked in Simulation Model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Coexistence Options and Performance Analysis of 100 Gbit/s Coherent PON in Brownfield DWDM Networks." pith.science (2026). https://pith.science/paper/GOT7NYLH

@misc{pith2026241215743,
  author       = {Pith},
  title        = {Pith review of: Coexistence Options and Performance Analysis of 100 Gbit/s Coherent PON in Brownfield DWDM Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GOT7NYLH}},
  note         = {Machine review of arXiv:2412.15743}
}

read the original abstract

We study system architectures for the coexistence of future coherent PON and DWDM networks. Considering deployed optical filters, we observe filtering penalties < 1dB at a laser frequency accuracy < 12GHz when using a cost-effective architecture.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

11 extracted references · 11 canonical work pages

[1]

ITU-T Standardization Q2/15 – Optical system for fibre access networks, 2022-2024, https://www.itu.int/net4/ITU-T/lists/q- text.aspx?Group=15&Period=17&QNo=2&Lang=en

work page 2022
[2]

Demonstration of a 32× 512 split, 100 km reach, 2× 32× 10 Gb/s hybrid DWDM-TDMA PON using tunable external cavity lasers in the ONUs

P. Ossieur, et al. "Demonstration of a 32× 512 split, 100 km reach, 2× 32× 10 Gb/s hybrid DWDM-TDMA PON using tunable external cavity lasers in the ONUs." Jour- nal of Lightwave Technology, 2011

work page 2011
[3]

Coherent Passive Optical Networks 100 Gbps Single-Wavelength PON,

CableLabs, "Coherent Passive Optical Networks 100 Gbps Single-Wavelength PON," CableLabs Specifica- tions, May 3rd 2023, https://www.cablelabs.com/specifi- cations/CPON-SP-ARCH accessed on 24 April 2024

work page 2023
[4]

Revolutionizing Access Networks: How TFDM Coherent PON Combines the Best of TDM and WDM PONs

Z. Jia, et.al., "Revolutionizing Access Networks: How TFDM Coherent PON Combines the Best of TDM and WDM PONs", NCTA technical papers, 2023

work page 2023
[5]

50-Gigabit-capable passive optical networks (50G- PON): Physical media dependent (PMD) layer specifica- tion,

“50-Gigabit-capable passive optical networks (50G- PON): Physical media dependent (PMD) layer specifica- tion,” ITU-T Recommendation G.9804.3, 2021, https://www.itu.int/rec/T-REC-G.9804.3/en

work page 2021
[6]

Optical interfaces for coarse wavelength division multi- plexing applications,

“Optical interfaces for coarse wavelength division multi- plexing applications,” ITU Recommendation ITU-T G.695, 2018, https://www.itu.int/rec/T-REC-G.695/en

work page 2018
[7]

Multi-wavelength transponders for high-capacity optical networks: a physi- cal-layer-aware network planning study,

J. Müller, O. Jovanovic, T. Fehenberger, G. Di Rosa, J.P. Elbers, and C. Mas-Machuca, "Multi-wavelength transponders for high-capacity optical networks: a physi- cal-layer-aware network planning study," Journal of Opti- cal Communication and Networking 15, 2023

work page 2023
[8]

Physical Layer Aspects of NG-PON2 Standards—Part 2: System Design and Technology Feasibility,

Y. Luo, H. Roberts, K. Grobe, M. Valvo, D. Nesset, K. Asaka, H. Rohde, J. Smith, J. Wey, and F. Effenberger, "Physical Layer Aspects of NG-PON2 Standards—Part 2: System Design and Technology Feasibility," Journal of Optical Communication and Networking,2016

work page 2016

Show all 11 references

[9]

Scaling laws for unamplified coherent transmission in next-generation short-reach and access networks,

G. R. Martella, A. Nespola, S. Straullu, F. Forghieri, and R. Gaudino, “Scaling laws for unamplified coherent transmission in next-generation short-reach and access networks,” Journal of Lightwave Technology 39, 5805– 5814, 2021, DOI: 10.1109/JLT.2021.3092523

2021
[10]

OpenZR+ Multi-source agreement Technical Specifica- tion

"OpenZR+ Multi-source agreement Technical Specifica- tion", [online] Available: http://openzrplus.org/docu- ments/
[11]

Local and remote laser frequency con- trol in point-to-multipoint networks using digital subcarri- ers,

S. Ranzini et al., "Local and remote laser frequency con- trol in point-to-multipoint networks using digital subcarri- ers," 49th European Conference on Optical Communica- tions (ECOC 2023), Hybrid Conference, Glasgow, UK, 2023, pp. 807-810, doi: 10.1049/icp.2023.2344

2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

ITU-T Standardization Q2/15 – Optical system for fibre access networks, 2022-2024, https://www.itu.int/net4/ITU-T/lists/q- text.aspx?Group=15&Period=17&QNo=2&Lang=en

work page 2022

[2] [2]

Demonstration of a 32× 512 split, 100 km reach, 2× 32× 10 Gb/s hybrid DWDM-TDMA PON using tunable external cavity lasers in the ONUs

P. Ossieur, et al. "Demonstration of a 32× 512 split, 100 km reach, 2× 32× 10 Gb/s hybrid DWDM-TDMA PON using tunable external cavity lasers in the ONUs." Jour- nal of Lightwave Technology, 2011

work page 2011

[3] [3]

Coherent Passive Optical Networks 100 Gbps Single-Wavelength PON,

CableLabs, "Coherent Passive Optical Networks 100 Gbps Single-Wavelength PON," CableLabs Specifica- tions, May 3rd 2023, https://www.cablelabs.com/specifi- cations/CPON-SP-ARCH accessed on 24 April 2024

work page 2023

[4] [4]

Revolutionizing Access Networks: How TFDM Coherent PON Combines the Best of TDM and WDM PONs

Z. Jia, et.al., "Revolutionizing Access Networks: How TFDM Coherent PON Combines the Best of TDM and WDM PONs", NCTA technical papers, 2023

work page 2023

[5] [5]

50-Gigabit-capable passive optical networks (50G- PON): Physical media dependent (PMD) layer specifica- tion,

“50-Gigabit-capable passive optical networks (50G- PON): Physical media dependent (PMD) layer specifica- tion,” ITU-T Recommendation G.9804.3, 2021, https://www.itu.int/rec/T-REC-G.9804.3/en

work page 2021

[6] [6]

Optical interfaces for coarse wavelength division multi- plexing applications,

“Optical interfaces for coarse wavelength division multi- plexing applications,” ITU Recommendation ITU-T G.695, 2018, https://www.itu.int/rec/T-REC-G.695/en

work page 2018

[7] [7]

Multi-wavelength transponders for high-capacity optical networks: a physi- cal-layer-aware network planning study,

J. Müller, O. Jovanovic, T. Fehenberger, G. Di Rosa, J.P. Elbers, and C. Mas-Machuca, "Multi-wavelength transponders for high-capacity optical networks: a physi- cal-layer-aware network planning study," Journal of Opti- cal Communication and Networking 15, 2023

work page 2023

[8] [8]

Physical Layer Aspects of NG-PON2 Standards—Part 2: System Design and Technology Feasibility,

Y. Luo, H. Roberts, K. Grobe, M. Valvo, D. Nesset, K. Asaka, H. Rohde, J. Smith, J. Wey, and F. Effenberger, "Physical Layer Aspects of NG-PON2 Standards—Part 2: System Design and Technology Feasibility," Journal of Optical Communication and Networking,2016

work page 2016