Pith. sign in

REVIEW 3 major objections 6 minor 12 references

Record-Breaking 1935.6 bit/s/Hz Spectral Efficiency in 19-Ring-Core Fiber Transmission of GMI-Estimated 25.24 Pb/s Capacity Using Low-Complexity 4x4 MIMO

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A 19-core fiber carrying 266 OAM modes reaches 1935.6 bit/s/Hz and an estimated 25.24 Pb/s using only 4x4 MIMO.

desk verdict The headline record is a bidirectional aggregate, so the 'record-breaking' claim is inflated by 2x; the underlying fiber and MIMO demo is still worth a look. read the letter →

arxiv 2506.04910 v1 pith:WOXG2NSA submitted 2025-06-05 physics.optics

Hualin Li , Junyi Liu , Jie Liu , Shuqi Mo , Haolin Zhou , Yuming Huang , Yining Huang , Lei Shen
show 5 more authors
Shuo Xu Lei Zhang Jie Luo Zhaohui Li Siyuan Yu
This is my paper
classification physics.optics
keywords space-divisionmultiplexingorbitalangularmomentum19-ring-corefiber4x4MIMOspectralefficiencygeneralizedmutualinformationC+Lbandsspatialdensity
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 claims a record spectral efficiency of 1935.6 bit/s/Hz in the C+L bands and a GMI-estimated aggregate capacity of 25.24 Pb/s after 10 km of a 19-ring-core fiber that supports 266 orbital-angular-momentum mode channels. Its point is that dense spatial multiplexing does not have to be paid for with large MIMO matrices: because the fiber suppresses inter-core and inter-mode-group crosstalk, equalizing just four modes at a time with a 35-tap 4x4 MIMO is enough. A sympathetic reader would care because it suggests a path to petabit-scale single-fiber links whose DSP cost stays low. The capacity and spectral-efficiency figures are estimated from generalized mutual information with the WDM comb mostly filled by ASE dummy channels, so they are upper bounds awaiting a fully loaded, decoded demonstration.

What carries the argument

The central object is the 19-ring-core fiber (19-RCF), whose 19 separate ring cores each carry orbital angular momentum (OAM) modes with topological charges $|l| = 0, 1, 2, 3$, producing 7 OAM modes per core and 266 mode channels after polarization multiplexing. The mechanism carrying the argument is a crosstalk hierarchy: the ring-core geometry keeps inter-core and inter-mode-group crosstalk weak enough that only the four modes within one mode group need joint equalization, so a 4x4 MIMO with 35-tap time-domain equalization suffices. Around this sit three experimental supports: sliding test wavelengths at high OSNR, ASE-derived dummy channels that populate the rest of the WDM comb, and GMI soft-decision estimation that converts received constellations into capacity numbers.

What would settle it

Load all 1240 wavelength channels with real modulated signals (or a statistically representative subset) through the same 10-km 19-ring-core fiber, measure per-channel GMI, and sum the capacities; if the total falls materially below 25.24 Pb/s or the C+L spectral efficiency below 1935.6 bit/s/Hz, the ASE-filled dummy channels did not reproduce the full-load environment. Independently, launch a single strong mode in one ring core and detect in neighbouring cores and mode groups to test whether inter-core and inter-mode-group crosstalk stays below what the 4x4 MIMO can equalize.

Watch

Extended reading notes

Core claim

The paper reports a 10-km transmission experiment in which 266 mode channels, formed from 19 ring cores times 7 OAM modes times 2 polarizations, carry 1240 wavelength channels across the S, C, and L bands. The central claim is that the C+L bands achieve 1935.6 bit/s/Hz spectral efficiency and a spatial efficiency density of $3.94\times10^{-2}$ bit/s/Hz/µm², with a GMI-estimated aggregate capacity of 25.24 Pb/s, while using only 4x4 MIMO with 35-tap time-domain equalization. GMI-estimated means the capacity is computed from per-channel generalized mutual information before forward-error correction, not measured as decoded error-free throughput. The authors argue this is the first experimental SDM system to combine high spectral efficiency, high spatial efficiency density, and more than 100 mode channels in one fiber while keeping MIMO complexity low.

Load-bearing premise

The result stands on the assumption that ASE-based dummy channels filling 1237 of the 1240 wavelengths reproduce the crosstalk and nonlinearity that real modulated channels would create, so the per-channel GMI values and the summed 25.24 Pb/s are only as good as that substitution.

Editorial extensions

If this is right

  • If the crosstalk hierarchy holds, future SDM fibers can add spatial channels without scaling MIMO size, keeping the DSP cost per bit near that of a 4x4 system.
  • A fully populated WDM comb with decoded forward-error correction would be the next test; GMI estimates are pre-FEC upper bounds, so realized error-free capacity would be lower after coding overhead.
  • The S-band contribution of about 491.8 Tb/s is limited by available amplifiers, not by the fiber, so improving S-band amplification should raise the aggregate capacity.
  • The same low-complexity approach could transfer to other fibers with negligible inter-core crosstalk, making SE and SED comparisons against strongly coupled multi-mode systems more direct.

Reading between the lines

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

  • A natural test the paper does not run is to replace the ASE dummy channels on a subset of wavelengths with real modulated channels and check whether neighboring-channel GMI changes; if it does, the 25.24 Pb/s estimate would need revision.
  • Because the demonstration is only 10 km long, the finding that 4x4 MIMO suffices may not survive longer links where inter-mode-group coupling accumulates; extending the same fiber to 100 km or more would test that scaling.
  • The SED metric rewards a small fiber cross-section, and the 250 µm cladding here suggests a trade-off with standard-cladding compatibility that the paper leaves implicit.
  • The wavelength uniformity claim rests on evaluating performance at 4 nm intervals plus sliding test channels; denser sampling or a fully loaded comb would show whether uniformity holds across all 1240 channels.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper reports an experimental SDM transmission over a 10-km 19-ring-core fiber supporting 266 OAM modes. Using 1240 WDM carriers in the S, C, and L bands, with PS-16QAM in C+L and QPSK in S, the authors measure GMI on three sliding test wavelengths and derive an aggregate GMI-estimated capacity of 25.24 Pb/s and a spectral efficiency of 1935.6 bit/s/Hz in the C+L bands, achieved with 4x4 time-domain MIMO with 35 taps. The paper claims this is a record SE/SED and the first high-SE, high-SED demonstration in a >100-mode fiber with low-complexity MIMO.

Significance. The result is potentially significant for SDM if validated: it suggests that dense spatial multiplexing can be achieved without large MIMO matrices when inter-core and inter-mode crosstalk are suppressed. The authors disclose that the capacity is GMI-estimated rather than decoded, and the sliding-test-channel methodology is standard in high-capacity experiments. However, the significance is substantially tempered by the bidirectional aggregation of the headline metrics, which are not comparable to the unidirectional results in Fig. 1, and by the use of ASE dummy channels for the vast majority of the WDM load.

major comments (3)
  1. [Abstract / Conclusion / Fig. 1] The headline '1935.6 bit/s/Hz' and '25.24 Pb/s' are bidirectional aggregates, yet the title and abstract present them without the qualifier 'bidirectional'. The Conclusion states that 'bidirectional C+L bands transmission achieves a SE of 1935.6 bit/s/Hz.' With 329,840 channels (1240 wavelengths × 266 modes) at 12 GBaud, a one-way capacity of 25.24 Pb/s would require an average GMI of about 6.38 bit/symbol per channel, exceeding the 4 bit/symbol entropy limit of PS-16QAM; halving to 12.62 Pb/s yields ~3.19 bit/symbol, which is plausible. Thus the per-direction SE is ~967.8 bit/s/Hz, not 1935.6 bit/s/Hz. Since Fig. 1 compares against unidirectional SDM experiments from Refs. [4-11], the 'record-breaking' claim is inflated by a factor of two. The authors should either report per-direction values or explicitly restrict the claim to bidirectional aggregate capacity and benchmark against bidirectional systems.
  2. [Experimental setup, second paragraph] The dummy channels used to fill the remaining 1237 of 1240 WDM slots are generated from ASE and shaped by optical processors, rather than modulated data. The paper assumes these reproduce the crosstalk and nonlinear environment of a fully loaded WDM system, but no evidence is given that ASE loading induces the same inter-modal or inter-core penalties as real modulated neighbors. Since the claimed capacity is a summation of GMI over all 1240 wavelengths, an optimistic GMI from incomplete loading would propagate directly into the headline number. The authors should justify this assumption quantitatively (e.g., by comparing GMI with and without modulated neighboring channels, or by measuring the crosstalk statistics).
  3. [Results, Fig. 3] The GMI-based capacity is reported without any decoded bit-error-rate (pre-FEC or post-FEC) validation, error bars, or repeated measurements. For a 'record-breaking' claim, GMI estimates are standard in the literature, but the absence of any companion BER measurement for the measured test channels and no indication of the number of independent measurements make it difficult to assess the margin against the PS-16QAM entropy limit. At minimum, the authors should report the distribution of GMI across repeated acquisitions for the primary test channel and across the 266 modes.
minor comments (6)
  1. [Abstract] The sentence 'GMI-estimated capacity of 25.24 Pb/s are transmitted' is ungrammatical ('capacity' is singular) and the phrase 'are transmitted' overstates what is an estimate; consider 'is' and 'was achieved'.
  2. [Results, SE calculation] The paper does not state the exact wavelength range used for the C+L band SE calculation; specify the total optical bandwidth (e.g., 1530-1625 nm) so the reader can verify the SE.
  3. [Abstract / Conclusion] The aggregate capacity 25.24 Pb/s includes the S-band (491.8 Tb/s), while the headline SE (1935.6 bit/s/Hz) refers only to C+L bands; clarify the band definitions in the abstract to avoid ambiguity.
  4. [Fig. 3(b)] There is a typo in the vertical axis label: 'data rata' should be 'data rate'.
  5. [Experimental setup / Results] It is not explained how GMI is obtained for all 1240 wavelengths from only three sliding test channels; if the curves in Fig. 3(a) are measurements at the edges of bands, clarify the interpolation or measurement procedure for intermediate channels.
  6. [Experimental setup, inter-core crosstalk] The paper relies on prior work [12] for negligible inter-core crosstalk but does not provide a measured crosstalk matrix for this fiber; a brief quantitative statement would strengthen the claim that the 4x4 MIMO is sufficient.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline capacity and spectral efficiency are measured GMI sums, not outputs of a fitted model that assumes the result.

full rationale

The paper's derivation chain is measurement-based: per-channel GMI is obtained from received constellations after 4x4 MIMO equalization, then multiplied by the symbol rate and summed over wavelengths, modes, cores, and propagation directions. No equation in the paper is defined in terms of the target record, and no parameter is fitted to the reported 25.24 Pb/s or 1935.6 bit/s/Hz figures. The use of ASE-generated dummy channels to emulate a loaded WDM environment is an experimental assumption about crosstalk and nonlinearity, not a circular one, and the GMI values are still measured on real test channels. The reliance on prior self-citations for negligible inter-core crosstalk and Rayleigh backscattering ([3], [12]) invokes separate earlier experimental results; even if those results are from the same group, they are external evidence with stated measurements, not an unverified ansatz that smuggles in the current claim. The most legitimate concern raised by the reader is that the headline SE and capacity aggregate two counter-propagating bidirectional transmissions, while many comparison points in Fig. 1 are per-direction, so the 'record' framing may overstate comparability. That is a metric-selection and reporting issue, not a circular derivation, and it does not change the fact that the underlying numbers come from measured GMI times symbol rate rather than from any self-referential construction. The reported bidirectional SE of 1935.6 bit/s/Hz is transparently labeled in the conclusion, and the same data imply roughly half that value per direction; the arithmetic is consistent and is not forced by the input assumptions. Overall, no circular step meeting the quoted-reduction standard is present, and the central claim retains independent experimental content.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

This is an experimental paper; no free parameters are fitted to data. The central claims rest on assumptions about the fiber's low crosstalk (inter-core and inter-mode-group) and about the validity of GMI estimation and ASE-based dummy channels. No new physical entities are postulated.

assumptions (5)
  • domain assumption Inter-core crosstalk is negligible, so each core can be treated independently when summing capacities.
    Invoked in the Experimental setup section: 'Prior studies confirm that inter-core crosstalk is negligible, ensuring minimal impact on communication signals [12]'.
  • domain assumption ASE-based dummy channels accurately represent the interference of real WDM channels.
    All channels except the sliding test wavelengths are filled with ASE noise shaped by optical processors; the paper does not discuss whether this underestimates nonlinear crosstalk from real modulated signals.
  • standard math GMI calculated by soft decision is a valid upper bound on achievable capacity with ideal FEC.
    The capacity and SE are reported as GMI estimates, not measured throughput after forward error correction.
  • domain assumption Bidirectional transmission has negligible Rayleigh backscattering.
    Invoked to justify bidirectional operation, cited to reference [3] in the Experimental setup section.
  • domain assumption Mode groups within a core are weakly coupled, so 4x4 MIMO per mode group is sufficient.
    The receiver demultiplexes one mode group at a time and uses only 4x4 MIMO; low inter-mode-group coupling is assumed from the fiber design and prior references [3], [12].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Record-Breaking 1935.6 bit/s/Hz Spectral Efficiency in 19-Ring-Core Fiber Transmission of GMI-Estimated 25.24 Pb/s Capacity Using Low-Complexity 4x4 MIMO." pith.science (2026). https://pith.science/paper/WOXG2NSA

@misc{pith2026250604910,
  author       = {Pith},
  title        = {Pith review of: Record-Breaking 1935.6 bit/s/Hz Spectral Efficiency in 19-Ring-Core Fiber Transmission of GMI-Estimated 25.24 Pb/s Capacity Using Low-Complexity 4x4 MIMO},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WOXG2NSA}},
  note         = {Machine review of arXiv:2506.04910}
}
read the original abstract

We achieve a record spectral efficiency of 1935.6 bit/s/Hz in the C+L bands in a 10-km 19-ring-core fiber supporting 266 OAM modes. GMI-estimated capacity of 25.24 Pb/s are transmitted using low-complexity 4x4 MIMO.

Figures

Figures reproduced from arXiv: 2506.04910 by the authors.

Figure 1
Figure 1. illustrates the trade-off between the SE and SED and the MIMO processing complexity quantified by the required number of complex multiplications (RNCM) per bit [3]. The data are drawn from recently reported SDM/WDM trans￾mission experiments where single-fiber capaci￾ties have surpassed 1 Pb/s. The figure reveals a clear trend: MIMO processing complexity in￾creases with higher SED. This presents a funda￾mental challe… view at source ↗
Figure 2
Figure 2. Fig.2. Experimental setup. ECL, external cavity laser; AWG, arbitrary waveform generator; EA, electrical amplifier; [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. presents a summary of data rate and trans￾mission channel measurements [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

12 extracted references · 9 canonical work pages

  1. [1]

    Space- division multiplexing in optical fibres,

    Richardson, D. J., Fini, J. M., Nelson, L. E. "Space- division multiplexing in optical fibres," Nature photonics ,no.5,pp.354-362, 2013. DOI: 10.1038/NPHOTON.2013.94

    work page doi:10.1038/nphoton.2013.94 2013

  2. [2]

    and Lin, S., “1-Pbps orbital Fig3

    Liu, J., Zhang, J., Liu, J., Lin, Z., Li, Z., Lin, Z., Zhang, J., Huang, C., Mo, S., Shen, L. and Lin, S., “1-Pbps orbital Fig3. (a) In the wavelength channels at the edges of the S, C, and L bands, the GMI-estimated data rate of 266 modes after transmission through 10 km of 19-RCF. (b)GMI-estimated data rata per 4nm range in the S, C and L bands. (c) The...

    work page 2022

  3. [3]

    High spectral- efficiency, ultra-low MIMO SDM transmission over a field-deployed multi-core OAM fiber,

    Liu, J., Mo, S., Xu, Z., Huang, Y., Huang, Y., Li, Z., Guo, Y., Shen, L., Xu, S., Gao, R. and Du, C., “High spectral- efficiency, ultra-low MIMO SDM transmission over a field-deployed multi-core OAM fiber,” Photonics Research, 13(1), pp.18-30,2024.DOI: 10.1364/PRJ.533993

    work page doi:10.1364/prj.533993 2024

  4. [4]

    2.05 Peta-bit/s super-nyquist-WDM SDM transmission using 9.8-km 6-mode 19-core fiber in full C band,

    Soma, D., Igarashi, K., Wakayama, Y., Takeshima, K., Kawaguchi, Y., Yoshikane, N., Tsuritani, T., Morita, I. and Suzuki, M., “2.05 Peta-bit/s super-nyquist-WDM SDM transmission using 9.8-km 6-mode 19-core fiber in full C band,” In 2015 European Conference on Optical Communication (ECOC) ,2015, pp. 1-3. DOI:10.1109/ECOC.2015.7341686

    arXiv 2015

  5. [5]

    10.16-Peta- B/s dense SDM/WDM transmission over 6-mode 19- core fiber across the C+ L band

    Soma, D., Wakayama, Y., Beppu, S., Sumita, S., Tsuritani, T., Hayashi, T., Nagashima, T., Suzuki, M., Yoshida, M., Kasai, K. and Nakazawa, M., “10.16-Peta- B/s dense SDM/WDM transmission over 6-mode 19- core fiber across the C+ L band”. Journal of Lightwave Technology, 36(6), pp.1362-1368,2018. DOI: 10.1109/JLT.2018.2799380

    work page arXiv

  6. [6]

    1.2 Pb/s Throughput Transmission Using a 160 μm Cladding, 4-Core, 3-Mode Fiber,

    Luí s, R.S., Rademacher, G., Puttnam, B.J., Eriksson, T.A., Furukawa, H., Ross-Adams, A., Gross, S., Withford, M., Riesen, N., Sasaki, Y. and Saitoh, K., “1.2 Pb/s Throughput Transmission Using a 160 μm Cladding, 4-Core, 3-Mode Fiber,” Journal of Lightwave Technology, 37(8), pp.1798- 1804.DOI:10.1109/JLT.2019.2902601

    arXiv

  7. [7]

    10.66 peta- bit/s transmission over a 38-core-three-mode fiber,

    Rademacher, G., Puttnam, B.J., Luí s, R.S., Sakaguchi, J., Klaus, W., Eriksson, T.A., Awaji, Y., Hayashi, T., Nagashima, T., Nakanishi, T. and Taru, T., "10.66 peta- bit/s transmission over a 38-core-three-mode fiber," In optical fiber communication conference, pp. Th3H-1

    work page

  8. [8]

    Peta-bit-per- second optical communications system using a standard cladding diameter 15-mode fiber,

    Rademacher, G., Puttnam, B.J., Luí s, R.S., Eriksson, T.A., Fontaine, N.K., Mazur, M., Chen, H., Ryf, R., Neilson, D.T., Sillard, P. and Achten, F., “Peta-bit-per- second optical communications system using a standard cladding diameter 15-mode fiber,” Nature Communications, 12(1), p.4238,2021.DOI: 10.1038/s41467-021-24409-w

    work page doi:10.1038/s41467-021-24409-w 2021

Show all 12 references
  1. [9]

    and Sillard, P., 1.53 peta-bit/s C-band transmission in a 55-mode fiber

    Rademacher, G., Luí s, R.S., Puttnam, B.J., Fontaine, N.K., Mazur, M., Chen, H., Ryf, R., Neilson, D.T., Dahl, D., Carpenter, J. and Sillard, P., 1.53 peta-bit/s C-band transmission in a 55-mode fiber. In 2022 European Conference on Optical Communication (ECOC) ,2022, pp. 1-4

    2022

  2. [10]

    22.9 Pb/s data-rate by extreme space-wavelength multiplexing,

    Puttnam, B.J., Van Den Hout, M., Di Sciullo, G., Luis, R.S., Rademacher, G., Sakaguchi, J., Antonelli, C., Okonkwo, C. and Furukawa, H., “22.9 Pb/s data-rate by extreme space-wavelength multiplexing,” In IET Conference Proceedings CP839 ,2023, pp. 1678-1681. DOI: 10.1049/icp.2023.2665

    2023

  3. [11]

    Randomly coupled 19-core multi-core fiber with standard cladding diameter,

    Rademacher, G., van den Hout, M., Luí s, R.S., Puttnam, B.J., Di Sciullo, G., Hayashi, T., Inoue, A., Nagashima, T., Gross, S., Ross-Adams, A. and Withford, M.J., 2023, March. “Randomly coupled 19-core multi-core fiber with standard cladding diameter,” In Optical Fiber Communi...

    2023 doi

  4. [12]

    Mode division multiplexing based on ring core optical fibers,

    Liu, J., Zhu, G., Zhang, J., Wen, Y., Wu, X., Zhang, Y., Chen, Y., Cai, X., Li, Z., Hu, Z. and Zhu, J., “Mode division multiplexing based on ring core optical fibers,” IEEE Journal of Quantum Electronics, 54(5), pp.1- 18.2018. DOI:10.1109/JQE.2018.2864561

    2018

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.