{"id":"b6ef24f1-8b36-4510-875e-6cd22844dfb4","arxiv_id":"1908.00453","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"In a 350 Gbit/s DP-64QAM experiment over 1610 km of SSMF, enumerative sphere shaping with blocklength 200 achieves a 15% reach increase over constant composition distribution matching at the same blocklength.","lead":"This paper reports the first experimental demonstration of probabilistic enumerative sphere shaping (ESS) in an optical fiber link, showing a 15% reach gain over constant composition distribution matching at equal short blocklength. The result suggests short-blocklength shaping could improve next-generation optical transceivers.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reach-gain claim hinges on HD-FEC BER threshold extrapolation: the 15% figure is an assumed-code benchmark, not a measured end-to-end result.","rationale":"The reader's weakest_assumption is exactly the FEC assumption, and I agree it is the most load-bearing because the headline '15% reach increase' is a single number computed by intersecting BER curves with the assumed staircase-code threshold. That said, the conditional verdict already flags this limitation, and the qualitative result (ESS outperforms CCDM at equal blocklength) is robust across AIR, effective SNR, and BER ordering. My proposed test -- actually decoding or replacing the assumed FEC with a verified transfer function -- would settle the quantitative claim. The concern does not change the verdict, because the central demonstration (first experimental ESS, short-blocklength advantage) does not depend on the exact 15% figure; the paper itself includes multiple supporting metrics. I didn't find an internal inconsistency or circular step; the weakest point is the benchmark-dependent quantitative extrapolation, which the reader already identified.","tokens_in":4892,"tokens_out":1421,"duration_ms":13000,"concrete_test":"Implement a full decode of the assumed staircase code (rate 0.9373) on the measured post-LDPC bit sequences, or alternatively use a documented and independently verified transfer curve for a staircase code with identical rate and threshold, and recompute the reach at the BER threshold of 4.5e-3. If the recomputed reach gain for ESS-200 vs CCDM-200 differs from 15% by more than a few percent, the headline quantitative claim would need to be revised. As a secondary check, re-measure or re-evaluate Fig. 3c at one or two additional distances near 1400 and 1610 km with error bars to confirm the crossing point is stable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central experimental claim is the 15% reach increase (210 km) of ESS-200 over CCDM-200 at 350 Gbit/s. This number is derived in Fig. 3c by intersecting measured post-LDPC BER curves with the BER threshold of an outer HD-FEC staircase code (rate 0.9373, BER 4.5e-3) taken from [14]. The staircase code is only assumed, not implemented or decoded; the paper states the threshold is from [14, Fig. 8]. The reach comparison is therefore an extrapolation from measured BER-with-LDPC data to an assumed 'post-HD-FEC zero-error' system. If the assumed staircase code's threshold or error-floor behavior differs for these shaped, short-blocklength PAS sequences, or if a different state-of-the-art FEC were used, the absolute reach values change. The paper's own AIR-based reach estimates (Fig. 3d) give 205 km gain, which is consistent with 210 km but does not validate the HD-FEC assumption. The absence of error bars and the use of a single optimal launch power, without re-optimization per distance, further weaken the quantitative '15%' figure. However, the qualitative ordering (ESS-200 better than CCDM-200, comparable to CCDM-3600) is supported by multiple metrics (AIR, effective SNR) and is not put in doubt by this concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental comparison of probabilistic enumerative sphere shaping (ESS) with blocklength 200 against constant-composition distribution matching (CCDM) with blocklengths 200 and 3600, and against uniform signaling, in a 350 Gbit/s/channel dual-polarization 64-QAM PAS system over standard single-mode fiber. The central claim is that ESS-200 achieves a 15% reach increase (210 km) over CCDM-200 at the same blocklength, while performing comparably to or slightly better than CCDM-3600. The results are presented via measured BER, effective SNR, and achievable information rate (AIR) as functions of launch power and distance.","tokens_in":5120,"tokens_out":6000,"duration_ms":58828,"significance":"If the results hold, this is the first experimental validation of ESS in optical fiber communications, showing that a short blocklength shaping scheme can approach the performance of long-blocklength CCDM while reducing complexity. The paper's strength is the direct experimental comparison with 1000 received LDPC blocks per setting, multiple performance metrics (BER, AIR, effective SNR), and consistency with the authors' prior simulations. The work is relevant to the design of low-latency, high-throughput optical transceivers using probabilistic shaping.","major_comments":[{"comment":"The central quantitative claim of a 15% reach increase (210 km) is obtained by intersecting the measured post-LDPC BER curves with the error-free threshold of an assumed outer staircase code (rate 0.9373, BER 4.5e-3) taken from Ref. [14]. This outer code is not implemented or decoded in the experiment. The validity of applying this threshold to the present shaped signals is not justified, because the error distribution at the input of the outer code depends on the inner LDPC code, the shaping algorithm, and the blocklength. To support the abstract claim of a 'demonstrated' reach increase, the authors should either implement the outer FEC or provide a sensitivity analysis showing how the reach gain changes with the chosen threshold, and explicitly state that the absolute reach values are based on an assumed FEC.","section":"Section IV, Fig. 3c"},{"comment":"No error bars or confidence intervals are reported for any measured BER or AIR value, despite the statement that 1000 received LDPC blocks were captured per launch power setting and per distance. The claimed 210 km gain is derived by interpolating between measurement points (e.g., between 1500 and 1800 km), and without a measure of statistical spread it is not possible to assess whether the observed difference between ESS-200 and CCDM-200 is significant. Please provide confidence intervals (e.g., based on binomial statistics for BER) and describe the interpolation/extrapolation procedure used to obtain reach values.","section":"Section IV, Figs. 3c and 3d"},{"comment":"The launch power is optimized at 1500 km (20 spans) and then fixed to 9.5 dBm for all transmission distances in Figs. 3c and 3d. If the optimal launch power depends on distance due to accumulated nonlinear effects, the comparison at other distances may not reflect each scheme's best performance, which could bias the relative reach. The authors should justify the use of a single launch power or show that small variations around 9.5 dBm do not affect the conclusions.","section":"Section III, Fig. 3"}],"minor_comments":[{"comment":"The word 'demonstrated' overstates the result given that the reach gain is computed via an assumed outer FEC; consider using 'estimated' or 'predicted' to match the body text.","section":"Abstract"},{"comment":"The term 'probabilistic enumerative sphere shaping' may be misleading because ESS as described is a bounded-energy sequence selection that induces a nonuniform distribution; consider clarifying that ESS is not inherently probabilistic in its encoding.","section":"Section I"},{"comment":"The text refers to 'black arrows' in Fig. 3a, but the arrows are not visible in the figure; please add or describe the annotations so the reader can identify the indicated performance gaps.","section":"Section III, Fig. 3a"},{"comment":"The notation H(CCC) is used in Eq. (1) but CCC is defined only in the equation itself; please define it in the surrounding text for clarity.","section":"Section IV, Eq. (1)"},{"comment":"The horizontal dashed line labeled 'HD-FEC with rate 0.9373' should be explicitly identified in the caption as the assumed threshold from Ref. [14], not a measured result.","section":"Section IV, Fig. 3c"},{"comment":"The conclusion calls ESS a 'low-complexity' alternative, but no complexity analysis is provided in the paper; please either provide a qualitative or quantitative complexity comparison or soften the claim.","section":"Section V"}],"recommendation":"major_revision","confidential_remarks":"The paper is a concise experimental letter and the first demonstration of ESS in optical transmission. The central quantitative claim, however, rests on an assumed outer FEC and lacks statistical uncertainty quantification. In my view, the manuscript requires a major revision to address these load-bearing concerns before it meets the journal's standards. I have no concerns about citation patterns or scope; the paper fits the journal's experimental communications category."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First: the paper is worth a serious referee. It gives the first experimental comparison of enumerative sphere shaping (ESS) against CCDM in optical fiber, and the main qualitative result—ESS at blocklength 200 beats CCDM at the same blocklength and lands close to CCDM at 3600—is well supported by the data. That is a genuinely new datapoint, not just a rerun of simulations.\n\nWhat it does well: the setup is clean. Four systems (uniform, CCDM-200, CCDM-3600, ESS-200) are run in the same loop, at the same 350 Gbit/s net rate, with enough captured blocks that the BER and AIR curves are smooth. The authors show both measured BER after LDPC/PAS decoding and an independent AIR-based reach estimate. Those two metrics agree on the ordering and roughly on the size of the ESS-200 over CCDM-200 gain (210 km vs 205 km). The citation to the earlier simulation [12] is appropriate; this is an experimental confirmation of that predicted behavior.\n\nWhere I part ways with the abstract: the '15% reach increase' is not a directly measured end-to-end result. It comes from intersecting the measured post-LDPC BER curves with a BER threshold of 4.5e-3 for an outer staircase code [14] that the authors did not implement. If that code's threshold or error floor behaves differently on short-blocklength shaped sequences, the absolute reach values and the percentage gain shift. This is an assumption, not a flaw that kills the paper, and the AIR-based estimate partly mitigates it—but the abstract's phrasing ('demonstrated') oversells a benchmarked extrapolation. The paper should say 'estimated using a model staircase code.'\n\nTwo smaller soft spots. There are no error bars or confidence intervals anywhere, so it is hard to tell whether the 210 km vs 205 km agreement is real or partly luck. And launch power is optimized only at 1500 km and then fixed for all distances; if the optimal power drifts with distance or shaping scheme, the comparison could be slightly tilted. Neither of these undermines the qualitative ordering, since the gaps in Fig 3 are large and consistent across AIR, effective SNR, and BER.\n\nMy bottom line: the paper deserves peer review and likely publication after a revision that quantifies uncertainty and relabels the reach claim as an estimate. The qualitative result—short-block ESS can replace CCDM at a fraction of the blocklength—is the real contribution and it looks solid.","headline":"A credible first experimental ESS-vs-CCDM comparison in fiber, with a robust qualitative ordering; the headline 15% reach gain is real but softer than the abstract implies because it flows through an assumed HD-FEC threshold.","tokens_in":5702,"tokens_out":2090,"would_cite":true,"duration_ms":21247,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"First experimental demonstration of probabilistic enumerative sphere shaping in optical fiber: ESS-200 transmits 350 Gbit/s over 1610 km, a 15% reach increase over CCDM-200.","keywords":["probabilistic shaping","enumerative sphere shaping","constant composition distribution matching","probabilistic amplitude shaping","optical fiber communications","64-QAM","reach increase","short blocklength"],"falsifier":"Implement the outer staircase decoder from reference [14] on the captured received data, count post-FEC block errors for ESS-200 and CCDM-200 across distances, and see whether the distance at which the 4.5e-3 BER threshold is crossed still yields a 210 km gap; a gap much smaller than 15% would show that the reach claim is an artifact of the assumed threshold.","tokens_in":4699,"feed_emoji":"📡","tokens_out":4927,"duration_ms":45344,"temperature":0.7,"pith_summary":"This paper reports the first experimental demonstration of probabilistic enumerative sphere shaping (ESS) in an optical fiber transmission system. The authors transmit dual-polarization 64-QAM at 350 Gbit/s per channel and compare ESS with blocklength 200 against constant-composition distribution matching (CCDM) at the same blocklength and against long-blocklength CCDM. They find that ESS-200 reaches 1610 km, a 15% (210 km) reach increase over CCDM-200 at the assumed hard-decision FEC threshold, and performs slightly better than even CCDM with blocklength 3600. The result matters because short blocklengths reduce the implementation complexity of probabilistic shaping, and ESS achieves this with a lower rate loss than CCDM.","feed_headline":"Sphere shaping beats CCDM by 15% reach in first fiber trial","feed_subtitle":"At 350 Gbit/s and blocklength 200, ESS-200 reaches 1610 km versus 1400 km for CCDM-200.","key_machinery":"The key object is enumerative sphere shaping (ESS): a shaping algorithm that selects output amplitude sequences whose total energy lies below a maximum bound, rather than sequences with exactly the same empirical amplitude distribution. Geometrically, CCDM addresses sequences on the shell of an n-sphere, while ESS addresses all sequences in that shell plus all inner shells. This lets ESS encode more input bits per block, lowering rate loss, and the authors show its short-blocklength advantage in AIR holds across launch powers. ESS is combined with probabilistic amplitude shaping, DVB-S2 LDPC encoding, and a finite-length AIR formula that subtracts rate loss from the bit-metric decoding rate.","core_discovery":"The central claim is that enumerative sphere shaping, which maps input bits to bounded-energy amplitude sequences rather than to a fixed composition, retains most of the shaping gain at short blocklengths and adds nonlinear tolerance. In the experiment, at the optimal launch power of 9.5 dBm and after 1610 km of standard single-mode fiber, ESS-200 yields a post-shaping BER below the 4.5e-3 threshold of the assumed staircase outer code, while CCDM-200 fails at 1400 km. The corresponding reach gain is 15%, or 210 km; an AIR-based calculation gives a 205 km gain. The authors also report that ESS-200 slightly outperforms CCDM-3600, so the trade-off between low complexity and performance does not require sacrificing reach.","pith_inferences":["Beyond the paper: because the ESS gain comes from lower rate loss rather than a code-specific property, the same class of gain should appear for other QAM orders and rates, provided the blocklength is short enough that CCDM's rate loss is large; this can be tested by repeating the loop experiment with 16-QAM or 256-QAM.","Beyond the paper: the bounded-energy constraint of ESS could be tuned span by span, effectively co-optimizing shaping and peak-power limits; the authors do not explore this, but the mechanism they validate makes it a natural next step.","Beyond the paper: the AIR-based reach estimate (205 km gain) is close to the BER-threshold estimate (210 km), which suggests the claimed 15% advantage is robust to the exact outer FEC choice, although only one FEC threshold was used in the paper.","Beyond the paper: since ESS-200 already matches CCDM-3600, an even shorter blocklength could be tested to find the complexity-performance frontier; at blocklengths near 100 the rate-loss advantage of ESS over CCDM may grow."],"forward_implications":["At equal blocklength 200, ESS-200 gives a 15% reach increase over CCDM-200, translating to 210 km at 350 Gbit/s over SSMF.","ESS-200 slightly outperforms CCDM-3600 in post-shaping BER, so short-block ESS offers the reach of long-block CCDM at lower implementation complexity.","ESS-200 offers roughly the same AIR-based reach as CCDM-3600, meaning an ideal PAS with short blocks loses no information-theoretic reach.","The shaping gain of ESS relative to CCDM-200 appears independent of launch power, so it applies in both linear and nonlinear transmission regimes.","At the chosen rate, ESS enables 350 Gbit/s per channel with blocklength 200, a regime relevant for low-latency optical transceivers."],"supporting_citations":[{"why":"The numerical split-step Fourier simulations that predicted ESS-200 would outperform CCDM-200; this experiment validates that prediction.","marker":"[12]"},{"why":"Introduces probabilistic amplitude shaping with CCDM, the baseline architecture this experiment builds on.","marker":"[4]"},{"why":"Recent extension of ESS to short-packet wireless systems, providing the bounded-energy sphere-coding formulation used here.","marker":"[11]"},{"why":"Original 1993 introduction of enumerative sphere shaping, the algorithm under test.","marker":"[10]"},{"why":"Supplies the outer staircase FEC rate 0.9373 and BER threshold 4.5e-3 that define the reach comparison.","marker":"[14]"},{"why":"Defines the rate loss of CCDM at finite blocklengths, the quantity ESS reduces.","marker":"[9]"},{"why":"Provides the finite-length bit-metric decoding AIR formula used to compute reach from measurements.","marker":"[5]"},{"why":"Provides the effective SNR calculation used to explain nonlinear tolerance differences.","marker":"[13]"}],"fun_headline_variants":["First fiber trial: ESS beats CCDM by 15% reach","ESS-200 reaches 1610 km, CCDM-200 stops at 1400 km","Enumerative sphere shaping gains 15% reach over CCDM in fiber","Short-block ESS outperforms CCDM by 15% reach in first demo","Probabilistic ESS-200: 15% reach gain vs CCDM-200"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 15% reach comparison assumes an outer hard-decision staircase code with rate 0.9373 and BER threshold 4.5e-3 taken from the literature, and that threshold is not measured in this experiment; with a different outer code the reach numbers and the size of the gain would change.","fun_headline_variants_meta":{"raw":{"variants":["First fiber trial: ESS beats CCDM by 15% reach","ESS-200 reaches 1610 km, CCDM-200 stops at 1400 km","Enumerative sphere shaping gains 15% reach over CCDM in fiber","Short-block ESS outperforms CCDM by 15% reach in first demo","Probabilistic ESS-200: 15% reach gain vs CCDM-200"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000207,"raw_usage":{"total_tokens":1300,"prompt_tokens":747,"completion_tokens":553,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":363,"completion_tokens_details":{"reasoning_tokens":448}},"tokens_in":363,"tokens_out":553,"duration_ms":5062,"temperature":1.0,"reasoning_tokens":448,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:54:54.204008+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Implement the outer staircase decoder from reference [14] on the captured received data, count post-FEC block errors for ESS-200 and CCDM-200 across distances, and see whether the distance at which the 4.5e-3 BER threshold is crossed still yields a 210 km gap; a gap much smaller than 15% would show that the reach claim is an artifact of the assumed threshold.","supporting_citations":[{"cited_title":"Bandwidth efficient and rate-matched low-density parity-check coded modulation,","cited_arxiv_id":null,"evidence_quote":"Introduces probabilistic amplitude shaping with CCDM, the baseline architecture this experiment builds on."},{"cited_title":"A pragmatic approach to shaped coded modulation,","cited_arxiv_id":null,"evidence_quote":"Original 1993 introduction of enumerative sphere shaping, the algorithm under test."},{"cited_title":"Staircase Codes: FEC for 100 Gb/s OTN,","cited_arxiv_id":null,"evidence_quote":"Supplies the outer staircase FEC rate 0.9373 and BER threshold 4.5e-3 that define the reach comparison."},{"cited_title":"Constant composition distribution matching,","cited_arxiv_id":null,"evidence_quote":"Defines the rate loss of CCDM at finite blocklengths, the quantity ESS reduces."},{"cited_title":"Multiset-partition distribution matching,","cited_arxiv_id":null,"evidence_quote":"Provides the finite-length bit-metric decoding AIR formula used to compute reach from measurements."},{"cited_title":"Experimental comparison of probabilistic shaping methods for unrepeated fiber transmission,","cited_arxiv_id":null,"evidence_quote":"Provides the effective SNR calculation used to explain nonlinear tolerance differences."}],"review_version":1}