{"id":"866c2f8f-b768-493f-bc2a-bf0ea9990e0e","arxiv_id":"2603.08422","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Nonlinear phase rotation plus short-block constellation shaping increases tolerable link loss by up to 6 dB in coherent optical satellite uplinks, with the HPOA reduced to a single nonlinear-power parameter.","lead":"This paper models how high-power fiber amplifiers distort laser signals sent from Earth to satellites, then proposes two low-complexity digital tricks—constellation shaping and nonlinear phase rotation—that recover up to 6 dB of link budget. Readers should care because optical satellite uplinks are power-starved, and this offers a cheap DSP-only path to more robust links.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Signal–ASE interaction inside the HPOA is asserted negligible without quantitative support; if it is not, the P_NL-only model and exact-inversion NLPC logic fail.","rationale":"The reader's conditional verdict is reasonable. I find no internal inconsistency in the analytical derivation: the PSD formula in the appendix checks out, and the simplified model is exact under the stated zero-dispersion/noiseless assumptions. The main soft spot is the lack of a quantitative regime check for signal–ASE interaction, which the reader grouped with dispersion but did not isolate. This does not overturn the paper; it reinforces the need for the conditional caveats already in the reader's verdict. The proposed test would settle whether the concern actually lands, so I keep the verdict unchanged rather than moving it.","tokens_in":16946,"tokens_out":16356,"duration_ms":167617,"concrete_test":"Use the §IV-A SSFM setup to repeat the maximum-acceptable-loss comparison with HPOA noise figure F_HPOA ∈ {4, 8, 12} dB and free-space loss L ∈ {40, 50} dB, keeping P_NL fixed, and compare full-model GMI against the simplified model (Eq. (4) plus receiver AWGN). If any point moves off the dispersionless P_NL line by more than ~0.5 dB, the single-parameter characterization and the split-NLPC design need an ASE-aware correction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on Eq. (4): propagation reduces to a memoryless SPM phase rotation, so the nonlinear channel is characterized by P_NL alone. This requires two conditions. The first, L << L_D, is well supported (§II-B2 gives L_D≈4.6 km vs L<100 m at 100 GBd). The second, negligible signal–ASE interaction inside the HPOA (§II-B3c), is only asserted with 'typically negligible.' The text compares ASE power to other noise sources, but that SNR comparison does not by itself bound the conversion of ASE intensity noise into nonlinear phase noise through Eq. (4). Moreover, Eq. (10) drops the HPOA ASE term only under L >> G_HPOA, which may not hold for compact or low-loss links, or near the lower end of acceptable-loss values. If the HPOA ASE–signal beat is non-negligible, it is signal-dependent phase noise that cannot be represented by the additive white Gaussian noise in Fig. 4, and the exact-inversion TX/RX phase rotations in (11)–(12) are no longer exact. The paper also does not report F_HPOA or the ASE profile used in the SSFM, so the reader cannot independently check the regime of validity. This is the least secure load-bearing step: dispersion is safely negligible, but the ASE neglect is unquantified and potentially regime-dependent.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies nonlinear propagation in the high-power optical amplifier (HPOA) stage of a coherent ground-to-satellite uplink. It argues that, because the fiber length is much shorter than the dispersion length and because signal–ASE interaction inside the HPOA is negligible, propagation can be modeled as a memoryless Kerr phase rotation, Eq. (4), fully characterized by a single parameter P_NL defined in Eq. (6). On this basis the authors propose two low-complexity DSP techniques: LUT-based short-block probabilistic amplitude shaping and TX/RX/split nonlinear phase compensation (NLPC). Simulations with a split-step Fourier solver are used to show that these techniques increase the maximum acceptable link loss by up to 6 dB, and that different HPOA configurations collapse onto the simplified dispersionless model when plotted against P_NL.","tokens_in":17237,"tokens_out":15310,"duration_ms":126518,"significance":"If the simplified model is valid, the paper makes a valuable contribution: it identifies a genuinely different operating regime from long-haul fiber, gives a clean single-parameter description of the HPOA nonlinearity, and demonstrates that very simple DSP can recover several dB of link budget. The analytical derivation of the SPM-modified autocorrelation function in the Appendix is a solid and useful result, and the numerical validation across several amplifier configurations is a good check. The paper also clearly explains why standard carrier-phase recovery is ineffective in this dispersionless regime. The main uncertainties are quantitative rather than conceptual: the neglect of signal–ASE interaction is asserted rather than demonstrated, and the simulation results lack error bars or Monte Carlo sample counts.","major_comments":[{"comment":"The reduction to the memoryless phase rotation (4) requires that signal–ASE interaction inside the HPOA be negligible. The only support is the statement that ASE is 'typically negligible' and the SNR approximation in Eq. (10), which assumes L >> G_HPOA. The paper does not report F_HPOA or the ASE profile used in the SSFM, so the regime of validity cannot be checked. If ASE co-propagating with the signal is not negligible, it produces signal-dependent phase noise that is not represented by the AWGN in Fig. 4, and the exact-inversion logic of Eqs. (11)–(12) breaks down. Please provide a quantitative estimate or, preferably, a numerical comparison of GMI with and without the distributed ASE term in Eq. (2). This is load-bearing for the central 'single-parameter P_NL' claim.","section":"Sec. II-B3c and Eq. (10)"},{"comment":"No number of Monte Carlo realizations or error bars is reported. Several quantitative conclusions rely on differences of about 0.5–1 dB: the short-block shaping advantage in Figs. 7–8, the superiority of κ≈0.6 over κ=1 in Fig. 9, and the baud-rate trade-off in Fig. 12. Without confidence intervals, these differences could be simulation noise. Please quantify the uncertainty (e.g., bootstrap intervals on GMI, or at least the number of symbols/realizations) and confirm that the claimed gains exceed it.","section":"Figs. 7–13 and Sec. IV-B"},{"comment":"The claim that performance is 'fully characterized by P_NL' is validated with only four configurations. These differ mostly in the effective γL product and do not independently vary the ASE noise figure or the shape of the longitudinal gain profile g(z). In the zero-dispersion noiseless limit the integral in Eq. (6) is indeed the only parameter, but the numerical validation should show that the conclusion is robust when F_HPOA and the g(z) profile are varied at fixed P_NL; otherwise the claim remains restricted to the particular amplifier design simulated.","section":"Fig. 13, Sec. IV-B"}],"minor_comments":[{"comment":"Equation (2) uses +jβ2/2, but standard SMF has a negative β2; the listed values β2=21.7 ps²/km and D=17 ps/nm/km are inconsistent in sign unless a different convention is intended. Please clarify.","section":"Sec. II-B2"},{"comment":"The SNR expression omits the factor 2 associated with per-polarization ASE PSD. Define the exact SNR convention used so the formula can be reproduced.","section":"Eq. (10)"},{"comment":"The legend 'ideal linear (MB)' is ambiguous: the curve is the linear-regime benchmark, not an ideal linear channel model. Rename to 'ideal MB (linear regime)' or similar.","section":"Fig. 7 and Sec. IV-B"},{"comment":"The launch power and link loss are swept in 1 dB steps. The reported gains and the optimal κ may be affected by this discretization; state the step size clearly or refine the sweep near the optimum.","section":"Sec. IV-B, Figs. 9–13"},{"comment":"The 'unlimited bandwidth' case is still limited by the oversampling factor n=8. The caption should note that this is a DSP-sampling limit, not a true continuous-time unlimited-bandwidth case.","section":"Fig. 9"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is likely acceptable once the ASE-interaction question and the simulation-uncertainty reporting are addressed. I see no fundamental error in the Appendix derivation, and the P_NL reduction is not circular because P_NL is defined from Eq. (6). The main risk is overclaiming 'fully characterized' based on a limited validation; the authors should either add the missing quantitative support or soften the claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First: this paper is worth refereeing. It is not a breakthrough, but it is a competent, clearly written engineering-science contribution with a genuinely new analytical result and a plausible, low-complexity solution to a real problem.\n\nWhat is actually new: the closed-form evolution of the autocorrelation/PSD for SPM in a dual-polarization Manakov system (Eq. 8 and the generalized M-mode Eq. 26), and the observation that very short block-length sphere shaping (N=4) is the sweet spot in a nearly dispersionless amplifier stage. The DSP recipes — LUT-based PAS and split nonlinear phase compensation — are standard building blocks, but the combination is tailored sensibly to the HPOA uplink and the complexity numbers are honest.\n\nThe analytical appendix is coherent; I checked the Gaussian integral and the determinant manipulations. The validation strategy in Fig. 13, where different HPOA configurations collapse onto the dispersionless model curve, is a good check and gives me confidence that the simplified model captures the essential physics in the tested regime.\n\nSoft spots, in approximate order of importance:\n\n1. The claim that signal–ASE interaction inside the HPOA is negligible (Sec. II-B3c) is asserted rather than demonstrated. Comparing ASE power to receiver noise bounds the direct SNR penalty, but does not by itself bound the conversion of ASE intensity noise into nonlinear phase noise through the Kerr term. The paper never reports the HPOA noise figure or the ASE profile used in the SSFM, so a reader cannot independently check where the approximation breaks. This is not fatal — the full-model simulations matching the simplified model is empirical evidence it holds in these cases — but it is a boundary condition on the central claim and should be quantified.\n\n2. No error bars or realization counts are given for the Monte Carlo GMI curves. The dB gains are headline numbers; they deserve a statistical guardrail.\n\n3. The \"first comprehensive investigation\" claim in the conclusion overstates things, given [16], [17], and the authors' own [19]. The novelty is real but incremental.\n\nNo code or data is released; for a simulation-only paper that is a limitation, though not disqualifying.\n\nBottom line: for an optical communications engineer working on satellite uplinks or high-power amplifiers, this is a useful paper. I would accept it for peer review with a request for moderate revision: quantify the ASE-interaction regime, add error bars, and tone down the novelty claim. The central argument holds up in the regime of interest.","headline":"A credible, practically useful study with a clean analytical kernel and a low-complexity DSP scheme; the single-parameter P_NL model is well validated inside the simulation, but the ASE-neglect assumption is under-quantified and the headline claims should be tempered.","tokens_in":17759,"tokens_out":3363,"would_cite":true,"duration_ms":30907,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A single parameter describes how high-power amplifiers distort satellite uplinks, and two low-complexity DSP tricks recover up to 6 dB of link loss.","keywords":["optical satellite communication","high-power optical amplifier","Kerr nonlinearity","self-phase modulation","nonlinear phase compensation","probabilistic constellation shaping","sphere shaping","look-up table"],"falsifier":"Measure the output power spectral density after the HPOA and compare with Eq. (8) at launch powers around and above P_NL (for the example, about 42.7 dBm, i.e., φ̄≈1). If the measured broadening deviates from the analytical prediction in a way that grows with the ASE noise level, or if using a longer fiber (L comparable to L_D) changes the GMI-versus-power curve, the model is falsified. Alternatively, repeat at a higher symbol rate so L/L_D is no longer small: the predicted linear scaling of maximum acceptable link loss with P_NL (Fig. 13) would fail.","tokens_in":16810,"feed_emoji":"🛰️","tokens_out":4676,"duration_ms":42144,"temperature":0.7,"pith_summary":"The paper aims to show that the dominant nonlinear impairment in a coherent optical satellite uplink is not a complicated distributed effect but a simple, memoryless phase rotation. Because the high-power optical amplifier (HPOA) stage is only a few tens of meters long, chromatic dispersion is negligible and the Kerr nonlinearity acts on each symbol independently; the whole amplifier can therefore be collapsed into one number, the characteristic nonlinear power P_NL. On top of that model, the paper proposes two nearly-free digital techniques — a 32-entry look-up-table shaping of short blocks and a nonlinear phase compensation split between the ground transmitter and the satellite receiver — and demonstrates by simulation that together they raise the maximum acceptable link loss by up to 6 dB. A sympathetic reader would care because this directly translates into higher launch power, longer link margins, or simpler system design for future optical satellite links.","feed_headline":"Low-complexity DSP adds 6 dB of loss budget for satellite uplinks","feed_subtitle":"A single parameter captures amplifier nonlinearity; shaping and split phase compensation push tolerable loss higher.","key_machinery":"The load-bearing object is the simplified channel model of Eq. (4): under negligible dispersion, HPOA propagation reduces to a memoryless phase rotation u(L,t) = u(0,t) exp(−j φ̄ |u(0,t)|^2), with φ̄ = P/P_NL. The single parameter P_NL (Eq. 6) captures all the fiber details (length, gain profile, Kerr coefficient) and is the only thing needed to predict performance. The paper's compensation machinery consists of (i) sphere shaping with block length N=4, implementable as a 32-entry look-up table, which reduces spectral broadening because it correlates the four quadratures within each 4D symbol; and (ii) NLPC, a phase rotation split between transmitter and receiver with optimal splitting ratio","core_discovery":"The paper establishes that in a coherent ground-to-satellite uplink, the nonlinear distortion introduced by the high-power optical amplifier (HPOA) stage is, to a good approximation, a memoryless self-phase modulation: each received symbol suffers a phase rotation proportional to its instantaneous power. Because the fiber section is short enough that chromatic dispersion is negligible, the entire amplifier stage is equivalent to a zero-dispersion noiseless fiber link and can be described by a single parameter, the characteristic nonlinear power P_NL = (∫ γ g(z) dz)^{-1}. On top of this model, the paper proposes two low-complexity DSP techniques — short-block (N=4) sphere shaping implemented","pith_inferences":["If the single-parameter model holds, the same low-complexity NLPC scheme should apply to any short high-power fiber stage, such as booster amplifiers in other free-space optical terminals or high-power laser delivery systems — a transfer the paper does not discuss.","The combination of LUT shaping and split NLPC could be turned into a rate-adaptive scheme: the LUT's rate granularity and the phase-compensation strength could be tuned on the fly in response to channel loss fluctuations, a practical robustness feature the paper only hints at.","The M-mode extension of the spectral-broadening formula (Eq. 26) suggests a testable path to few-mode or multi-mode HPOAs, where the same P_NL characterization might still apply — though the paper does not simulate that case.","Because the distortion is memoryless, advanced sequence-based equalizers that exploit inter-symbol correlations are unlikely to beat the simple phase rotation; this also means the nonlinearity cannot be averaged out by temporal filtering, so compensation must be symbol-level."],"forward_implications":["Uplink power budgets can be increased by up to 6 dB without changing the HPOA hardware, directly enabling higher throughput or more margin against atmospheric attenuation.","The single-parameter P_NL lets designers compare different HPOA implementations purely by a number, without full propagation simulation.","The optimal shaping block length is very short (N=4), making LUT-based shaping with negligible complexity and fine rate granularity the natural choice for this channel, unlike long-haul fiber systems.","Because the nonlinearity is memoryless, conventional carrier phase recovery cannot mitigate it; split NLPC with κ≈0.6 provides about 1 dB over TX-only NLPC, and the total NLPC gain is roughly 4–5 dB.","The simplified dispersionless model can replace the full physical model for performance estimation in the considered scenario, which is validated by the collapse of all simulated HPOA configurations onto a single P_NL curve."],"fun_headline_variants":["6 dB more link loss with simple DSP","Satellite uplinks gain 6 dB via low-cost fix","One parameter tames distortion, adds 6 dB margin","Low-complexity DSP boosts uplink budget by 6 dB","Shaping and phase rotation: 6 dB win for space links"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The model rests on two claims: that chromatic dispersion is negligible over the amplifier's length at the chosen symbol rate and bandwidth, and that the ASE generated inside the amplifier does not noticeably interact with the signal; if either fails, propagation is no longer a memoryless phase rotation and the single-parameter description breaks down.","fun_headline_variants_meta":{"raw":{"variants":["6 dB more link loss with simple DSP","Satellite uplinks gain 6 dB via low-cost fix","One parameter tames distortion, adds 6 dB margin","Low-complexity DSP boosts uplink budget by 6 dB","Shaping and phase rotation: 6 dB win for space links"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000158,"raw_usage":{"total_tokens":1046,"prompt_tokens":715,"completion_tokens":331,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":459,"completion_tokens_details":{"reasoning_tokens":248}},"tokens_in":459,"tokens_out":331,"duration_ms":3629,"temperature":1.0,"reasoning_tokens":248,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T18:31:59.804849+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the output power spectral density after the HPOA and compare with Eq. (8) at launch powers around and above P_NL (for the example, about 42.7 dBm, i.e., φ̄≈1). If the measured broadening deviates from the analytical prediction in a way that grows with the ASE noise level, or if using a longer fiber (L comparable to L_D) changes the GMI-versus-power curve, the model is falsified. Alternatively, repeat at a higher symbol rate so L/L_D is no longer small: the predicted linear scaling of maximum acceptable link loss with P_NL (Fig. 13) would fail.","supporting_citations":[],"review_version":2}