{"id":"068d77fa-b0d2-4355-b884-25ff2455061a","arxiv_id":"2506.08053","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"PD-SDCMA uses sparse dimension mapping over OFDM carriers to support more access groups and higher-order modulation in flexible PONs, according to simulation.","lead":"This paper proposes PD-SDCMA, a multiple access scheme that assigns each user group a sparse set of OFDM frequency dimensions to reduce interference in passive optical networks. Simulations in a 25 km fiber link suggest it can serve more user groups with lower bit error rate than two existing NOMA variants.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline BER/power-budget comparison is not made at equal spectral efficiency: PD-SDCMA uses more dimensions per user than PD-NOMA in the 3- and 5-group cases, so the apparent gains may be a bandwidth-for-interference trade rather than a capacity improvement.","rationale":"The reader's concern about subspace divisibility is real, but it concerns implementation details that could be patched by a defined mapping or code release. The spectral-efficiency confound directly affects whether the central claim 'supports more users' is true in any meaningful sense: the simulations appear to compare unequal amounts of information per OFDM symbol. This issue is independent of code availability and cannot be fixed by adding error bars. I therefore regard the equal-SE normalization as the most load-bearing issue, while agreeing that the divisibility ambiguity is a secondary reproducibility flaw. The verdict should remain conditional pending either an equal-SE comparison or a clear statement that the comparison is per-user-rate at fixed power rather than per-bit capacity.","tokens_in":7805,"tokens_out":14230,"duration_ms":182326,"concrete_test":"Compute the aggregate spectral efficiency (bits/s/Hz) at the HD-FEC operating point for Fig. 5(a)-(c) and re-run the comparison at equal spectral efficiency, e.g., match PD-NOMA's total bits per OFDM symbol by adding unused/zero dimensions to PD-SDCMA or by reducing PD-NOMA's modulation order. If PD-SDCMA no longer shows a BER or power-budget gain at equal SE, the central capacity claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central BER comparisons in Fig. 5 are not normalized for spectral efficiency. For the 2-group 16QAM case, Table II's S2D=[1 2;2 3] spreads the two groups over three dimensions, giving 2×4/3 = 2.67 bit/dim, whereas PD-NOMA gives 2×4/2 = 4 bit/dim. For 3-group QPSK, PD-SDCMA uses P=3 dimensions and carries 3×(N/P)=N QPSK symbols per OFDM symbol (2 bit/dim for N=512), while PD-NOMA, sharing the same two dimensions for all groups, carries 3×N/2=768 symbols (3 bit/dim). For 5 groups the gap widens: 2 bit/dim versus 5 bit/dim for PD-NOMA. The reported BER improvement and the 0.6 dB power-budget gain may therefore reflect a lower information load per dimension rather than a genuine interference-mitigation gain. The paper never reports aggregate throughput or bits/s/Hz at the HD-FEC operating point, so 'supporting more users' cannot be distinguished from 'spending more bandwidth per user.'","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes power-domain sparse dimensional constellation multiple access (PD-SDCMA) for flexible passive optical networks. The scheme uses a signal-space dimension selection (S2D) strategy: each access group's low-dimensional constellation is mapped to a sparse subset of dimensions in a high-dimensional OFDM signal space, and groups are combined in the power domain. The authors argue that this reduces multi-user interference relative to PD-NOMA and 3D-NOMA. Simulation results over a 25 km single-mode fiber are reported for two access groups with 16QAM and for three and five access groups with QPSK, and the paper claims that PD-SDCMA supports more access groups and significantly reduces BER compared with PD-NOMA and 3D-NOMA.","tokens_in":8075,"tokens_out":8364,"duration_ms":93208,"significance":"The idea of using sparse dimension allocation to reduce multi-user interference in power-domain NOMA for PONs is interesting and, if validated, could be practically relevant because it builds on the existing OFDM/IFFT architecture. The paper is not circular: the S2D matrices and power ratios are fixed design choices, and the comparison is made against independent PD-NOMA and 3D-NOMA baselines. However, the evidence is currently insufficient: the BER comparisons are not spectral-efficiency normalized, the subspace partition is not reconciled with the OFDM carrier count, and the simulation basis is very small. The qualitative concept may survive a careful re-analysis, but the quantitative claims of supporting more users and significantly reducing BER are not established as stated.","major_comments":[{"comment":"The central BER comparisons are not made at equal spectral efficiency. For the 3-group QPSK case, a P=3 partition would carry N QPSK symbols per OFDM symbol (2 bit per real dimension), whereas PD-NOMA, sharing two dimensions for all groups, carries 3N/2 QPSK symbols (3 bit per real dimension); for the 5-group case the gap is 2 vs 5 bit per real dimension. The reported BER improvement and the 0.6 dB power-budget gain may therefore reflect a lower information load per dimension rather than a genuine interference-mitigation gain. The paper does not report aggregate throughput or bits/s/Hz at the HD-FEC operating point, so the claim of supporting more users cannot be distinguished from spending more bandwidth per user. Please provide equal-spectral-efficiency comparisons or an explicit throughput accounting.","section":"Section III.B, Fig. 5, Table II"},{"comment":"The described partition of the N-dimensional signal space into N/P subspaces is incompatible with the stated OFDM parameters. With N0=256 orthogonal carriers, the I/Q signal space has 512 real dimensions, which is not divisible by P=3 or P=5 as used in Table II. Moreover, the mapping from S2D matrix rows to actual OFDM tones is never specified: Eq. (9) leaves the number of carriers per subspace blank, and Eq. (10) sums over N0 carriers without relating the subspace index i to the carrier index n. The simulation may therefore not implement the scheme as described. Please specify the exact tone-to-subspace mapping, for example by padding with unused dimensions or by choosing N0 appropriately, and confirm that the simulation uses that mapping.","section":"Section II.B, Eq. (7), Table I"},{"comment":"The headline conclusions rest on a single simulation with 1000 symbols and no error bars. At BER values near the HD-FEC threshold of 3.8e-3, 1000 symbols yields limited precision (only a few tens of errors), so the claimed significant BER reduction and the 0.6 dB sensitivity advantage are not statistically supported. The receiver is also assumed ideal. Please provide Monte Carlo repetitions or confidence intervals, and state explicitly how BER is averaged over access groups.","section":"Section III.A, Table I"},{"comment":"The abstract claims support for higher-order modulation formats, but the only higher-order result is the 2-group 16QAM case in Fig. 5(a), where no 3D-NOMA comparison is shown, and no analysis of minimum Euclidean distance or power-control range is given. The claim as stated is broader than the evidence. Please either add the missing comparisons or analysis, or narrow the contribution claims to what is actually demonstrated.","section":"Abstract and Section III.B"}],"minor_comments":[{"comment":"The integral in Eq. (1) is missing the differential dt; it should read ∫ f_n(t) f_m(t) dt = δ_{mn}.","section":"Section II.A, Eq. (1)"},{"comment":"The IFFT definition in Eq. (12) uses W^{-nk} and a 1/N factor, while Eq. (11) uses positive exponents and no normalization; please align the conventions.","section":"Section II.B, Eq. (12)"},{"comment":"The statement that the upper limit of the number of access groups for PD-NOMA PON is two is presented as a general limit but is only evidenced by one simulation scenario; please rephrase as an empirical observation for the tested parameters.","section":"Section III.B"},{"comment":"There are numerous typos and spacing errors (e.g., 'V ector', 'Recommendati-ons', 'correspond-ding', 'P rep resents'), and Figs. 2 and 3 are not described in enough detail; in particular, the xoy/yoz planes in Fig. 3 do not match the (1,2,3) subspace dimensions used in Table II.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is arguably more aligned with optical-access journals such as JLT or Optics Express than with cs.ET, though that alone is not a reason to reject. No code or data are provided, so reproducibility rests entirely on the written description. The spectral-efficiency issue is the main risk: if the authors cannot demonstrate an advantage at equal spectral efficiency, the contribution should be repositioned as a trade-off study rather than a capacity improvement. The dimension-partition inconsistency is serious but may be fixable with a precise mapping specification and appropriate padding of dimensions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a conditional paper: the S2D idea is plausible and the simulations are a reasonable first test, but the headline comparison is not made at equal spectral efficiency, and the mapping from the S2D matrix to actual OFDM tones is never spelled out.\n\nWhat is actually new: a cyclic sparse dimension allocation for NOMA in PONs, where each access group’s 2D constellation is placed on a different pair of dimensions inside a P-dimensional subspace. The paper compares this against PD-NOMA and 3D-NOMA in a 25 km SMF simulation, reports a 0.6 dB sensitivity gain at three groups, and shows five groups working where the others fail. That is a legitimate extension of the cited PSCD-NOMA and 3D-NOMA work, even if the underlying principle is close to SCMA. The vector-space motivation is clear, the S2D matrix formalism is easy to follow, and the scheme is OFDM-compatible. The power ratios and mapping patterns are design choices, not fitted to force conclusions.\n\nThe soft spots are real. The biggest one is spectral efficiency. For the 3-group QPSK case, PD-SDCMA sends one QPSK symbol per dimension, while PD-NOMA sends 1.5 symbols per dimension on the shared two dimensions. For 5 groups it is 1 versus 2.5 symbols per dimension. The paper never reports bits/s/Hz or aggregate throughput at the FEC threshold, so the 0.6 dB gain and the \"supports more users\" claim may just be bandwidth-for-interference trade. This is load-bearing, not a cosmetic issue. Second, the paper says the N-dimensional space is divided into N/P subspaces, but with N0=256 carriers there are 512 dimensions, which is not divisible by P=3 or P=5. Equation (7) doesn't fix the leftover dimensions, and the exact carrier-to-dimension assignment is absent. As written, I can't verify that the simulated signal matches the described scheme. Third, smaller but worth saying: 1000 symbols, no error bars, ideal SIC, no code. That is normal for a sim paper, but the lack of statistical confidence matters more once the equal-efficiency comparison is missing.\n\nWho is this for? Researchers working on NOMA-based flexible PONs, especially anyone comparing sparse dimension allocation against PD-NOMA or 3D-NOMA. It is not a breakthrough, but it is a usable building block.\n\nRecommendation: send it to peer review. A serious editor should not desk-reject it. The reviewers should push for an equal-spectral-efficiency comparison, a precise carrier-to-dimension mapping, and code/data release. If those are fixed, this could be a solid conference or journal paper.","headline":"A plausible sparse-dimension NOMA extension for PONs whose headline BER gains are not tested at equal spectral efficiency and whose dimension mapping is under-specified.","tokens_in":8596,"tokens_out":2373,"would_cite":false,"duration_ms":31757,"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":"The paper claims PD-SDCMA, a sparse-dimension constellation superposition scheme, lets flexible PONs serve more access groups with higher-order modulation and lower BER than PD-NOMA or 3D-NOMA, while staying compatible with OFDM.","keywords":["PD-SDCMA","S2D-strategy","Flexible PON","power-domain NOMA","OFDM-PON","passive optical network","sparse constellation superposition","non-orthogonal multiple access"],"falsifier":"Ask for the exact mapping of the S2D matrix rows to frequency tones in the 256-carrier simulation. Since 512 I/Q dimensions is not a multiple of 3 or 5, an equal partition into P-dimensional subspaces is impossible; if the simulation uses a non-equal partition or a modified mapping, then the claimed BER comparisons are not a test of the described PD-SDCMA scheme. Re-running the experiment with the explicit mapping would settle the claim.","tokens_in":7603,"feed_emoji":"📡","tokens_out":8212,"duration_ms":87798,"temperature":0.7,"pith_summary":"This paper introduces PD-SDCMA, a multiple-access scheme for flexible passive optical networks that superimposes groups' constellations not across all signal dimensions but only onto a carefully chosen pair of dimensions per subspace. The aim is to reduce the multi-user interference that limits PD-NOMA, so that the network can use higher-order modulation (16QAM) and more access groups (five) without extra power-control hardware. If it works, it would let operators increase PON capacity and flexibility by a DSP-only change inside an OFDM transmitter, and would beat both PD-NOMA and 3D-NOMA in user count and bit-error rate in a 25 km fiber link. The paper supports this with simulations and a derivation that the scheme is an IFFT/FFT-based process, hence OFDM-compatible.","feed_headline":"Dim-sparse coding lifts PON NOMA beyond two user groups","feed_subtitle":"In a 25 km fiber simulation, it carries 16QAM and five access groups where PD-NOMA and 3D-NOMA fail.","key_machinery":"The central object is the S2D-strategy matrix $S_{g\\times2}$, with one row per access group: for $g$ groups, row $i$ lists the two subspace dimensions that group $i$'s constellation occupies, cycling $(1,2),(2,3),\\dots,(g,1)$. In each $P$-dimensional subspace, a group's symbol fills only the two selected I/Q coordinates and sets all others to zero. Because each dimension is selected by exactly two rows, any given dimension carries the power-domain superposition of only two groups, so successive interference cancellation sees less cross-group interference. The transmitted signal is built by summing $g$ IFFT outputs with power weights $P_1+\\cdots+P_g=1$, and the receiver uses FFT plus SIC, keeping the scheme inside an OFDM modem.","core_discovery":"PD-SDCMA claims to lower the serial interference of power-domain NOMA by making the superposition sparse in the signal-space dimension. Under the S2D-strategy, an N-dimensional space is divided into P-dimensional subspaces; the access groups are arranged so that group i occupies dimensions (i, i+1) in each subspace (cyclically), and every other dimension is zero. Consequently, any single dimension carries the power superposition of only two groups, whereas PD-NOMA superposes all groups on the same two dimensions. The receiver's SIC then demaps the joint constellation onto the desired user's plane; the paper gives the example of two QPSK groups producing a 16-point cuboid joint constellation that collapses to an 8QAM-like constellation on the xoy plane, so the z-axis interference does not affect the first user. Simulation over 25 km single-mode fiber is claimed to show: 16QAM with two groups reaches the HD-FEC threshold under PD-SDCMA while PD-NOMA cannot; three groups reach BER 3.8e-3 at -10.8 dBm ROP versus -10.2 dBm for 3D-NOMA; and five QPSK groups still work, while 3D-NOMA fails.","pith_inferences":["The paper leaves implicit the exact mapping from S2D rows to physical OFDM tones; with 256 carriers giving 512 I/Q dimensions, an equal partition into P=3 or P=5 subspaces is arithmetically impossible, so a testable extension is to state the precise tone assignment used in simulation.","The sparse-dimension principle is not tied to the downlink; applying it to uplink Flexible PON, where user launch powers differ, could test whether the reduced per-dimension interference relaxes SIC ordering constraints.","The paper's claim of compatibility with any constellation suggests swapping the QPSK/16QAM building blocks for larger multidimensional constellations, which would require quantifying the trade-off between per-group dimensions and interference."],"forward_implications":["If the simulation holds, PD-SDCMA raises the access-group ceiling of Flexible PON from two to at least five under QPSK, directly increasing the number of users a single optical line terminal can serve.","16QAM becomes usable for two coupled access groups, so operators can trade modulation order against group count without needing higher-sensitivity receivers or tighter power control.","Because the transmit chain is an IFFT with zeroed unselected dimensions and the receiver uses FFT, existing OFDM-PON DSP can adopt the scheme as a software update.","Reducing per-dimension interference to two groups weakens error propagation in SIC, which is the mechanism that lets the scheme beat PD-NOMA's group-count limit.","The scheme's constellation-agnostic design means future multidimensional constellations can be plugged in for further spectral-efficiency gains."],"supporting_citations":[{"why":"Supplies the sparse-coding constellation predecessor that PD-SDCMA extends to dimension-sparse superposition.","marker":"[17]"},{"why":"Provides the 3D-NOMA baseline and its constellation minimum-distance results that the simulations must beat.","marker":"[18]"},{"why":"Establishes PD-NOMA as a candidate architecture for Flexible PON, the baseline scheme the paper improves.","marker":"[13]"},{"why":"Gives power-division NOMA in flexible optical access, the direct comparison target for user capacity and BER.","marker":"[14]"},{"why":"Motivates the sparse-superposition design as a cure for SIC error propagation in NOMA.","marker":"[19]"},{"why":"Shows OFDM for optical communications, the architecture PD-SDCMA claims compatibility with.","marker":"[10]"}],"fun_headline_variants":["Sparse dimensions unlock five-user NOMA in PONs","Power-domain NOMA gets sparse to cut interference","PD-SDCMA: sparse superposition for flexible PONs","Sparse NOMA doubles PON access groups over old schemes","Sparse power-domain coding lifts PON NOMA to 16QAM"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scheme depends on splitting all available signal dimensions into equal P-dimensional subspaces with no leftovers so that each dimension is shared by exactly two user groups; the paper does not say how this split is made for its 256-carrier (512-dimensional) simulations with P=3 or P=5.","fun_headline_variants_meta":{"raw":{"variants":["Sparse dimensions unlock five-user NOMA in PONs","Power-domain NOMA gets sparse to cut interference","PD-SDCMA: sparse superposition for flexible PONs","Sparse NOMA doubles PON access groups over old schemes","Sparse power-domain coding lifts PON NOMA to 16QAM"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000384,"raw_usage":{"total_tokens":2085,"prompt_tokens":1049,"completion_tokens":1036,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":665,"completion_tokens_details":{"reasoning_tokens":950}},"tokens_in":665,"tokens_out":1036,"duration_ms":10812,"temperature":1.0,"reasoning_tokens":950,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:35:17.937596+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Ask for the exact mapping of the S2D matrix rows to frequency tones in the 256-carrier simulation. Since 512 I/Q dimensions is not a multiple of 3 or 5, an equal partition into P-dimensional subspaces is impossible; if the simulation uses a non-equal partition or a modified mapping, then the claimed BER comparisons are not a test of the described PD-SDCMA scheme. Re-running the experiment with the explicit mapping would settle the claim.","supporting_citations":[{"cited_title":"Power Sparse Code Division Non-Orthogonal Multiple Access Scheme for Next-Generation Flexible Optical Access,","cited_arxiv_id":null,"evidence_quote":"Supplies the sparse-coding constellation predecessor that PD-SDCMA extends to dimension-sparse superposition."},{"cited_title":"Performance improvement of non-orthogonal multiple access with a 3D constellation and a 2D IFFT modulator,","cited_arxiv_id":null,"evidence_quote":"Provides the 3D-NOMA baseline and its constellation minimum-distance results that the simulations must beat."},{"cited_title":"Non-Orthogonal Multiple Access (NOMA) for Cellular Future Radio Access,","cited_arxiv_id":null,"evidence_quote":"Establishes PD-NOMA as a candidate architecture for Flexible PON, the baseline scheme the paper improves."},{"cited_title":"Power-Division Non-Orthogonal Multiple Access (NOMA) in Flexible Optical Access With Synchronized Downlink/Asynchronous Uplink,","cited_arxiv_id":null,"evidence_quote":"Gives power-division NOMA in flexible optical access, the direct comparison target for user capacity and BER."},{"cited_title":"Solution for error propagation in a NOMA-based VLC network: symmetric superposition coding,","cited_arxiv_id":null,"evidence_quote":"Motivates the sparse-superposition design as a cure for SIC error propagation in NOMA."},{"cited_title":"OFDM for Optical Communications,","cited_arxiv_id":null,"evidence_quote":"Shows OFDM for optical communications, the architecture PD-SDCMA claims compatibility with."}],"review_version":1}