{"id":"ed92d2c8-15ed-47b5-9c34-7dcb10c8dc6f","arxiv_id":"2505.03589","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"AFBM merges DAFT precoding with filter-bank multicarrier filtering, yielding lower PAPR and out-of-band emissions than AFDM while retaining quasi-orthogonality in doubly-dispersive channels.","lead":"The paper introduces AFBM, a wireless waveform that combines chirp-based precoding with filter-bank spectral shaping to handle fast-moving channels. It claims lower peak-to-average power and sharper spectrum than the existing AFDM waveform, which could help future integrated sensing and communication systems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The -100 dB OOBE claim rests on PHYDYAS O=4, a configuration for which the paper's own Section II-B says complex orthogonality restoration is not guaranteed; without BER/SIR results, the abstract's simultaneous claims are unsupported.","rationale":"The reader's weakest assumption identifies the most load-bearing concern: the AFBM claim requires simultaneous DD robustness and low OOBE, but the only configuration that yields the advertised -100 dB OOBE (PHYDYAS O=4) is outside the O<=1.5 regime where the paper guarantees complex orthogonality. This is not a minor implementation detail; it directly determines whether the central novelty claim holds. Section I promises a BER comparison against AFDM, yet Section III reports only PAPR and OOBE, so the quasi-orthogonality claim is supported only by an effective-channel structure plot, not by any detection metric. The K=1 simplification in Section II-C compounds this issue because the filter-bank matrix G couples adjacent symbols for K>1, making the 'without loss of generality' claim questionable. These gaps are addressable rather than demonstrations that the waveform is false, so the existing CONDITIONAL verdict remains appropriate; the paper needs quantified SIR/BER evidence before the headline claims can be accepted.","tokens_in":7833,"tokens_out":5770,"duration_ms":59268,"concrete_test":"Run an end-to-end Monte Carlo BER simulation using the Section III parameters (L=128, P=192, N=256, K=8, three-path DD channel, PHYDYAS O=4, matched-filter receiver with perfect CSI) and compare AFBM versus AFDM at equal spectral efficiency. In parallel, compute the noiseless SIR from the off-diagonal energy of the combined matrix C_f^H Q_P^H G^H H^H H G Q_P C_f averaged over channel realizations. If the O=4 SIR is below about 15 dB, or if AFBM's required SNR for a target BER (e.g. 10^-2) is more than 1-2 dB worse than AFDM's, the quasi-orthogonality claim fails in the OOBE configuration; if SIR is high and BER tracks AFDM, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline promise is that AFBM simultaneously delivers AFDM-like DD robustness and -100 dB OOBE. The OOBE result is obtained with PHYDYAS O=4 (Fig. 4), but Section II-B states that the filter-compensation construction restores complex orthogonality only for overlap factors O <= 1.5; for larger O, ``interference components outside the main diagonal'' reduce the SIR. The paper never quantifies this SIR reduction and provides no BER results for any configuration, despite the introduction announcing a BER comparison. The effective-channel analysis in Section II-C is also restricted to K=1, which by construction omits the inter-symbol overlap that the block-Toeplitz filter matrix G introduces for K>1. Consequently, the two headline properties are never demonstrated in a single configuration, and the quasi-orthogonality claim rests on structure plots rather than detection performance.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes Affine Filter Bank Modulation (AFBM), a waveform that combines a DAFT (chirp) precoding stage, a frequency-domain zero-padded IDAFT, and a filter-bank transmit structure, with a compensation vector intended to restore complex orthogonality. The authors claim that AFBM simultaneously provides quasi-orthogonality in doubly-dispersive channels comparable to AFDM, PAPR about 3 dB lower than AFDM, and OOBE as low as -100 dB when a PHYDYAS prototype filter is used. The manuscript develops a matrix model of the transceiver, derives a compensation condition, presents an effective-channel analysis for a single symbol, and gives simulation results for PAPR and OOBE. It does not provide BER results, despite announcing them in the introduction, and it does not quantify the signal-to-interference ratio degradation for high-overlap filters.","tokens_in":7931,"tokens_out":3267,"duration_ms":34491,"significance":"If the simultaneous claims were fully demonstrated, AFBM would be a meaningful contribution to high-mobility and ISAC waveform design, since the cited prior art (AFDM, DAFT-s-AFDM, FBMC) does not jointly achieve DD robustness, low PAPR, and low OOBE. The paper contributes a clear matrix formulation of the proposed transceiver, an explicit compensation construction, and initial PAPR/OOBE simulation evidence that the waveform inherits the expected filter-bank spectral containment and low-PAPR properties. However, the central quasi-orthogonality claim is currently supported only by structural plots and a K=1 analysis, with no detection-performance or interference-quantification evidence, so the significance is conditional on substantial additional validation.","major_comments":[{"comment":"The paper does not deliver the promised BER comparison. The introduction states that simulation results will compare AFBM against AFDM 'in terms of bit error rate (BER), PAPR, and OOBE,' but Section III contains only PAPR and OOBE figures. Since the abstract's central claim is that AFBM 'maintains quasi-orthogonality similar to that of AFDM' in doubly-dispersive channels, this claim is currently unsupported by any detection-performance result. The authors should add BER versus SNR curves for AFBM and AFDM over the doubly-dispersive channel, or explicitly remove the quasi-orthogonality claim.","section":"Section III / Abstract"},{"comment":"The compensation construction is only approximate and is guaranteed only for O <= 1.5, as stated immediately after Eq. (13). The O=4 PHYDYAS configuration used to obtain the -100 dB OOBE result in Fig. 4 falls exactly in the regime where the text says interference components outside the main diagonal reduce the SIR. Since no SIR or BER quantification is provided for O=4, the paper never demonstrates the simultaneous realization of -100 dB OOBE and quasi-orthogonality in a single configuration. The authors should quantify the SIR as a function of O and provide BER results for the O=4 case, or restrict the OOBE claim to configurations for which the orthogonality restoration is established.","section":"Section II-B, Eqs. (11)-(13)"},{"comment":"The effective-channel analysis is restricted to K=1 and is said to be 'without loss of generality,' but the full transmit model in Eq. (8) uses a block-Toeplitz filter matrix G that introduces inter-symbol overlap for K>1. Setting K=1 removes all overlapping blocks, so the analysis cannot capture inter-symbol interference that the filter bank introduces in the actual K=8 simulation configuration. The quasi-orthogonality conclusion therefore rests on a structural plot rather than on a metric for the full block system. The authors should either extend the effective-channel analysis to K>1, provide a rigorous argument that K=1 is representative, or give a quantitative interference-energy metric for the block system.","section":"Section II-C, Eq. (19)"},{"comment":"The abstract claims PAPR levels '3 dB lower' than AFDM, but the simulation result in Fig. 3 reports 'an advantage of 2 dB with respect to regular AFDM.' These numbers should be reconciled. If the 3 dB figure refers to a different operating point or parameter set, that should be stated explicitly; otherwise the abstract overstates the simulation evidence.","section":"Abstract and Fig. 3"}],"minor_comments":[{"comment":"The block-Toeplitz matrix display in Eq. (7) is garbled and appears to have missing entries and misplaced zeros; it should be redrawn with a clear definition of the blocks G_p and their positions.","section":"Eq. (7)"},{"comment":"The introduction announces a BER comparison, but no BER results appear in Section III; either add the comparison or correct the description of the simulation section.","section":"Section I"},{"comment":"The 'approximately equals' in Eq. (11) should be replaced by a precise design criterion or an explicit equality under the stated conditions, since the subsequent compensation derivation relies on diagonal dominance of the bracketed matrix.","section":"Eq. (11)"},{"comment":"The caption contains the typo 'PHYDYAS4' and should read 'PHYDYAS' with a space before the overlap factor.","section":"Fig. 2 caption"},{"comment":"The notation for the filter matrix, e.g., G_p, \\tilde G, and the relationship between N, P, and L, is introduced quickly and is hard to follow; a short table of dimensions and symbols would improve readability.","section":"Notation throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a plausible architecture and some encouraging PAPR/OOBE trends, but the headline claim of simultaneous DD quasi-orthogonality and very low OOBE is not currently demonstrated. The missing BER results and the unquantified SIR degradation for O=4 are the key gaps; both are fixable with additional simulations and analysis, so I do not recommend rejection. The authors should also be encouraged to strengthen the K>1 analysis rather than relying on a 'without loss of generality' statement that is misleading for overlapping filter banks."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my read on arXiv:2505.03589. The paper proposes AFBM, which is basically pruned DFT-spread FBMC with the DFT-spreading stage replaced by a DAFT, with chirp parameters chosen as in AFDM. That is a real, named new waveform, and the compensation vector design in Section II-B is the piece that makes the combination nontrivial: for a given prototype filter, they compute a diagonal pre/post-equalizer that restores complex orthogonality up to O≤1.5. The effective-channel analysis in Section II-C is reasonable and shows the expected band-diagonal structure. The PAPR and OOBE simulations are consistent with the ingredients: low PAPR from the single-carrier-like spreading, low OOBE from the filter bank.\n\nThe soft spots are real but not fatal. First, the abstract claims -100 dB OOBE and quasi-orthogonality simultaneously, but the -100 dB curve is obtained with PHYDYAS O=4, which is exactly the regime where the paper says the compensation construction is not guaranteed to work (O≤1.5) and that SIR will drop. The SIR loss is never quantified and no BER curves are shown, even though the introduction says a BER comparison will appear. Second, the effective-channel analysis is done for K=1 and called \"without loss of generality\"; with overlapping filter-bank symbols, K>1 introduces inter-symbol filter taps, so this is a gap. Third, the abstract says PAPR 3 dB lower while the text and Figure 3 report about 2 dB. The comparison set is also narrow: only AFDM, not DAFT-s-AFDM, despite the abstract referencing it.\n\nNone of these shows the waveform cannot work. They are missing evidence, not demonstrated failures. The design is knowledgeable, the derivation of the compensation vector is explicit, and the limitations are partially acknowledged in Section II-B (the O≤1.5 condition is stated). The paper just does not close the loop between the best-OOBE configuration and detection performance.\n\nThis paper deserves a serious referee. The right outcome is major revision: add BER results for at least the O=4 and O=1.5 configurations, quantify SIR as a function of O, and fix the K=1 generality. After that, it could be a solid contribution to the 6G waveform literature.","headline":"AFBM is a sensible combination of DAFT precoding and DFT-spread FBMC, but the abstract's simultaneous claims are not supported because the best-OOBE configuration (PHYDYAS O=4) violates the paper's own O≤1.5 orthogonality-restoration condition and no BER curves are shown.","tokens_in":8612,"tokens_out":2262,"would_cite":false,"duration_ms":21363,"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 AFBM achieves AFDM-like quasi-orthogonality in doubly-dispersive channels while reporting PAPR 3 dB lower (2 dB in simulation) and OOBE down to -100 dB.","keywords":["affine filter bank modulation","AFBM","doubly-dispersive channels","peak-to-average power ratio","out-of-band emissions","DAFT precoding","filter bank multicarrier","integrated sensing and communications"],"falsifier":"Simulate AFBM with the PHYDYAS $O=4$ filter over the paper's three-path doubly-dispersive channel, compare BER against AFDM with identical chirp parameters, and inspect the off-diagonal energy of $C_f^H Q_P^H \\tilde{G}^T \\tilde{G} Q_P C_f$. A multi-dB SNR loss or an error floor in that configuration would disprove the simultaneous claim, because the -100 dB OOBE result is only achieved at $O=4$.","tokens_in":7559,"feed_emoji":"📡","tokens_out":8668,"duration_ms":74308,"temperature":0.7,"pith_summary":"The paper introduces Affine Filter Bank Modulation (AFBM), a waveform built by inserting a DAFT precoding stage into an FBMC filter bank, and argues that it combines properties no prior waveform in this line has delivered together: robustness to doubly-dispersive channels comparable to AFDM, peak-to-average power ratio several decibels lower (the abstract says 3 dB; the simulation figure shows about 2 dB), and out-of-band emissions as low as -100 dB when a PHYDYAS prototype filter with overlap factor 4 is used. A sympathetic reader would care because high-mobility and integrated-sensing links need exactly this combination: DD robustness for reliable transmission, low PAPR for efficient amplification, and low OOBE for spectral coexistence. The paper supports the claim with an effective-channel analysis showing that the delay-Doppler shifts of the channel map to deterministic diagonal shifts, as in AFDM, and with PAPR and OOBE simulations; it does not report BER curves.","feed_headline":"AFBM beats AFDM on PAPR and OOBE in high-mobility links","feed_subtitle":"AFBM fuses DAFT precoding with FBMC filtering to cut PAPR and OOBE while keeping near-AFDM orthogonality.","key_machinery":"The load-bearing object is the compensation precoder $C_f = W_L \\operatorname{diag}\\{\\tilde{b}\\}$, an $L$-point discrete affine Fourier transform multiplied by a diagonal filter-compensation vector whose nonzero entries ($\\tilde{b}_{\\tilde{l}} = 1/\\sqrt{\\tilde{c}_{\\tilde{l}}}$) are set only in the first and last $L/4$ positions, where the data actually live. This $C_f$ is what turns a conventional FBMC filter bank into a complex-orthogonality-preserving structure, because the DAFT spreads each data symbol over chirp subcarriers while the compensation cancels the interference introduced by the prototype filter. The second essential element is the truncated IDAFT plus frequency-domain zero padding (the $Q_P$ matrix) that keeps the chirp sampling below Nyquist and contains the spectrum. Together they make the effective channel $\\mathbf{H}_{\\mathrm{eff}}$ behave like AFDM's: each propagation path produces a deterministic diagonal shift determined by its delay and Doppler indices.","core_discovery":"On its own terms, the paper claims that AFBM is the first waveform to simultaneously offer AFDM-like quasi-orthogonality in doubly-dispersive channels, a low PAPR close to that of DAFT-spread AFDM, and FBMC-grade spectral containment. The mechanism is a precoding matrix $C_f = W_L \\operatorname{diag}\\{\\tilde{b}\\}$ that combines an $L$-point DAFT with a filter compensation vector, followed by a truncated IDAFT, frequency-domain zero padding, and a block-Toeplitz prototype filter; the compensation is designed so that $C_f^H Q_P^H \\tilde{G}^T \\tilde{G} Q_P C_f \\approx U$ holds, restoring complex orthogonality for overlap factors $O \\le 1.5$. The effective channel $\\mathbf{H}_{\\mathrm{eff}} = Q_P^H G^H \\left(\\sum_{r} h_r \\Phi_r Z^{f_r} \\Pi^{\\ell_r}\\right) G Q_P$ retains the deterministic band-diagonal shift structure that gives AFDM its DD robustness. Simulation comparisons against AFDM report a PAPR advantage (about 2 dB in the figure, 3 dB in the abstract) and OOBE down to about -100 dB with the PHYDYAS filter at $O=4$.","pith_inferences":["The paper leaves open whether the -100 dB OOBE configuration and the quasi-orthogonality claim can coexist: the unquantified SIR loss at $O=4$ could be small enough to hide in BER figures, or large enough to split AFBM into two regimes, one for spectral containment and one for high-rate transmission.","One testable implication beyond the paper: comparing BER at $O=1.5$ vs $O=4$ on the same channel would quantify the SIR penalty and show whether an iterative receiver, which the authors list as future work, is necessary.","Because the effective channel's diagonal shifts encode delay and Doppler directly, AFBM's structure suggests radar parameter estimation similar to AFDM should be possible; the paper defers sensing evaluation, so this is a projection.","A natural extension the authors do not pursue is optimizing the chirp parameters of the DAFT stage for both PAPR and channel orthogonality simultaneously, which they mention as future work on PAPR specifically."],"forward_implications":["If AFBM delivers on its claims, ISAC systems can use one waveform for both communication and sensing in high-mobility channels without choosing between spectral containment and PAPR efficiency.","The $O \\le 1.5$ compensation guarantee gives a concrete low-latency operating point with controlled self-interference, while larger overlap factors trade SIR for deeper OOBE suppression.","Because the effective channel preserves AFDM's deterministic path-dependent diagonal structure, existing AFDM channel estimation and equalization techniques can be adapted to AFBM rather than designed from scratch.","The low PAPR means high-power amplifiers need less back-off, so the good OOBE is less likely to be destroyed by nonlinear distortion in realistic transmitters."],"supporting_citations":[{"why":"Defines AFDM, the baseline waveform whose quasi-orthogonality and high PAPR AFBM is compared against.","marker":"[7]"},{"why":"Supplies the doubly-dispersive channel model with chirp-cyclic prefix and the effective-channel analysis approach AFBM adapts.","marker":"[6]"},{"why":"The pruned DFT-spread FBMC scheme whose low-PAPR property AFBM inherits in affine form.","marker":"[12]"},{"why":"Establishes the overlap-factor condition O<=1.5 required for complex-orthogonality restoration used by the compensation stage.","marker":"[14]"},{"why":"Cited as the source of the pre-established prototype filter coefficients used in the filter bank.","marker":"[15]"}],"fun_headline_variants":["AFBM trims PAPR by 3 dB, OOBE to -100 dB vs AFDM","New AFBM waveform improves both PAPR and OOBE for high-mobility links","AFBM combines DAFT and FBMC to cut PAPR and OOBE in DD channels","AFBM: lower PAPR, -100 dB OOBE, near-AFDM orthogonality","High-mobility waveform AFBM beats AFDM on PAPR and OOBE"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central claim stands on the unquantified assumption that the signal-to-interference loss caused by using a high-overlap PHYDYAS filter ($O=4$) is small enough that AFBM keeps near-AFDM bit-error-rate performance; the paper guarantees orthogonality only for $O \\le 1.5$ and gives no BER curves or SIR numbers.","fun_headline_variants_meta":{"raw":{"variants":["AFBM trims PAPR by 3 dB, OOBE to -100 dB vs AFDM","New AFBM waveform improves both PAPR and OOBE for high-mobility links","AFBM combines DAFT and FBMC to cut PAPR and OOBE in DD channels","AFBM: lower PAPR, -100 dB OOBE, near-AFDM orthogonality","High-mobility waveform AFBM beats AFDM on PAPR and OOBE"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000268,"raw_usage":{"total_tokens":1640,"prompt_tokens":989,"completion_tokens":651,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":605,"completion_tokens_details":{"reasoning_tokens":530}},"tokens_in":605,"tokens_out":651,"duration_ms":5956,"temperature":1.0,"reasoning_tokens":530,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:47:53.554734+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate AFBM with the PHYDYAS $O=4$ filter over the paper's three-path doubly-dispersive channel, compare BER against AFDM with identical chirp parameters, and inspect the off-diagonal energy of $C_f^H Q_P^H \\tilde{G}^T \\tilde{G} Q_P C_f$. A multi-dB SNR loss or an error floor in that configuration would disprove the simultaneous claim, because the -100 dB OOBE result is only achieved at $O=4$.","supporting_citations":[{"cited_title":"Affine frequency division multiplexing for next generation wireless communications,","cited_arxiv_id":null,"evidence_quote":"Defines AFDM, the baseline waveform whose quasi-orthogonality and high PAPR AFBM is compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the doubly-dispersive channel model with chirp-cyclic prefix and the effective-channel analysis approach AFBM adapts."},{"cited_title":"Pruned DFT-spread FBMC: Low PAPR, low la- tency, high spectral efficiency,","cited_arxiv_id":null,"evidence_quote":"The pruned DFT-spread FBMC scheme whose low-PAPR property AFBM inherits in affine form."},{"cited_title":"A generalized DFT precoded filter bank system,","cited_arxiv_id":null,"evidence_quote":"Establishes the overlap-factor condition O<=1.5 required for complex-orthogonality restoration used by the compensation stage."},{"cited_title":"A two-dimensional FFT precoded filter bank scheme,","cited_arxiv_id":null,"evidence_quote":"Cited as the source of the pre-established prototype filter coefficients used in the filter bank."}],"review_version":1}