{"id":"2dd133df-3f42-43be-baa7-8499111059e5","arxiv_id":"2501.13793","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"GF-OTFS applies global subband filtering in the frequency-Doppler domain to SC-FDMA-based OTFS, reducing Doppler leakage and out-of-band emissions while keeping BER competitive or better than existing OTFS variants.","lead":"The paper presents GF-OTFS, a modulation scheme that applies universal filtered multicarrier filtering to the frequency-Doppler bins of an SC-FDMA implementation of OTFS, aiming to suppress Doppler-induced interference. Simulations show comparable or better bit error rates than windowed OTFS while cutting out-of-band emissions enough to meet a 5G spectral mask.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"All-ones UFMC predistortion calibration leaves off-diagonal ICI and per-bin power imbalance unverified; these could break the claimed BER gains.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the all-ones calibration of the UFMC predistortion. This is indeed the most critical unverified step because the central claim of comparable or better BER relative to RW-OTFS rests on the UFMC modem being near-transparent after predistortion. If the calibration fails to suppress ICI or introduces a power imbalance, the BER results in Fig. 9 could be artifacts of the chosen normalization rather than properties of the modulation. Alternative concerns, such as the interleaving adjacency claim or the absence of a CP, are less load-bearing: the interleaving effect is empirically supported by the Doppler-spread impulse responses in Fig. 6, and the no-CP design is handled by the receiver's full-channel equalization in the MMSE and LSMR-IC detectors. The predistortion, by contrast, is justified only by a single heuristic calibration without quantitative analysis of the resulting effective matrix. The proposed concrete test would directly measure whether the all-ones predistortion achieves diagonal dominance and negligible EVM, thus settling whether the concern lands. The reader's conditional verdict is appropriate; the paper needs this verification before the BER gains can be trusted.","tokens_in":10732,"tokens_out":18033,"duration_ms":158888,"concrete_test":"For the parameters in Table I (Nsc=512, N_rb_sc=4, Chebyshev filter length 129), compute the effective modem matrix A = Ru Tn P. Measure diagonal dominance via ρ = max_k ( ∑_{j≠k} |A[k,j]| ) / |A[k,k]|, and simulate the modem-only cascade with random 16-QAM symbols over an ideal channel (H=I, no noise) to compute the error vector magnitude (EVM). If ρ > 0.1 or EVM > -30 dB, residual ICI or per-bin distortion is significant enough to dominate high-SNR BER, and the reported gains would not hold without additional equalization or power renormalization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The predistortion stage in Eqs. (9)-(10) is the only mechanism that makes the UFMC modem approximately transparent, and its validity is load-bearing for the central BER claim. It is calibrated with a single all-ones probe: ~s_f,0 = Ru T0 1_Nsc, then P = diag{c / ~s_f,0}. This only forces the diagonal of the effective modem matrix A = Ru Tn P to a constant; it does not cancel off-diagonal inter-carrier interference from subband filtering, because the all-ones vector measures row sums, not individual columns. With the chosen filter length L_f = Nsc/4 + 1 and subband size N_rb_sc = 4, transition-band leakage between adjacent subbands can be non-negligible. Moreover, P divides by the per-bin modem response; at subband edges where |~s_f,0[k]| is small, |P_k| is large, creating a per-bin transmit power imbalance that, under a fixed total power constraint, lowers the effective SNR of weak bins and can degrade high-SNR BER. The paper does not report the diagonal dominance of A, the EVM of the modem-only cascade (H=I, no noise), or the transmit power normalization used in the BER simulations. Since the claimed up-to-5 dB gain over RW-OTFS and the absence of an error floor depend on this near-transparent modem assumption, this unverified step is the weakest load-bearing point.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes GF-OTFS, a modulation scheme that replaces the OFDM modem inside the SC-FDMA-based implementation of OTFS with a UFMC modem. The authors derive the transmit and receive signal models, introduce a predistortion matrix calibrated by an all-ones probe, define an equivalent delay-Doppler channel, and compare BER, out-of-band emissions, and Doppler-spread impulse responses against OTFS, RW-OTFS, and DR-UFMC. The central claims are that the UFMC filtering of adjacent frequency-Doppler bins mitigates inter-Doppler interference and that GF-OTFS achieves comparable or better BER than the baselines while significantly improving spectral containment.","tokens_in":11013,"tokens_out":9618,"duration_ms":91770,"significance":"The paper is well organized, and the algebraic development of the SC-FDMA-based OTFS representation and its extension to UFMC is coherent. The proposed architecture is a reasonable way to combine OTFS's Doppler resilience with UFMC's spectral containment, and the simulations include relevant baselines as well as 3GPP spectral-mask compliance. However, the central BER claim rests on an unverified predistortion assumption and on under-specified simulation details, so the contribution is not yet fully supported. If the predistortion step is validated and the simulation methodology is clarified, the work would be a useful step toward low-OOB Doppler-resilient waveforms.","major_comments":[{"comment":"The predistortion matrix P is derived by passing a single all-ones vector through the UFMC modem, i.e., \\tilde{s}_{f,0} = R_u T_0 1_{Nsc} and P = diag{1_{Nsc} \\oslash \\tilde{s}_{f,0}} times a scalar. This procedure only equalizes the row sums of the equivalent modem matrix A = R_u T_n for a constant input; it does not constrain the off-diagonal entries of A for arbitrary QAM data. With the chosen filter length L_f^{GF} = Nsc/4+1 and subband size N_{rb}^{sc}=4, transition-band leakage between adjacent subbands can introduce non-negligible off-diagonal inter-carrier interference that the all-ones probe cannot measure. Moreover, at subband edges where |\\tilde{s}_{f,0}[k]| is small, |P_k| is large, which can create per-bin transmit power imbalance; under a fixed total power constraint this lowers the effective SNR of weak bins. The paper should prove that A P is a scaled identity or report diagonal-dominance metrics, and should include a modem-only evaluation (e.g., EVM or BER with and without predistortion). Without this, the no-error-floor and up-to-5 dB BER claims in Fig. 9 are not supported.","section":"Section III, Eqs. (9)-(10)"},{"comment":"The simulation methodology for the central BER result is insufficiently specified. The text does not state whether the predistortion is active, how the transmit power is normalized after applying P, whether perfect channel state information is assumed, how H_GF_DD is obtained by the LSMR-IC detector, or which SNR definition (e.g., E_b/N_0 versus E_s/N_0) is used. The paper also derives an MMSE receiver in Eq. (14) but reports BER with LSMR-IC, leaving the actual receiver chain unclear. These details are essential because the claimed gain over RW-OTFS depends on the modem being approximately transparent and on a consistent power-normalization rule across all compared techniques.","section":"Section IV, Fig. 9"},{"comment":"The claim that UFMC filtering of adjacent frequency-Doppler bins mitigates inter-Doppler interference requires a precise statement of how the interleaver orders the vector s_f = \\Psi(I_N \\otimes F_M)\\Omega d. The paper never specifies whether a subband groups consecutive Doppler indices for a fixed frequency, consecutive frequencies for a fixed Doppler, or some other ordering. Without this, the physical mechanism by which subband filtering reduces Doppler leakage, as visualized in Fig. 6, cannot be checked, and the distinction from DR-UFMC's delay-domain filtering is not made rigorous.","section":"Section III, before Eq. (7)"}],"minor_comments":[{"comment":"The abstract says the technique achieves 'comparable BER performance' while Section IV reports up to 5 dB improvement over RW-OTFS; align the wording to avoid inconsistency.","section":"Abstract and Conclusion"},{"comment":"The expression for P is ambiguous because the scalar factor (1/Nsc)\\sum_k |\\tilde{s}_{f,0}[k]| appears without parentheses; rewrite the formula so the division and multiplication are unambiguous.","section":"Eq. (10)"},{"comment":"The sentence 'and we consider a scenario with ideal pulse shaping along the delay dimension' is grammatically incomplete and should be moved to Section IV with a precise statement of what ideal delay-domain pulse shaping means for each baseline.","section":"Introduction"},{"comment":"The assertion that Zak OTFS and ODDM achieve the same BER as OTFS is made without a citation or supporting figure; either add a reference or qualify the statement.","section":"Section IV"},{"comment":"The 'Doppler spread reduction' mentioned in the abstract is not quantified; Fig. 6 is illustrative only, so provide a quantitative metric such as effective interference power or spread width if the claim is to be used as a headline result.","section":"Abstract and Fig. 6"},{"comment":"The text says only half of the subbands are activated, but does not explain how the per-subband sidelobes are computed or which window is used for RW-OTFS; clarify the procedure so the comparison is reproducible.","section":"Fig. 7"}],"recommendation":"major_revision","confidential_remarks":"The main gap is not the mathematics but the validation of the predistortion step, which is load-bearing for the BER claims. The authors should be encouraged to add a modem-only verification and to report full simulation details; this is feasible within a revision. The paper also leans heavily on the authors' own prior work [11] and on [9], so the revision should more clearly state the incremental contribution of the UFMC replacement relative to those works."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Honest take: GF-OTFS is a real new idea—UFMC filtering across adjacent frequency-Doppler bins instead of along delay—and the paper does a decent job of deriving the system model and showing via simulation that the OOB emissions meet the 5G mask and BER is competitive. The math is coherent and the comparison against DR-UFMC and RW-OTFS is fair.\n\nI agree with the reader and the stress-test note that the predistortion step is the load-bearing weak point. Calibrating P with an all-ones vector only forces the diagonal of the modem response; it does not cancel off-diagonal subband leakage. With L_f = Nsc/4 + 1 and N_rb_sc = 4, the transition bands between subbands are wide enough that this off-diagonal term is very likely non-negligible. The paper never reports the diagonal dominance of the effective modem matrix, the modem-only EVM, or the transmit power normalization. If the predistortion is imperfect, the transmitted signal carries residual distortion and the reported up-to-5 dB gain over RW-OTFS may shrink.\n\nThat said, the stress-test note may overstate the danger: even if P does not fully diagonalize, the MMSE detector in (14) operates on the true equivalent channel H_GF_DD, which includes the actual modem response. So detection can partly compensate for residual filtering distortion, provided the receiver knows the channel. The issue is more about per-bin power imbalance and whether the all-ones calibration introduces an SNR penalty on weak bins, not an immediate collapse of the whole scheme.\n\nOther soft spots are minor: a single simulation scenario with no statistical error bars, no shipped code or data, and the SC-FDMA representation leans on the authors' own prior work. Self-citation is not a flaw here; [11] is the natural foundation and the current result is evaluated by independent simulation against external baselines.\n\nWho this is for: people working on high-mobility waveforms, OTFS variants, and UFMC. The novel combination and clean derivation justify peer review; the predistortion analysis needs more than an all-ones probe. I would send it to review with a request that the authors report diagonality, EVM, and power normalization, and ideally add a second BER scenario. It deserves a serious referee.","headline":"GF-OTFS is a genuinely new combination—UFMC filtering across frequency-Doppler bins—with a coherent derivation and promising simulations, but the all-ones predistortion calibration leaves off-diagonal subband leakage unverified and the BER gains rest on that assumption.","tokens_in":11574,"tokens_out":2138,"would_cite":true,"duration_ms":21712,"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 new modulation scheme, GF-OTFS, replaces the OFDM modem in SC-FDMA-based OTFS with a UFMC modem that filters adjacent frequency-Doppler bins, mitigating inter-Doppler interference and improving spectral containment.","keywords":["OTFS","UFMC","SC-FDMA","inter-Doppler interference","spectral containment","delay-Doppler domain","fractional Doppler","5G NR spectral mask"],"falsifier":"Pass random 16-QAM symbols through the GF-OTFS transmitter and compare the received constellation after the UFMC demodulator with the prediction of Eq. (13) at high SNR; a larger error-vector magnitude than the all-ones probe predicts would falsify the predistortion assumption. Equivalently, measure whether the empirical BER matches the MMSE bound from Eq. (14) as SNR grows, since a residual filtering distortion would create an error floor.","tokens_in":10499,"feed_emoji":"📡","tokens_out":8178,"duration_ms":60649,"temperature":0.7,"pith_summary":"The paper proposes GF-OTFS, a modulation technique that replaces the OFDM modem in SC-FDMA-based OTFS with a universal filtered multicarrier (UFMC) modem. The central claim is that filtering adjacent frequency-Doppler bins as subbands suppresses inter-Doppler interference and sharpens spectral containment, something earlier UFMC-based OTFS variants did not achieve because they filtered along the delay dimension. In simulations over a TDL-C channel at 500 km/h with 16-QAM, GF-OTFS shows no error floor and gains up to about 5 dB in bit error rate over receiver-windowed OTFS at high SNR, while keeping out-of-band emissions below the 3GPP 5G spectral mask. The paper derives an equivalent delay-Doppler channel for the filtered system and demonstrates the BER performance with an MMSE detector.","feed_headline":"Global filtering cuts OTFS Doppler leakage and OOB emissions","feed_subtitle":"New GF-OTFS scheme matches BER while keeping out-of-band emissions under the 5G mask.","key_machinery":"The load-bearing object is the UFMC modem inserted where an OFDM modem normally sits in the SC-FDMA implementation of OTFS. The Cooley-Tukey factorization $F_{MN} = \\Psi (I_N \\otimes F_M) \\Omega (F_N \\otimes I_M)$ rewrites the OTFS transform as a twiddle-factor multiplication followed by an SC-FDMA modulator, which places the frequency-Doppler bins in adjacency and lets them be grouped into UFMC subbands. A predistortion matrix $P$, obtained by passing an all-ones vector through the UFMC modem and pointwise inverting the received response, is applied before filtering to compensate the modem's distortion. The equivalent channel $H^{\\mathrm{GF}}_{\\mathrm{DD}} = \\Gamma^H H^{\\mathrm{GF}}_{\\mathrm{FD}} \\Gamma$ then maps transmitted delay-Doppler symbols to received ones through the filtered frequency-Doppler channel.","core_discovery":"On the paper's own terms, the discovery is that the interleaving stage of SC-FDMA-based OTFS exposes the frequency-Doppler bins as contiguous samples, so a UFMC modem can act as a global pulse-shaping filter along the Doppler dimension. By grouping four consecutive frequency-Doppler bins into subbands and applying an FIR filter to each, the scheme reduces the Doppler sidelobes that cause inter-Doppler interference and steepens the spectral roll-off beyond what time-domain windowing achieves. The resulting equivalent delay-Doppler channel matrix allows standard MMSE detection, and in simulation the scheme outperforms DR-UFMC and receiver-windowed OTFS in BER while meeting the 5G spectral mask.","pith_inferences":["A natural extension, not tested in the paper, is to sweep the subband size $N^{\\mathrm{rb}}_{\\mathrm{sc}}$ beyond 4; if the mechanism is correct, wider subbands should suppress more Doppler leakage but may worsen the predistortion residual.","The predistortion method could be adapted to use a data-driven estimate from pilots instead of an all-ones probe, which would make the scheme robust to filter mismatch and hardware imperfections.","The same logic may transfer to other delay-Doppler multiplexing schemes, such as Zak-OTFS or ODDM, provided their transform also exposes contiguous frequency-Doppler bins to permit global filtering.","The equivalent-channel derivation suggests that any modifiable linear modem placed in the SC-FDMA path could be compensated with the same all-ones predistortion recipe, not only UFMC."],"forward_implications":["If GF-OTFS performs as claimed, high-mobility links can transmit QAM constellations without an error floor under fractional Doppler shifts.","Compliance with the 3GPP 5G spectral mask means GF-OTFS could be used in shared spectrum without additional out-of-band filtering.","The BER gain over receiver-windowed OTFS reaches roughly 5 dB at high SNR in the simulated setup, so high-rate coded links operating at high SNR would benefit most.","The shorter transition band relative to DR-UFMC reduces interference between adjacent subbands, which matters for multiuser spectrum sharing."],"supporting_citations":[{"why":"provides the SC-FDMA-based delay-Doppler representation that GF-OTFS replaces the OFDM modem in","marker":"[11]"},{"why":"supplies the UFMC modulator and demodulator structure with subband filtering used as the global filter","marker":"[12]"},{"why":"defines the receiver-windowed OTFS baseline that GF-OTFS is compared against in BER and Doppler spread","marker":"[10]"},{"why":"defines the Doppler-resilient UFMC baseline that filters along the delay dimension, the contrast for GF-OTFS","marker":"[8]"},{"why":"motivates the predistortion stage for filter-bank-style modems that GF-OTFS adopts","marker":"[16]"},{"why":"supplies the LSMR-IC symbol detector used in the BER simulations","marker":"[21]"}],"fun_headline_variants":["GF-OTFS: interleaved bins enable global Doppler filtering","UFMC global filter tames OTFS Doppler sidelobes","GF-OTFS cuts OOB emissions and BER vs state-of-the-art","SC-FDMA interleaving plus UFMC gives OTFS clean Doppler spectrum","GF-OTFS: global filter curbs Doppler interference and OOB"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the predistortion matrix computed from an all-ones probe cancels the UFMC modem's distortion for arbitrary QAM data, leaving the equivalent channel in Eq. (13) as a faithful delay-Doppler channel; if the filtering distortion depends on the data or the inversion amplifies subband-edge distortion, the reported BER gains would not hold.","fun_headline_variants_meta":{"raw":{"variants":["GF-OTFS: interleaved bins enable global Doppler filtering","UFMC global filter tames OTFS Doppler sidelobes","GF-OTFS cuts OOB emissions and BER vs state-of-the-art","SC-FDMA interleaving plus UFMC gives OTFS clean Doppler spectrum","GF-OTFS: global filter curbs Doppler interference and OOB"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001094,"raw_usage":{"total_tokens":4548,"prompt_tokens":905,"completion_tokens":3643,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":521,"completion_tokens_details":{"reasoning_tokens":3547}},"tokens_in":521,"tokens_out":3643,"duration_ms":21992,"temperature":1.0,"reasoning_tokens":3547,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:35:32.615959+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Pass random 16-QAM symbols through the GF-OTFS transmitter and compare the received constellation after the UFMC demodulator with the prediction of Eq. (13) at high SNR; a larger error-vector magnitude than the all-ones probe predicts would falsify the predistortion assumption. Equivalently, measure whether the empirical BER matches the MMSE bound from Eq. (14) as SNR grows, since a residual filtering distortion would create an error floor.","supporting_citations":[{"cited_title":"Universal-filtered multi-carrier technique for wireless systems beyond LTE,","cited_arxiv_id":null,"evidence_quote":"supplies the UFMC modulator and demodulator structure with subband filtering used as the global filter"},{"cited_title":"Low complex inter-Doppler inte rference mitigation for OTFS systems via global receiver windowing,","cited_arxiv_id":null,"evidence_quote":"defines the receiver-windowed OTFS baseline that GF-OTFS is compared against in BER and Doppler spread"},{"cited_title":"Doppler-res ilient uni- versal filtered multicarrier (DR-UFMC): A beyond-OTFS modu lation,","cited_arxiv_id":null,"evidence_quote":"defines the Doppler-resilient UFMC baseline that filters along the delay dimension, the contrast for GF-OTFS"},{"cited_title":"Low-comp lexity symbol detection and interference cancellation for OTFS sy stem,","cited_arxiv_id":null,"evidence_quote":"supplies the LSMR-IC symbol detector used in the BER simulations"}],"review_version":1}