{"id":"3dfb980f-62d9-4bc4-962f-ce09a4670515","arxiv_id":"2412.07074","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A CSF-inspired estimator recovers the full OFDM channel from scattered pilots by exploiting delay-Doppler stationarity, matching MMSE accuracy in simulations.","lead":"The paper proposes an OFDM channel estimation method that first estimates the delay-Doppler channel spreading function from scattered pilots, then reconstructs the time-frequency channel for all data symbols. If it holds, high-mobility links could approach ideal-estimator accuracy without needing channel statistics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2's 'exact' off-grid Doppler recovery is not supported by Algorithm 1: Eq. (19)(a) uses a finite-N approximation, so the exact-recovery claim is overstated even though the method works approximately in simulations.","rationale":"The reader's conditional verdict is appropriate. I partially agree with the reader's weakest-assumption analysis: the on-grid delay and one-path-per-delay-bin assumptions are real practical limitations, and the paper should disclose them. However, the more load-bearing problem is internal to the claimed exactness: Theorem 2 says 'exactly estimated' for off-grid Doppler, but Algorithm 1's fractional-Doppler estimator relies on the large-N approximation in Eq. (19)(a). For finite N this approximation is biased, so the exact statement is not established even under the paper's idealized channel model. This does not destroy the paper's practical contribution: the simulations demonstrate near-MMSE behavior in the tested regime, and Theorem 1 for on-grid Doppler is sound. The appropriate fix is to soften Theorem 2 to approximate recovery with a bounded error, or to provide a non-asymptotic error bound, and to state the on-grid delay and one-path-per-delay-bin assumptions as limitations. Since the reader already recommended conditional acceptance, the verdict does not need to change; the conditions should be sharpened to include the Eq. (19)(a) approximation error.","tokens_in":9652,"tokens_out":8531,"duration_ms":93033,"concrete_test":"Run a noiseless single-path experiment with N=64, M=128, dt=df=4. Set an on-grid delay, e.g., l_i=3, and a fractional Doppler k_i=2.3. Compute the pilot-derived periodic CSF from Eq. (10), run Algorithm 1 with P=1, and compare \\hat{kF} and the reconstructed \\hat{hDD} with the true values. Repeat for kF in {-0.4, -0.3, ..., 0.4}. If the error is nonzero at machine precision, Theorem 2's exact statement is disproved; the measured bias can be compared with the O(1/N) error of Eq. (19)(a).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the original CSF can be exactly recovered from one period of the pilot-derived periodic CSF is only established in the on-grid Doppler case (Theorem 1). Theorem 2 extends this to off-grid Doppler, but Algorithm 1 cannot deliver exactness. In Eq. (19), the fractional Doppler estimate is obtained from the peak-to-adjacent-bin ratio, and step (a) replaces sin(pi*d_t*(kF_i-(k'_0-k0))/N)/sin(pi*d_t*kF_i/N) by (kF_i-(k'_0-k0))/kF_i. This is an approximation for finite N, not an identity, so the estimated \\hat{kF}_i carries an O(1/N) bias even in the noiseless limit. Consequently, the reconstructed CSF and CTF in Eqs. (21) and (16) are not exactly equal to the true channel functions, and the 'exactly estimated' wording of Theorem 2 is false unless an explicit error bound or a limiting statement is supplied. A secondary, also load-bearing limitation is structural: Section II-A assumes on-grid delays, Algorithm 1 requires the number of paths \\hat P as an input, and Step 6 nulls an entire delay row, so two paths sharing the same delay value cannot both be recovered. The simulations impose exactly these favorable conditions, so they do not exercise the regime where Theorem 2 is claimed to hold generally.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a CSF-inspired CTF estimation method for OFDM systems in high-mobility scenarios. Pilots placed on a TF lattice are used to form a discrete CTF, whose 2D DFT/IDFT yields a periodic CSF. Theorem 1 states that under on-grid delay and Doppler conditions with compact support (Doppler within [-1/(2d_t T), 1/(2d_t T) - 1/(NT)] and delay within [0, 1/(d_f Δf) - 1]), one period of the pilot-derived periodic CSF exactly equals the original CSF. Theorem 2 claims exact recovery also for off-grid Doppler using Algorithm 1, which estimates fractional Doppler from the ratio of adjacent bins. The CTF for data symbols is then reconstructed by inverse transforms. Numerical results compare the proposed estimator with linear interpolation and MMSE, reporting near-MMSE performance at lower complexity and without requiring channel statistics.","tokens_in":9925,"tokens_out":7103,"duration_ms":70950,"significance":"The on-grid derivation from Eqs. (10) to (16) is clean and correctly identifies the scaling factors (N/d_t^2, M/d_f^2) and the support limits, providing a useful design guideline for pilot spacing in doubly selective channels. The method does not require channel statistical information and offers a complexity advantage over MMSE if the stated complexity holds. However, the claimed exactness for off-grid Doppler is not supported by the current analysis: Eq. (19) relies on a finite-N small-angle approximation, and Algorithm 1 depends on on-grid delays, a one-path-per-delay-bin assumption, and an externally supplied number of paths. These limitations are load-bearing for Theorem 2 and are not exercised by the simulations, which use on-grid delays and likely known P. With these caveats addressed, the approximate estimator is a plausible and practically interesting contribution.","major_comments":[{"comment":"The 'exactly estimated' claim in Theorem 2 is not supported by the derivation. Step (a) in Eq. (19) replaces the ratio of sine functions by (k_i^F - k'_0 + k_0)/k_i^F, which is a small-angle approximation valid only for sufficiently large N, not an identity. Consequently, the estimated fractional Doppler \\hat{k}_i^F in Eq. (20) carries an O(1/N) bias even in the noiseless limit, and the reconstructed CSF in Eq. (21) and CTF in Eq. (16) are not exactly equal to the true channel functions. The theorem should be restated as an approximate statement with an explicit error bound, or at minimum with a limit N → ∞ qualifier.","section":"Section III-B, Eq. (19), Theorem 2"},{"comment":"Theorem 2 omits the on-grid delay assumption that is explicit in Section II-A, namely that l_i = M Δf τ_i is an integer. Algorithm 1 Step 4 estimates the delay as l_0, an integer bin, and Step 6 nulls the entire delay row; this is valid only when each path delay is on-grid and at most one path occupies each resolvable delay bin. As written, Theorem 2 claims exact estimation under conditions only on the ranges of ν_i and τ_i, which is misleading for practical wideband channels with continuous delays and multiple paths per delay bin. The theorem statement and Remark 4 must list all underlying assumptions and restrict the claims accordingly.","section":"Section II-A and Theorem 2"},{"comment":"The numerical results do not test the regime in which Theorem 2 is claimed to hold. The captions of Figs. 2 and 3 specify on-grid delays, and the extended vehicular A channel is not used with its natural continuous delays. No experiment with off-grid delays or with multiple paths sharing one delay bin is reported, and the number of paths \\hat{P} is presumably known or obtained by an unspecified energy threshold. The conclusion that the proposed design 'offers near-optimal performance' is therefore validated only under conditions that match the favorable assumptions, not under the general conditions implied by Theorem 2.","section":"Section IV, Figs. 2 and 3"}],"minor_comments":[{"comment":"In Step 6, the index set for k is written as S^1_M; since k is the Doppler index of length N, it should be S^1_N.","section":"Algorithm 1, Step 6"},{"comment":"Remark 3 attributes the off-grid error to 'DD domain aliasing,' but the effect is spectral leakage of the sinc-like Dirichlet kernel; consider using 'leakage' for clarity.","section":"Remark 3"},{"comment":"The claim that the method works 'without requiring prior information' is too strong: Algorithm 1 requires \\hat{P} as an input and relies on the one-path-per-delay-bin structural assumption. The claim should be qualified.","section":"Abstract and Section V"},{"comment":"Algorithm 1 is taken from prior work [12] with minimal adaptation; the specific choices for the energy threshold and the handling of multiple paths in the same delay bin are not described, which makes the method difficult to reproduce.","section":"Section III-B, Algorithm 1"},{"comment":"The claimed complexity O(N log N + M log M + P) should be justified for Eq. (21) and Algorithm 1: the reconstruction in Eq. (21) as written is a sum over paths and, if evaluated directly for all (k,l), would cost more than the stated complexity.","section":"Section IV, complexity claim"},{"comment":"The set notation S^α_β is described as the 'positive integer set from 0 to α−1'; since 0 is included, 'nonnegative integers' would be more accurate.","section":"Notation"}],"recommendation":"major_revision","confidential_remarks":"The core idea is promising and the on-grid case is solid, but the central exactness claim for off-grid Doppler needs to be corrected or substantially qualified. The paper would benefit from an explicit error bound for the off-grid estimation, a complete statement of all assumptions in Theorem 2, and simulations with off-grid delays and multiple paths per delay bin. With those revisions, the contribution could be suitable for publication in a communications letters venue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Best read as: The paper's central result is a pair of pilot-spacing theorems for recovering the delay-Doppler CSF of an OFDM channel from regularly spaced pilots. That is a genuine, if modest, advance over [9] and [10], which estimate the CSF without giving compactness-based spacing rules. The on-grid Doppler case (Theorem 1) is derived cleanly, with scaling factors that fall out of the DFT relations rather than being tuned, and the simulations support the 'near-MMSE' conclusion. Credit where due: the comparison against linear interpolation and MMSE is standard and fair, and the idea of using CSF stationarity to skip channel statistics is worth having on the record.\n\nThe soft spots, in order. (1) Theorem 2 is worded as exact recovery for off-grid Doppler, but Eq. (19)(a) exchanges a ratio of sine functions for a ratio of arguments using 'N is relatively large.' That is a finite-N approximation, so the estimated fractional Doppler carries an O(1/N) bias even without noise. The theorem needs either a bound or a softer statement. (2) Algorithm 1 assumes one path per resolvable delay bin and takes P as an input from an energy threshold; step 6 nulls an entire delay row, so two paths in the same delay bin will not both be recovered. The simulations use on-grid delays and do not exercise that scenario, so the practical envelope is smaller than the theorem suggests. (3) The complexity claim 'O(N log N + M log M + P)' omits the per-path peak search over the (N/dt) by (M/df) grid; the actual cost includes O(P N M/(dt df)) for that search, still dramatically cheaper than the MMSE inversion, but the stated complexity is incomplete.\n\nThese are fixable issues, not a broken method. The on-grid result stands, and the off-grid approach works approximately in realistic settings. The paper deserves peer review and, after tightening the language, I'd use it as a reference. My recommendation: go to review, with a request for a bound or a softened Theorem 2 and a corrected complexity account.","headline":"Solid on-grid CSF recovery and a useful pilot-spacing rule; off-grid 'exact' claim needs softening to approximate.","tokens_in":10505,"tokens_out":3076,"would_cite":true,"duration_ms":33200,"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":"Under compact delay-Doppler conditions, one period of a pilot-derived periodic CSF recovers the OFDM channel exactly, and the data-symbol CTF follows by inverse transform.","keywords":["OFDM","channel transfer function estimation","channel spreading function","high mobility","delay-Doppler domain","pilot design","doubly selective channel","fractional Doppler"],"falsifier":"Simulate a two-path channel whose delays fall into the same resolvable delay bin but whose Dopplers differ, with on-grid delays and high SNR; if Algorithm 1 cancels the two paths together and the CTF MSE departs from the MMSE curve, the exact-recovery claim fails in exactly the regime the theorems exclude. A second test simulates continuously random delays and checks whether the MSE gap to MMSE appears precisely when a delay moves off the grid.","tokens_in":9436,"feed_emoji":"📡","tokens_out":6888,"duration_ms":65290,"temperature":0.7,"pith_summary":"This paper tries to show that an OFDM receiver can obtain the full channel transfer function (CTF) under high mobility from pilot symbols alone, without interpolating across fading. The route is to estimate the channel spreading function (CSF) in the delay-Doppler domain, which stays quasi-static over intervals much longer than the channel coherence time, and then transform it back to the time-frequency domain for every data symbol. The paper proves that, when all path delays lie on the delay grid and the delay and Doppler supports fit inside ranges set by pilot spacing, one period of the periodic CSF computed from pilots exactly equals the original CSF, so no resolution is lost. In simulations the resulting CTF estimator matches the MMSE estimator's accuracy without needing channel statistics and runs at roughly DFT-level complexity. If correct, this gives high-mobility OFDM a practical channel estimator that keeps the current waveform and only changes where pilots sit.","feed_headline":"One pilot period can recover the full OFDM channel in high mobility","feed_subtitle":"A delay-Doppler snapshot from pilots restores the whole channel, matching MMSE accuracy without channel statistics.","key_machinery":"The central object is the periodic channel spreading function $h_{\\mathrm{DD}}^{\\mathrm{Periodic}}[k,l]=\\mathrm{DFT}_N\\{\\mathrm{IDFT}_M\\{h_{\\mathrm{TF}}^{\\mathrm{Discrete}}[m,n]\\}\\}$, computed from pilot-only CTF samples spaced by $d_t$ in time and $d_f$ in frequency. It is two-dimensionally periodic with periods $N/d_t$ and $M/d_f$, so one period reproduces the original CSF exactly when the channel's delay and Doppler supports fit inside those periods; the proof compares the geometric-series kernels $R_{\\mathrm{Doppler}}^{\\mathrm{Periodic}}(k_i,k)$ and $R_{\\mathrm{delay}}^{\\mathrm{Periodic}}(l_i,l)$ with their full-grid counterparts and fixes the scaling factors $\\tilde{M}=M/d_f^2$, $\\tilde{N}=N/d_t^2$. In the off-grid Doppler case, Algorithm 1 locates each path by the peak of this kernel and estimates the fractional part of the Doppler from the magnitude ratio of the two neighboring samples, using the large-$N$ approximation $\\sin(\\pi d_t x/N)\\approx \\pi d_t x/N$.","core_discovery":"The central claim is that the delay-Doppler channel spreading function of an OFDM link can be recovered exactly from one period of the periodic CSF obtained by applying a length-$N$ DFT along time and a length-$M$ IDFT along frequency to the CTF sampled only at pilot resource elements. Theorem 1 states that if the normalized Doppler lies in $[-1/(2d_t T),\\,1/(2d_t T)-1/(NT)]$ and the normalized delay lies in $[0,\\,1/(d_f \\Delta f)-1]$, with delays on-grid, then $h_{\\mathrm{DD}}^{\\mathrm{Periodic}}[k,l]$ exactly represents the original CSF on that period, and the CTF for all data symbols follows from the inverse transform in (16). For off-grid Doppler, Algorithm 1 estimates each path's fractional Doppler from the ratio of the two largest adjacent peaks of the sinc-like kernel, reconstructs the CSF via (21), and Theorem 2 asserts exact recovery under the same compactness bounds. The numerical results show MSE close to the MMSE estimator and BER close to the ideal-channel benchmark in both on-grid and off-grid Doppler cases.","pith_inferences":["A practical corollary the paper leaves implicit is a closed-form pilot-spacing rule: choose $d_t$, $d_f$ so that $1/(2d_t T)\\geq \\nu_{\\max}$ and $1/(d_f \\Delta f)\\geq \\tau_{\\max}$, which could be adapted per mobility class in a standards implementation.","Because the estimated CSF contains path delays and Dopplers, the same pilot snapshot could feed integrated sensing and communication, turning channel estimation into radar parameter estimation without extra pilots.","The fractional-Doppler step could be iterated: after each path is subtracted in Algorithm 1, a second peak-ratio pass or a Newton refinement of $\\hat{k}_i$ would likely remove the small off-grid MSE gap to MMSE; this is a testable extension, not claimed in the paper.","The one-path-per-delay-bin assumption could be relaxed by oversampling the delay axis or applying a subspace method inside each occupied bin; the paper's exact statements do not cover that case."],"forward_implications":["In high-mobility OFDM links satisfying the compactness bounds, pilot spacing $d_t$, $d_f$ carries enough information to recover every data-symbol CTF exactly, so time-frequency interpolation across fast fading is unnecessary.","The estimator's complexity is $O(N\\log N + M\\log M + P)$, versus the MMSE estimator's $O((N/d_t)^3(M/d_f)^3)$, and it needs no covariance or channel-statistics information.","With a single-tap equalizer, the proposed method's BER approaches the ideal-channel benchmark and tracks the MMSE estimator in both on-grid and off-grid Doppler simulations.","The MSE is slightly higher in the off-grid Doppler case because of the fractional-Doppler approximation, but the BER penalty is negligible.","The pilot arrangement is a standard lattice with $d_t$ and $d_f$ chosen from the CSF compactness conditions, so the approach is compatible with existing OFDM frames."],"supporting_citations":[{"why":"Supplies the doubly selective channel model and the quasi-stationary CSF property that motivates estimating the CSF instead of the CTF.","marker":"[8]"},{"why":"Provides the on-grid delay normalization $l_i = M\\Delta f \\tau_i$ and the phase-shift/cyclic-shift OFDM input-output model used in (2).","marker":"[6]"},{"why":"Justifies the ICI-neglecting approximation in (3), the starting point for the pilot-only CTF model under high mobility.","marker":"[11]"},{"why":"Provides the fractional-Doppler peak-ratio estimation and the energy-threshold multipath count used by Algorithm 1.","marker":"[12]"},{"why":"Defines the lattice pilot arrangement and the linear-interpolation baseline whose error floor the proposed method removes.","marker":"[2]"},{"why":"Defines the MMSE estimator baseline whose accuracy the proposed method matches without channel statistics.","marker":"[3]"}],"fun_headline_variants":["Single pilot period recovers all OFDM channel coefficients in high mobility","Delay-Doppler CSF from one pilot period restores full CTF for OFDM","Exact CTF recovery from single-pilot-period CSF in fast fading","One pilot period yields full OFDM channel from delay-Doppler CSF"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The exact-recovery theorems assume every path's delay lands exactly on the integer delay grid, that at most one path occupies each resolvable delay bin, and that the number of paths is known from an energy threshold.","fun_headline_variants_meta":{"raw":{"variants":["Single pilot period recovers all OFDM channel coefficients in high mobility","Delay-Doppler CSF from one pilot period restores full CTF for OFDM","Exact CTF recovery from single-pilot-period CSF in fast fading","One pilot period yields full OFDM channel from delay-Doppler CSF"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0008,"raw_usage":{"total_tokens":3520,"prompt_tokens":949,"completion_tokens":2571,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":565,"completion_tokens_details":{"reasoning_tokens":2490}},"tokens_in":565,"tokens_out":2571,"duration_ms":18650,"temperature":1.0,"reasoning_tokens":2490,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:10:14.674977+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate a two-path channel whose delays fall into the same resolvable delay bin but whose Dopplers differ, with on-grid delays and high SNR; if Algorithm 1 cancels the two paths together and the CTF MSE departs from the MMSE curve, the exact-recovery claim fails in exactly the regime the theorems exclude. A second test simulates continuously random delays and checks whether the MSE gap to MMSE appears precisely when a delay moves off the grid.","supporting_citations":[{"cited_title":"Hlawatsch and G","cited_arxiv_id":null,"evidence_quote":"Supplies the doubly selective channel model and the quasi-stationary CSF property that motivates estimating the CSF instead of the CTF."},{"cited_title":"Interference cancellation and iterative detection for orthogonal time frequency space modulation,","cited_arxiv_id":null,"evidence_quote":"Provides the on-grid delay normalization $l_i = M\\Delta f \\tau_i$ and the phase-shift/cyclic-shift OFDM input-output model used in (2)."},{"cited_title":"On the effective- ness of OTFS for joint radar parameter estimation and communication,","cited_arxiv_id":null,"evidence_quote":"Justifies the ICI-neglecting approximation in (3), the starting point for the pilot-only CTF model under high mobility."},{"cited_title":"Deterministic pilot design and channel estimation for downlink massive MIMO-OTFS systems in presence of the fractional doppler,","cited_arxiv_id":null,"evidence_quote":"Provides the fractional-Doppler peak-ratio estimation and the energy-threshold multipath count used by Algorithm 1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the lattice pilot arrangement and the linear-interpolation baseline whose error floor the proposed method removes."},{"cited_title":"Deep learning-based channel estimation,","cited_arxiv_id":null,"evidence_quote":"Defines the MMSE estimator baseline whose accuracy the proposed method matches without channel statistics."}],"review_version":1}