{"id":"038a76bf-89b8-42cc-a88a-be6833b53fb7","arxiv_id":"2506.07256","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"J-BMOCZ adds one asymmetric zero to BMOCZ and estimates CFO by template matching in the Fourier domain, avoiding cyclically permutable codes.","lead":"Researchers propose a small change to a non-coherent radio modulation called BMOCZ: move one zero slightly outward so the receiver can estimate a carrier frequency offset from the signal's own spectrum, with no pilots and no special channel code. This could make BMOCZ practical with off-the-shelf codes like BCH or LDPC in short-packet links.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing conjugation in the AACF definition makes the message-independence proof invalid for complex BMOCZ zeros; the K=2 worked example uses real zeros and cannot expose it.","rationale":"The paper's central claim is that J-BMOCZ enables pilot-free CFO estimation by correlating the received magnitude spectrum against a message-independent template. The key technical requirement is the BMOCZ property that every codeword has the same AACF, hence the same |X(e^{jω})|. That property is true for the standard conjugated AACF and for the proposed conjugate-reciprocal zero pairs, but the manuscript's own equations do not state it correctly: Eq. (4) omits conjugation, and Eq. (5) writes X(z)X(1/z) instead of X(z)X^*(1/z^*). For non-real zeros, which occur whenever K≥4 with uniform phase spacing, the un-conjugated product is not invariant under the bit choices, so the literal derivation does not support the template. This is an internally inconsistent proof step, not merely a disagreement with prior literature, and it is the weakest load-bearing point in the paper. The Le=1 flat-fading assumption is explicit and consistent with the simulation setup, so I do not treat it as a separate objection. A concrete all-codeword numerical check would settle the issue quickly: the corrected identity should pass, and if the authors confirm that the implemented template uses the conjugated form, the central mechanism stands. The existing CONDITIONAL verdict is appropriate: the idea is likely sound but the submitted text needs a substantive correction to the AACF definition and the zero-domain identity.","tokens_in":9732,"tokens_out":19492,"duration_ms":209155,"concrete_test":"Compute, for all 2^8 codewords of J-BMOCZ with K=8, R=1.176, ζ=1.15, and ψ_k=2πk/8, the quantity |X(e^{j2πn/N})| for N=256 and also the literal z-domain product X(z)X(1/z) from Eq. (5). If every codeword gives the same |X| to numerical precision, the intended message-independence holds under the corrected conjugated AACF; if the literal un-conjugated products differ, Eq. (5) as printed is false and the proof of the template's invariance must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central estimator needs |X(e^{jω})| to be identical across all codewords, so t_N in Eq. (10) is a known message-independent template and the argmax in Eq. (13) is meaningful. The manuscript derives this from the 'identical AACF' claim, but the derivation as printed is invalid. Definition 1 (Eq. 4) defines the AACF coefficients without conjugation, while Eq. (9) sets A(e^{jω})=X(e^{jω}) overline{X(e^{jω})}=|X|^2, which corresponds to the conjugate correlation. The zero-domain formula in Eq. (5), A(z)=X(z)X(1/z), produces zeros at α and 1/α for each zero α; the conjugate-reciprocal pair used by BMOCZ is {α, 1/\\bar{α}}. These coincide only for real zeros. For K≥4, take zeros at ψ_k=2πk/K with k not 0 or K/2: choosing r_k e^{jψ_k} versus r_k^{-1} e^{jψ_k} changes the un-conjugated product X(z)X(1/z), so all codewords do not share the literal A(z) of Eq. (5). The paper's only worked example (K=2, phases 0 and π) has real zeros and therefore cannot reveal the error. The corrected statement is A(z)=Σ_ℓ (Σ_i x_i \\bar{x}_{i+ℓ}) z^ℓ and A(z)=X(z)X^*(1/z^*); with that correction, the message-independence of |X| is standard and true for J-BMOCZ. The concern is thus that the central claim, as written, lacks a valid proof; the simulations must have used the corrected identity implicitly, and this needs to be stated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes J-BMOCZ, a modification of Huffman BMOCZ in which the zero associated with the first message bit is placed at radius ζR (or its reciprocal) instead of the common radius R, thereby breaking the rotational symmetry that makes the integer CFO ambiguous in Huffman BMOCZ. It then derives a pilot-free CFO estimator: because BMOCZ codewords share a message-independent magnitude spectrum |X(e^{jω})|, the receiver cross-correlates the magnitude spectrum of the received block with this known template over candidate rotations, using an iterative IDFT-based search, and then applies standard DiZeT decoding. The paper includes a fully worked K=2 example and validates the method by BER and BLER simulations in AWGN and flat fading, reporting a roughly 1 dB loss in AWGN (2 dB in fading) for uncoded J-BMOCZ under CFO relative to Huffman BMOCZ without CFO, and better coded BER than Huffman BMOCZ with an ACPC under CFO.","tokens_in":10139,"tokens_out":34332,"duration_ms":320759,"significance":"If the corrected derivation holds, the paper removes a real limitation of Huffman BMOCZ: CFO estimation without a cyclically permutable code, so arbitrary channel codes and message lengths can be used in conjunction with the non-coherent DiZeT decoder. The approach is simple, the algorithm is explicit, and the K=2 example is arithmetically checkable; the equal-rate comparison between coded J-BMOCZ and (31,16)-ACPC Huffman BMOCZ is a fair benchmark. These strengths support the paper's interest for short-packet non-coherent communication. The main caveats are that the central proof as printed (Eqs. (4)-(5)) is internally inconsistent and must be restated with the conjugate AACF, and that the headline 1 dB figure is obtained at a ζ tuned on the reported BER metric, so sensitivity to ζ should be documented.","major_comments":[{"comment":"The AACF in Definition 1 is defined without conjugation (a_ℓ = Σ_i x_i x_{i+ℓ}), and Eq. (5) gives A(z)=X(z)X(1/z), but Eq. (9) evaluates A(e^{jω}) as |X(e^{jω})|^2, which corresponds to the conjugate correlation A(z)=X(z)\\overline{X(1/\\bar{z})}. The un-conjugated product in Eq. (5) is message-dependent whenever any ψ_k is not 0 or π: the two bit choices for zero k produce the root sets {r_k e^{jψ_k}, r_k^{-1} e^{-jψ_k}} and {r_k^{-1} e^{jψ_k}, r_k e^{-jψ_k}}, which coincide only for real zeros. For J-BMOCZ with K≥4, ψ_k=2πk/K includes complex phases, so the printed derivation does not establish that all codewords share one magnitude template. The claim is nevertheless correct after the standard fix a_ℓ=Σ_i x_i\\overline{x}_{i+ℓ} (equivalently A(z)=X(z)\\overline{X(1/\\bar{z})}), under which the conjugate-reciprocal pair {α_k, 1/\\bar{α}_k} is independent of b_k; the authors should correct Definition 1 and Eq. (5) accordingly and state that the simulations use this conjugate form. The only worked example (K=2, ψ_k ∈ {0,π}) has real zeros and therefore cannot expose the inconsistency. Relatedly, the parenthetical after Eq. (1) states (r_k e^{jψ_k})^{-1}=r_k^{-1}e^{jψ_k}; this is the ordinary reciprocal only if r_k e^{jψ_k} is real, and the intended conjugate-reciprocal identity requires conjugation of r_k e^{jψ_k} before inversion.","section":"Section II-A, Def. 1, Eqs. (4)-(5); Section III-A, Eq. (9)"},{"comment":"The asymmetry factor is chosen by sweeping its value and selecting that which minimizes the BER under a CFO, and the reported 1 dB (AWGN) and 2 dB (fading) losses, as well as the coded comparisons, are all evaluated at the resulting ζ=1.15. Because ζ is tuned on the same BER metric that is then reported, the quantitative claims are conditional on an in-sample optimum. The authors should report BER versus ζ (e.g., at a representative Eb/N0), state the sensitivity of the 1 dB figure to ζ and to K, and indicate whether ζ=1.15 is fixed across SNR, channel, and block length or re-tuned per simulation.","section":"Section IV, parameter selection for ζ"}],"minor_comments":[{"comment":"The abstract states that the BER loss under CFO is just 1 dB, but Section IV-A reports a 2 dB loss in the fading channel; the abstract should qualify the figure as the AWGN result.","section":"Abstract vs. Section IV-A"},{"comment":"The manuscript calls the transform used for t_N and for the columns of \\tilde{Y}_N the 'N-point IDFT'; with the standard DFT/IDFT convention these are N-point DFT evaluations of the zero-padded sequence. Please define the transform convention once so that the sign in Eq. (12) is unambiguous.","section":"Section III-A, Eqs. (10) and (12)"},{"comment":"The sentence 't_N is equivalent to the magnitude of the N-point IDFT of x, which is known at the receiver' conflates the unknown transmitted codeword with the known template; rephrase to say the template is known because, by the corrected AACF property, |X(e^{jω})| is independent of the message.","section":"Section III-A, sentence after Eq. (10)"},{"comment":"The manuscript does not report CFO estimation error (e.g., RMSE of \\hat{φ} versus Eb/N0, N, or ζ); such a curve would directly validate the Section III claim that the CFO is estimable over [0,2π).","section":"Section III-IV"},{"comment":"The DiZeT rule in Eq. (6) applies the per-zero scaling r_k^K; for J-BMOCZ r_k differs for k=0, so state explicitly that the scaling uses r_k from (1)/(14) and note that this is the natural generalization of the Huffman rule.","section":"Eq. (6), DiZeT scaling"},{"comment":"Please state the number of Monte Carlo realizations used for the BER/BLER curves in Figs. 4-5, or add confidence intervals, so that the reported 1 dB and fractional-dB differences are meaningful.","section":"Section IV, simulation details"},{"comment":"The refinement window half-width in iteration ι is δ/ι; for the true CFO to remain in the window after the coarse pass, δ should exceed the coarse bin width 2π/N (here δ=0.2 with N=64 does so). State this condition or comment on the lock-loss behavior.","section":"Algorithm 1"}],"recommendation":"major_revision","confidential_remarks":"The conjugation error in Definition 1 and Eq. (5) looks like a notation slip: the message-independence of the BMOCZ magnitude spectrum is a standard result from [2], and the K=2 example is consistent with the corrected (conjugate) version. Since the entire CFO estimator depends on this property, I would ask the authors to restate it cleanly and confirm that the simulations use the conjugate correlation. The ζ sweep is a legitimate concern for the quantitative claim, but it is addressable by a sensitivity study. The paper fits eess.SP well; my recommendation is major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new idea is simple and clever: jut one zero out radially to break Huffman's rotational symmetry, then estimate CFO by correlating the received magnitude spectrum against a message-independent template. This removes the need for cyclically permutable codes and opens the door to standard channel codes, a real benefit for this niche. The worked K=2 example checks out, and the simulation results (1 dB loss for uncoded J-BMOCZ under CFO vs Huffman without CFO; coded J-BMOCZ beating BMOCZ-ACPC) are plausible.\n\nThe soft spots: first, the AACF in Definition 1 (Eq. 4) is written without conjugation, but Eq. 9 evaluates it as |X|^2, which corresponds to the conjugate correlation. The zero-domain formula in Eq. 5 is accordingly wrong for complex zeros: it gives zeros at alpha and 1/alpha, not alpha and 1/bar{alpha}. For the uniform-phase Huffman/J-BMOCZ constellations with K >= 4, most zeros are off the real axis, so the identity 'all sequences share the same AACF' does not follow from the printed equations. The K=2 example uses real zeros only, so it cannot catch the error. The corrected identity A(z)=X(z)X^*(1/z^*) is standard and true for conjugate-reciprocal zeros, so I expect the estimator is fine, but the paper needs to fix the notation and re-derive.\n\nSecond, zeta is selected by sweeping to minimize BER on the same metric report; that's a tuned parameter, and without cross-validation or an analysis of sensitivity it's hard to know how robust the 1 dB claim is. Also, no code or data are shared, and the flat-fading assumption (Le=1) is assumed throughout, so the claim that it works in fading rests on that idealization. These are presentation and reproducibility concerns, not fatal flaws.\n\nBottom line: this is a solid contribution to a niche subfield, the math just needs a correction. I'd send it to review, with a request to fix the AACF definition, report the zeta sweep details, and release simulation code. It's worth a serious referee.","headline":"A clever twist on BMOCZ with a fixable math error in the AACF definition; worth reviewing.","tokens_in":10605,"tokens_out":1763,"would_cite":true,"duration_ms":15263,"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":"By moving one zero outward in the BMOCZ constellation, a receiver can estimate the whole carrier frequency offset from the received spectrum alone, removing the need for pilots or cyclically permutable codes.","keywords":["carrier frequency offset (CFO)","binary modulation on conjugate-reciprocal zeros (BMOCZ)","Huffman BMOCZ","aperiodic autocorrelation function","IDFT-based CFO estimation","non-coherent communication","zero modulation","pilot-free synchronization"],"falsifier":"Run Algorithm 1 on a simulated received signal with a known CFO $\\varphi_0$ and a two-tap channel ($L_e = 2$). If the inner-product peak in Eq. (13) shifts away from $\\varphi_0$ or develops multiple comparable peaks, the pilot-free claim is confined to the flat-fading case. Separately, evaluate Eq. (5) with $X(z)\\overline{X(1/\\bar{z})}$ and compare it to $|X(e^{j\\omega})|^2$ for the jutted radii: the equality in Eq. (9) is load-bearing, and any numerical mismatch changes the template $\\mathbf{t}_N$ and the CFO estimate.","tokens_in":9553,"feed_emoji":"📡","tokens_out":10548,"duration_ms":93431,"temperature":0.7,"pith_summary":"This paper claims that a small, deliberate asymmetry in the zero constellation of BMOCZ — moving one zero outward along the real axis — lets a receiver estimate a carrier frequency offset anywhere in $[0, 2\\pi)$ without pilots and without channel coding. The reason the trick works is that every BMOCZ codeword shares the same aperiodic autocorrelation function, so the receiver knows the magnitude spectrum template in advance; a CFO simply rotates all zeros and shifts that template, and the receiver finds the shift by an IDFT-based correlation. This matters because existing Huffman BMOCZ CFO correction forces a cyclically permutable code, which fixes the coding scheme and constrains message length. If the claim holds, the constellation shape alone resolves the CFO ambiguity, at a cost of roughly 1 dB of BER relative to Huffman BMOCZ with no CFO in AWGN, and it opens BMOCZ to standard codes and soft-decision decoding.","feed_headline":"A jutted zero estimates carrier offset without pilots or coding","feed_subtitle":"Asymmetric BMOCZ constellation finds the full CFO by correlation, at a BER cost near 1 dB.","key_machinery":"The load-bearing object is the identical aperiodic autocorrelation function $A(z)$, defined for the BMOCZ sequence $\\mathbf{x}$ and expressible from its zeros as $X(z)X(1/z)$; evaluated on the unit circle it should equal $|X(e^{j\\omega})|^2$, which makes the template vector $\\mathbf{t}_N$ (the magnitude of the $N$-point IDFT of a codeword) known to the receiver independent of the message. The 'jutted' zero — radius $\\zeta R$ for $k=0$, radius $R$ for the other $K-1$ zeros — is the second piece: it breaks the $K$-fold rotational symmetry of the Huffman constellation so the template is non-periodic and the correlation has a unique peak. Algorithm 1 carries the estimation: it generates a modulation matrix $\\mathbf{M}$ of candidate rotations, forms $N$ modulated copies of the received sequence, takes their $N$-point IDFTs, and picks the rotation whose magnitude column has the largest inner product with $\\mathbf{t}_N$, then refines the search interval over iterations.","core_discovery":"The paper's central discovery is that the CFO ambiguity of Huffman BMOCZ is an artifact of rotational symmetry, not of non-coherent zero modulation itself. In Huffman BMOCZ the zeros sit on two concentric circles, so any rotation by the base angle $2\\pi/K$ maps a valid message to another valid message. J-BMOCZ breaks that symmetry by setting the radius of the first zero pair to $\\zeta R$ instead of $R$, creating a 'jutted' zero while leaving all other zeros at the Huffman radii. Because the AACF — and hence the magnitude spectrum — of BMOCZ sequences is message-independent, the receiver can build the template $\\mathbf{t}_N$ and estimate the full CFO $\\varphi$ by aligning $\\mathbf{t}_N$ with the magnitude spectrum of the received sequence through the modulation matrix and inner-product maximization in Eq. (13). The paper reports that uncoded J-BMOCZ under a CFO loses about 1 dB BER over CFO-free Huffman BMOCZ in AWGN (about 2 dB in fading), and that J-BMOCZ with a (31,16)-BCH code outperforms Huffman BMOCZ with an ACPC under CFO by about 1.75 dB in BER in fading, while keeping similar BLER.","pith_inferences":["Editorial inference: the template correlation should remain partially functional for short channels with more than one tap when the channel zeros sit away from the unit circle, since they add ripple to $|Y(e^{j\\omega})|$ rather than a clean shift; this can be tested by running Algorithm 1 on a two-tap channel with known zeros.","Editorial inference: $\\zeta$ is a tunable trade-off — larger $\\zeta$ sharpens the template peak and the CFO estimate but should deepen the baseline BER penalty — so an analytic $\\zeta(K)$ rule, which the paper leaves to future work, would let operators choose the operating point in advance.","Editorial inference: the same correlation logic applies to any message-independent spectral feature, so variants of J-BMOCZ with multiple jutted zeros could trade a larger template peak for more spectrum-shaping flexibility.","Editorial inference: the no-interference assumption means the practical regime is small CFOs relative to bandwidth, as the paper itself notes; an OFDM implementation could equally use this estimator for a timing offset, which would make it a joint synchronization primitive rather than only a CFO fix."],"forward_implications":["Pilot-free and channel-coding-free CFO correction becomes possible for BMOCZ in flat-fading links, releasing the message length and coding constraints imposed by cyclically permutable codes.","Uncoded J-BMOCZ keeps a roughly 1 dB BER loss relative to CFO-free Huffman BMOCZ in AWGN under a CFO uniformly drawn from $[0, 2\\pi)$, with the loss growing to about 2 dB in a fading channel.","Coded J-BMOCZ with a (31,16)-BCH code beats Huffman BMOCZ with the (31,16)-ACPC by about 1.75 dB in BER under CFO in fading, at similar BLER, while also avoiding the ACPC's fixed outer code.","Because no cyclic permutability constraint is needed, J-BMOCZ is compatible with soft-decision decoding and standard polar or LDPC codes, which the paper identifies as a likely performance improvement.","The iterative search with $N=64$, $\\delta=0.2$, and two iterations resolves the full CFO range, so the added receiver complexity over the baseline DiZeT decoder is a modest number of small IDFTs."],"supporting_citations":[{"why":"Defines BMOCZ, the Huffman zero constellation, the DiZeT decoder, and the identical AACF property that the estimator relies on.","marker":"[2]"},{"why":"Supplies the existing oversampled DiZeT plus ACPC CFO correction that J-BMOCZ replaces, and the performance baseline for comparisons.","marker":"[3]"},{"why":"Supplies the impulse-equivalent pulse-train characterization and the zero-radius rule used by both constellations.","marker":"[9]"},{"why":"Provides the OFDM subcarrier-mapping argument that lets the system assume a flat-fading channel with no channel zeros.","marker":"[8]"},{"why":"Introduces the modulation-on-zeros concept and the zero-preservation property that underlies non-coherent detection.","marker":"[1]"}],"fun_headline_variants":["J-BMOCZ breaks zero symmetry to estimate CFO without pilots","Asymmetric zeros unlock CFO estimation in BMOCZ, no pilots needed","Jutted binary modulation estimates CFO with 1 dB penalty","CFO-free BMOCZ with jutted zeros beats coded Huffman under offset","One jutted zero removes CFO ambiguity without coding or pilots"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The estimator assumes every codeword shares a single known magnitude-spectrum template — the identical AACF — and that the channel contributes no zeros ($L_e = 1$), so the received spectrum is the template shifted by the CFO.","fun_headline_variants_meta":{"raw":{"variants":["J-BMOCZ breaks zero symmetry to estimate CFO without pilots","Asymmetric zeros unlock CFO estimation in BMOCZ, no pilots needed","Jutted binary modulation estimates CFO with 1 dB penalty","CFO-free BMOCZ with jutted zeros beats coded Huffman under offset","One jutted zero removes CFO ambiguity without coding or pilots"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000218,"raw_usage":{"total_tokens":1459,"prompt_tokens":987,"completion_tokens":472,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":380}},"tokens_in":603,"tokens_out":472,"duration_ms":5601,"temperature":1.0,"reasoning_tokens":380,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:39:26.628419+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run Algorithm 1 on a simulated received signal with a known CFO $\\varphi_0$ and a two-tap channel ($L_e = 2$). If the inner-product peak in Eq. (13) shifts away from $\\varphi_0$ or develops multiple comparable peaks, the pilot-free claim is confined to the flat-fading case. Separately, evaluate Eq. (5) with $X(z)\\overline{X(1/\\bar{z})}$ and compare it to $|X(e^{j\\omega})|^2$ for the jutted radii: the equality in Eq. (9) is load-bearing, and any numerical mismatch changes the template $\\mathbf{t}_N$ and the CFO estimate.","supporting_citations":[{"cited_title":"MOCZ for blind short-packet communication: Basic p rinciples,","cited_arxiv_id":null,"evidence_quote":"Defines BMOCZ, the Huffman zero constellation, the DiZeT decoder, and the identical AACF property that the estimator relies on."},{"cited_title":"MOCZ for blind short- packet communication: Practical aspects,","cited_arxiv_id":null,"evidence_quote":"Supplies the existing oversampled DiZeT plus ACPC CFO correction that J-BMOCZ replaces, and the performance baseline for comparisons."},{"cited_title":"The generation of impulse-equivalent puls e trains,","cited_arxiv_id":null,"evidence_quote":"Supplies the impulse-equivalent pulse-train characterization and the zero-radius rule used by both constellations."},{"cited_title":"Short-message communi cation and FIR system identification using Huffman sequences,","cited_arxiv_id":null,"evidence_quote":"Introduces the modulation-on-zeros concept and the zero-preservation property that underlies non-coherent detection."}],"review_version":1}