{"id":"f9ed0afd-33b0-44c3-ae0a-ee1c6efe58f6","arxiv_id":"2501.11330","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Mixing the modulo-folded signal with a periodic comb before low-pass filtering and sampling yields samples that match ideal modulo sampling, enabling modulo recovery with realistic ADCs.","lead":"This paper introduces a hardware scheme that tames the high-frequency artifacts created by modulo folding, so ordinary ADCs can be used instead of expensive wideband ones. The authors prove the samples from their mixer-and-filter pipeline match ideal modulo samples and validate it with a physical prototype.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The finite-comb replacement for the delta comb in Theorem 1 is never bounded; the paper's practical equivalence claim rests entirely on empirical N=2000 curves, so the central hardware claim is not yet established.","rationale":"After reading the full text, I find the reader's weakest-assumption identification persuasive. The theorem itself is clean and the hardware prototype is a genuine contribution, but the bridge from the delta comb to the physical comb is the single most load-bearing step: without a bound on E_mod-HF, the central claim that realistic ADCs produce samples equivalent to an ideal wideband ADC is not established. I considered other possible objections, such as the nonstandard quantization-step definitions in Eqs. (7)–(9), the bit-accounting for the extra side-information bit, and the nonideal hardware LPF, but these are secondary or can be absorbed by the same empirical validation. The finite-comb gap is more fundamental because it affects the theoretical equivalence itself. The recommended verdict remains CONDITIONAL, matching the reader, since the issue is a missing quantification rather than a demonstrated falsehood.","tokens_in":12094,"tokens_out":9263,"duration_ms":96160,"concrete_test":"Run the simulation of Section IV-A with an ideal LPF and p_N(t) for N ∈ {100, 500, 1000, 2000, 4000, 8000}, keeping T_s, λ, and the same 500-signal ensemble, and plot E_mod-HF versus N together with E_mod-Q for 6 and 8 bits; report the worst-case E_mod-HF as well as the average. If the worst-case E_mod-HF does not fall below E_mod-Q at N=2000 for every signal, or does not decay with N at a predictable rate, then the claimed equivalence to an ideal modulo sampler is not supported in the demonstrated regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1 is mathematically correct for the ideal comb p(t)=Σ δ(t−nT_s), because multiplying by a delta train and low-pass filtering recreates the sample-and-reconstruct identity. The actual prototype and all simulations replace this comb by a finite Fourier sum p_N(t)=Σ_{k=−N}^N e^{j2πkt/T_s} with N=2000, as stated in Section II-C. Since M_λ x is not bandlimited, the low-pass output contains aliased high-frequency components from the shifted copies with |k|≤N and misses all copies with |k|>N; the difference from M_λ x[nT_s] is E_mod-HF. Section II-C defines E_mod-HF only through Eq. (10) and gives no theoretical bound as a function of N, signal bandwidth, oversampling ratio, or λ. Figures 13 and 14 show empirically that without the comb the error is orders of magnitude above quantization noise and with N=2000 it drops below it, but this is one signal ensemble, one N, and no error bars or worst-case analysis. The abstract's statement that samples are 'equivalent' to an ideal high-spec ADC, and the claim that 'any existing modulo recovery method can be applied', are therefore supported only by an unquantified numerical choice. This is the load-bearing gap: if E_mod-HF is not guaranteed below E_mod-Q for realistic N, the proposed hardware does not actually deliver the ideal modulo samples.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes and prototypes a hardware front-end for modulo sampling in which the folded signal is multiplied by a periodic comb and low-pass filtered before sampling, so that a narrowband ADC can capture samples claimed to equal those of an ideal wideband modulo sampler. Theorem 1 proves this equality for an ideal delta comb and an ideal LPF. The paper then describes an SRD-based comb generator, an AD834 mixer, and a 7th-order Butterworth LPF, and reports simulations and hardware experiments comparing the proposed approach with ideal modulo sampling, direct LPF-modulo sampling, and classical sampling.","tokens_in":12403,"tokens_out":7166,"duration_ms":76426,"significance":"If the ideal equivalence were robust to finite comb length and nonideal filtering, the paper would be a genuine practical step toward modulo ADCs built from standard components. The clean ideal-theoretic result in Theorem 1 is correct and useful, and the hardware effort is concrete, with a component-level prototype and reproducible-looking experiments. The strength of the paper is therefore the combination of a valid sampling-theoretic identity with a real hardware demonstration. However, the central practical claim that the samples are 'equivalent' to an ideal high-spec ADC is not established, because the finite-comb error E_mod-HF is only defined and never bounded, and the nonideal filter is not analyzed.","major_comments":[{"comment":"The statement that the practical system yields samples equivalent to those of an ideal high-spec ADC is not established, because E_mod-HF is only defined and never bounded. Theorem 1 applies to p(t) = Σ_n δ(t − nT_s), while the practical system uses the finite Fourier sum p_N(t) = Σ_{k=−N}^{N} e^{j2πkt/T_s}. Since M_λx is not bandlimited, the output differs from M_λx[nT_s] by a term that depends on the high-frequency tail of M_λx and on the missing harmonic copies with |k| > N. The paper provides no upper bound on E_mod-HF as a function of N, T_s, λ, the input bandwidth, or the oversampling ratio, and no condition ensuring E_mod-HF is below E_mod-Q. Figures 13 and 14 show a single simulation ensemble at N=2000 with no error bars or worst-case analysis. This is load-bearing, because the abstract's 'equivalent' claim and the statement that 'any existing modulo recovery method can be applied effectively' depend on this unquantified numerical choice.","section":"Section II-C, Eq. (10)"},{"comment":"The prototype uses a 7th-order Butterworth LPF with a −3 dB cutoff at 25 kHz, whereas Theorem 1 assumes an ideal LPF with cutoff π/T_s. The nonideal filter's passband ripple and finite stopband attenuation change y[nT_s] relative to the ideal value, and this deviation is not modeled in E_mod-HF. Moreover, the comb generator's output harmonic amplitudes are not reported, so it is not verified that the hardware implements the equal-amplitude Fourier sum assumed in Section II-C. The paper should quantify the filter-induced error and the comb spectral error, or provide design criteria such as required stopband attenuation and harmonic flatness in terms of the target MSE.","section":"Section III-D and Fig. 11"},{"comment":"There is an internal inconsistency in the hardware description. Section III-A defines Sampler 3 as classical sampling of the input and Sampler 4 as the folded signal passed through a 25 kHz LPF, but Section IV-B says Sampler 3 records the folded signal after passing through a LPF, while the caption of Fig. 15 labels the LPF path as Sampler 4. Additionally, Section IV-B states that the hardware recovery used 'straightforward unwrapping' [45], [46], not the quantization-robust algorithm of [33] that Section II-C introduces and the abstract credits for the results. The sampler labeling should be corrected, and the text should clarify which recovery algorithm was used in each hardware experiment, because the current description makes the hardware validation difficult to interpret.","section":"Section IV-B versus Section III-A"}],"minor_comments":[{"comment":"The statement contains a typo: 'than y[nT_s]' should be 'then y[nT_s]'.","section":"Theorem 1 statement"},{"comment":"The finite-comb definition writes p(t) = Σ_{k=−N}^{N} e^{jk/T_s t}; the exponent is missing the factor 2π and should read e^{j2πkt/T_s}.","section":"Section II-C"},{"comment":"The text says that the inclusion of the comb generator 'significantly reduces E_mod-Q', but E_mod-Q is the quantization error and is independent of the comb; the sentence should refer to E_mod-HF.","section":"Section IV-A"},{"comment":"The claim that the errors are independent because quantization occurs after analog processing is not by itself a justification for adding the two MSE terms; independence of the error processes should be stated or derived, and the MSE should be defined over the relevant randomness.","section":"Equation (10)"},{"comment":"There are minor grammatical errors throughout, such as 'The theorem above provide' and 'which shown in the figure'; a careful proofreading pass would improve clarity.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The core ideal-theoretic theorem is correct, and the hardware prototype is a meaningful engineering contribution. The main blocker is the missing quantification of E_mod-HF and of the nonideal LPF/comb behavior. I would encourage the authors to add a rigorous bound, or at least a carefully quantified numerical study across N, bandwidth, and oversampling ratio, and to soften the 'equivalent' wording unless such a bound is supplied. The sampler-label inconsistency in Section IV-B should also be fixed before the paper is accepted. I do not see a fundamental flaw that would justify rejection, but the practical claim needs substantial additional support."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read. The paper's core trick is to apply the MWC mixing identity to modulo sampling: multiply the folded signal by a periodic comb, low-pass filter, and sample at the lower rate, getting exactly the ideal modulo samples in the limiting case of an ideal delta comb. Theorem 1 is correct. That is a neat observation, and the engineering follow-through is real: they built a board with an SRD comb generator and an AD834 multiplier, and the oscilloscope shots show the samples lining up. Credit where due: the paper is honest that the practical implementation uses a truncated comb and defines E_mod-HF, the error from that truncation.\n\nThe soft spot is that E_mod-HF is defined, not bounded. There is no analysis of how the error scales with N, signal bandwidth, oversampling ratio, or the range λ. The empirical curves in Figures 13 and 14 show one ensemble of 500 random sinc signals at N=2000, and for that ensemble the error drops below quantization noise. That is a decent proof-of-concept, but it is not a guarantee. The abstract's claim that samples are 'equivalent' to an ideal high-spec ADC is too strong for the finite-comb hardware. Also, the hardware comb is not exactly the truncated Fourier sum the paper analyses; it is an SRD-generated pulse train whose harmonic content is not characterized, so there is a second unquantified step. The LPF is 7th-order Butterworth, not the ideal brick-wall in the theorem. Minor: the hardware demo uses simple unwrapping, not the advertised Bernardo recovery with the 1-bit side channel, so that specific robustness claim is not demonstrated on the bench.\n\nNone of these are fatal. The central idea is sound, and the prototype is a step forward. But the paper needs more work before it's ready: either a theoretical bound on E_mod-HF or a much more thorough empirical study across parameter ranges, plus an honest characterization of the actual comb spectrum. I agree with the reader's conditional verdict. I'd send it to peer review, but I'd ask for a major revision to address the missing error analysis and to temper the abstract's language.","headline":"A sound and useful adaptation of MWC to modulo sampling, backed by a real hardware prototype, but the load-bearing error from the finite comb is defined and never bounded, so the practical 'equivalent to an ideal ADC' claim outruns the evidence.","tokens_in":725,"tokens_out":1706,"would_cite":false,"duration_ms":45121,"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":"Multiplying the folded signal by a delta comb before low-pass filtering lets a realistic ADC produce exactly the samples of an ideal modulo sampler.","keywords":["modulo sampling","dynamic range","analog-to-digital converter","high-frequency components","low-pass filter","comb generator","quantization noise","hardware prototype"],"falsifier":"Measure $E\\left[|y[nT_s]-M_\\lambda x[nT_s]|^2\\right]$ for the finite-comb system with $N=2000$ while increasing the input signal's maximum frequency or lowering the oversampling ratio at fixed bit depth; if this high-frequency error grows to dominate the quantization error and the recovered signal degrades well below the ideal-sampler baseline, the claimed practical equivalence fails.","tokens_in":11911,"feed_emoji":"📡","tokens_out":7757,"duration_ms":69423,"temperature":0.7,"pith_summary":"Modulo (folding) sampling lets an ADC capture signals much larger than its dynamic range, but the modulo operation creates high-frequency components that ordinary ADCs filter out, corrupting the samples. This paper proposes an analog mixer plus low-pass filter ahead of a realistic ADC: multiply the folded signal by a periodic comb, filter, and sample. Theorem 1 proves that the resulting samples equal the ideal pointwise samples of the folded signal, so any existing modulo recovery method applies unchanged. Simulations with 6-bit and 8-bit quantizers and a hardware prototype show that the comb-based pipeline tracks the ideal modulo sampler and beats classical sampling at oversampling rates of at least 5. The paper's central claim is that realistic, low-bandwidth ADCs can therefore replace the high-specification ADCs that modulo sampling previously required.","feed_headline":"Realistic ADCs can match ideal modulo sampling with a comb mixer","feed_subtitle":"No wideband ADC needed: mixing the folded signal with a comb and low-pass filtering reproduces ideal modulo samples.","key_machinery":"The load-bearing object is the delta-comb multiplier $p(t)=\\sum_{n\\in\\mathbb{Z}} \\delta(t-nT_s)$ placed between the analog modulo operator and the low-pass filter. Multiplying by the comb periodizes the folded signal's spectrum with spacing $2\\pi/T_s$; the low-pass filter with cutoff $\\pi/T_s$ then selects the baseband copy, and Poisson's summation formula equates the DTFT of the samples $y[nT_s]$ with the DTFT of the ideal modulo samples $M_\\lambda x[nT_s]$. This identity is what lets a narrowband ADC emulate a wideband one. In the hardware realization the infinite comb is approximated by a finite Fourier sum with $N=2000$ harmonics, and the resulting mismatch is tracked as $E_{\\text{mod-HF}}$ in the total error budget $E_{\\text{mod}}=E_{\\text{mod-HF}}+E_{\\text{mod-Q}}$.","core_discovery":"The paper's central discovery is an exact equivalence: if the folded signal $M_\\lambda x(t)$ is multiplied by a delta comb $p(t)=\\sum\\delta(t-nT_s)$ and then passed through a low-pass filter with cutoff $\\pi/T_s$ before sampling at rate $T_s$, the output samples satisfy $y[nT_s]=M_\\lambda x[nT_s]$. The proof uses the fact that multiplying by the comb periodizes the spectrum of the folded signal with period $2\\pi/T_s$; Poisson's formula identifies the DTFT of the desired samples with this periodized spectrum, and the filter keeps exactly the baseband copy. Consequently, the samples from a realistic ADC are identical to those an ideal wideband pointwise sampler would produce from the folded signal. The paper then decomposes the practical error into $E_{\\text{mod-HF}}$, caused by replacing the infinite comb with a finite Fourier sum of $N=2000$ terms, plus the quantization error $E_{\\text{mod-Q}}$, and shows experimentally that the comb brings $E_{\\text{mod-HF}}$ below the quantization error. A hardware prototype built from a step-recovery-diode comb generator, an analog multiplier, and a 25 kHz low-pass filter confirms the sample alignment and supports recovery with a 1-bit side-information algorithm.","pith_inferences":["The paper does not provide a theoretical bound on $E_{\\text{mod-HF}}$ as a function of $N$, signal bandwidth, and oversampling ratio; deriving one would turn the empirical choice $N=2000$ into a design rule and is a natural next step.","The same comb-then-filter identity only relies on periodization and baseband selection, so it likely carries over to shift-invariant and finite-rate-of-innovation modulo recovery, where the same high-frequency problem arises.","A simpler periodic waveform, such as a sine wave, could replace the full comb at the price of weighting the folded spectrum; the trade-off between hardware simplicity and spectral content is not explored in the paper.","Real low-pass filters have passband ripple and finite stopband rejection, so quantifying how filter nonideality translates into $E_{\\text{mod-HF}}$ is a testable extension of the hardware results."],"forward_implications":["Any existing modulo recovery algorithm for bandlimited signals, including quantization-aware ones, can be applied directly to the samples from a realistic ADC.","The ADC's bandwidth requirement drops back to the input signal's Nyquist rate; the analog mixer and filter absorb the high-frequency burden.","At oversampling rates of at least 5 and with 6- or 8-bit quantizers, the proposed pipeline approaches ideal modulo sampling and outperforms a classical infinite-dynamic-range ADC.","Because $E_{\\text{mod-HF}}$ and quantization error are independent, improving the comb's harmonic content is a direct route to better high-precision recovery.","A hardware prototype validates the theoretical equivalence, making unlimited-dynamic-range sampling viable with off-the-shelf components."],"supporting_citations":[{"why":"Introduces the residual recovery approach for bandlimited modulo sampling that the proposed system relies on.","marker":"[28]"},{"why":"ICASSP version of the residual recovery algorithm; establishes the baseline for recovering the signal from modulo samples.","marker":"[29]"},{"why":"Describes the analog modulo board used in the prototype and establishes the hardware context for wideband modulo sampling.","marker":"[32]"},{"why":"Provides the 1-bit side-information recovery method with quantization-noise guarantees used in the experiments.","marker":"[33]"},{"why":"Supplies the modulated wideband converter architecture that the mixer-plus-low-pass-filter sampling chain is inspired by.","marker":"[43]"},{"why":"Xampling implementation of the modulated wideband converter, the practical predecessor of the proposed analog front end.","marker":"[44]"},{"why":"Supplies the sampling-theory tools (CTFT/DTFT, Poisson's formula) used in the proof of Theorem 1.","marker":"[1]"},{"why":"Establishes the unlimited-sampling paradigm that the paper extends to realistic ADCs.","marker":"[15]"}],"fun_headline_variants":["Realistic ADCs match ideal modulo sampling via comb mixer","Comb mixer lets standard ADCs handle modulo sampling","Modulo sampling goes practical with comb and LPF","Hardware trick makes modulo sampling work with cheap ADCs","Folding signals no longer demand wideband ADCs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The practical equivalence rests on the unquantified assumption that replacing the delta comb by a finite Fourier sum with $N=2000$ harmonics makes $E_{\\text{mod-HF}}$ negligible; the paper supports this only with empirical curves in Figures 13 and 14, not with a theoretical bound.","fun_headline_variants_meta":{"raw":{"variants":["Realistic ADCs match ideal modulo sampling via comb mixer","Comb mixer lets standard ADCs handle modulo sampling","Modulo sampling goes practical with comb and LPF","Hardware trick makes modulo sampling work with cheap ADCs","Folding signals no longer demand wideband ADCs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000184,"raw_usage":{"total_tokens":1372,"prompt_tokens":1052,"completion_tokens":320,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":668,"completion_tokens_details":{"reasoning_tokens":243}},"tokens_in":668,"tokens_out":320,"duration_ms":3468,"temperature":1.0,"reasoning_tokens":243,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:23:20.967797+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure $E\\left[|y[nT_s]-M_\\lambda x[nT_s]|^2\\right]$ for the finite-comb system with $N=2000$ while increasing the input signal's maximum frequency or lowering the oversampling ratio at fixed bit depth; if this high-frequency error grows to dominate the quantization error and the recovered signal degrades well below the ideal-sampler baseline, the claimed practical equivalence fails.","supporting_citations":[{"cited_title":"Unlimited sampling beyond modulo,","cited_arxiv_id":null,"evidence_quote":"Introduces the residual recovery approach for bandlimited modulo sampling that the proposed system relies on."},{"cited_title":"Residual recovery algorithm for modulo sampling,","cited_arxiv_id":null,"evidence_quote":"ICASSP version of the residual recovery algorithm; establishes the baseline for recovering the signal from modulo samples."},{"cited_title":"A hardware prototype of wideband high-dynamic range analog- to-digital converter,","cited_arxiv_id":null,"evidence_quote":"Describes the analog modulo board used in the prototype and establishes the hardware context for wideband modulo sampling."},{"cited_title":"Modulo sampling with 1-bit side information: Performance guarantees in the presence of quantization,","cited_arxiv_id":null,"evidence_quote":"Provides the 1-bit side-information recovery method with quantization-noise guarantees used in the experiments."},{"cited_title":"From theory to practice: Sub-Nyquist sampling of sparse wideband analog signals,","cited_arxiv_id":null,"evidence_quote":"Supplies the modulated wideband converter architecture that the mixer-plus-low-pass-filter sampling chain is inspired by."},{"cited_title":"Xampling: Analog to digital at sub-Nyquist rates,","cited_arxiv_id":null,"evidence_quote":"Xampling implementation of the modulated wideband converter, the practical predecessor of the proposed analog front end."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the sampling-theory tools (CTFT/DTFT, Poisson's formula) used in the proof of Theorem 1."},{"cited_title":"On unlimited sampling and reconstruction,","cited_arxiv_id":null,"evidence_quote":"Establishes the unlimited-sampling paradigm that the paper extends to realistic ADCs."}],"review_version":1}