Pith. sign in

REVIEW 4 major objections 6 minor 24 references

A Gridless Compressive Sensing Based Channel Estimation for Millimeter Wave MIMO OFDM Systems with One-Bit Quantization

T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proposes a gridless atomic-norm method for estimating sparse mmWave MIMO-OFDM channels from one-bit quantized measurements.

desk verdict The proposed sign-consistency constraints do not follow from the paper's own one-bit model, and the error is central enough to sink the recovery claim. read the letter →

arxiv 1911.06700 v1 pith:YYSCTH5Q submitted 2019-08-19 eess.SP

Mahdi Eskandari , Hamidreza Bakhshi This is my paper · ORCID
classification eess.SP
keywords atomicnormminimizationone-bitquantizationmillimeterwaveMIMOOFDMchannelestimationgridlesscompressivesensingreweightedmassivesparse
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

One-bit analog-to-digital converters are the cheapest and least power-hungry way to digitize the very wide bandwidths used in millimeter-wave communication, but a one-bit receiver discards all magnitude information, and previous compressive-sensing estimators put the channel's angles and delays on a fixed grid, so a channel that falls between grid points is estimated with bias. This paper proposes a gridless estimator, Binary Atomic Norm Minimization (BiANM), and a reweighted version, ReBiANM, that recover the normalized sparse mmWave MIMO-OFDM channel directly from one-bit measurements without any grid. The central claim is that atomic-norm minimization, formulated as a semidefinite program, can enforce the sign constraints imposed by one-bit quantization while exploiting the continuous angle-delay sparsity of the channel, and that reweighting makes the estimate more accurate. If correct, the method gives low-power one-bit receivers a channel-estimation path that avoids grid mismatch while keeping the estimation problem convex.

What carries the argument

The load-bearing object is the atomic norm over the atom set $\mathcal{A} = \{\mathbf{a}(\omega) : \omega \in [0,1]^2\}$, where $\mathbf{a}(\omega)$ is the Kronecker product of an angle steering vector and a delay Vandermonde vector; the atomic norm is the convex hull of these atoms and is evaluated by a semidefinite program involving a two-level Hermitian Toeplitz matrix $T_{\mathrm{2D}}(\mathbf{u})$. BiANM minimizes this norm subject to the sign-consistency inequalities $\Re(\mathbf{R}'\mathbf{\Phi}\mathbf{h})\ge 0$ and $\Im(\mathbf{R}'\mathbf{\Phi}\mathbf{h})\ge 0$ elementwise, plus the unit-norm constraint $\|\Re(\mathbf{\Phi}\mathbf{h})\|^2 + \|\Im(\mathbf{\Phi}\mathbf{h})\|^2 = 1$ that prevents the zero solution. ReBiANM changes the objective to $\operatorname{trace}(\mathbf{\Theta}_j T_{\mathrm{2D}}(\mathbf{u}))$ with $\mathbf{\Theta}_j$ updated from the previous iterate, which lets the solver move between $\ell_0$-style sparsity and $\ell_1$-style convexity. The two-level Toeplitz Vandermonde decomposition is what turns continuous parameter recovery into a finite-dimensional semidefinite program.

What would settle it

A concrete test is to take a fixed sparse channel, generate noiseless one-bit signs, then deliberately flip a known fraction of them (or add noise until that fraction flips), and run BiANM and ReBiANM against the true normalized channel: if the normalized error jumps abruptly at any nonzero flip fraction, or if removing the majority-vote oversampling destroys low-SNR recovery, the sign-consistency assumption is doing load-bearing work that the paper does not justify.

Watch

Extended reading notes

Core claim

The paper's central claim is that one-bit channel estimation can be posed as a gridless convex optimization: replace the non-convex atomic $\ell_0$ norm by its convex envelope, the atomic norm, and add elementwise sign-consistency constraints that require each quantized measurement to agree with the sign of the corresponding noiseless linear measurement. In this formulation, named BiANM, the channel is estimated as $\hat{\mathbf{h}}$ from $\mathbf{r} = \operatorname{sign}(\mathbf{\Phi}\mathbf{h}+\mathbf{n})$ with $\mathbf{\Phi} = (\mathbf{X}^T\otimes \mathbf{I}_M)$; because the atoms are continuous in angle and delay, no grid is introduced. ReBiANM iterates the same semidefinite program with a weight $\mathbf{\Theta}_j$ built from the previous solution, interpolating between $\ell_0$-like sparsity enhancement and the convex $\ell_1$-like atomic norm. The paper states, and supports by Monte-Carlo simulation, that both methods estimate the channel normalized to unit norm, with ReBiANM improving accuracy.

Load-bearing premise

The method assumes the recorded one-bit signs are consistent with the true noiseless sign pattern, so the sign-consistency constraints in the optimization remain feasible, even though the measurement model includes noise.

Editorial extensions

If this is right

  • Grid mismatch is removed as a source of error: angles and delays are recovered on the continuum rather than from a finite dictionary, so a path falling between grid points no longer biases the estimate.
  • One-bit receivers can support channel estimation at low power, because the estimator needs only the sign of each measurement instead of high-resolution samples.
  • ReBiANM is claimed to give more accurate normalized channel estimates than plain BiANM, with the paper's simulations showing the largest gains at low SNR where noise flips signs.
  • Because the optimization is a semidefinite program, the method has a tractable convex solver despite the NP-hardness of the original atomic $\ell_0$ problem.
  • The recovered channel is normalized to unit norm, so tasks that depend on channel direction, such as beam alignment, can proceed without amplitude recovery.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the hard sign constraints could be replaced by a probabilistic sign-flip model to make the estimator robust to noisy measurements; the paper's majority-vote oversampling is an ad hoc step toward this but no probabilistic model is given.
  • Editorial inference: the two-level Toeplitz semidefinite-program structure is not specific to one-bit quantization, so a similar gridless formulation could in principle handle other quantized or nonlinear measurement models once the constraints are adapted.
  • Editorial inference: because the output is only the normalized channel, a real link would need a separate amplitude-recovery stage to obtain channel gains; the paper leaves that open.
  • Editorial inference: a direct comparison with a grid-based one-bit compressive sensing method at matched complexity would separate the benefit of going gridless from the benefit of reweighting, since the paper's simulations compare only gridless variants.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This paper considers channel estimation for millimeter-wave massive MIMO OFDM systems with one-bit ADCs. The received signal is modeled as r = sign(Φh + n), where h is a sparse channel vector composed of a small number of propagation paths. The authors propose a gridless method, Binary Atomic Norm Minimization (BiANM), which estimates the normalized sparse channel by minimizing an atomic norm subject to sign-consistency constraints, and a reweighted version (ReBiANM). They claim that the method avoids the grid mismatch of prior grid-based one-bit compressive sensing approaches. The only simulation, Fig. 1, compares BiANM with ReBiANM, with and without a majority-vote oversampling scheme.

Significance. If the method were correct, a gridless one-bit channel estimator for mmWave MIMO OFDM would be a useful contribution to the low-resolution ADC literature. The paper, however, does not provide a valid derivation of the optimization constraints, and the experimental evidence is limited to a self-comparison of the proposed variants. The central recovery claim is therefore unsupported, and the paper's contribution is limited to a suggestion that such a technique might be possible.

major comments (4)
  1. [Section III, Eqs. (9) and (11)] The sign-consistency constraints are not valid under the stated measurement model. The paper argues that Re(R' Φ h) ≥ 0 and Im(R' Φ h) ≥ 0 follow from the product of each quantized measurement with the measurement being nonnegative, but this is false for the model r = sign(Φh + n). With additive noise, sign(Φh + n)_i can differ from sign(Φh)_i whenever noise changes the quadrant of the argument. Even in the noiseless case, the paper's own definition of sign (applied componentwise to real and imaginary parts) does not imply nonnegativity of those products; for example, z = 1 + 3j gives sign(z) = 1 + j and Re(sign(z) z) = -2. Hence the true normalized h need not satisfy the constraints, the feasible set may be empty, and the BiANM recovery claim is unsupported.
  2. [Section IV, oversampling paragraph] The majority-vote oversampling procedure does not repair the sign-consistency issue. With finite SNR, sign(Φh + n) disagrees with sign(Φh) at a nonzero fraction of coordinates; a majority vote over five noise realizations reduces but does not eliminate these disagreements, and the majority-voted r is still not guaranteed to be consistent with any h. The paper provides no analysis of the residual probability of infeasibility, so the proposed workaround does not restore the validity of the constraints.
  3. [Section IV, Fig. 1] The only reported experiment compares BiANM with ReBiANM under two oversampling conditions, but it does not compare against any existing grid-based one-bit compressive sensing method, a classical least-squares estimator, or a lower bound. Without such baselines, the NMSE curves cannot be interpreted as evidence of accuracy, and the claimed advantage over grid-based approaches is untested. In addition, the number of Monte-Carlo runs is not stated and no error bars are provided, making the reported curves unverifiable.
  4. [Section III, Eqs. (10)-(11)] The equivalence between the atomic norm of the two-level Vandermonde channel and the SDP (10) is asserted but not established. The paper neither defines T_2D(u) explicitly nor verifies that the two-level Toeplitz/SDP conditions hold for the channel vector h in (4). Since the convex reformulation is the foundation of both BiANM and ReBiANM, this gap leaves the numerical solver and the claim of convexity unsubstantiated.
minor comments (6)
  1. [Abstract] The phrase 'recover sparse channel form one-bit measurements' should read 'from one-bit measurements'.
  2. [Introduction] The first paragraph contains typographical errors, including 'scalating demand', 'etracted', and 'trigged'; a careful proofread is needed.
  3. [Section II and Section IV] Section II presents the uplink scenario, whereas Section IV describes the simulation as 'downlink mmWave channel estimators'; the system model should be consistent.
  4. [References] Reference [10] is cited with a year of 2007 for an Asilomar conference paper on one-bit quantization, which appears to be incorrect; reference [12]'s title also seems unrelated to the channel-estimation topic. Please verify the references.
  5. [Section IV, Eq. (12)] The definition of the reweighting matrix Θ_j is garbled and the label 'BiReANM' is inconsistent with the text's 'ReBiANM'; please clarify the update rule and unify the notation.
  6. [Section IV, simulation setup] The simulation section states results are averaged over 'many independent realizations' but does not specify the number of realizations or error bars; please provide these for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: BiANM/ReBiANM are assembled from published atomic-norm SDP machinery, and the simulation data are independent of the method's fitted parameters.

full rationale

The derivation chain is not circular. The BiANM program in equations (9) and (11) is a convex optimization formed from externally published atomic-norm SDP results (references [20]-[23]) plus one-bit sign constraints; it is not obtained by fitting to the simulation outputs. The ReBiANM reweighting schedule (zeta = 1 halved across J = 5 iterations) is a heuristic initialization stated before the experiments, not a parameter calibrated on the tested channels, so the reported NMSE is not a renamed fit. There are no self-citations and no uniqueness theorem imported from the authors. The paper's main caveat is the sign-consistency constraints in equations (9) and (11): the claim that 'the product of each quantized measurement with the measurement is always non-negative' is not valid under the complex componentwise sign model with noise, so the true normalized channel can be infeasible. That is a correctness and modeling limitation, not a circularity: the optimization does not define its output as equal to an input by construction, and the synthetic measurements are generated from an independent channel-plus-noise model. The majority-vote oversampling in Section IV is an acknowledged workaround for noisy measurements, and it likewise does not make the estimator predict its own fitted values. Overall, no load-bearing step reduces to its own inputs, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the sparse channel model, the SDP representation of the atomic norm, and the validity of sign-consistency constraints in noise. Only the first two are supported by prior literature; the third is an ad hoc assumption. The algorithm introduces two hand-tuned parameters (zeta schedule and oversampling count) but no fitted constants.

free parameters (3)
  • zeta (reweighting regularization parameter) = initialized to 1, halved at each of 5 iterations
    Controls the tradeoff between atomic 0-norm and 1-norm in ReBiANM; chosen by hand, not derived.
  • Number of oversampling samples = 5
    Majority vote across 5 noisy measurements is used to handle noise, chosen without justification.
  • Iteration count J = 5
    Number of reweighting iterations in ReBiANM, chosen arbitrarily.
assumptions (3)
  • domain assumption The channel is sparse with a small number of continuous AoAs and delays
    The entire sparse recovery framework relies on this, stated in Section II equation (1).
  • standard math The two-level Hermitian Toeplitz matrix T_2D(u) characterizes the atomic norm for the vectorized MIMO-OFDM channel
    Invoked from [22] and [23], but not proven for this problem; equation (10) uses it without verifying conditions.
  • ad hoc to paper Sign measurements are consistent with the noiseless channel output, r_i (Phi h)_i >= 0
    Used to build constraints (9) and (11); not valid under noise and not statistically justified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Gridless Compressive Sensing Based Channel Estimation for Millimeter Wave MIMO OFDM Systems with One-Bit Quantization." pith.science (2026). https://pith.science/paper/YYSCTH5Q

@misc{pith2026191106700,
  author       = {Pith},
  title        = {Pith review of: A Gridless Compressive Sensing Based Channel Estimation for Millimeter Wave MIMO OFDM Systems with One-Bit Quantization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YYSCTH5Q}},
  note         = {Machine review of arXiv:1911.06700}
}
read the original abstract

This paper considers the problem of estimating the sparse mmWave massive multiple input - multiple output (MIMO) OFDM channel from one-bit quantized measurements. Unlike previous grid-based one-bit compressive sensing approaches, we present a gridless convex method to recover sparse channel form one-bit measurements via Binary Atomic Norm Minimization (BiANM) and Reweighted Binary Atomic Norm Minimization (ReBiANM). Simulation results verify the accuracy of the binary and reweighted binary atomic norm minimization techniques.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 23 canonical work pages

  1. [1]

    Massive hybrid antenna array for millimeter -wave cellular communications,

    J. A. Zhang, X. Huang, V. Dyadyuk, and Y. J. Guo, “Massive hybrid antenna array for millimeter -wave cellular communications,” IEEE Trans. Wireless Commun., vol. 22, no. 1, pp. 79–87, Feb. 2015

    work page 2015

  2. [2]

    V. W. S. Wong, R. Sch ober, D. W. K. Ng, and L. -C. Wang, Key Technologies for 5G Wireless Systems. Cambridge Univ. Press, 2017

    work page 2017

  3. [3]

    Massive multiuser MIMO in heterogeneous cellular networks with full duplex s mall cells,

    S. Akbar, Y. Deng, A. Nallanathan, M. Elkashlan, and G. K. Karagiannidi, “Massive multiuser MIMO in heterogeneous cellular networks with full duplex s mall cells,” IEEE Trans. Commun., vol. PP, no. 99, pp. 1–1, 2017

    work page 2017

  4. [4]

    What will 5G be?

    J. Andrews, S. Buzzi, C. Wan, S. Hanly, A. L. A. Soong, and J. Zhang, “What will 5G be?” IEEE J. Select. Areas Commun ., vol. 32, no. 6, pp. 1065–1082, Jun. 2014

    work page 2014

  5. [5]

    Energy limits in A/D converters,

    E. B. Murmann,“Energy limits in A/D converters,” in Proc. IEEE Faible Tension Faible Consommation, , Jun. 2013, pp. 1–4

    work page 2013

  6. [6]

    B. Le, T. Rondeau, J. Reed, and C. Bostian, ”Analog -to-digital converters,” IEEE Signal Process. Mag., vol. 22, no. 6, pp. 69 –77, Nov. 2005

    work page 2005

  7. [7]

    R. H. Walden, ”Analog -to-digital converter survey and an alysis,” IEEE J. Sel. Areas Commun, vol. 17, no. 4, pp. 539–550, Apr. 1999

    work page 1999

  8. [8]

    C. Risi, D. Persson, and E. G. Larsson, ”Massive MIMO with one-bit ADC,” arXiv preprint arXiv:1404.7736, Apr. 30, 2014

    arXiv 2014

Show all 24 references
  1. [9]

    S. Wang, Y. Li, and J. Wan g, ”Multiuser detection in massive spatial modulation MIMO with low -resolution ADCs,” IEEE Trans. Wireless Commun., vol. 14, no. 4, pp. 2156–2168, Apr. 2015

    2015

  2. [10]

    J. Mo, P. Schniter, N. G. Prelcic, and R. W. Heath, Jr. ”Channel estimation in millimeter wave MIMO systems with one-bit quantization,” in Proc. Asilomar Conf. Signals, Syst. Comput.,Feb. 2007, pp. 1–7

    2007

  3. [11]

    M. T. Ivrlac and J. A. Nossek, ”On MIMO channel estimation with singlebit signal-quantization” in Proc. ITG Workshop Smart Antennas , Aug. 2015, pp. 2909–2913

    2015

  4. [12]

    Mezghani, F

    A. Mezghani, F. Antreich, and J. Nossek, ”Compressive twodimensional harmonic retrieval via atomic norm minimization,” in Proc. ITG Workshop Smart Antennas, 2010, pp. 143–150

    2010

  5. [13]

    C. K. Wen, C. J. Wang, S. Jin, K. K. Wong, and P. Ting, ”Bayes-optimal joint channel-and-data estimation for massive MIMO with low-precision ADCs,” IEEE Trans. Signal Process , vol. 64, no. 10, pp. 2541 –2556, 2015

    2015

  6. [14]

    H. Wang, T. Liu, C. K. Wen, and S. Jin, ”Optimal data detection for OFDM system with low -resolution quantization,” in Proc. IEEE Int. Conf. Commun. Syst., Dec. 2016, pp. 1–6

    2016

  7. [15]

    T. C. Zhang, C. K. Wen, S. Jin, and T. Jiang, ”Mixed -ADC massive MIMO detectors: Performance analysis and design optimization,” IEEE Trans. Wireless Commun., vol. 15, no. 11, pp. 7738–7752, Nov. 2016

    2016

  8. [16]

    J. Mo, P. Schniter, and R. W. Heath, Jr., ”Channel Estimation in Broadband Millimeter Wave MIMO Systems With Few -Bit ADCs,” IEEE Trans. Signal Process., vol. 66, no. 5, Feb. 2018

    2018

  9. [17]

    Adaptive One -bit Compressive Sensing with Application to LowPrecision Receivers at mmWave,

    C. Rusu, R. Mendez -Rial, N. Gonzalez-Prelcic, and R. W. Heath, “Adaptive One -bit Compressive Sensing with Application to LowPrecision Receivers at mmWave,”in 2015 IEEE Global Communications Conference (GLOBECOM), Feb. 2016. pp. 1-6

    2015

  10. [18]

    Z. Chen, C. Yang, ”Pilot Decontamination in Wideband Massive MIMO Systems by Exploiting Channel Sparsity,” IEEE Trans. Wireless Commun., vol. 15, no. 7, Feb. 2016

    2016

  11. [19]

    C. Zhou, Z. Zhang, F. Liu, and B. Li, ”Gridless compressive sensing method for line spectral estimation from one-bit measurements,” to appear in Digit. Signal Process., 2016

    2016

  12. [20]

    Chandrasekaran, B

    V. Chandrasekaran, B. Recht, P. A. Parrilo, a nd A. S. Willsky, ”The convex geometry of linear inverse problems,” Foundations of Computational mathematics, vol. 12, no. 6, pp. 805-849, 2012

    2012

  13. [21]

    Yang and L

    Z. Yang and L. Xie, ”Exact Joint S parse Frequency Recovery via Optimization Methods,” arXiv preprint arXiv: 1405.6585

    arXiv

  14. [22]

    G. Tang, B. Bhaskar, P. Shah, and B. Recht,” Compressed sensing off the grid,” IEEE Trans. Inf. Theory, vol. 59, no. 11, pp. 7465 -7490, Nov 2013

    2013

  15. [23]

    Zai Yang, Lihua Xie, and Petre Stoica, ”Vandermonde decomposition of multilevel Toeplitz matrices with application to multidimensional superresolution,” IEEE Trans. Inf. Theory, vol. 62, no. 6, pp. 3685-3701, 2016

    2016

  16. [24]

    Yang and L

    Z. Yang and L. Xie,” Enhancing Sparsity and Resolution via Reweighted Atomic Norm Minimization,” IEEE Trans. Signal Process., vol. 64, no. 4, Feb. 2016

    2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.