{"id":"b540bdc4-4754-4a12-998a-d94dd9125794","arxiv_id":"2502.05016","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Using the far-field model in the near-field under-estimates MUSIC DOA errors, while near-field MUSIC estimates range accurately only below the Fraunhofer distance.","lead":"This paper simulates how wrong assumptions about near-field versus far-field propagation affect MUSIC direction-of-arrival estimation. It finds that mistakenly using the simpler far-field model in the near-field produces large errors and, worse, makes the estimator appear more accurate than it is.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reported Fraunhofer distances cannot be reproduced from the stated antenna count and frequency; unspecified array geometry may invalidate the claimed near-field/far-field boundary and the range-accuracy result.","rationale":"The paper's central claims are simulation-based, and the primary quantitative results are comparisons of MUSIC error across distances and models. In stress-testing, I first looked for a logical flaw in the 'underestimation' argument: the paper compares hFF with far-field data (zero error) to hFF with near-field data (large error) and calls this 'underestimation.' While the wording is loose, the intended practical point—that evaluating MUSIC under a wrong propagation model gives an overly optimistic error—is clear and not internally inconsistent. The more serious, concrete problem is that the simulation setup is under-specified in a way that directly undermines the Fraunhofer-distance-based narrative. The reported df values cannot be derived from NU and frequency without the array aperture D, and standard array geometries give df values that differ by orders of magnitude. The text even misstates the 80 GHz wavelength as 3.7 cm instead of 3.75 mm. If the actual array is not a standard half-wavelength array, then the Fraunhofer boundary used to separate 'near' and 'far' regimes is arbitrary, and the claim that range accuracy degrades beyond df is not a general result. This concern is load-bearing because it affects both the DOA and range conclusions. The reader's weakest assumption (missing simulation details) is correct but under-specified; I sharpen it to the specific unresolved issue of array geometry and Fraunhofer distance. A concrete analytical check can settle this without new experiments: solve for D from the reported df and test whether any physically reasonable NU-element array can realize it. If it cannot, the numerical results are not reproducible, and the paper should be revised to report the array layout and all simulation parameters before acceptance.","tokens_in":7639,"tokens_out":16606,"duration_ms":172451,"concrete_test":"Re-derive the array aperture D from df = 2D^2/λ for each reported (NU, fc) pair, then check whether any standard geometry (ULA, UPA, UCA) with NU elements and physically plausible spacing (λ/4 to λ) yields that D. For NU=144, fc=3 GHz, df=14.4 m implies D=0.849 m. Test a 12×12 uniform planar array at half-wavelength spacing: D=11*(λ/2)=0.55 m, df=6.05 m, a factor of 2.4 below the reported value. If no geometry matches within 10%, the reported Fraunhofer distances are implausible and the boundary-dependent conclusions (Figs. 1–5) must be re-evaluated with the actually used array.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central numerical claims—(i) that DOA error is 'underestimated' when the far-field model is used in the near-field, and (ii) that range estimates are accurate only inside the Fraunhofer distance—both depend on the reported Fraunhofer distances (df = 14.4 m for NU=144 at fc=3 GHz, df = 0.54 m for NU=144 at fc=80 GHz). However, df = 2D^2/λ requires the array aperture D, which is never specified. The paper states only NU and frequency. For a standard half-wavelength ULA with 144 elements, D = 143*(λ/2) = 7.15 m at 3 GHz, giving df = 1022 m, not 14.4 m. To obtain df = 14.4 m, D must be 0.849 m, implying an inter-element spacing of about 0.77λ in a 12×12 planar array—an unusual geometry that is not described. Likewise, the 80 GHz case states λ = 3.7 cm, but fc=80 GHz implies λ = 3.75 mm; the reported df = 0.54 m is consistent with the millimeter wavelength, not 3.7 cm. Without the array layout, the test distances (0.2–30 m) cannot be classified as inside or outside the near-field region, so the claimed transition at the Fraunhofer distance and the 'underestimation' factor are not reproducible from the information given. This is not a mere formatting issue: if a conventional ULA had been used, all test distances would lie in the near-field, and the observed error growth beyond 10 m would not correspond to a far-field effect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a simulation study of MUSIC-based DOA and range estimation under model mismatch between the assumed propagation regime and the actual one. The authors consider four combinations: near-field received signal with near-field, approximate near-field (ANM), and far-field beamforming models, plus the matched far-field case. Simulations are run for fc=3GHz and fc=80GHz, with NU=144 or 256 antennas, SNR=30dB (10dB for the range figure), and source distances from 0.2m to 30m. The main claims are that using the far-field model in the near-field causes a performance drop, that incorrectly assuming far-field propagation leads to an underestimation of the DOA estimation error, and that near-field MUSIC gives accurate range estimates inside the Fraunhofer distance but degrades beyond it.","tokens_in":7933,"tokens_out":3453,"duration_ms":36060,"significance":"If the results are robust, this is a useful cautionary study for 6G localization: it demonstrates that model mismatch in MUSIC is not benign, and it identifies the Fraunhofer distance as a meaningful boundary for range estimation. A clear strength is that the paper contains no fitted parameters; the plotted curves are direct outputs of the MUSIC algorithm on synthetic data, so the qualitative trends are not obtained by curve fitting. However, the quantitative claims are currently limited by missing reproducibility details (array geometry, search-grid resolution, Monte Carlo trial counts), and the stated Fraunhofer distances cannot be reconstructed from the parameters given. The work is therefore a promising and relevant contribution, but it needs additional reporting before the quantitative conclusions can be accepted.","major_comments":[{"comment":"The Fraunhofer distances reported in the paper are not reproducible from the stated parameters (NU and fc) because the array aperture D is never specified. With df = 2D^2/λ, the value df = 14.4 m for NU=144 at fc=3GHz implies D ≈ 0.849 m, whereas a conventional half-wavelength ULA with 144 elements would have D = 7.15 m and df ≈ 1022 m, placing all test distances (0.2-30 m) in the near-field. Since the classification of test distances as inside or outside the near-field drives the central claims of a transition at df and of an 'underestimated' DOA error, the array geometry must be stated explicitly so that the simulation setup can be reproduced and the claimed boundary verified.","section":"Section IV, Figs. 1-4"},{"comment":"The simulation results are presented as single curves with no error bars, no stated number of Monte Carlo trials, and no sensitivity analysis over array geometry or MUSIC search-grid resolution. The abstract and conclusion say that the loss in performance is 'quantified,' but with only one deterministic-looking trace per configuration, the numerical magnitudes — including the amount of error underestimation — are not yet supported. Please report the number of trials, the grid spacings for azimuth, elevation, and range, and at least one error-bar or percentile representation so that the quantitative strength of the claims can be assessed.","section":"Section IV, Figs. 1-5"},{"comment":"The text states that the operating wavelength is 3.7 cm for fc = 80 GHz, but fc = 80 GHz corresponds to λ = 3.75 mm, and the reported Fraunhofer distance df = 0.54 m for NU = 144 is consistent with the millimetre wavelength, not with 3.7 cm. This inconsistency must be corrected because the reader cannot otherwise verify the Fraunhofer-distance calculations that underlie the near-field/far-field classification in those figures.","section":"Section IV, Figs. 3-4"}],"minor_comments":[{"comment":"In the definition of the far-field array response, the subscript Δ_ku should be Δ_kU for consistency with the rest of the equation; the current notation mixes the element index u with the destination index U.","section":"Section II, Eq. (2)"},{"comment":"The phrase 'maximum diameter among the source and destination surface diameters' is imprecise for an antenna array. The diameter D in df = 2D^2/λ should be defined explicitly, for example as the maximum distance between any two antenna elements of the receive array.","section":"Section II, Fraunhofer distance definition"},{"comment":"The range-estimation figure would be more informative if the paper stated the range search grid spacing and the number of candidate range values used in the near-field MUSIC search, since the accuracy of the range estimate can depend on that discretization.","section":"Section IV, Fig. 5"}],"recommendation":"major_revision","confidential_remarks":"The topic is appropriate for the journal and the central qualitative findings are plausible. The main barrier to acceptance is reproducibility: the array geometry, search grid, and Monte Carlo trial counts are missing, and the Fraunhofer distances cannot be checked without them. The wavelength typo at 80 GHz should also be fixed. I see no novelty or citation concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Julian—quick read of arXiv:2502.05016. The useful core is the question: what happens when MUSIC uses the wrong propagation model? The paper gives simulation curves showing that far-field processing in the near-field underestimates DOA error, and that near-field processing gives accurate range only inside the Fraunhofer distance. That is a practical issue for 6G localization, and the qualitative trends are plausible.\n\nWhat's new: prior work on near/far model mismatch (refs [2] and [8]) looks at spectral efficiency or broader implications; this is a MUSIC-specific quantification with ANM and far-field beamforming matrices. It is entirely simulation-based and the paper is honest about that. No claim of new theory or algorithm.\n\nThe soft spots are real and load-bearing. The Fraunhofer distances in the figures do not follow from the stated antenna counts and frequencies. For NU=144 at 3 GHz, df=14.4 m implies an array aperture D=0.849 m, which is not described and does not match a half-wavelength ULA (which would give df≈1000 m). At 80 GHz the text says λ=3.7 cm, which is off by a factor of ten; that alone changes df by 100x. Without the array geometry, the classification of test distances as near or far is meaningless, and the claimed transition at the Fraunhofer distance is not reproducible. The paper also reports no error bars, trial counts, or search-grid spacing, so I have no idea how much the curves depend on implementation.\n\nThat said, the central qualitative claim—mismatch hurts, and assuming far-field hides the error—is almost certainly correct and consistent with the prior literature. The paper is not conceptually confused; it has a clear setup and the simulations support the broad picture. But a load-bearing numerical inconsistency like this needs to be fixed before the numbers can be trusted.\n\nWho this is for: engineers choosing between near- and far-field processing in 6G arrays. They'd get general guidance, not exact thresholds. Does it deserve a serious referee? Yes—it is relevant and the flaws look fixable, but the review must demand the missing simulation details and corrected wavelength. If I were the editor, I'd send it to review and expect a major revision.","headline":"Quantifies a real practical risk in MUSIC model mismatch, but the central numbers don't reproduce: the reported Fraunhofer distances and 80 GHz wavelength are internally inconsistent.","tokens_in":8414,"tokens_out":3507,"would_cite":false,"duration_ms":37319,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"When a receiver in the near-field wrongly assumes far-field propagation, MUSIC underestimates its own direction-of-arrival error, even as true accuracy degrades.","keywords":["MUSIC","direction-of-arrival estimation","near-field","far-field","model mismatch","Fraunhofer distance","range estimation","6G localization"],"falsifier":"Run the same mismatch experiment with a fully documented setup: a half-wavelength-spaced uniform linear or planar array, at least 10,000 Monte Carlo trials per distance, and a fine search grid (for example 0.01 degrees in angle and 0.01 m in range). If below the Fraunhofer distance the DOA root-mean-square error using the far-field beamformer does not exceed the error the far-field model predicts for itself, or if the near-field range error does not rise steeply beyond $2D^2/\\lambda$, the paper's central claims would be contradicted.","tokens_in":7467,"feed_emoji":"📡","tokens_out":8566,"duration_ms":82573,"temperature":0.7,"pith_summary":"This paper asks what happens when a receiver in the near-field of a transmitter runs MUSIC, a standard direction-of-arrival estimator, using a beamforming model meant for the far-field. It reports, from simulations, that the mismatch degrades DOA accuracy, and worse, that the far-field model's own error estimate is too optimistic, so an operator would underestimate how badly the algorithm is doing. It also reports that MUSIC with the full near-field model estimates source range very accurately when the true distance is below the Fraunhofer distance, but that accuracy collapses beyond it. The approximate near-field model sits in between: better than far-field, worse than exact near-field. The reason this matters is that 6G systems with many antennas have enlarged near-field regions, so the regime assumption is no longer automatically safe.","feed_headline":"Wrong propagation model hides true direction-finding error","feed_subtitle":"Simulations show MUSIC's far-field model underestimates its own DOA error inside the near-field region.","key_machinery":"The machinery is the MUSIC (multiple signal classification) algorithm run with a beamforming matrix built from three different assumed models: the exact near-field model, an approximate near-field model (a parabolic-wave approximation of the spherical wavefront), and the far-field model (a planar wavefront). MUSIC estimates directions by scanning candidate array-response vectors and selecting those that are most nearly orthogonal to the noise subspace of the received-signal covariance matrix; the orthogonality test is only as good as the assumed model that generates those vectors. The paper's analysis of mismatch is organized around the Fraunhofer distance, $d_f = 2D^2/\\lambda$, with $D$ the array diameter and $\\lambda$ the wavelength, which marks where the near-field spherical wavefront begins to look planar to the array. When the receiver is inside this distance, the far-field beamforming vector is the wrong probe, and the paper's simulations show the consequences for DOA and range estimates.","core_discovery":"The paper's central claim is that feeding MUSIC a propagation model that does not match the actual propagation regime produces a specific, quantifiable performance drop rather than a generic degradation. In its simulations, when the receiver is in the near-field, the MUSIC beamformer built from the far-field planar-wave model yields large DOA error at short distances, and this error only falls to the level of the exact near-field model once the distance exceeds the Fraunhofer distance. At the same time, using the far-field beamformer against the far-field signal model gives essentially perfect DOA estimates, which is why the paper concludes that an incorrect far-field assumption underestimates the DOA estimation error: the algorithm's own error measure looks small while the true error against the actual near-field signal is large. For range, the exact near-field model estimates distance almost perfectly below the Fraunhofer distance and degrades sharply above it, while the approximate near-field model is less accurate even inside that distance and the far-field model cannot estimate range at all.","pith_inferences":["The 'underestimation' result implies a practical hazard the paper does not discuss: a near-field localization system that self-reports uncertainty from the far-field model will be overconfident, which matters for scenarios like autonomous navigation where a wrong angle with a small reported error can drive the wrong control action.","Because the simulation details behind the error magnitudes (array geometry, number of trials, search-grid resolution) are not reported, the qualitative direction of the results is likely robust, but the specific factor of underestimation should be treated as provisional until replicated with a documented grid.","The sharp boundary behavior around the Fraunhofer distance suggests a testable design rule: a receiver could compare the MUSIC spectrum residual under the near-field and far-field models and use the model with the smaller residual as a regime detector; the paper's data show the two models are distinguishable inside the near-field.","For 6G, the results imply that range estimation via MUSIC should carry a validity flag based on the Fraunhofer distance, since the algorithm itself does not indicate when its range output stops being reliable."],"forward_implications":["Inside the near-field, an operator who assumes far-field propagation and uses the far-field MUSIC beamformer will see a small reported error while the true DOA error is large, so the far-field model cannot be trusted to certify localization accuracy.","The approximate near-field model improves on the far-field model but still loses accuracy in the near-field; it only becomes reliable for DOA at distances beyond the Fraunhofer distance.","Range estimation with the exact near-field model is essentially perfect inside the Fraunhofer distance and degrades sharply outside it, so range estimates should be treated as regime-dependent.","At distances greater than the Fraunhofer distance, the far-field received-signal model matches the near-field model's performance for DOA, confirming that the simpler model is sufficient there.","The exact near-field model used in the near-field delivers perfect DOA estimates in the simulations, matching the far-field model's performance in the far-field."],"supporting_citations":[{"why":"Supplies the Fraunhofer distance definition used to separate the near-field and far-field regions in the simulations.","marker":"[5]"},{"why":"Provides the reduced-dimension MUSIC approach for joint DOA and range estimation that the paper's near-field beamforming search builds on.","marker":"[9]"},{"why":"Establishes near/far-field model mismatch as an open issue for 6G localization and sensing, which is the problem this paper addresses.","marker":"[2]"},{"why":"Shows spherical-wavefront model mismatch affects spectral efficiency, the prior mismatch result the paper extends to direction-of-arrival estimation.","marker":"[8]"}],"fun_headline_variants":["MUSIC's far-field model hides true near-field error","Wrong propagation model underestimates MUSIC's DOA error","Near-field mismatch makes MUSIC overconfident in DOA","Fraunhofer boundary marks MUSIC's range accuracy limit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative error curves rest on simulation details the paper does not report—the antenna array geometry, the number of Monte Carlo trials, and the angle and range search-grid spacings—so the size of the predicted error drop is not yet shown to be robust.","fun_headline_variants_meta":{"raw":{"variants":["MUSIC's far-field model hides true near-field error","Wrong propagation model underestimates MUSIC's DOA error","Near-field mismatch makes MUSIC overconfident in DOA","Fraunhofer boundary marks MUSIC's range accuracy limit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000265,"raw_usage":{"total_tokens":1669,"prompt_tokens":1069,"completion_tokens":600,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":685,"completion_tokens_details":{"reasoning_tokens":530}},"tokens_in":685,"tokens_out":600,"duration_ms":6757,"temperature":1.0,"reasoning_tokens":530,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T20:33:56.439194+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same mismatch experiment with a fully documented setup: a half-wavelength-spaced uniform linear or planar array, at least 10,000 Monte Carlo trials per distance, and a fine search grid (for example 0.01 degrees in angle and 0.01 m in range). If below the Fraunhofer distance the DOA root-mean-square error using the far-field beamformer does not exceed the error the far-field model predicts for itself, or if the near-field range error does not rise steeply beyond $2D^2/\\lambda$, the paper's central claims would be contradicted.","supporting_citations":[{"cited_title":"RIS-Aided Localization under Position and Orientation Offsets in the Near and Far Field","cited_arxiv_id":"2210.03599","evidence_quote":"Supplies the Fraunhofer distance definition used to separate the near-field and far-field regions in the simulations."},{"cited_title":"Localiza tion of near-field sources: A reduced-dimension music algorithm ,","cited_arxiv_id":null,"evidence_quote":"Provides the reduced-dimension MUSIC approach for joint DOA and range estimation that the paper's near-field beamforming search builds on."},{"cited_title":"Near and Far Field Model Mismatch: Implications on 6G Communications, Localization, and Sensing","cited_arxiv_id":"2310.06604","evidence_quote":"Establishes near/far-field model mismatch as an open issue for 6G localization and sensing, which is the problem this paper addresses."},{"cited_title":"Spherical w avefronts improve MU-MIMO spectral efficiency when using electricall y large arrays,","cited_arxiv_id":null,"evidence_quote":"Shows spherical-wavefront model mismatch affects spectral efficiency, the prior mismatch result the paper extends to direction-of-arrival estimation."}],"review_version":1}