{"id":"94316a95-eb9d-4385-9d2c-d587a790d3e6","arxiv_id":"2411.12888","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"At 6.5 GHz a target adds distinguishable multipath components, while at 8.75 GHz it blocks more paths, based on indoor FR3 measurements.","lead":"This paper measures indoor radio channels at 6.5 and 8.75 GHz with and without a metal target, and reports that the lower frequency gains extra reflected paths while the higher frequency suffers more blockage. The results are offered as a first benchmark for integrated sensing and communication (ISAC) in the FR3 mid-band spectrum.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The frequency-dependent MPC-count claim rests on an uncalibrated model-order estimator whose SNR sensitivity can produce exactly the reported mode shifts; synthetic calibration is required.","rationale":"The reader's weakest assumption identifies exactly the load-bearing point: the reported frequency-dependent behavior is inferred from histograms of model-order estimates produced by a heuristic elbow method that is never calibrated. I agree, and I sharpen the concern by noting that Section II-B itself assumes a frequency-invariant true number of delays L_a, so the only defensible interpretation of the histograms is as estimates of resolvable/detectable components; that interpretation requires the estimator to be unbiased or at least monotonically related to the true count across SNR, frequency, and target presence. The antenna being used outside its rated band at 8.75 GHz couples frequency to SNR and makes this an active confound. A synthetic-channel calibration and an SNR-degradation check on real 8.75 GHz data would settle whether the mode shifts reflect propagation physics or numerical behavior. This does not require changing the reader's verdict: the paper remains a useful benchmark contribution, but its central interpretive claim should stay conditional on such validation. I therefore recommend no adjustment beyond the original conditional acceptance.","tokens_in":8797,"tokens_out":5187,"duration_ms":63216,"concrete_test":"Simulate synthetic channels with known numbers of MPCs (L = 2, 3, 4, 5) at 6.5 and 8.75 GHz using the exact Pi-Radio parameters (Non = 521, FFT size 1024, bandwidth 500 MHz, same smoothing size) and SNR values estimated from the measured noise floors, including the out-of-band antenna gain at 8.75 GHz. Run the Section III-A elbow/MUSIC pipeline and compare the histograms of \\hat L with the true L at each SNR and frequency. Then take the measured 8.75 GHz no-target CIRs and artificially degrade their SNR to match the target-present SNR; if the 4→2 mode shift is reproduced without adding any blockage, the central claim is not supported. If \\hat L recovers L across SNR and frequency, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central physical claim—6.5 GHz gains distinguishable multipath components under target presence while 8.75 GHz loses them—is read directly from histograms of the elbow-estimated model order \\hat L (Section V-B). Section III-A introduces the second-order elbow method without any calibration or synthetic validation, and the system model (Section II-B) even assumes the true number of delays L_a is the same at both frequencies for a given geometry. The meaningful quantity is therefore \"distinguishable\" components, and \\hat L must be a reliable, condition-invariant proxy for that quantity. The measurements contain exactly the conditions where such estimators can fail: the antenna is used at 8.75 GHz outside its specified 6.0–8.5 GHz band (Section IV-A), so gain, impedance match, and SNR differ by frequency; and target presence changes not only the number of paths but also their amplitudes and the eigenvalue profile of the smoothed covariance matrix. If the elbow method undercounts at lower SNR or with weaker late paths, the observed mode reduction at 8.75 GHz (4→2 for Beta, 5→3 for Alpha, 4→3 overall) could be a numerical artifact of estimator bias rather than physical blockage, and the corresponding 6.5 GHz mode increases could reflect threshold effects rather than new reflections. No independent check—synthetic channels with known MPC counts, or comparison with a ray-tracer ground truth—separates these alternatives. This is load-bearing because every stated ISAC insight, new late paths at lower FR3 frequencies and clutter-blockage sensing at higher FR3 frequencies, is a restatement of these histogram modes.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports on an indoor multi-band channel measurement campaign at 6.5 GHz and 8.75 GHz, conducted with and without a metallic target in the environment, using the Pi-Radio SDR platform. The processing chain consists of frequency-domain smoothing, a second-order elbow method for model-order selection, and MUSIC-based delay estimation, followed by K-means clustering of delay estimates and an illustrative positive/negative region analysis. The central empirical claim is that the presence of the target increases the number of distinguishable multipath components at 6.5 GHz, while at 8.75 GHz it reduces that number, which is interpreted as frequency-dependent blockage.","tokens_in":9132,"tokens_out":4894,"duration_ms":47563,"significance":"If the frequency-dependent effect is real, the finding would be a useful input to ISAC design in FR3, suggesting that lower FR3 frequencies can exploit newly created late reflections while higher FR3 frequencies are better served by detecting clutter suppression. The paper provides a valuable measurement setup description and a novel dataset; the authors are also transparent about the antenna being used outside its rated band (Section IV-A). However, the strength of the evidence is currently insufficient: the central claim is read from histograms of a model-order estimate whose bias is not calibrated, and there is no statistical uncertainty quantification or validation on synthetic channels. These issues are addressable and do not invalidate the potential value of the measurements, but they must be resolved before the claim is accepted as established.","major_comments":[{"comment":"The second-order elbow method used to estimate the number of multipath components, bLa, is not validated or calibrated. The central result—the histogram mode shifts in Figs. 4 and 5—depends entirely on this estimator being a condition-invariant proxy for the number of distinguishable paths. The measurements include a frequency-dependent SNR difference: the antenna is used at 8.75 GHz outside its specified 6.0–8.5 GHz band (Section IV-A), and target presence changes path amplitudes. An uncalibrated elbow method can systematically undercount at lower SNR or when late paths are weak, which would produce the observed reduction at 8.75 GHz and the apparent increase at 6.5 GHz as a numerical artifact. Please add validation on synthetic channels with known numbers of MPCs, including the SNR and eigenvalue regimes of the measurements, or otherwise demonstrate that bLa is not biased differentially across frequency and target conditions.","section":"Section III-A and Section V-B"},{"comment":"The reported mode shifts are not accompanied by any uncertainty quantification or statistical test. The histograms in Figs. 4 and 5 have broad, overlapping distributions; for example, in Fig. 5 (Beta), the 'no target' distribution has many counts at 4 and 5, while the 'with target' distribution has a mode at 2 but counts across the range. It is not clear that these distributions are reliably different. Provide confidence intervals on the modes or perform a permutation test on the difference in estimated MPC counts, and specify how many independent channel snapshots contribute to each histogram (the text mentions 100 channel estimates per measurement but not how many positions/orientations/antennas are used and whether the 100 are independent).","section":"Section V-B"},{"comment":"The positive/negative region analysis in Fig. 7 is used to reinforce the central claim, but the criteria for defining a P-region or N-region are not stated. Without a quantitative definition (e.g., a threshold on the difference between MUSIC-PDPs or on path gains), the selection of green and red regions appears anecdotal. Please specify the procedure and report, across all measurement locations, the frequency with which new paths appear at 6.5 GHz versus the frequency with which paths are suppressed at 8.75 GHz, rather than only one illustrative case.","section":"Section V-C"}],"minor_comments":[{"comment":"The sampling frequency is stated as B = 983.04 Hz, but it should presumably be 983.04 MHz given the occupied bandwidth of approximately 500 MHz; please correct the units.","section":"Section IV-C"},{"comment":"The axis label 'Occurence' is misspelled; it should be 'Occurrence'. Also, the x-axis ranges are inconsistent across the subplots (e.g., Fig. 5 top starts at 2 while the others start at 0), which makes visual comparison harder; please unify them.","section":"Figures 4 and 5"},{"comment":"Please specify the frequency-smoothing subarray size and overlap used in the measurements, since these parameters affect both the model-order estimate and the MUSIC resolution.","section":"Section III-A"},{"comment":"After Eq. (2), the text states that the total number of delays La is the same for both frequencies, but the observed bLa differs. Please clarify explicitly that bLa is the number of distinguishable (resolvable) components, which may differ from the geometric La, and that the mode shifts in Section V-B reflect resolvability rather than physical path count.","section":"Section II-B"},{"comment":"The expectation in C_{a,f_c} is not defined operationally; please specify that a sample covariance over the 100 channel estimates is used and how the smoothing is applied.","section":"Equation (5)"},{"comment":"Reference [9] has incomplete page information ('pp. 1-1'); please update to the final published details if available.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a conference-style measurement report; the central claim is interesting but rests on an unvalidated model-order estimator and lacks statistical tests. The authors should be asked to add synthetic calibration and significance testing, as detailed in the major comments. If the data can be made available, that would also strengthen the reproducibility of the benchmark claimed in the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: a genuinely useful first FR3 multi-band ISAC measurement set, but the headline claim leans on an uncalibrated model-order estimator, so treat the histogram modes as suggestive, not established.\n\nThe new thing here is the dataset: an indoor comparison of 6.5 and 8.75 GHz channels with and without a metal target, using a 2x1 MISO setup and MUSIC-refined PDPs. I'm not aware of prior work doing exactly this in FR3, so as a benchmark it has value. The processing is standard, but the P/N-region framing (new late paths vs. blocked paths) is a nice way to think about ISAC opportunities. The qualitative pattern is consistent across both the histograms and the per-location MUSIC-PDP plots, which gives it some credibility independent of the elbow counter.\n\nThe soft spots are the ones you'd expect. The second-order elbow method is never calibrated or validated on synthetic channels, and the whole quantitative comparison is built on its estimates of L. That matters because the antenna is used at 8.75 GHz outside its rated band, which likely changes SNR and eigenvalue spread. If the elbow undercounts at lower SNR, the mode drops at 8.75 GHz could be a numerical side effect rather than physical blockage. There are also no error bars or significance tests on the mode shifts, and the sample is one lab, five RX positions, three orientations. The footnote that the antenna 'can be used' at 8.75 GHz is inadequate; a return-loss plot would have settled it.\n\nI want to be fair: the physics is plausible, the paper doesn't oversell to the point of being wrong, and the P/N plots in Fig. 7 give a direct look at the effect. But the phrasing 'results reveal' is stronger than the evidence supports. For a conference paper, this is acceptable as a benchmark if the authors add synthetic calibration or soften the conclusions.\n\nWho should read it: anyone starting FR3 channel measurements or ISAC work in the upper mid-band. It deserves a serious referee; I would accept it conditionally, asking for a calibration of the elbow method or at least a comparison with a fixed model order, plus a clearer antenna characterization.\n\nMy recommendation: engage with it as a useful starting point, but don't base system design on the specific mode numbers without checking the estimator first.","headline":"Useful first FR3 multi-band ISAC measurements, but the central mode-shift claim depends on an uncalibrated estimator and needs synthetic validation.","tokens_in":9664,"tokens_out":2928,"would_cite":true,"duration_ms":30958,"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":"The paper claims that in the low FR3 band, adding a reflective target increases the number of resolvable multipath components at 6.5 GHz and decreases it at 8.75 GHz.","keywords":["FR3","upper mid-band","channel modeling","multi-band measurements","integrated sensing and communication","MUSIC delay estimation","multipath components","indoor propagation"],"falsifier":"Run the same measurement and processing pipeline on synthetic channels with a known number of multipath components, matched to the SNR and bandwidth of both frequency bands, with and without a simulated target. If the estimated path-count modes shift in the same direction as the measurements despite no physical target-induced path creation or blockage, the reported frequency dependence is an artifact of the estimator.","tokens_in":8626,"feed_emoji":"📡","tokens_out":5902,"duration_ms":51763,"temperature":0.7,"pith_summary":"This paper reports a two-frequency indoor channel measurement campaign in the low FR3 band, at 6.5 GHz and 8.75 GHz, designed to see how the presence of a reflective target changes the multipath channel. The central claim is that the target affects the two frequencies in opposite ways: at 6.5 GHz new distinguishable reflections appear, while at 8.75 GHz existing paths are blocked. This matters for integrated sensing and communication because it suggests that the best sensing strategy depends on which FR3 frequency is used: lower frequencies can exploit newly created late-arriving paths, while higher frequencies should detect targets by looking for path suppression. The paper also offers a processing methodology, based on MUSIC delay refinement with frequency smoothing and clustering, as a benchmark for future FR3 multi-band studies.","feed_headline":"Targets add paths at 6.5 GHz but block them at 8.75 GHz","feed_subtitle":"New FR3 measurements show lower band gains reflections while upper band suffers blockage for ISAC.","key_machinery":"The central machinery is a frequency-smoothed MUSIC power-delay-profile estimator. Because multipath components are coherent, the channel covariance is rank-one; the paper applies frequency-domain smoothing to restore rank, uses a second-order elbow method on the smoothed eigenvalues to select the number of paths, and then evaluates a MUSIC pseudo-spectrum to refine delays. K-means clustering with a silhouette criterion groups the per-frame delay estimates into stable clusters. The path-count histograms and the positive/negative region analysis are all computed from these estimates, so the entire frequency-dependent conclusion rests on this pipeline.","core_discovery":"On its own terms, the paper claims that the target's effect on the background channel is frequency-dependent. At 6.5 GHz, histograms of the estimated number of multipath components shift to higher counts when the target is present, such as a mode increasing from 3 to 4 across all orientations. At 8.75 GHz, the mode shifts to lower counts, from 4 to 3 across all orientations, and from 4 to 2 for one orientation. The paper interprets this as lower frequencies diffracting around the target to create new multi-bounce paths, while higher frequencies are more susceptible to blockage and reflection losses. It further identifies positive regions, where new target-related reflections appear, and negative regions, where the target suppresses existing clutter paths, and argues that each supports a distinct sensing modality.","pith_inferences":["A testable extension would be to sweep intermediate frequencies between 6.5 and 8.75 GHz to locate where the target's effect switches from path creation to path suppression.","Because the result is derived from an uncalibrated model-order estimator, an SNR-matched synthetic-channel experiment is needed to separate physical propagation effects from estimator artifacts.","If the pattern holds, a dual-band ISAC system could fuse the two modalities: use the lower band's new reflections for detection and the higher band's suppression for localization or shadow estimation.","The observed stability of the 8.75 GHz delay clusters hints that higher FR3 frequencies may provide more precise delay estimation once blockage is modeled, a hypothesis worth testing with a larger antenna array."],"forward_implications":["At 6.5 GHz, a sensing receiver should expect additional late-arriving multipath components created by the target; these positive-region paths carry target information without overlapping main clutter.","At 8.75 GHz, target presence is better detected as a reduction in the number of paths; negative-region blockage analysis is the more promising sensing mode.","The two frequencies cannot share a single multipath-count model, so FR3 channel models and ISAC algorithms should treat sub-bands separately.","The methodology, combining smoothed MUSIC with elbow-based model order selection and K-means clustering, gives a repeatable benchmark for comparing future FR3 multi-band measurements."],"supporting_citations":[{"why":"Provides the FR3 software-defined radio front-end that enables the 6.5 GHz and 8.75 GHz measurements.","marker":"[6]"},{"why":"Supplies the MUSIC-type super-resolution delay estimation approach that the paper adapts.","marker":"[12]"},{"why":"Gives the MISO channel estimation via windowing used to separate the two transmit antennas.","marker":"[13]"},{"why":"Explains why the multipath covariance is rank-deficient due to coherent signals, motivating the smoothing step.","marker":"[15]"},{"why":"Provides the space-frequency smoothing technique that restores the rank for eigenmode analysis.","marker":"[16]"},{"why":"Supplies the second-order statistics and principal-component background behind the elbow model-order selection.","marker":"[17]"}],"fun_headline_variants":["Target adds paths at 6.5 GHz, blocks at 8.75 GHz","6.5 GHz gains paths with target, 8.75 GHz loses","Target flips multipath effect between 6.5 and 8.75 GHz","FR3 target: extra paths at 6.5 GHz, blockage at 8.75","At 6.5 GHz target adds paths, at 8.75 GHz blocks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole frequency-dependent pattern depends on the second-order elbow method correctly counting the number of multipath components from the smoothed covariance matrix at both frequencies, with and without the target; if that estimator shifts with SNR or eigenvalue structure, the reported mode changes could be numerical artifacts rather than propagation effects.","fun_headline_variants_meta":{"raw":{"variants":["Target adds paths at 6.5 GHz, blocks at 8.75 GHz","6.5 GHz gains paths with target, 8.75 GHz loses","Target flips multipath effect between 6.5 and 8.75 GHz","FR3 target: extra paths at 6.5 GHz, blockage at 8.75","At 6.5 GHz target adds paths, at 8.75 GHz blocks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001302,"raw_usage":{"total_tokens":5259,"prompt_tokens":841,"completion_tokens":4418,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":457,"completion_tokens_details":{"reasoning_tokens":4307}},"tokens_in":457,"tokens_out":4418,"duration_ms":29432,"temperature":1.0,"reasoning_tokens":4307,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:03:38.654005+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same measurement and processing pipeline on synthetic channels with a known number of multipath components, matched to the SNR and bandwidth of both frequency bands, with and without a simulated target. If the estimated path-count modes shift in the same direction as the measurements despite no physical target-induced path creation or blockage, the reported frequency dependence is an artifact of the estimator.","supporting_citations":[{"cited_title":"A frequency hopping software-defined radio platform for communications and sens- ing in the upper mid-band,","cited_arxiv_id":null,"evidence_quote":"Provides the FR3 software-defined radio front-end that enables the 6.5 GHz and 8.75 GHz measurements."},{"cited_title":"Experimental validation of superresolution delay estimation algorithm using a 26 GHz radar setup,","cited_arxiv_id":null,"evidence_quote":"Supplies the MUSIC-type super-resolution delay estimation approach that the paper adapts."},{"cited_title":"Channel estimation for MIMO space time coded OTFS under doubly selective channels,","cited_arxiv_id":null,"evidence_quote":"Gives the MISO channel estimation via windowing used to separate the two transmit antennas."},{"cited_title":"A statistical multipath detector for antenna array based GNSS receivers,","cited_arxiv_id":null,"evidence_quote":"Explains why the multipath covariance is rank-deficient due to coherent signals, motivating the smoothing step."},{"cited_title":"On spatio-frequential smoothing for joint angles and times of arrival estimation of multipaths,","cited_arxiv_id":null,"evidence_quote":"Provides the space-frequency smoothing technique that restores the rank for eigenmode analysis."}],"review_version":1}