{"id":"5a05501e-5d4d-4be4-8ff6-0d15cfaa4d8d","arxiv_id":"2512.00707","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"New 4.85 GHz urban double-directional channel measurements in Japan yield LSP statistics and an indicative 4–28 GHz log-log scaling model for delay and angular spreads.","lead":"This paper reports new urban macro- and micro-cell radio channel measurements at 4.85 GHz in Yokohama, extracting path loss, delay/angular spreads, K-factor, and spatial-consistency statistics. It then combines these with literature data to sketch how those spreads scale across the 4–28 GHz range, relevant to 5G/6G spectrum planning around the FR1/FR3 boundary.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table V cross-band models contradict the paper's own 4.85 GHz measurements (UMi LoS DS: 39 ns predicted vs 189 ns measured), so the 'measurement-anchored' claim is unsupported.","rationale":"The reader's weakest assumption (scenario-matching of literature anchors) is a plausible cause, but the paper's own tables provide stronger and more direct evidence: the fitted models in Table V do not reproduce the measured 4.85 GHz values for several key parameters. This is not a speculative concern about hidden anchor mismatch; it can be checked by simple arithmetic. It is load-bearing because the cross-band trends are the claimed FR1/FR3 bridge and the 'implementation-ready' parameter set. If the 4.85 GHz anchors are effectively outliers to the fit, the frequency slopes and intercepts in Table V are artifacts of the robust regression and the chosen non-positive slope constraint, not of the in-house measurements. I still do not recommend rejection: the 4.85 GHz path-loss, delay/angular spread, and spatial-consistency measurements may be valuable, and the paper explicitly labels the cross-band extrapolation as indicative. The appropriate verdict remains CONDITIONAL, requiring either recomputation of Table V with explicit anchor weights/confidence intervals or a clear relabeling of the cross-band results as literature-only. The Table III inconsistencies noted by the reader reinforce this but are not the load-bearing issue I emphasize here.","tokens_in":18475,"tokens_out":20047,"duration_ms":194097,"concrete_test":"Run a single reproducibility script for Eq. (23) with exactly the Table IV inputs, reporting (i) the predicted LSP at fc=4.85 GHz, (ii) the bisquare weight assigned to each 4.85 GHz row, and (iii) the Table V coefficients after leave-one-out removal of the 4.85 GHz UMi point. If the UMi LoS DS prediction deviates from the measured 188.81 ns mean by more than the bootstrap 5-95% CI from Section IV-C (or >20%), the model is not measurement-anchored and Table V should be relabeled as literature-only.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the 4.85 GHz campaign anchors the cross-band LSP models. Evaluating the Table V models at fc=4.85 GHz (using the stated UMi convention log10(1+fc)) against Table IV gives: UMi LoS DS 39 ns vs 188.81 ns (4.8x low), UMi NLoS DS 78 ns vs 268.85 ns, UMi LoS ASA 43 deg vs 68.3 deg, UMi LoS ASD 36 deg vs 73.4 deg, and UMa NLoS DS 127 ns vs 241.38/176.80 ns. The robust regression in Eq. (23) can therefore downweight the large 4.85 GHz residuals, so the fits are effectively driven by the heterogeneous literature anchors rather than by the in-house statistics. This contradicts the abstract's statement that calibrated 4.85 GHz statistics and literature anchors jointly enforce smooth evolution across 4-28 GHz. The abstract also promises a leave-one-out sensitivity analysis of the UMi DS fit, but no such analysis appears in Section V. As printed, Table V is not measurement-anchored at 4.85 GHz; it is a literature-only fit drawn through the 4.85 GHz points.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports 4.85 GHz double-directional MIMO measurements in three Yokohama urban routes (two UMa, one UMi), extracting path loss, delay spread, azimuth spreads, Rician K-factor, and spatial-consistency statistics from SAGE-extracted multipath components. It then combines the measured 4.85 GHz LSP means with literature anchor points from 6–28 GHz in a constrained robust log-log regression to obtain DS/ASA/ASD frequency trends across the FR1/FR3 boundary, and compares these trends with 3GPP TR 38.901 parameterizations. The paper claims that the 4.85 GHz statistics anchor the cross-band models and that the resulting parameter set provides an implementation-ready basis for 5G/6G simulations around the upper-FR1/FR3 transition.","tokens_in":18817,"tokens_out":7631,"duration_ms":65049,"significance":"The 4.85 GHz measurement campaign itself is a valuable contribution: the paper documents an 8×8 full-MIMO double-directional sounder, SAGE-based MPC extraction, K-power-means clustering, adaptive distance binning with bootstrap confidence intervals, and spatial-decorrelation estimation, all in a band where outdoor urban measurements are scarce. If the reported LSP statistics were internally consistent, the paper would be a useful reference for upper-FR1/FR3 channel modeling. However, the manuscript as written contains physically impossible standard deviations in Table III, internal inconsistencies between Table III and Table IV, and a cross-band model that does not reproduce the 4.85 GHz anchor values it claims to be anchored to. These are load-bearing issues that must be resolved before the central claims can be accepted.","major_comments":[{"comment":"The log-domain standard deviations for ASD and ASA in Table III are physically impossible. For example, Area2 NLoS ASD has μ = -0.3454 and σ = 5.26, and Area2 NLoS ASA has μ = -0.0840 and σ = 5.36. A log10 standard deviation greater than 5 implies that the angular spread varies over many orders of magnitude, far exceeding the 0–360° physical range. These values cannot result from a lognormal fit to valid angular-spread data and indicate an error in the reported statistics. Because Table III is presented as the 4.85 GHz model parameter reference, this error undermines the central measurement characterization.","section":"Table III"},{"comment":"The mean values in Table III do not match the 4.85 GHz entries in Table IV, despite both tables purporting to summarize the same measurements. For example, UMi LoS DS: Table III log10 mean = -6.9338 corresponds to 116.5 ns, whereas Table IV lists 188.81 ns. Area1 UMa LoS DS: Table III gives 101.6 ns versus Table IV's 142.40 ns. Area1 UMa LoS ASD: Table III μ = 1.4100 corresponds to 25.7°, whereas Table IV lists 36.3°. These discrepancies are too large to be explained by lognormal-to-linear conversion and are not discussed in the text. One of the two tables is incorrect, and this inconsistency directly affects the cross-band anchor values used in Section V.","section":"Table III vs. Table IV"},{"comment":"The cross-band models in Table V are not anchored to the paper's own 4.85 GHz measurements, contradicting the abstract's claim that 'calibrated 4.85 GHz statistics and scenario-specific literature anchors jointly enforce smooth evolution.' Evaluating the Table V UMi models at fc = 4.85 GHz gives roughly 39–42 ns for LoS DS versus 188.81 ns in Table IV, about 78–85 ns for NLoS DS versus 268.85 ns, about 43–49° for LoS ASA versus 68.3°, and about 36–39° for LoS ASD versus 73.4°. Because Eq. (23) minimizes a robust bisquare loss without any constraint that the fit pass through or even preferentially weight the 4.85 GHz anchor, the large 4.85 GHz residuals are downweighted and the fitted lines are effectively determined by the heterogeneous literature anchors. As printed, Table V is a literature-driven fit drawn near, not through, the 4.85 GHz data. To support the 'measurement-anchored' fram","section":"Section V, Eq. (23), Table V"},{"comment":"The abstract promises that 'the sensitivity of the UMi DS fit was examined via leave-one-out analysis,' but no leave-one-out analysis appears anywhere in Section V or elsewhere in the manuscript. This omission is material: with only six UMi DS anchors in Table IV, the fitted slope and intercept are potentially controlled by a single anchor, and the claimed robustness cannot be assessed without the promised sensitivity check. The leave-one-out analysis should be added, or the abstract claim removed.","section":"Abstract and Section V"}],"minor_comments":[{"comment":"The note says 'UMi uses log10(1 + fc), while UMa uses log10(fc),' but the 'This work' UMi formulas in the same table appear to use log10(fc). Clarify whether the note applies only to the 3GPP columns or to both, and make the formulas internally consistent.","section":"Table V note"},{"comment":"The last bullet of the UMi paragraph contains an incomplete sentence: 'In contrast, hows substantial growth and variability in NLoS...' — the subject (likely ASD) appears to be missing. Please correct.","section":"Section IV-C"},{"comment":"The x-axis labels in Fig. 8 read '20 60' and are unclear; they should explicitly indicate the log10 scale, e.g., 'log10(ASD [deg])' with appropriate tick labels.","section":"Fig. 8"},{"comment":"The text states that the ITU-R ABG model is not fitted to the measurement data, but Table II includes columns for ITU-R α, β, γ, and σ. Clarify what these entries represent, or remove them to avoid implying a fit.","section":"Table II"}],"recommendation":"major_revision","confidential_remarks":"The internal inconsistencies in Table III and Table IV, and the failure of Table V to reproduce the 4.85 GHz anchor values, are serious and central. I recommend major revision rather than rejection because these issues are potentially correctable: the Table III entries may be fixable, and the cross-band model could be re-fit with an explicit anchor constraint or honestly reframed as literature-driven with 4.85 GHz as a comparison point. However, if the Table III/IV discrepancies cannot be traced to a simple correction, the paper's primary 4.85 GHz parameter reference would not be trustworthy, and rejection would be warranted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my take. The 4.85 GHz campaign is the real payload. New double-directional UMa/UMi measurements in Yokohama with careful SAGE-based MPC extraction, distance-binned LSP trends, and spatial-consistency analysis. That fills a genuine gap near the WRC-27 band and is worth having as a reference. Path-loss and K-factor analysis are solid. The cross-band synthesis is a different story. Table V claims to be measurement-anchored, but evaluating the fitted lines at 4.85 GHz gives UMi LoS DS ~39 ns against the measured 188.8 ns, NLoS ~78 ns vs 268.9 ns, and similar misses on angular spreads. Because the robust regression uses bisquare weights, the large residuals at the paper's own anchor point are downweighted and the lines are effectively driven by the literature anchors. So 'jointly enforce smooth evolution' is not what the fits do. 'Measurement-informed' would be fair; 'measurement-anchored' is not. The internal tables also need attention: Table III lists log-domain ASD/ASA standard deviations of 3.82–5.26 with means around 0.5–0.6, which is physically impossible, and Table III DS means disagree with Table IV (e.g., Area3 LoS 116 ns vs 188.8 ns). Finally, the abstract promises a leave-one-out sensitivity analysis of the UMi DS fit; I could not find it anywhere in Section V. If it exists, it belongs in the body; otherwise the claim should be dropped. The paper is still worth a serious referee. The measurement portion is novel and useful, and the cross-band issues are fixable: correct the tables, add confidence intervals, present Table V as a literature-referenced trend rather than an anchored model, and clean up the promises. I'd cite the 4.85 GHz statistics once the tables are corrected, and I'd bring this to reading group as a case study in how robust regression can quietly decouple a 'reference' model from its anchor points. Recommendation: send to peer review, major revision.","headline":"The 4.85 GHz measurement set is valuable; the cross-band model as fitted is not measurement-anchored, and the paper's own tables are internally inconsistent.","tokens_in":19365,"tokens_out":5341,"would_cite":true,"duration_ms":51323,"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":"A 4.85 GHz measurement-based channel model links upper FR1 to FR3 frequency bands.","keywords":["4.85 GHz","channel measurement","frequency-continuous model","FR1-FR3","urban macrocell","urban microcell","large-scale parameters","spatial consistency"],"falsifier":"Re-running the frequency-continuous regression with a different set of high-frequency anchors—for instance, leaving out all 24–28 GHz points or replacing them with independent measurements from another urban campaign—and checking whether the fitted slopes and intercepts change substantially would settle whether the trends are robust or an artifact of dataset mixing.","tokens_in":18311,"feed_emoji":"📡","tokens_out":1255,"duration_ms":15545,"temperature":0.7,"pith_summary":"This paper presents a double-directional channel measurement campaign at 4.85 GHz in urban macrocell (UMa) and urban microcell (UMi) environments, and uses those measurements as an anchor to build frequency-continuous large-scale parameter (LSP) models for delay spread, azimuth spread of arrival, and azimuth spread of departure across roughly 4–28 GHz. The central assertion is that combining the new 4.85 GHz statistics with literature-reported measurements at higher frequencies yields smooth, physically consistent log-log trends that avoid the discontinuities seen when extrapolating standardized 3GPP models across the 7.125 GHz FR1–FR3 boundary. A sympathetic reader would care because this directly addresses a measurement gap in the under-explored upper-FR1 band targeted by WRC-27, and because the resulting parameters are proposed as implementation-ready for 5G/6G system-level simulations, beam management, and spectrum planning. The paper also provides route-specific path loss, K-factor, and spatial-consistency statistics, showing that standardized model assumptions often overestimate delay spread in UMa and underestimate it in UMi.","feed_headline":"One measured band anchors channel model from 4 to 28 GHz","feed_subtitle":"New 4.85 GHz urban measurements plus literature anchors yield smooth FR1–FR3 delay and angle spread trends.","key_machinery":"The central object is a constrained robust log-log regression model log10 X(f) = a log10 f + b, with a ≤ 0, fitted jointly to route-wise means at 4.85 GHz and literature anchor points up to 28 GHz for each LSP X ∈ {DS, ASA, ASD}. This power-law-in-frequency model is what enforces smooth evolution across the 7.125 GHz FR1–FR3 boundary. The 4.85 GHz anchor values come from an 8×8 MIMO channel sounder with SAGE-based MPC extraction, and the paper also uses adaptive distance binning with bootstrap confidence intervals and exponential spatial autocorrelation fitting to quantify spatial consistency.","core_discovery":"The paper claims that a parameterized, frequency-continuous LSP model can be anchored at a single well-calibrated 4.85 GHz measurement band and extended to 28 GHz by fitting log-log regressions to a combination of in-house route means and scenario-matched literature anchors. The fitted models for DS, ASA, and ASD show systematically weaker dispersion in UMa and stronger frequency-dependent compaction in UMi than the 3GPP reference parameterizations over the same 4–28 GHz interval. The authors further claim that their 4.85 GHz measurements themselves reveal significant deviations from 3GPP defaults: measured delay spread is smaller than 3GPP in UMa and larger in UMi, ASD is generally underest","pith_inferences":["Independent 4–8 GHz measurements in other cities could test whether the frequency slopes in Table V generalize beyond Yokohama, or whether site-specific geometry dominates the fitted trends.","The constraint a ≤ 0 in the regression imposes a physically motivated but unverified assumption that all angular and delay spreads must non-increase with frequency; negative slopes might be artifacts of high-frequency anchors with narrower measurement bandwidths rather than true physical trends.","The paper's claim of smoother cross-band continuity could be tested by ray-tracing simulations at 6, 8, 10, 15, and 20 GHz in the same three areas, producing synthetic anchor points without the heterogeneity of literature data.","The low measured K-factor values suggest that sub-6 GHz urban channels are richer in diffuse multipath than standard models assume; this has implications for beam-tracking algorithms that rely on a strong specular component."],"forward_implications":["If the fitted log-log trends are correct, system-level simulators for 5G/6G can use a single continuous LSP parameterization from 4–28 GHz without artificial jumps at 7.125 GHz.","The 4.85 GHz measurement reference provides a missing data point for the WRC-27 upper-FR1 band, enabling more reliable calibration of standardized models in that range.","Route-specific spatial-consistency distances, especially the long PL residual decorrelation in UMi, suggest that geometry-based stochastic channel models need scenario-specific spatial correlation rather than generic defaults.","The observed deviations from 3GPP (e.g., UMi NLoS DS exceeding standard values) imply that guard interval and cyclic prefix designs may need to be re-optimized for mid-band urban deployments.","The UMa NLoS ASD model is explicitly flagged as incomplete due to insufficient anchors, indicating a need for more UMa measurements in the 4–24 GHz range before a definitive scenario-wide model is possible."],"fun_headline_variants":["4.85 GHz urban measurements drive 4–28 GHz channel trends","Single-band anchor: 4.85 GHz data plus literature spans FR1–FR3","One frequency, wide scope: 4.85 GHz anchors cross-band model","Yokohama 4.85 GHz data extends to 28 GHz via literature anchors","Measured 4.85 GHz plus literature yields FR1–FR3 LSP trends"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The cross-band frequency trends assume that literature anchor points from different campaigns are scenario-matched and statistically comparable to the in-house 4.85 GHz measurements, despite differences in equipment, bandwidth, array geometry, MPC extraction, and LoS/NLoS definitions.","fun_headline_variants_meta":{"raw":{"variants":["4.85 GHz urban measurements drive 4–28 GHz channel trends","Single-band anchor: 4.85 GHz data plus literature spans FR1–FR3","One frequency, wide scope: 4.85 GHz anchors cross-band model","Yokohama 4.85 GHz data extends to 28 GHz via literature anchors","Measured 4.85 GHz plus literature yields FR1–FR3 LSP trends"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000224,"raw_usage":{"total_tokens":1378,"prompt_tokens":902,"completion_tokens":476,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":646,"completion_tokens_details":{"reasoning_tokens":381}},"tokens_in":646,"tokens_out":476,"duration_ms":5117,"temperature":1.0,"reasoning_tokens":381,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T19:22:02.943478+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-running the frequency-continuous regression with a different set of high-frequency anchors—for instance, leaving out all 24–28 GHz points or replacing them with independent measurements from another urban campaign—and checking whether the fitted slopes and intercepts change substantially would settle whether the trends are robust or an artifact of dataset mixing.","supporting_citations":[],"review_version":2}