Urban Macro/Microcellular Channel Characterization at 4.85 GHz With Literature-Referenced Upper FR1-to-FR3 Cross-Band Analysis

Inocent Calist; Minseok Kim

REVIEW 4 major objections 4 minor 1 cited by

Urban Macro/Microcellular Channel Characterization at 4.85 GHz With Literature-Referenced Upper FR1-to-FR3 Cross-Band Analysis

T0 review · 4 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A 4.85 GHz measurement-based channel model links upper FR1 to FR3 frequency bands.

desk verdict 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. read the letter →

arxiv 2512.00707 v5 pith:YBEFV6O7 submitted 2025-11-30 eess.SP

Inocent Calist , Minseok Kim This is my paper · ORCID

classification eess.SP

keywords 4.85GHzchannelmeasurementfrequency-continuousmodelFR1-FR3urbanmacrocellmicrocelllarge-scaleparametersspatialconsistency

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

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.

What carries the argument

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.

What would settle it

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.

Watch

Extended reading notes

Core claim

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

Load-bearing premise

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.

Editorial extensions

If this is right

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.

Reading between the lines

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

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.

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.

Desk Editor's Note

read the 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.

Referee Report

4 major / 4 minor

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.

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 (4)

[Table III] 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.
[Table III vs. Table IV] 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 V, Eq. (23), Table V] 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
[Abstract and Section V] 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.

minor comments (4)

[Table V note] 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 IV-C] 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.
[Fig. 8] 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.
[Table II] 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.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the 4.85 GHz characterization is an independent measurement, and the cross-band trends are transparent regressions over measured and literature anchors, not disguised predictions.

full rationale

The derivation chain is not circular. The 4.85 GHz results are obtained from an independent double-directional measurement campaign: path loss (Eqs. 9-12), RMS delay spread (Eq. 13), azimuth spreads (Eq. 14), Rician K-factor (Eqs. 15-17), and spatial-consistency decorrelation distances (Eqs. 19-21) are computed directly from measured MPCs, with 3GPP TR 38.901 used only as an external comparison baseline. The frequency-continuous LSP models in Table V are explicitly fitted by robust linear regression (Eqs. 22-24) to the 4.85 GHz statistics plus the literature anchors in Table IV; the paper states they are 'measurement-informed and indicative rather than a definitive multi-band model.' Thus the fitted trends are presented as regressions of their inputs, not as independent predictions, so there is no hidden reduction of a derived result to an input. Self-citations appear for the sounder design [18], clustering methodology [21], and some literature anchors [21,40], but these are independent measurements or methods, not a self-referential uniqueness argument, and none is load-bearing by itself since multiple external anchors are used. Two issues are flagged but are not circularity: the abstract promises a leave-one-out sensitivity analysis of the UMi DS fit, but Section V does not report it; and the Table V UMi LoS DS model evaluates to about 42 ns at 4.85 GHz versus the measured 188.81 ns in Table IV, indicating the robust fit downweights the in-house point. These are validity/reporting concerns, not circular derivation.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

No new physical entities are proposed. The free parameters are standard fit outputs for path-loss and LSP models, plus user-chosen processing thresholds. The most consequential assumptions are the log-log frequency-scaling functional form, the non-increasing slope constraint, and the comparability of heterogeneous literature anchors.

free parameters (7)

CI path-loss exponent n per route/state = Table II: Area1 UMa LoS n=2.19, NLoS n=2.81; Area3 UMi LoS n=1.78, NLoS n=2.54
Fitted by least squares to the measured path-loss data at 4.85 GHz; standard model output, not a hidden choice.
FI path-loss slope α, intercept β, and shadowing σ = Table II values per route/state, e.g., Area3 UMi NLoS α=5.45, β=-25.98, σ=7.48
Fitted to the same 4.85 GHz path-loss data as an alternative model.
LSP log-domain means and standard deviations (Table III) = e.g., Area1 UMa LoS log10 DS μ=-6.9931, σ=0.18; printed ASD/ASA values
These are the central measured parameter set; some printed σ values are physically implausible, which is a data-integrity concern.
Distance-binning thresholds = Nmin=20 snapshots, Δdmax=50 m
Chosen by authors to balance statistical reliability and spatial locality; affects route-dependent trends.
SAGE MPC count = 300 MPCs per snapshot
Chosen by authors; residual power ratio after extraction is reported as <5% in LoS, but the count is a user-set parameter.
Cross-band log-log slope aX and intercept bX (Table V) = e.g., UMa LoS DS: -0.19 log10(fc) - 6.71; UMi NLoS DS: -0.45 log10(fc) - 6.76
Fitted via constrained robust regression to the 4.85 GHz point plus literature anchors; these are the output model parameters.
Spatial decorrelation distance Dcorr per LSP/route/state = Fig. 11 values, e.g., PL residual decorrelation distances of a few meters to tens of meters
Fitted by nonlinear least squares to empirical spatial autocorrelation functions.

assumptions (6)

domain assumption Plane-wave superposition of Dirac MPCs (Eq. 7)
The propagation channel is modeled as a finite sum of discrete plane-wave components with independent delays and angles; this is standard in channel modeling but is an assumption about physical reality.
domain assumption Wide-sense stationarity for spatial autocorrelation (Eq. 19)
The empirical ACF computation assumes stationarity of LSP traces over the route segments, though the paper itself notes strong non-stationarity in some environments.
domain assumption Antenna pattern compensation via beamforming (Eq. 3)
The measured MIMO matrix is compensated using assumed known antenna radiation patterns; sidelobe interference is acknowledged but not fully removed.
domain assumption Log-log power-law frequency dependence of LSPs (Eq. 22)
The cross-band model assumes DS, ASA, and ASD vary linearly with log10(f); this is empirically motivated but not derived and may not hold across mixed scenarios.
ad hoc to paper Non-increasing LSP constraint a ≤ 0 (Eq. 23)
The robust regression enforces a non-positive frequency slope to suppress non-physical positive slopes; this biases the fit and can hide genuine frequency increases if they exist.
domain assumption Scenario-matched comparability of literature anchors (Table IV)
Anchors from different campaigns at 6–28 GHz are treated as comparable to the Yokohama UMa/UMi routes; differences in equipment, algorithms, and route geometry are not corrected for.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Urban Macro/Microcellular Channel Characterization at 4.85 GHz With Literature-Referenced Upper FR1-to-FR3 Cross-Band Analysis." pith.science (2026). https://pith.science/paper/YBEFV6O7

@misc{pith2026251200707,
  author       = {Pith},
  title        = {Pith review of: Urban Macro/Microcellular Channel Characterization at 4.85 GHz With Literature-Referenced Upper FR1-to-FR3 Cross-Band Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YBEFV6O7}},
  note         = {Machine review of arXiv:2512.00707}
}

abstract

The transition from 5G to 6G requires frequency-dependent, physically consistent radio channel models across the upper-FR1/FR3 transition region, particularly in the under-explored $4$--$8$~GHz region targeted in the current WRC-$27$ studies, where outdoor urban channel measurements and characterizations remain scarce. This paper presents a $4.85$~GHz measurement-anchored study of urban channels and a literature-referenced cross-band analysis. Double-directional measurements were conducted at $4.85$~GHz in urban macrocell (UMa) and urban microcell (UMi) routes in Yokohama, Japan, from which path loss, delay spread (DS), azimuth spread of arrival/departure (ASA/ASD), $K$-factor, and route-dependent spatial-consistency statistics were extracted. To align these results in a broader cross-band context, the measured $4.85$~GHz large-scale parameter (LSP) means were combined with scenario-matched literature anchors to derive log-log trends for DS, ASA, and ASD over an approximately $4$--$28$~GHz range around the $7.125$~GHz upper-FR1/FR3 cross-band boundary. The resulting trends were compared with 3GPP UMa/UMi reference parameterizations over the same interval, and the sensitivity of the UMi DS fit was examined via leave-one-out analysis. Because the cross-band analysis still relies on a single in-house measurement band alongside heterogeneous anchors from different campaigns, it is presented as measurement-informed and indicative rather than as a definitive multi-band model. The paper therefore contributes both a detailed, parameterized $4.85$~GHz urban measurement reference and a bounded literature-referenced view of channel behavior near the upper-FR1/FR3 transition.

Figures

Figures reproduced from arXiv: 2512.00707 by the authors.

**Figure 1.** Measurement system. sounding. This configuration yields a delay resolution of approximately 10 ns, with a tone spacing of 195 kHz and a total delay span of 5.12 µs. The transmitter (Tx) and receiver (Rx) were configured in an 8 × 8 full-MIMO architecture, enabling double-directional wideband channel characterization. The Tx (base station) employed a uniformly spaced linear array (ULA) with eight vertically polarized… view at source ↗

**Figure 2.** Three urban cellular scenarios. TABLE I: Measurement setups. Scenarios Area BS Antennas MS Antennas Urban Macro (UMa) Kannai Area (Area1), Yokohama Daiichi Yuraku Bldg. 8-elem ULA Height: 33 m Direction: N 8-elem UCA Height: 2.7 m (on vehicle roof) Chinatown (Area2), NTTCom Yokohama Yamashita Bldg. 8-elem ULA Height: 34 m Direction: SE Urban Micro (UMi) Streets (Area3), Yokohama World Porters 8-elem UCA Height: 3 m … view at source ↗

**Figure 3.** Measurement routes [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (9 more)

**Figure 4.** Figure 4: Power delay profiles (PDPs) computed from the measured, MPC-reconstructed, and residual MIMO channel matrices using (5). and only the first snapshot in each group was retained for analysis. III. SYSTEM-INDEPENDENT CHANNEL EXTRACTION A. MIMO Channel Matrix and Multidime…

**Figure 5.** Figure 5: Path loss modeling. #1 #2 #3 ©2025 Esri (a) Area1 (UMa) Route. #1 #2 ©2025 Esri (b) Area2 (UMa) Route. #1 #2 #3 ©2025 Esri (c) Area3 (UMi) Route [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: Measurement routes PL characterization. TABLE II: Model parameter fitting at 4.85 GHz. The ITU-R site-general models for the above-rooftop and below-rooftop categories correspond to UMa and UMi, respectively. Route Scenario State CI model fitting FI model fitting ITU-R…

**Figure 7.** Figure 7: Delay spread. 20 60 log10(ASD [deg]) 0 0.2 0.4 0.6 0.8 1 CDF 20 60 log10(ASA [deg]) 0 0.2 0.4 0.6 0.8 1 CDF LoS Meas NLoS Meas LoS 3GPP NLoS 3GPP (a) Area1 (UMa). 20 60 log10(ASD [deg]) 0 0.2 0.4 0.6 0.8 1 CDF 20 60 log10(ASA [deg]) 0 0.2 0.4 0.6 0.8 1 CDF LoS Meas NLo…

**Figure 8.** Figure 8: Angle spreads. TABLE III: Model parameters obtained from measurements at 4.85 GHz (Mean µ and Std. Dev. σ). Parameter Area1 (UMa) Area2 (UMa) 3GPP (UMa) Area3 (UMi) 3GPP (UMi) LoS NLoS LoS NLoS LoS NLoS LoS NLoS LoS NLoS RMS DS log10(DS/1 s) µ −6.9931 −6.8815 −6.9953 −…

**Figure 9.** Figure 9: Empirical CDFs of the Rician K-factor measured at 4.85 GHz for UMa and UMi scenarios, compared against the frequencyconstant 3GPP reference model . 3) Rician K-factor The Rician K-factor quantifies the power ratio between the dominant (typically LoS) specular componen…

**Figure 10.** Figure 10: LSP distance dependency. LoS/NLoS median curves with shaded 5-95% CI bands for DS, ASA, and ASD versus distance. each bin and parameter X, we generate B = 1000 bootstrap samples by sampling with replacement from the bin contents, recompute the median for each sample, …

**Figure 11.** Figure 11: Spatial consistency. Bars show the estimated decorrelation distance. The vertical error bars denote the 95% confidence intervals of the decorrelation distance. index blocks Ij , a set of indices {j, j + 1, . . . , j + L − 1}, selected uniformly with boundary wrap-arou…

**Figure 12.** Figure 12: Frequency continuity. Markers represent the measured data (blue circles for LoS and red squares for NLoS, and the filled markers for this work), while the lines indicate the corresponding fitted frequency-dependent models (blue for LoS and red for NLoS); the dashed li…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Environment-Aware Channel Prediction for Vehicular Communications: A Multimodal Visual Feature Fusion Framework
cs.CV 2026-04 unverdicted novelty 6.0 of 10

A three-branch multimodal fusion network predicts path loss, delay spread, angular spreads, and angular power spectrum from vehicle camera images and GPS in urban V2I scenarios.

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.