{"id":"977cc9f6-a551-4abc-9916-703c3e57802f","arxiv_id":"2412.20755","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"An 8 GHz-bandwidth double-directional channel measurement campaign in an urban microcell at 6-14 GHz produces statistical models for path loss, delay spread, and angular spread, and highlights vegetation-obstructed LoS as a distinct category.","lead":"This paper reports a 6-14 GHz outdoor radio channel measurement campaign in an urban street, using rotating directional antennas at both link ends to capture signal directions and delays. It fits statistical models for path loss, delay spread, and angular spread and argues that tree-blocked 'OLoS' links need their own propagation category, which matters for designing 5G and 6G networks in the upper midband.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The pseudo-omni PDP in Eq. (5) takes the maximum over azimuth per delay bin rather than summing all directions; in OLoS this can discard comparable multipath power and bias the path-loss and delay-spread statistics that the paper's conclusions rest on.","rationale":"The paper is a solid measurement report: the setup is described in detail, the Friis check on the LoS point supports the calibration, and the limitations (11 locations, 1 LoS point, no data release) are disclosed. The central quantitative claims—path loss exponents, shadowing, delay spread—rest on the pseudo-omni PDP of Eq. (5). The max-over-azimuth operation is not equivalent to an omni antenna: at each delay it keeps only the strongest azimuth bin, so any comparable power arriving from other azimuths in the same delay bin is discarded. The applied gain correction accounts only for the elevation-sum pattern, not for this truncation. In LoS the direct component dominates, so the Friis check passes; in OLoS, where the direct path is heavily attenuated and reflections are comparable, the truncation could bias the total path gain and the delay weighting. This directly affects Tables III-IX and the OLoS-specific conclusions. The concern is testable with the existing data: sum over all azimuth pairs instead of taking the max and compare. No new measurements are needed. If the sum-based path gain differs by more than about 1 dB at OLoS locations, the fitted exponents and the RMSDS frequency-stability claim need revision. The paper would be strengthened by reporting this check, and the conditional verdict should require it. The reader's weakest assumption identified the same Eq. (5) construction, and this stress-test concurs; the appropriate verdict remains conditional acceptance pending that validation.","tokens_in":22484,"tokens_out":10209,"duration_ms":111433,"concrete_test":"Using the saved double-directional transfer functions, recompute the omni PDP for all 11 locations as P_sum(τ;d) = Σ_{φTx} Σ_{φRx} Σ_k P(τ, φTx, φRx, θk; d) with the appropriate normalization and elevation-gain correction, then compare the resulting path gain and RMSDS with the max-based Eq. (5) values. Report the per-location and per-band differences in dB and dB s. If the omni path gain at any OLoS location increases by more than 1 dB, the PL exponents in Tables III/IV are biased; if the mean RMSDS shifts by more than 1 dB s, the frequency-stability conclusion needs re-examination. Also recompute the Friis check at Rx1 with the sum construction to confirm the calibration still holds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is the pseudo-omni PDP construction in Eq. (5). For each delay bin the code selects the strongest Tx-Rx azimuth pair (after summing the five Rx co-elevations) rather than summing over azimuth, and only a constant gain correction is subtracted. A true omni-directional antenna would sum the power from all directions at each delay; the max operation discards the power in non-best azimuth pairs. In a sparse LoS channel this bias is negligible, which is why the Friis check at Rx1 passes, but in OLoS, where the direct path is attenuated and many comparable reflected components arrive, the discarded azimuthal power can be a substantial fraction of the total. The bias affects the omni path gain used in the PL fits (Tables III-V) and the delay weighting used in the RMSDS fits (Tables VI-IX); the fitted exponents around 3.9-4.6 and the near-frequency-flat RMSDS conclusions would both be affected. The paper does not report any validation of Eq. (5) against a sum-over-azimuth synthesis in a multipath-rich environment, so the magnitude of the bias is unknown. A secondary effect is that the Tx azimuth scan only covers ±60°, so paths departing outside this sector are missing from the 'omni' PDP; this also biases the path loss and is not quantified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an ultra-wideband (6-14 GHz) double-directional channel measurement campaign in an urban microcellular environment, using a VNA with radio-over-fiber haul and rotating horn antennas at both link ends. The authors analyze more than 25,000 directional power delay profiles over 11 receiver locations (1 LoS, 10 OLoS), and derive statistical models for path loss, shadowing, RMS delay spread, and angular spread, reporting 95% confidence intervals for all fitted parameters. They also provide sample PDP and APS analyses, and argue that OLoS must be treated as a separate propagation category in the upper midband. The paper claims to be the first UWB double-directional outdoor measurement campaign in this frequency range.","tokens_in":22753,"tokens_out":7356,"duration_ms":73917,"significance":"If the measurement and processing are sound, the dataset is a valuable contribution to FR3 channel modeling, spectrum allocation, and beamforming studies, because it fills a gap in double-directional UWB outdoor data between 6 and 14 GHz. The paper is transparent about its calibration procedure (daily OTA reference checked against Friis), reports confidence intervals throughout, and explicitly lists limitations in the conclusions. The descriptive PDP and APS results are clearly presented and illustrate the importance of vegetation-induced OLoS effects. However, two load-bearing issues need attention: the pseudo-omni PDP construction and the very small number of independent locations; both directly affect the fitted statistics that support the paper's main claims.","major_comments":[{"comment":"The pseudo-omni PDP is constructed by taking, for each delay bin, the maximum over Tx and Rx azimuth of the elevation-summed power. In an OLoS environment with multiple comparable reflections arriving from different azimuths, this discards power that a true omni-directional antenna would collect, and the constant gain correction shown in Fig. 3 only accounts for the elevation combination, not the azimuthal selection. Because the omni path gain (Table III) and RMSDS (Table VI) are computed from this PDP, the fitted path loss exponents (approx. 3.9-4.6) and the conclusion of frequency-flat RMSDS could be systematically biased. The very fine delay resolution (0.125 ns) makes it plausible that each delay bin contains at most one specular MPC, but the paper does not justify or validate this construction for a vegetation-rich, diffuse-scattering environment, nor does it quantify the impact of the Tx azimuth mask of only ±60° on the 'omni' path gain. Please either validate Eq. (5) against a sum-over-azimuth synthesis (with appropriate beam-pattern compensation) at representative OLoS points, or provide a sensitivity analysis showing that the bias is negligible; the current evidence is insufficient.","section":"Sec. III-A, Eq. (5), Tables III and VI"},{"comment":"The statistical models are fitted to only 10 OLoS points and 1 LoS point (Table II). The 95% confidence intervals are extremely wide (e.g., α from -5.40 to 49.91 dB at 6-7 GHz in Table III; β from -1.80 to 1.91 dB s for omni RMSDS in Table VI). The paper acknowledges the small number of locations in the conclusions, but the abstract and Section I.C present the fitted models as a central contribution without this caveat. In addition, the statement in Section IV-D that RMSDS is 'relatively stable across frequency bands' is not strongly supported by the data: the CIs for the mean μ in Tables VII and IX overlap substantially across bands, and the distance slopes β in Tables VI and VIII cross zero in most bands. Please rephrase the model claims as preliminary, add a statistical power analysis, or reduce the emphasis on specific parameter values as general models.","section":"Sec. IV-C, Tables II, III, VI, VII"},{"comment":"The paper claims to be 'the first UWB double-directional measurement campaign in this frequency range.' This is contradicted by the authors' own companion paper [39], which reports ultra-wideband double-directionally resolved channel measurements of LoS microcellular scenarios in the upper midband. The more specific claim in Section V (first double-directional UWB campaign in outdoor OLoS scenarios) may be correct, but the broader claims in the abstract and Section I.C should be qualified to avoid overstatement. Please update the novelty statement to be consistent with the existing companion work.","section":"Abstract and Sec. I.C"}],"minor_comments":[{"comment":"The text 'Fig. 3 uses a threshold with fixed absolute level' appears to be a cross-reference error; Fig. 3 shows the antenna elevation pattern, while the threshold discussion concerns the PDP figures (Fig. 4). Please correct the reference.","section":"Sec. IV-A"},{"comment":"The notation 'P_Max-Dir(τ) = arg max_{i,j,k} ...' is inconsistent because the left-hand side is a PDP while arg max returns indices. Please rewrite the equation to first define the selected index triple and then set P_Max-Dir to the corresponding directional PDP.","section":"Eq. (4)"},{"comment":"The abstract contains a grammatical error: 'We analyze over 25,000 directional power delay profiles and providing key insights' should read 'and provide key insights'.","section":"Abstract"},{"comment":"The caption contains the typo 'Elvation Pattern'; it should be 'Elevation Pattern'.","section":"Fig. 3 caption"},{"comment":"The 'All Bands' mean RMSDS for the omni case is -84.54 dB s, whereas the band-wise means in the same table are around -82 dB s; please clarify why the all-bands mean differs from the band-wise means by about 2 dB, or correct the value.","section":"Table VII"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope and the raw measurement campaign is valuable. The main revision burden is the validation of Eq. (5) and the tightening of statistical claims given the small sample size. Please also ensure the novelty claim is aligned with the authors' companion paper [39]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's real contribution is the OLoS double-directional dataset in 6–14 GHz outdoor urban microcells: 10 OLoS locations, 8 GHz bandwidth, 25k directional PDPs, with path loss, shadowing, delay spread, and angular spread fits. That is genuinely new, and the measurement work itself is credible. The Friis check on the LoS point, the daily OTA calibration, and the honest listing of limitations in Sec. V all suggest the authors know what they are doing. If you need a first-cut OLoS model for FR3 system design, this is the best public source right now.\n\nThe soft spots are real but not fatal. First, the “first” claim needs qualification: their companion paper [39] already reported UWB double-directional LoS in the same band, so the novelty is specifically OLoS. Second, the statistical fits rest on 10 OLoS points plus 1 LoS point; the 95% confidence intervals in Tables III–V are wide enough that the point estimates (e.g., β ≈ 3.9–4.6) should be read as environment-specific initial values, not generalizable constants. Third, and most important, the pseudo-omni PDP in Eq. (5) does not actually emulate an omni antenna: it takes the maximum over Tx/Rx azimuth per delay bin rather than summing the power from all directions. In a sparse LoS channel the direct path dominates, so the Friis check passes at Rx1. But in OLoS, where several comparable reflected components can arrive at similar delays from different azimuths, the max operation systematically discards power. That biases the omni path loss fits and the RMSDS values, and the Tx azimuth scan only covers ±60°, so paths leaving outside that sector are missing entirely. The paper does not validate Eq. (5) against a sum-over-azimuth synthesis in a multipath-rich case, so the magnitude of the bias is unknown. It is a load-bearing assumption for the numbers in Tables III–VII, though the qualitative LoS/OLoS contrast likely survives.\n\nWho is this for? People building initial 3GPP-style upper-midband channel models and looking for a sanity-check dataset. It deserves a serious referee, with the expectation of major revision: qualify the novelty, add a sensitivity analysis of the pseudo-omni construction (or at least bound the bias), and either release the data or plan to. As is, it's a useful but provisional measurement report.","headline":"Useful first OLoS double-directional dataset for 6–14 GHz urban microcells, but the pseudo-omni PDP construction and the small location count make the fitted numbers provisional rather than definitive.","tokens_in":23433,"tokens_out":2912,"would_cite":true,"duration_ms":28771,"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 the 6–14 GHz urban microcell channel has its own propagation character, and that vegetation-obstructed sightlines must be modeled as a separate category from clear line-of-sight and non-line-of-sight links.","keywords":["FR3","upper midband","ultra-wideband measurements","double-directional channel sounding","obstructed line-of-sight","path loss modeling","delay spread","angular spread"],"falsifier":"Run the same double-directional scan at one location while recording the channel with a true omni-directional antenna; if the constructed pseudo-omni PDP's total power differs from the omni measurement by more than the calibration uncertainty, the Eq. (5) construction is biased and the fitted path-loss and delay-spread statistics would need recomputation.","tokens_in":22204,"feed_emoji":"📡","tokens_out":8492,"duration_ms":81739,"temperature":0.7,"pith_summary":"The paper tries to establish that the 6–14 GHz upper-midband channel in urban microcells behaves differently from what existing sub-6 GHz and millimeter-wave data would suggest, and that this band needs its own measurements. It reports what it describes as the first outdoor ultra-wideband double-directional measurement campaign in this frequency range, analyzing more than 25,000 directional power delay profiles across eleven locations. The measurements show that vegetation-obstructed line-of-sight links deviate strongly from free-space expectations and from clear LoS links, so the paper argues OLoS should be its own category. If the claim is right, regulators and system designers gain the first statistical basis for path loss, shadowing, delay spread, and angular spread in FR3.","feed_headline":"First 6–14 GHz double-directional channel data for cities","feed_subtitle":"25,000 directional profiles put vegetation-blocked links in their own channel class.","key_machinery":"The load-bearing construction is the pseudo-omni power delay profile, Eq. (5), which for each delay bin sums the five Rx elevation tilts and then picks the strongest Tx–Rx azimuth pair, followed by a frequency-dependent gain correction that converts the virtual antenna pattern to omni-equivalent power. This reduces the 2340 directional transfer functions measured at each location to a single omni-directional PDP, and every fitted quantity in the paper—path loss, shadowing, RMS delay spread, and angular spread—is derived from it. The contrasting Max-Dir PDP, Eq. (4), selects the single beam pair with the highest total power and supplies the directional statistics.","core_discovery":"On the paper's own terms, the central discovery is that the 6–14 GHz urban microcell channel cannot be treated as an interpolation of lower- and higher-band behavior. Using the new campaign, it reports omni-directional path-loss exponents around 3.6–4.6 and Max-Dir exponents around 3.6–4.3, both well above free space; root-mean-square delay spread that stays roughly flat across frequency, with omni mean near −84.5 dB s and Max-Dir mean near −89.8 dB s; and angular spreads that shrink only modestly with distance and vary little with frequency. The OLoS points, where foliage blocks the direct path, show received powers 25 dB or more below Friis predictions, with reflected clusters carrying much of the energy. The paper concludes that OLoS must be separated from LoS and NLoS in upper-midband modeling.","pith_inferences":["An untested extension of the paper's logic is that vegetation, not geometry, drives most of the OLoS excess loss; a campaign that varies foliage density while holding distance fixed would separate the two.","It follows implicitly that the pseudo-omni max-per-delay-bin construction could overestimate omni power if two equal-strength paths arrive from different azimuths in the same delay bin; validating against a true omni antenna at one point would quantify this.","If the frequency-stable delay-spread result generalizes, the standard-model assumption that delay spread decreases with frequency would need revision for FR3.","With only one clear-LoS location, the paper's LoS-versus-OLoS contrast rests on a single point; additional LoS routes would show whether the reported exponent gap is universal."],"forward_implications":["Upper-midband system design must use measured FR3 parameters rather than interpolating a standard model built from sub-6 GHz and above-24 GHz data, since the 6–14 GHz channel has its own path-loss and dispersion behavior.","OLoS should enter channel models as its own category; mixing vegetation-blocked links with clear LoS and NLoS would bias path-loss exponents and shadowing variances.","Equalizer and scheduling designs can treat RMS delay spread as roughly flat across 6–14 GHz in this environment, since the measured means change by only about 2 dB from 6–7 GHz to 13–14 GHz.","Beam management can rely on a transmit angular spread that shrinks slowly with distance and a wider receive angular spread from the 360-degree scan, with weak frequency dependence.","Sub-band-specific path-loss fits give regulators and operators a frequency-resolved basis for deciding which parts of FR3 to allocate or aggregate."],"supporting_citations":[{"why":"Supplies the prior outdoor UWB SISO measurements in 3–18 GHz that this campaign extends by adding double-directional resolution.","marker":"[20]"},{"why":"Provides the closest prior outdoor double-directional microcell measurements at 11 GHz, which are wideband but not UWB.","marker":"[33]"},{"why":"Companion study of vegetation impact in the upper midband, used to attribute the OLoS excess attenuation to foliage.","marker":"[5]"},{"why":"Standard model that claims validity across 0.6–100 GHz but is based on sub-6 GHz and above-24 GHz data, motivating the 6–14 GHz gap this paper fills.","marker":"[2]"},{"why":"Supplies the max-over-azimuth construction approach that the pseudo-omni PDP follows in Eq. (5).","marker":"[53]"},{"why":"Provides the textbook definitions of path loss, delay spread, and angular spread used throughout the processing.","marker":"[47]"}],"fun_headline_variants":["First 6–14 GHz city channel map from 25k profiles","Upper-midband: 6–14 GHz OLoS needs distinct channel stats","City 6–14 GHz: first double-directional channel sweep","Foliage shifts 6–14 GHz channel behavior in this first study","6–14 GHz urban channels: OLoS and LoS diverge sharply"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Every fitted statistic rests on Eq. (5)'s assumption that a delay-bin-wise maximum over azimuth directions, after elevation summing and gain correction, represents the total power a true omni-directional antenna would receive.","fun_headline_variants_meta":{"raw":{"variants":["First 6–14 GHz city channel map from 25k profiles","Upper-midband: 6–14 GHz OLoS needs distinct channel stats","City 6–14 GHz: first double-directional channel sweep","Foliage shifts 6–14 GHz channel behavior in this first study","6–14 GHz urban channels: OLoS and LoS diverge sharply"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000739,"raw_usage":{"total_tokens":3273,"prompt_tokens":888,"completion_tokens":2385,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":2286}},"tokens_in":504,"tokens_out":2385,"duration_ms":18887,"temperature":1.0,"reasoning_tokens":2286,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:12:46.071347+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same double-directional scan at one location while recording the channel with a true omni-directional antenna; if the constructed pseudo-omni PDP's total power differs from the omni measurement by more than the calibration uncertainty, the Eq. (5) construction is biased and the fitted path-loss and delay-spread statistics would need recomputation.","supporting_citations":[{"cited_title":"Outdoo r wideband channel measurements and modeling in the 3–18 ghz band,","cited_arxiv_id":null,"evidence_quote":"Supplies the prior outdoor UWB SISO measurements in 3–18 GHz that this campaign extends by adding double-directional resolution."},{"cited_title":"11 ghz band mimo chan nel char- acteristics in a street micro-cell environment,","cited_arxiv_id":null,"evidence_quote":"Provides the closest prior outdoor double-directional microcell measurements at 11 GHz, which are wideband but not UWB."},{"cited_title":"An Ultra-Wideband Study of Vegetation Impact on Upper Midband / FR3 Communication","cited_arxiv_id":"2412.17864","evidence_quote":"Companion study of vegetation impact in the upper midband, used to attribute the OLoS excess attenuation to foliage."},{"cited_title":"5G; Study on channel model for frequencies from 0. 5 to 100 GHz,","cited_arxiv_id":null,"evidence_quote":"Standard model that claims validity across 0.6–100 GHz but is based on sub-6 GHz and above-24 GHz data, motivating the 6–14 GHz gap this paper fills."},{"cited_title":"Syn- chronous channel sounder using horn antenna and indoor meas urements on 28 GHz,","cited_arxiv_id":null,"evidence_quote":"Supplies the max-over-azimuth construction approach that the pseudo-omni PDP follows in Eq. (5)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the textbook definitions of path loss, delay spread, and angular spread used throughout the processing."}],"review_version":1}