{"id":"9dbed54a-6472-404f-9145-50fb174ba312","arxiv_id":"2412.12306","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"First double-directional ultra-wideband measurements in 6-18 GHz show below-free-space path loss and frequency-stable delay and angular spreads in an urban LoS microcell.","lead":"This paper reports the first ultra-wideband double-directional channel measurements in the 6-18 GHz upper mid-band, taken in an urban street canyon with a 20 meter tall transmitter. It finds path loss below free-space predictions and stable delay and angular spreads across 1 GHz sub-bands, informing 5G/6G system design for Frequency Range 3.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (5) builds the 'omni' PDP by per-delay max-hold over azimuth directions, not by summing power over directions, so the below-free-space path-loss claim is not established for a true omnidirectional response; the max-hold bias is downward relative to a power sum, not upward as the reader states.","rationale":"The reader identified the right equation but mischaracterized the direction of the bias: max over directions is not larger than the sum over directions, so the constructed 'omni' path gain is a lower bound on a true omni power sum, not an upward bias. Nevertheless, the load-bearing concern is real: Eq. (5) defines the dependent variable for the paper's central path-loss claim, and that variable is a max-hold composite, not an omnidirectional response. This makes the fitted omni path-loss model in Table III and the comparison to free space in Figure 4 processing-dependent and not representative of a standard omni antenna. The manuscript also acknowledges broad confidence intervals and deferred calibration details, but the Eq. (5) processing is the most concrete and checkable threat to the headline result. A straightforward recomputation from the stored directional PDPs can settle whether the below-free-space conclusion survives for a true omni synthesis. The reader's CONDITIONAL verdict remains appropriate, so no verdict change is recommended; the condition should be reworded around true omnidirectional synthesis rather than an upward bias.","tokens_in":11120,"tokens_out":10963,"duration_ms":112752,"concrete_test":"Recompute the omni PDP for each Rx location and each 1-GHz sub-band as a power sum over all measured Tx/Rx/elevation bins, e.g., P_true(tau; d) = sum over phi_Tx, phi_Rx, theta_i of P(tau, phi_Tx, phi_Rx, theta_i; d), using the same calibrated H and noise threshold as in Eqs. (1)-(3). Refit the omni path-loss model of Table III and compare to Friis. If the power-summed omni path loss remains below Friis at the same distances and the fitted alpha and beta shift by less than the 95% confidence intervals, the qualitative below-free-space claim stands; if the shift is large or some points cross above Friis, the conclusion should be restated as specific to the max-hold metric. Also report the ratio of summed power to max-hold power per delay bin to quantify the processing dependence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline quantitative result -- omni path loss below free space due to multipath -- is computed from P_omni in Eq. (5), which is not an omnidirectional PDP. For each delay bin the formula selects the largest power over all Tx/Rx azimuth pairs after summing the five Rx co-elevations. A true omnidirectional receive response is an integral/sum of incident power over all directions; since max is less than or equal to the sum, P_omni(tau) is at most P_true_omni(tau). Thus the reported omni path gain is a lower bound on true omni power, not an upper-biased estimate, contrary to the reader's wording. The consequence is that the fitted alpha, beta, and shadowing sigma in Table III are parameters of a max-hold beam-selection metric, not of an omni antenna; they depend on the 10-degree angular grid, the horn beamwidth, the thresholding in Eq. (3), and the fact that the Tx scans only a 120-degree sector. Comparing these fitted values to Friis free-space predicts a property of a hypothetical angle-tracking receiver, not of an omni terminal. The qualitative claim that multipath lowers path loss below free space may survive -- a true power sum can only add power -- but the manuscript does not establish it with the metric actually used, and the plotted and modeled 'omni' path-loss curves are not directly portable to system analysis. The fix is to recompute the omni PDP as a power sum over all measured directions, with the same reference-gain normalization as Eq. (1), and re-fit Table III and Figure 4.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes a 6-18 GHz double-directional channel sounding campaign in an urban LoS microcellular street canyon. The authors sweep Tx/Rx azimuth and Rx elevation angles, construct directional PDPs, combine them into 'omni' and 'best-beam' PDPs, and fit path-loss, shadowing, delay-spread, and angular-spread statistics for the full band and for 1-GHz sub-bands. The headline claims are that this is the first double-directional UWB measurement campaign in the upper mid-band; that both omni and best-beam path loss fall below the Friis free-space prediction due to multipath; and that delay spread and angular spread are stable across the 1-GHz sub-bands.","tokens_in":11475,"tokens_out":8642,"duration_ms":78523,"significance":"If the data and processing are reliable, the campaign provides a scarce resource: double-directional ultra-wideband measurements in the 6-18 GHz FR3 candidate band, with enough bandwidth to study sub-band dependence. The visual validation of identified MPCs against the RTH and GER building reflections and the LoS distance in Figs. 2-3 is credible and gives confidence in the angular association and the measured delays. The paper is also honest that all model parameters are fits to the measured data and does not present derivation-based predictions. However, the 'omni' quantity defined in Eq. (5) is not a true omnidirectional response, so the fitted 'omni' path-loss parameters in Table III are not directly portable to system analysis. This is the main load-bearing issue and it can be addressed by recomputing the omni PDP as a power sum over directions.","major_comments":[{"comment":"The 'omni-directional' PDP is defined by taking, for each delay bin, the maximum over Tx/Rx azimuth pairs after summing the five Rx co-elevations, rather than by summing power over all directions. Since a maximum is no larger than the corresponding sum, P_omni(tau) is a lower bound on a true omni-directional PDP, and the resulting path gain is a lower bound on the true omni path gain; the stress-test concern therefore lands, but the direction is the opposite of the reader's phrasing of an upward bias. The substantive problem stands: the fitted alpha, beta, and sigma in Table III and Fig. 4a describe a per-delay best-beam-selection metric, not an omnidirectional antenna, and they depend on the 10-degree angular grid, the horn beamwidth, the thresholding in Eq. (3), and the fact that the Tx scans only a 120-degree sector. Please recompute P_omni as a power sum over all measured directions with the same reference-gain normalization as Eq. (1), or relabel the metric and restrict all 'omni' claims accordingly. The qualitative below-free-space claim would become stronger, not weaker, under a true power-sum omni PDP, so that part of the abstract may survive the correction; however, the numerical model in Table III is not an omni model.","section":"III, Eq. (5)"},{"comment":"The absolute path-loss comparison in Fig. 4 and the fitted alpha/beta values require an explicit definition of path gain and of how the horn antenna gains and the OTA reference are removed. The manuscript only states that the parameters are defined in prior work [21], [22]. Without the formula linking the calibrated transfer function of Eq. (1) to PG and PL, and without stating the reference-gain basis of the Friis comparison, the reader cannot verify that the reported path loss is on an isotropic-gain basis. Please include the explicit path-gain/path-loss definition and the calibration normalization used in Fig. 4 and Tables III-V.","section":"II-A and III, Eq. (1)"},{"comment":"The 'All Bands' row of Table VII reports mu = -97.14 dB s and sigma = 2.40 dB s, while every 1-GHz sub-band row reports mu near -87.3 dB s and sigma near 11.7 dB s. Pooling the sub-band samples should not shift the mean by 10 dB or shrink the standard deviation by roughly a factor of five; this entry appears erroneous. The full-band statistics feed the stability claim in the abstract and conclusions, so this entry should be corrected or the discrepancy explained.","section":"IV-D, Table VII"}],"minor_comments":[{"comment":"The expression [P_calc(tau) : (tau <= tau_gate) AND (P_calc(tau) >= P_lambda)] does not specify what happens to samples that fail the conditions; please state explicitly that they are discarded or set to zero.","section":"III, Eq. (3)"},{"comment":"In Eq. (4), 'arg max' returns the indices of the maximizing beam pair, but the left-hand side is a function of delay; please write P_Max-Dir(tau) as the value of P at the maximizing indices.","section":"III, Eq. (4)"},{"comment":"There is a typo in the paragraph on Rx6: 'partial obstruction pf the LoS' should read 'partial obstruction of the LoS'; similarly, in Section IV-B, 'attenuated it is partially hitting' is ungrammatical and should be rephrased.","section":"IV-C"},{"comment":"Because the path-loss fits use only six Rx locations, the 95% confidence intervals are very broad; the text acknowledges this, but the table captions should state the number of locations so that readers do not over-interpret the fitted exponents.","section":"Tables III-IV"},{"comment":"The statement that the lower bound of the measured angular spread follows from the beamwidth of the directional antenna should be quantified, since the 10-degree angular grid also imposes a discretization limit on the resolvable angular spread.","section":"IV-E"}],"recommendation":"major_revision","confidential_remarks":"The paper is a measurement-report contribution whose main value is the dataset and the physical validation of the identified MPCs; the number of locations is small, and the authors should be encouraged to state the preliminary nature of the fitted parameters more prominently. The 'first double-directional UWB campaign in the upper mid-band' novelty claim should be checked against recent FR3 work, including [20], and softened if any overlapping double-directional UWB dataset has appeared. The self-citations [21]-[23] for processing methodology are appropriate and not a concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First: this is a genuinely new measurement. Prior double-directional upper-mid-band campaigns were wideband (up to 1 GHz); this one gives 12 GHz instantaneous bandwidth across 6–18 GHz with both ends directionally resolved, over 14,000 directional PDPs. The visual validation of PDP/APS peaks against known building reflections (RTH, GER) is convincing and is the kind of sanity check that makes propagation data trustworthy. The paper is honest about its limitations: it flags broad confidence intervals, notes the Tx scans only a 120° sector, and labels all fitted parameters as fits, not predictions.\n\nThe main soft spot is Eq. (5). The 'omni' PDP is not an omni PDP. It is a per-delay max-hold over Tx/Rx azimuth pairs after summing the five Rx elevations. That is a beam-selection metric, not a power sum. The stress-test note correctly catches a mistake in the reader's take: max is ≤ sum, so the max-hold power is a lower bound on the true omni power, and the reported omni path loss is an upper bound. That means the headline claim—path loss below free space due to multipath—is actually conservative and likely survives a true power-sum recomputation. What does not survive is the portability of Table III's α, β, and σ: those parameters describe a hypothetical angle-tracking receiver with a 10° grid and a 120° Tx sector, not an omni terminal. The fix is easy: recompute P_omni as a sum over all measured directions (or at least over all Rx azimuths), re-fit the tables, and relabel the original metric as something like 'max-hold directional PDP.' This is a moderate revision, not a fundamental flaw.\n\nSecond soft spot: the absolute calibration is deferred to prior papers. Dividing by a daily OTA measurement at 44 m is described in one sentence; the reader cannot tell whether the reported path-loss intercepts include antenna gains or which free-space reference is used. Given the tables are the main quantitative output, that needs a fuller treatment.\n\nThird: six Rx locations. With six points, the path-loss fits have enormous confidence intervals (Table III shows α from ~40 to ~75 in some bands). The authors acknowledge this, but the frequency-stability claim—spread and angular spread roughly constant across sub-bands—rests on overlapping CIs from one location set. Fine as a preliminary observation; not a settled statistical statement.\n\nWho is this for: people building FR3 channel models and MIMO design studies, and propagation researchers who need a reference dataset. It deserves a serious referee; the measurement effort is real, the dataset is new, and the main weakness (Eq. 5) is fixable without new measurements. Send it to review with a request to clarify the omni construction and the calibration.","headline":"First real double-directional UWB dataset in 6–18 GHz, worth refereeing, but the 'omni' PDP is a max-hold beam-selection metric, not a power sum—so the fitted omni path-loss parameters are not directly portable, even though the below-free-space result is conservative.","tokens_in":12062,"tokens_out":3775,"would_cite":true,"duration_ms":33586,"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":"First double-directional UWB channel maps of the 6–18 GHz upper mid-band show street-canyon multipath pulling path loss below free space.","keywords":["upper mid-band","FR3","double-directional channel measurements","ultra-wideband channel sounding","path loss","delay spread","angular spread","street canyon"],"falsifier":"Rebuild the omni PDP by summing received power over all measured azimuth and elevation directions at each delay bin instead of selecting the strongest azimuth, then recompute path gain and path loss; if the summed path loss rises to or above the free-space curve, the below-free-space result depends on the max-hold processing rather than on true omnidirectional reception. A complementary check would be a control measurement in an open field with no building reflections.","tokens_in":10937,"feed_emoji":"📡","tokens_out":9650,"duration_ms":82680,"temperature":0.7,"pith_summary":"This paper sets out to give the first double-directional (direction-resolved at both link ends) ultra-wideband channel characterization of the 6–18 GHz upper mid-band, the spectrum being considered for future FR3 systems, using measurements in an urban line-of-sight microcellular street canyon. It reports three headline findings: multipath from buildings on both sides of the street makes path loss lower than the free-space prediction for both the best-beam and the constructed omni-directional profiles; RMS delay spread and angular spread stay nearly unchanged across every 1 GHz sub-band from 6 to 18 GHz; and shadowing is small. These results matter because system designers deciding whether FR3 bands can reuse one channel model, and whether free-space link budgets are safe, currently lack double-directional UWB data in this range. The campaign's over 14,000 directional power delay profiles provide a first reference point for such modeling.","feed_headline":"Street canyons beat free-space path loss at 6–18 GHz","feed_subtitle":"New double-directional measurements in a Los Angeles microcell show delay and angular spread barely move across the band.","key_machinery":"The machinery is a double-directional ultra-wideband channel sounder: a VNA that records the complex transfer function over 6–18 GHz in 12,001 frequency points, front-ended by high-gain horn antennas on precision positioners, so every link is resolved in Tx azimuth, Rx azimuth, and Rx elevation. A time-gated over-the-air calibration isolates system and antenna effects from the channel response. From the calibrated transfer functions the paper forms directional power delay profiles by inverse Fourier transform, picks the strongest beam pair (Max-Dir) by total power, and builds an 'omni' PDP by summing over the five Rx co-elevations while selecting the strongest Tx and Rx azimuth per delay bin. Path loss, RMS delay spread, and the angular-spread metric are then fitted with power-law and Gaussian models per 1 GHz sub-band.","core_discovery":"In a line-of-sight microcellular street-canyon scenario, over the 6–18 GHz upper mid-band, the paper reports the first double-directional ultra-wideband measurement campaign, built from more than 14,000 directional power delay profiles collected with a 12 GHz-bandwidth RF-over-fiber VNA sounder and horn antennas rotated in azimuth and elevation at both link ends. It finds that path loss for both the best-beam (Max-Dir) and the constructed omni-directional power delay profile is below the free-space value, which the authors attribute to multipath contributions from buildings on both sides of the street; the only exception is a receiver partially obstructed by tree branches, which shows loss above free space. Path loss still rises with distance and frequency, but remains below the free-space curve over the observed range. It further finds that RMS delay spread and angular spread at the transmitter and receiver are nearly the same across every 1 GHz sub-band as across the full band, indicating that the channel's time and angular dispersion are stable over this frequency range.","pith_inferences":["Editorial inference: re-summing power over all measured azimuth directions at each delay bin, instead of taking the strongest one, would lower the constructed omni path gain; the below-free-space omni path loss may shrink or disappear under true power summation.","Editorial inference: with only six receiver points spanning 59–185 m, the distance dependence of delay and angular spread is weakly constrained; the paper's own linear fits have confidence intervals straddling zero, so the stability claim is better supported across frequency than across distance.","Editorial inference: the same double-directional dataset could be reprocessed to extract per-cluster angles and delays for ray-tracing validation at FR3, or to test whether a single stochastic cluster model holds across the 12 GHz bandwidth."],"forward_implications":["LoS microcellular links in street canyons can expect path loss below the free-space value across 6–18 GHz, so free-space-based link budgets are conservative for these deployments.","Best-beam reception spatially filters late reflections, so beamformed links see lower RMS delay spread than the omni profile; the delay-spread gap between Max-Dir and omni is a direct consequence of this filtering.","Delay spread and angular spread statistics change little across 1 GHz sub-bands, meaning one set of model parameters could serve the whole 6–18 GHz range for LoS microcells.","Fitted shadowing standard deviations are small, mostly under 3 dB, so LoS coverage predictions can use a tight lognormal margin.","The one tree-obstructed receiver shows loss above free space, indicating vegetation can reverse the multipath gain and deserves separate treatment in FR3 models."],"supporting_citations":[{"why":"Prior wideband outdoor measurements at 3–18 GHz; provides the baseline for comparing delay-spread stability.","marker":"[11]"},{"why":"Earlier double-directional measurements at 11 GHz indoors; the 'first UWB double-directional' claim positions itself against this class.","marker":"[17]"},{"why":"Recent urban outdoor propagation measurements at 6.75 and 16.95 GHz that this campaign extends with continuous UWB double-directional coverage.","marker":"[20]"},{"why":"Supplies the RF-over-fiber VNA channel-sounder architecture used in this campaign.","marker":"[21]"},{"why":"Supplies the measurement setup and parameter definitions for path gain, path loss, delay gating, and shadowing.","marker":"[22]"},{"why":"Supports the choice of noise and delay thresholds in the PDP processing.","marker":"[23]"},{"why":"Provides the max-hold construction used to build the 'omni' PDP by selecting the strongest azimuth per delay bin.","marker":"[24]"},{"why":"Defines path gain, path loss, RMS delay spread, and the lognormal shadowing model used in the fits.","marker":"[25]"},{"why":"Supplies the angular-spread metric used for the transmitter and receiver angular statistics.","marker":"[26]"}],"fun_headline_variants":["6–18 GHz path loss below free-space in street canyon","Delay and angular spread stable across 6–18 GHz","First double-directional UWB channel map of upper mid-band","Multipath beats free-space path loss in microcell at 6–18 GHz","Upper mid-band microcell: spreads steady, loss below free space"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The below-free-space path loss result rests on defining the 'omni-directional' profile by selecting, at each delay, the strongest azimuth direction rather than adding power from all directions; that max-hold choice can only raise collected power, so the conclusion depends on this processing proxy rather than on a true omnidirectional antenna.","fun_headline_variants_meta":{"raw":{"variants":["6–18 GHz path loss below free-space in street canyon","Delay and angular spread stable across 6–18 GHz","First double-directional UWB channel map of upper mid-band","Multipath beats free-space path loss in microcell at 6–18 GHz","Upper mid-band microcell: spreads steady, loss below free space"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000463,"raw_usage":{"total_tokens":2303,"prompt_tokens":926,"completion_tokens":1377,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":1286}},"tokens_in":542,"tokens_out":1377,"duration_ms":12547,"temperature":1.0,"reasoning_tokens":1286,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:13:08.965384+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rebuild the omni PDP by summing received power over all measured azimuth and elevation directions at each delay bin instead of selecting the strongest azimuth, then recompute path gain and path loss; if the summed path loss rises to or above the free-space curve, the below-free-space result depends on the max-hold processing rather than on true omnidirectional reception. A complementary check would be a control measurement in an open field with no building reflections.","supporting_citations":[],"review_version":1}