{"id":"cc087d42-2247-40b8-9a3e-44b63b482013","arxiv_id":"2502.00699","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"At 28 GHz, fitting scattering models with only in-plane receiver data mispredicts out-of-plane power, so building-surface scattering parameters should be fitted with 3D receiving positions.","lead":"Researchers measured 28 GHz radio scattering from four real building surfaces at multiple angles and in three-dimensional receiver space, then fitted two standard scattering models to the data. The paper concludes that millimeter-wave channel models should treat building-surface scattering as an angle-dependent, three-dimensional effect rather than a fixed material property.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section IV-A-4 reports FVU=0.6742 for the in-plane fit evaluated out of plane, but no out-of-plane FVU for the 3D-fit parameters; the claim that 3D fitting partially corrects the error is therefore quantitatively unsupported.","rationale":"The reader's weakest_assumption concerns stray multipath contamination. That is a real risk, but it is not the most decisive gap. Even if the 35 dB suppression of non-wall paths is accepted, Section IV-A-4 still lacks the single quantity needed to substantiate the paper's central claim: the out-of-plane FVU for the 3D-fit parameters. Reporting in-plane FVU=0.6259 for the 3D fit is irrelevant to the error it claims to correct; the in-plane fit already had good in-plane performance by construction. The visual comparison in Fig. 6 is suggestive but not quantified, and the FVU values are high enough (0.63-0.67) that a small absolute change could decide the conclusion. The proposed test is cheap: it requires no new measurements, only a re-run of the existing ray-tracing model and measured power data. Because the missing number can be supplied by the authors and the verdict should remain conditional pending that number, I recommend UNCHANGED. Agreement with the reader is partial: they noted this omission in the rationale but selected the multipath assumption as weakest. My concern is more directly tied to the central claim and is internally checkable from the paper.","tokens_in":8531,"tokens_out":5348,"duration_ms":47829,"concrete_test":"Re-run the Wireless InSite ray tracing with the 3D-fit parameters (S=0.42, alpha_R=6, alpha_i=4, Lambda=0.2) for the rough-wall scene, evaluate FVU over the same out-of-plane Rx positions (delta_h = 10, 20, 30 cm) used in Section IV-A-4, and compare it with the reported 0.6742 for the in-plane-fit parameters. If the out-of-plane FVU is not substantially below 0.6742, the central claim that 3D fitting partially corrects the out-of-plane error is unsupported; if it is substantially lower, the claim is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central 3D-correction claim in Section IV-A-4 rests on a one-sided comparison. The authors report that the incident-plane-only parameter set (S=0.60, alpha_R=1, alpha_i=10, Lambda=0.2) yields FVU=0.6742 when evaluated at out-of-plane Rx heights, and that the 3D-fit parameter set (S=0.42, alpha_R=6, alpha_i=4, Lambda=0.2) 'can partially correct this error.' However, the only FVU given for the 3D fit is 0.6259 'within the incident plane,' which is not the error being corrected. Without the out-of-plane FVU for the 3D-fit parameters, the claim is supported only by visual inspection of Fig. 6. The parameter shift is large (S: 0.60 to 0.42; alpha_R: 1 to 6; alpha_i: 10 to 4), and the reported in-plane FVU is still high (0.6259), so it is not obvious that the 3D fit reduces out-of-plane error. Both fits are evaluated in-sample on the same data used for fitting, so FVU differences can also reflect overfitting rather than genuine generalization. If the 3D-fit parameters do not reduce out-of-plane FVU relative to 0.6742, the headline conclusion is not supported by the data as presented.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports 28 GHz bistatic diffuse-scattering measurements from a metal sheet and three building-wall surfaces (marble, smooth, rough) at incident angles of 20°, 30°, and 40°. The authors parameterize a directive or dual-lobe scattering model in the Wireless InSite ray tracer, using a roughness-based initial scattering coefficient from Eq. (4) and then minimizing the FVU metric of Eq. (7) to obtain S, αR, αi, and Λ for each surface. The central new claim, made in Section IV-A-4, is that fitting with receiver data at multiple heights (3D Rx space) yields different scattering parameters from fitting with incident-plane data alone, and that the 3D fit 'can partially correct' the out-of-plane error of the in-plane-only fit. The paper concludes that 3D receiving positions should be considered when parameterizing scattering from building surfaces.","tokens_in":8824,"tokens_out":6070,"duration_ms":55858,"significance":"The measurement campaign is valuable: real-world building surfaces at 28 GHz, multiple incident angles, and receiver positions outside the incident plane are underrepresented in the literature, and the paper provides a useful parameterization table for the four surfaces. The roughness-based initial S computed from Eq. (4) gives a physically grounded starting point. However, the paper's headline quantitative claim is not supported by the numbers as reported (see major comments), and the high in-sample FVU values mean the model fits are of limited accuracy. If the missing out-of-plane FVU for the 3D fit and a proper validation split confirm the claimed improvement, the conclusion would be an important practical message for mmWave channel modelers.","major_comments":[{"comment":"The claim that the 3D-fitted parameter set (S=0.42, αR=6, αi=4, Λ=0.2) 'can partially correct' the out-of-plane error is not quantitatively supported, because the out-of-plane FVU for this parameter set is not reported. The paper reports only FVU=0.6742 for the in-plane-only fit evaluated out of plane, and FVU=0.6259 for the 3D fit evaluated in the incident plane. The latter is not the metric being corrected. Please report the out-of-plane FVU for both parameter sets, ideally per receiver height, and state whether the 3D fit actually reduces out-of-plane FVU relative to 0.6742; otherwise the conclusion rests on visual inspection of Fig. 6.","section":"Section IV-A-4, Fig. 6"},{"comment":"All scattering parameters are fitted to the same measured positions that are later used for evaluation by minimizing FVU, so the agreement in Figs. 5 and 6 is in-sample. The reported minimum FVU values (0.2223 to 0.6273) are high, meaning a large fraction of the power variation remains unexplained even after fitting. To support the claim that the 3D fit generalizes better than the in-plane fit, please add a cross-validation or out-of-sample evaluation (e.g., fit on in-plane data and evaluate out of plane, or fit on a subset of heights and evaluate on the rest), and compare the fitted model against a baseline with specular reflection only. Also provide an estimate of the uncertainty of the fitted parameters.","section":"Section IV-A-2 and IV-A-4"},{"comment":"The assumption that one-hop reflection and one-hop scattering from the wall under test dominate the received power is load-bearing for the parameter fitting, but it is only stated, not demonstrated, for the out-of-plane receiver positions. The paper says other paths are more than 35 dB below the main path and are treated as noise, but no measured power delay profiles or ray-tracing path lists are shown to verify this at the non-zero receiver heights. If ground or adjacent-structure multipath is not negligible, the fitted S, αR, αi, and Λ will absorb the error, biasing the in-plane versus 3D comparison. Please show the residual multipath level at representative receiver positions or carry out a sensitivity analysis by adding and removing paths in the ray tracer.","section":"Section IV-A-1"}],"minor_comments":[{"comment":"The sentence following Eq. (2) contains a typo: 'the new reflection coefficient the for rough surface' should read 'the new reflection coefficient for the rough surface.'","section":"Section II-A"},{"comment":"The heading 'Method for Calculating Initial Scatting Coefficient S' and the Fig. 6 caption 'Recevied power' contain typos; they should read 'Scattering' and 'Received', respectively.","section":"Section II-B and Fig. 6"},{"comment":"The quantity defined in Eq. (7) is the square root of the fraction of unexplained variance (normalized RMSE), not the usual FVU; please clarify the definition or rename the metric to avoid confusion.","section":"Eq. (7)"},{"comment":"The notation 'αR, αi=1,2,...,10' should be written as 'αR, αi ∈ {1,2,...,10}' to avoid ambiguity about whether the two parameters are equal.","section":"Section II-C"},{"comment":"Please clarify the relationship between the 5 ns maximum excess delay and the 1.5 m path-length-difference criterion, since at the speed of light these two thresholds correspond to the same delay; state which one is used to select paths.","section":"Section IV-A-1"}],"recommendation":"major_revision","confidential_remarks":"The central issue is easily addressable with data the authors already have: report the out-of-plane FVU for the 3D fit and ideally a held-out evaluation. If the authors cannot provide the out-of-plane FVU, they should soften the 'partially corrects' claim to a qualitative observation. The paper is otherwise within scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. First, this is a real measurement paper: 28 GHz scattering from actual building surfaces (metal, marble, smooth wall, rough wall) at multiple incident angles and, for the rough wall, at four receiver heights. That alone is useful—most prior scattering parameterization work stayed in the incident plane at one incident angle, and there is very little public data like this. Second, the paper's headline claim—that 3D receiver sampling is necessary—is qualitatively convincing but quantitatively under-supported. The authors show that the in-plane-only parameter set (S=0.60, alpha_R=1, alpha_i=10, Lambda=0.2) gives FVU=0.6742 out of plane, but they never report the out-of-plane FVU for their 3D-fitted set (S=0.42, alpha_R=6, alpha_i=4, Lambda=0.2). The only number given for that set is FVU=0.6259 in-plane, which is not the error being corrected. The claim that 3D fitting 'partially corrects this error' rests on the visual difference in Fig. 6. That is a fixable gap, but it is the load-bearing number of the paper.\n\nThe fitting is in-sample and FVU values are high (0.22–0.63), so the model doesn't describe the data tightly. No error bars, and the constant-Lambda=0.2 choice in Fig. 7 conflicts with Table II. The single-path assumption (other paths >35 dB below) is plausible given the sounding setup, but it is stated, not demonstrated.\n\nWhat the paper does well: the theoretical S from Eq. (4) using measured roughness and Fresnel coefficients provides partial independent grounding, and the parameter trends across materials and incident angles are coherent. The measurement setup is described clearly enough to reproduce. This is the kind of anchor data that ray-tracing parameter libraries need.\n\nBottom line: the measurement contribution is real and the qualitative conclusion (in-plane-only fitting mispredicts out-of-plane scattering) is probably right. The missing out-of-plane FVU for the 3D fit should be added, and the sensitivity of the parameters to the fitting procedure discussed. For readers working on mmWave channel modeling or ray tracing, this is worth reading and citing. A serious referee can push on the numbers without dismissing the data.","headline":"Real 28 GHz building-scattering measurements with 3D receiver positions; the 3D-correction claim needs the missing out-of-plane FVU before it fully lands.","tokens_in":9384,"tokens_out":1861,"would_cite":true,"duration_ms":17490,"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":"Building-scatter models at 28 GHz need 3D receiver data, not just in-plane fits.","keywords":["channel measurement","diffuse scattering model","millimeter wave","ray tracing","28 GHz","building surfaces","3D receiver positions","air-to-ground networks"],"falsifier":"Extend the rough-wall measurement to receiver heights above 2.0 m (for example 2.1 to 2.5 m) and compare the measured power against the two fitted parameter sets: the incident-plane-only set ($S=0.60$, $\\alpha_R=1$, $\\alpha_i=10$, $\\Lambda=0.2$) and the 3D set ($S=0.42$, $\\alpha_R=6$, $\\alpha_i=4$, $\\Lambda=0.2$). If the incident-plane-only set matches the new out-of-plane heights as well as or better than the 3D set, the central claim would be refuted; if the 3D set clearly performs better, the claim is supported.","tokens_in":8287,"feed_emoji":"📡","tokens_out":9696,"duration_ms":80452,"temperature":0.7,"pith_summary":"This paper measures 28 GHz scattering from four building surfaces—metal sheet, marble wall, smooth wall, and rough wall—at several incident angles and with the receiver moved through three-dimensional positions. Its central claim is that scattering-model parameters for real building surfaces should be fitted against received power across multiple heights, not only in the incident plane. For the rough wall, an incident-plane-only fit ($S=0.60$, $\\alpha_R=1$, $\\alpha_i=10$, $\\Lambda=0.2$) leaves a fraction of variance unexplained of 0.6742 at out-of-plane positions, while a fit that includes three additional receiver heights ($S=0.42$, $\\alpha_R=6$, $\\alpha_i=4$, $\\Lambda=0.2$) reduces that error to 0.6259. The broader point matters for millimeter-wave channel modeling in air-to-ground networks, where scattering off building surfaces is a significant propagation mechanism and is sensitive to both material and geometry.","feed_headline":"3D receiver data fixes 28 GHz wall-scatter models","feed_subtitle":"In-plane-only fits leave 0.67 unexplained variance outside the plane; adding three heights cuts it to 0.63.","key_machinery":"The load-bearing mechanism is the dual-lobe (backscattering) diffuse scattering model, which writes the scattered power at a receiving point as a weighted sum of a lobe centered on the specular-reflection direction and a lobe centered on the incident direction, with the balance set by $\\Lambda$. The model parameters—scattering coefficient $S$, lobe-width factors $\\alpha_R$ and $\\alpha_i$, and balance $\\Lambda$—are tuned by minimizing the fraction of variance unexplained (FVU) between measured and simulated total received power. An initial $S$ is obtained from surface roughness and the Fresnel reflection coefficient through $S = \\sqrt{(1-R^2)\\Gamma^2}$, using the roughness loss factor $R$. The argument's force comes from adding out-of-plane receiver positions to this fitting procedure, which changes the fitted parameters and improves predictions of backscatter outside the incident plane.","core_discovery":"The paper's discovery, stated on its own terms, is that the diffuse scattering parameters of real building surfaces are three-dimensional quantities: they vary with incident angle and with receiver position in space, not just with material type. On the rough wall at 28 GHz, fitting the dual-lobe backscattering model to measured power only in the incident plane produced $S=0.60$, $\\alpha_R=1$, $\\alpha_i=10$, $\\Lambda=0.2$, and that parameterization predicted out-of-plane backscatter power poorly, with FVU equal to 0.6742. Adding receiver positions at heights 10, 20, and 30 cm above the transmit height changed the fit to $S=0.42$, $\\alpha_R=6$, $\\alpha_i=4$, $\\Lambda=0.2$ and lowered the unexplained variance to 0.6259, partially correcting the out-of-plane error while keeping the in-plane fit intact. For smoother surfaces, the single-lobe directive model with $\\alpha_R=4$ and $S$ between roughly 0.11 and 0.28 matched the measurements; for the rough surface, the dual-lobe model was required, with $\\Lambda$ increasing from 0.1 to 0.3 as the incident angle grew from 20° to 40°.","pith_inferences":["If the 3D-fit claim holds, single-plane scattering measurements at other bands (for example sub-THz and THz) likely have the same out-of-plane blind spot, and repeating this measurement protocol at those bands would test whether the effect persists as the wavelength-to-roughness ratio changes.","Because the paper integrates all power within a 5 ns excess-delay window, the fitted $S$, $\\alpha_R$, $\\alpha_i$, and $\\Lambda$ are effective parameters that merge specular and diffuse arrivals; a higher-delay-resolution sounder could separate the two and test whether the dual-lobe model then splits into distinct components.","The procedure suggests a practical calibration recipe for ray-tracing tools: a small number of receiver arcs at several heights around one wall point may be sufficient to fix the four scattering parameters, shortening outdoor measurement campaigns."],"forward_implications":["Channel models for air-to-ground millimeter-wave links should parameterize building-scatter models with receiver positions in three dimensions, not a single planar arc, because the incident-plane-only fit leaves FVU equal to 0.6742 at out-of-plane positions.","The scattering coefficient of a given surface is angle-dependent: for the rough wall it rises from 0.45–0.55 at 40° incidence to 0.65–0.75 at 20°, so a single material label without incidence angle is not a sufficient description.","For smoother surfaces, the single-lobe directive model with $\\alpha_R=4$ suffices, while rough surfaces require the dual-lobe model with a backscattering component.","The backscattering-lobe balance $\\Lambda$ increases with incident angle (0.1 at 20°, 0.2 at 30°, 0.3 at 40°), so the relative strength of backscatter grows as incidence becomes more oblique.","Parameter sets derived from one incident angle should not be transferred to other geometries; measurements at multiple angles and heights are required."],"supporting_citations":[{"why":"Supplies the directive and backscattering-lobe diffuse scattering models and the energy-conservation relation used to compute initial scattering coefficients.","marker":"[6]"},{"why":"Provides the roughness loss factor R that converts Fresnel reflection into the initial scattering coefficient S.","marker":"[13]"},{"why":"Gives the rough-surface scattering theory that motivates the small-incident-angle initial-S formula.","marker":"[5]"},{"why":"Provides the energy-conservation assumption and the received-power formula used in the numerical simulations.","marker":"[7]"},{"why":"Supplies the scattering-mechanism taxonomy and the guidance that Lambda should decrease as incident angle increases.","marker":"[8]"},{"why":"Establishes the parameterization and tuning process for diffuse scattering models that the paper adapts to 3D receiver data.","marker":"[11]"},{"why":"Shows that scattering-model parameters for a material can transfer across scenarios once tuned, motivating the search for reliable parameter sets.","marker":"[10]"},{"why":"Represents the prior single-incident-angle, incident-plane measurement approach the paper argues is incomplete.","marker":"[9]"}],"fun_headline_variants":["3D data tightens 28 GHz wall-scatter fits","3D receiver angles improve 28 GHz wall-scatter model","Wall-scatter prediction goes 3D at 28 GHz","3D scattering data boosts 28 GHz model accuracy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusions rest on the assumption that each measured power reading comes almost entirely from one specular reflection and one scattered path off the wall under test, with all other paths (ground, neighboring surfaces, higher-order reflections) more than 35 dB weaker and therefore treated as noise.","fun_headline_variants_meta":{"raw":{"variants":["3D data tightens 28 GHz wall-scatter fits","3D receiver angles improve 28 GHz wall-scatter model","Wall-scatter prediction goes 3D at 28 GHz","3D scattering data boosts 28 GHz model accuracy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000239,"raw_usage":{"total_tokens":1520,"prompt_tokens":959,"completion_tokens":561,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":575,"completion_tokens_details":{"reasoning_tokens":492}},"tokens_in":575,"tokens_out":561,"duration_ms":5416,"temperature":1.0,"reasoning_tokens":492,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T18:02:14.738446+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extend the rough-wall measurement to receiver heights above 2.0 m (for example 2.1 to 2.5 m) and compare the measured power against the two fitted parameter sets: the incident-plane-only set ($S=0.60$, $\\alpha_R=1$, $\\alpha_i=10$, $\\Lambda=0.2$) and the 3D set ($S=0.42$, $\\alpha_R=6$, $\\alpha_i=4$, $\\Lambda=0.2$). If the incident-plane-only set matches the new out-of-plane heights as well as or better than the 3D set, the central claim would be refuted; if the 3D set clearly performs better, the claim is supported.","supporting_citations":[{"cited_title":"Measurement and modeling of scattering from buildings,","cited_arxiv_id":null,"evidence_quote":"Supplies the directive and backscattering-lobe diffuse scattering models and the energy-conservation relation used to compute initial scattering coefficients."},{"cited_title":"Boithias et al., Radio wave propagation","cited_arxiv_id":null,"evidence_quote":"Provides the roughness loss factor R that converts Fresnel reflection into the initial scattering coefficient S."},{"cited_title":"Beckmann et al ., The Scattering of Electromagnetic Waves From Rough Surfaces","cited_arxiv_id":null,"evidence_quote":"Gives the rough-surface scattering theory that motivates the small-incident-angle initial-S formula."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the energy-conservation assumption and the received-power formula used in the numerical simulations."},{"cited_title":"Scattering mechanisms and modeling for terahertz wireless communications,","cited_arxiv_id":null,"evidence_quote":"Supplies the scattering-mechanism taxonomy and the guidance that Lambda should decrease as incident angle increases."},{"cited_title":"Diffuse scattering directive model parameterization method for construction materials at mmWave frequencies,","cited_arxiv_id":null,"evidence_quote":"Establishes the parameterization and tuning process for diffuse scattering models that the paper adapts to 3D receiver data."},{"cited_title":"On the importance of diffuse scattering model parameterization in indoor wireless channels at mmWave frequencies,","cited_arxiv_id":null,"evidence_quote":"Shows that scattering-model parameters for a material can transfer across scenarios once tuned, motivating the search for reliable parameter sets."},{"cited_title":"Guo et al., ”Diffuse scattering analysis of indoor propagation channel at terahertz frequency,” Proc","cited_arxiv_id":null,"evidence_quote":"Represents the prior single-incident-angle, incident-plane measurement approach the paper argues is incomplete."}],"review_version":1}