{"id":"7fd34d86-5d42-4326-b89c-a2b5a442f20f","arxiv_id":"2504.20514","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A 7.8 GHz measurement campaign shows U6G channels are sparse under line-of-sight and that dividing the ELAA into subarrays preserves a far-field approximation with an NMSE near -19 dB.","lead":"Researchers measured wireless channels in the upper 6 GHz band using a virtual 64-antenna array in indoor, outdoor, and outdoor-to-indoor scenes. The results characterize small-scale fading and show that splitting the large array into subarrays keeps channel accuracy high enough for distributed processing.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The -20 dB subarray-wise far-field NMSE is computed from a 100-MPC NOMP fit, not from the DFT-based processing the paper uses to motivate distributed low-complexity design; the key quantitative evidence may not validate the actual claim.","rationale":"The paper contributes a useful first U6G ELAA measurement dataset and a plausible qualitative framework, but the central quantitative claim rests on Eq. (14). The reader's weakest assumption was the virtual-array staticity/phase-drift issue. That concern is legitimate but partly mitigated by the static-environment protocol and Rb-clock synchronization, and the per-snapshot 8-element subarrays are internally coherent. The more direct gap is that the reported NMSE is computed from a high-dimensional NOMP fit with L=100 MPCs, which can absorb near-field curvature and measurement noise, whereas the paper's practical motivation is low-complexity DFT-based processing. This is an internal evidence-to-claim gap rather than a failure of the measurement setup. It is fixable by re-evaluating the NMSE with a DFT-basis reconstruction and by reporting Q and error bars, so the verdict remains CONDITIONAL. I do not change the reader's verdict, hence UNCHANGED.","tokens_in":17723,"tokens_out":8327,"duration_ms":90755,"concrete_test":"Recompute the NMSE of Eq. (14) with H_recon,b constructed from a DFT beamspace representation of each subarray, selecting the strongest K DFT beams for K = 1, 2, and 4 (and K = N_b, the subarray size), instead of the NOMP 100-MPC fit. If the DFT-based NMSE is materially worse than the reported -18 dB, the claim that the far-field subarray approximation supports DFT-based distributed processing is not validated. Also report Q and the per-observation error distribution for each B so the B=8 vs B=16 comparison is statistically meaningful.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in Sec. IV.E.5 is that the near-field ELAA channel can be effectively approximated as far-field subarray-wise, with NMSE about -20 dB, and that this supports DFT-based distributed processing. The only quantitative support is Eq. (14), but the reconstruction H_recon,b used there is obtained via the NOMP estimator with L=100 MPCs per subarray, as described in Sec. II.C and the footnote. This is a continuous-angle far-field model with up to 100 degrees of freedom per subarray, not the fixed, low-complexity DFT beamspace representation that Sec. IV.F.1 explicitly promotes. A 4- or 8-element subarray with an aperture of roughly 6-14 cm at a 5.6 m distance is small enough that many superposed plane waves can fit the observed subarray channel almost arbitrarily well; the NMSE therefore measures the residual of a very flexible oracle fit rather than the error incurred by practical DFT-based subarray processing. The reported improvement from B=8 to B=16 (-18.09 dB to -18.82 dB) is also reported without Q, per-observation variance, or statistical significance, so it is not clear whether the improvement is meaningful. The qualitative DFT-energy plots in Fig. 13 show concentration but are not converted into an NMSE or beam-count requirement. Thus the headline quantitative evidence conflates 'the subarray channel is well fit by 100 far-field MPCs' with 'the subarray channel is well approximated by the low-complexity DFT structures the paper recommends.'","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a channel measurement campaign at a center frequency of 7.8 GHz using a 100 MHz OFDM sounder and a virtual 64-element array built from an 8-element physical array. The authors characterize small-scale fading statistics (APAS, RMS AS, PDP, RMS DS, Rician K, sparsity) in indoor, outdoor, and outdoor-to-indoor scenarios, then analyze ELAA channel properties from a distributed-processing perspective: subarray-wise non-stationarities and consistencies, sub-band characteristics, and the accuracy of a subarray-wise far-field approximation quantified by an NMSE of about -18 dB. The paper concludes that these measurements validate low-complexity DFT-based distributed processing for U6G ELAA systems.","tokens_in":18008,"tokens_out":8986,"duration_ms":82481,"significance":"The measurement dataset and scenario comparisons are potentially useful for channel modeling in the upper mid-band, and the subarray-wise far-field approximation is directly relevant to reducing ELAA processing complexity. The paper's strengths include the construction of a custom sounder, the use of NOMP for MPC extraction, and the breadth of analyzed metrics. However, several technical issues—an incorrect Gini index formula, a mismatch between the claimed U6G band and the actual 7.8 GHz measurement frequency, and a validation gap between the NOMP-based NMSE and the DFT-based processing recommendation—currently undermine the support for the central claims.","major_comments":[{"comment":"The Introduction (Section I) defines U6G as the 6245–7125 MHz band, but the system described in Section II.A up-converts the signal to a 7.8 GHz center frequency. All subsequent claims about 'U6G channel characteristics' therefore concern a frequency band outside the defined U6G range. Please either reclassify the reported band as FR3/upper mid-band and revise the title, abstract, and contributions accordingly, or provide a clear justification for treating 7.8 GHz as part of the U6G band.","section":"I and II.A"},{"comment":"The Gini index formula in Eq. (12) does not satisfy the stated property that G=0 for equal powers. For L MPCs with equal power, the expression gives G = 1 - 1/L (approximately 0.99 for L=100), not 0. This contradicts the text that states G=0 represents equal power. Please correct the formula or the normalization, and recompute the reported Gini-index values and CDFs in Fig. 9.","section":"III.F, Eq. (12)"},{"comment":"The NMSE of the subarray-wise far-field approximation is computed with H_recon obtained from the NOMP estimator using L=100 far-field MPCs per subarray (Section II.C), not from the low-complexity DFT beamspace representation recommended in Section IV.F.1. The reported -18.82 dB (B=16) and -18.09 dB (B=8) therefore quantify the fit of a highly flexible, oracle-like far-field model, not the error of the practical DFT-based subarray processing the paper promotes. Please provide an NMSE for DFT-based reconstructions with a limited number of beams per subarray, or clearly separate the goodness-of-fit claim from the DFT-processing claim.","section":"IV.C and IV.E.5"},{"comment":"The virtual 64-element array is synthesized from eight snapshot positions of an 8-element physical array. The paper states that the environment was static, but it does not provide any phase-coherence validation between the snapshots, such as repeated reference measurements, drift statistics, or calibration residuals. Phase drift, clock instability, or any environmental change during the mechanical movement would directly contaminate the spatial channel and all subsequent subarray comparisons, including the NMSE values. Please provide evidence of temporal stability or describe a phase-compensation procedure.","section":"II.B and IV"},{"comment":"The delay term p_nf(τℓ) = exp(−j2πη_f Δf τℓ / λ) in Eq. (2) is dimensionally inconsistent: the exponent should be a phase (e.g., −j2πη_f Δf τℓ after carrier removal), and the division by λ makes the argument of the exponential carry units of 1/m. This appears to be a typo, but it should be corrected and the reconstruction formula in Eq. (3) checked for consistency with the corrected phase model.","section":"II.C, Eq. (2)"}],"minor_comments":[{"comment":"After correcting Eq. (12), the Gini-index values and the associated sparsity discussion in Section III.F and Fig. 9 should be revisited, as the reported range of 0.91 to 0.96 will shift under a correct formula.","section":"III.F"},{"comment":"The Gaussian fits for RMS AS, RMS DS, and Rician K are presented without goodness-of-fit statistics (e.g., RMSE or Kolmogorov–Smirnov tests); adding such statistics would strengthen the modeling claims.","section":"III.B, III.D, III.E"},{"comment":"The claim that a wider band yields a larger RMS AS is based on four very similar values (18.85°, 18.13°, 19.25°, 18.73°); a statistical comparison or error bars are needed to support this trend.","section":"IV.D"},{"comment":"The statement 'The NMSE of this approximation is about -20 dB' is inconsistent with the reported values -18.82 dB and -18.09 dB; please use the exact figures and, if possible, report the number of observations Q and per-observation variability.","section":"IV.E.5"},{"comment":"The DFT energy-concentration analysis is qualitative; consider reporting the fraction of channel energy captured by the first one or two DFT beams per subarray to quantify the support for DFT-based processing.","section":"IV.C, Fig. 13"}],"recommendation":"major_revision","confidential_remarks":"The frequency-band mismatch (claiming U6G while measuring at 7.8 GHz) is a serious framing issue that affects the title and core contributions. If the authors cannot justify 7.8 GHz as U6G, the manuscript would need substantial rewriting to target the FR3/upper-mid-band instead. The virtual-array phase-stability concern is also nontrivial, as the entire ELAA analysis rests on the synthesized array faithfully representing a real 64-element array. The NMSE evidence is weaker than claimed because it validates a flexible NOMP fit rather than the DFT-based structures that the paper recommends. I recommend major revision rather than rejection because the underlying measurement campaign and several of the observed characteristics are valuable, but the listed issues are load-bearing and must be addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuinely useful measurement paper—new U6G (7.8 GHz) ELAA channel data, with a sensible distributed subarray and sub-band analysis. The headline that a near-field ELAA channel can be approximated as far-field subarray-wise is credible. But the quantitative support for the -20 dB NMSE needs work: that number comes from a 100-MPC NOMP fit, not from the DFT-based processing the paper is pitching, and there is a bug in the Gini formula that should be fixed.\n\nWhat is new: they built a 7.8 GHz sounder, ran indoor/outdoor/O2I campaigns, synthesized a 64-element ELAA by moving an 8-element array, and characterized APAS, AS, PDP, DS, K factor, and sparsity. The subarray-wise consistency/non-stationarity analysis and the subband angular-consistency results are the most original part. The DFT-based visualization in Fig. 13 is a nice qualitative demonstration that subarray channels concentrate in angle.\n\nWhere it is soft. First, Eq. (12) as written does not give G=0 for equal powers; for L=100 it gives roughly 0.99. The measured Gini values around 0.91–0.96 then fall outside the possible range of the formula, which strongly suggests a typo, likely a missing factor L in the denominator. The paper should correct the equation and recheck the numbers. Second, the central NMSE claim: H_recon,b is built from NOMP with 100 far-field MPCs per subarray. That is a flexible continuous-angle fit, not a fixed DFT beamspace reconstruction. So the -18.82 dB and -18.09 dB numbers show the subarray channel can be well explained by superposed plane waves, which does support the far-field assumption in principle, but they do not directly measure the error of the low-complexity DFT-based processing that Sec. IV.F motivates. The paper should either compute NMSE using a DFT projection with a small number of beams, or explicitly label the current value as a best-case bound for the far-field model class. Third, the NMSE is reported without observation count, error bars, or a baseline; the 0.73 dB improvement with more subarrays may be noise. Minor: the distribution fits are visual only, and there is no quantitative comparison with the closely related mid-band ELAA measurements they cite. The virtual-array assumption is standard and they did keep the environment static, but a sentence on phase-drift checking would help.\n\nBottom line: the data and the main qualitative insights are worth having, and the paper deserves a serious referee. It should be sent to review, but with a request to fix the Gini formula, reframe or re-derive the NMSE evidence, and add error bars and comparisons. I would accept after those revisions; the central far-field approximation claim itself holds up.","headline":"Useful U6G ELAA measurement paper with credible far-field subarray claim, but the -20 dB NMSE evidence comes from a flexible NOMP fit rather than the DFT processing it motivates, and the Gini formula has a fixable typo.","tokens_in":18627,"tokens_out":8229,"would_cite":true,"duration_ms":77576,"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":"Measured U6G extra-large-array channels can be split into subarrays and treated as far-field channels, supporting distributed 6G processing.","keywords":["channel measurement","U6G","extra-large scale antenna array","ELAA","distributed processing","small-scale fading","spatial non-stationarity","subarray-wise far-field approximation"],"falsifier":"Measure the same geometry with a physical 64-element ELAA, or repeat the virtual-array scan in random order while monitoring phase drift, and check whether the subarray-wise far-field NMSE stays near -18 to -20 dB and whether the LoS azimuth still follows the spherical-wave model; if the error rises substantially or the pattern is not reproducible, the virtual-array static-environment assumption is what produced the result.","tokens_in":17452,"feed_emoji":"📡","tokens_out":7680,"duration_ms":76965,"temperature":0.7,"pith_summary":"This paper reports channel measurements at 7.8 GHz with 100 MHz bandwidth in indoor, outdoor, and outdoor-to-indoor scenarios, and asks whether an extremely large antenna array of 64 virtual elements can be handled by distributed subarray processing. It finds that although the full array sees a non-stationary, near-field channel, each subarray of 4 or 8 elements is well approximated as a far-field channel, with reconstruction NMSE of about -20 dB (measured -18.82 dB for 16 subarrays and -18.09 dB for 8). It also finds consistent angular and delay structure across subarrays and across sub-bands. A sympathetic reader would care because these are measurement-based justifications for using low-complexity, distributed DFT-based processing for U6G ELAA systems rather than full-array near-field processing.","feed_headline":"Splitting huge U6G antenna arrays tames near-field effects","feed_subtitle":"64-element virtual arrays keep each subarray far-field-like at about -19 dB error, easing 6G distributed processing.","key_machinery":"The argument is carried by three linked instruments. First, a virtual-array channel sounder synthesizes a 64-element ELAA by horizontally moving an 8-element physical array while all objects are kept stationary, giving measured snapshots of the full ELAA. Second, the Newtonized orthogonal matching pursuit (NOMP) estimator decomposes the received OFDM signal into 100 multipath components with delay and azimuth parameters and reconstructs the channel matrix $\\mathbf{H}$. Third, a two-dimensional DFT transform $\\mathbf{H}_{\\mathrm{D-S}} = \\mathbf{U}_F \\mathbf{H} \\mathbf{U}_A^H$, together with a normalized mean-square error (NMSE) comparing the virtual full-array channel with a far-field reconstruction built subarray-by-subarray, provides the concrete evidence: the full array spreads over several spatial DFT directions while each subarray concentrates on nearly one direction.","core_discovery":"The central discovery is that the near-field non-stationarity of an ELAA at upper 6 GHz does not force full-array near-field processing: dividing the 64-element array into subarrays makes each subarray's channel nearly a single far-field direction in the DFT domain, and the subarray-wise far-field reconstruction achieves an NMSE around -20 dB, improving as the number of subarrays grows. The measurement evidence is that LoS azimuth angles across eight subarrays track the spherical-wave model, while RMS angular spread and Rician K factor vary across subarrays, showing non-stationarity, yet dominant angular directions, delay spread, and LoS arrival time remain consistent, showing common structure. In the frequency domain, four sub-bands yield nearly identical average power angular spectra and RMS angular spreads from 18.13 to 19.25 degrees, supporting sub-band distributed processing. The paper states its own conclusion as: the near-field non-stationary ELAA channel can be effectively approximated as far-field subarray-wise under the U6G frequency band.","pith_inferences":["Editorial inference: Because each subarray behaves as a far-field channel, the same DFT-based hybrid beamforming and compressed channel-estimation routines built for conventional massive MIMO can likely be reused per subarray, making the hardware and baseband savings concrete; the paper argues this direction but does not implement a full system.","Editorial inference: The measured consistency of RMS angular spread across sub-bands suggests a testable frequency-extrapolation scheme in which spatial covariance estimated on one sub-band initializes beamforming on another; the paper reports the supporting statistics but does not run that algorithm.","Editorial inference: A true 64-element array measurement, rather than a virtual array assembled from eight static snapshots, would be the natural stress test of the -18 to -20 dB approximation claim, since any temporal drift in the virtual scan could masquerade as spatial non-stationarity."],"forward_implications":["A U6G ELAA base station can process the channel subarray by subarray under a far-field assumption, so per-subarray algorithms from conventional massive MIMO carry over.","Increasing the number of subarrays from 8 to 16 improves the subarray-wise far-field reconstruction NMSE from -18.09 dB to -18.82 dB, giving a tunable complexity-accuracy tradeoff.","Because subarrays share dominant angular directions, delay spread, and LoS arrival time, CSI learned on one subarray can be reused for others, lowering channel-estimation overhead.","Because spatial characteristics are similar across sub-bands, with RMS angular spread between 18.13 and 19.25 degrees, sub-band or out-of-band CSI can support distributed estimation without full-band sounding.","Under line-of-sight, the U6G channel is sparse, with Gini index between 0.91 and 0.96, so compressive estimation and low-rank processing remain effective for the full array."],"supporting_citations":[{"why":"Provides the virtual-array measurement methodology and the L=100 MPC setting used for mmWave non-stationarity; this paper extends it to U6G.","marker":"[17]"},{"why":"Supplies the Newtonized orthogonal matching pursuit (NOMP) estimator used to extract multipath delays and angles from the measured OFDM signals.","marker":"[23]"},{"why":"Gives the spherical-wave geometric model against which the measured subarray-wise LoS azimuth angles are compared.","marker":"[36]"},{"why":"Establishes that DFT vectors form bases for far-field channels in the delay and angular domains, used to compare full-array versus subarray concentration.","marker":"[37]"},{"why":"Also supports the DFT-domain representation of far-field channels, grounding the transform used in the approximation analysis.","marker":"[38]"},{"why":"Supplies the Saleh-Valenzuela cluster model used to fit measured power delay profiles and extract cluster and ray decay factors.","marker":"[30]"},{"why":"Provides the in-band and out-of-band channel reconstruction concept that motivates the sub-band CSI consistency analysis.","marker":"[26]"},{"why":"Extends the multiband reconstruction idea used to argue that similar spatial CSI across sub-bands enables distributed channel estimation.","marker":"[39]"}],"fun_headline_variants":["Subarrays turn U6G near-field into far-field","Divide and conquer: U6G arrays get far-field subchannels","U6G array split: near-field becomes far-field","Subarrays fix U6G near-field for distributed processing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim rests on treating a virtual 64-element array as a real one: the 8-element physical array is moved through eight positions under the assumption that the channel is perfectly static, so any phase drift, clock error, or environmental change between snapshots would be misread as spatial variation and could bias every subarray comparison.","fun_headline_variants_meta":{"raw":{"variants":["Subarrays turn U6G near-field into far-field","Divide and conquer: U6G arrays get far-field subchannels","U6G array split: near-field becomes far-field","Subarrays fix U6G near-field for distributed processing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00115,"raw_usage":{"total_tokens":4787,"prompt_tokens":984,"completion_tokens":3803,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":3733}},"tokens_in":600,"tokens_out":3803,"duration_ms":28663,"temperature":1.0,"reasoning_tokens":3733,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:26:59.169912+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the same geometry with a physical 64-element ELAA, or repeat the virtual-array scan in random order while monitoring phase drift, and check whether the subarray-wise far-field NMSE stays near -18 to -20 dB and whether the LoS azimuth still follows the spherical-wave model; if the error rises substantially or the pattern is not reproducible, the virtual-array static-environment assumption is what produced the result.","supporting_citations":[{"cited_title":"Multi-frequency mmWave massive MIMO channel measurements and characterization for 5G wireless communication systems,","cited_arxiv_id":null,"evidence_quote":"Provides the virtual-array measurement methodology and the L=100 MPC setting used for mmWave non-stationarity; this paper extends it to U6G."},{"cited_title":"Newtonized orthog- onal matching pursuit: Frequency estimation over the continuum,","cited_arxiv_id":null,"evidence_quote":"Supplies the Newtonized orthogonal matching pursuit (NOMP) estimator used to extract multipath delays and angles from the measured OFDM signals."},{"cited_title":"Spherical-wave model for short-range MIMO,","cited_arxiv_id":null,"evidence_quote":"Gives the spherical-wave geometric model against which the measured subarray-wise LoS azimuth angles are compared."},{"cited_title":"Channel estimation via orthogonal matching pursuit for hybrid MIMO systems in millimeter wave commu- nications,","cited_arxiv_id":null,"evidence_quote":"Establishes that DFT vectors form bases for far-field channels in the delay and angular domains, used to compare full-array versus subarray concentration."},{"cited_title":"Channel estimation for TDD/FDD massive MIMO systems with channel covariance computing,","cited_arxiv_id":null,"evidence_quote":"Also supports the DFT-domain representation of far-field channels, grounding the transform used in the approximation analysis."},{"cited_title":"A statistical model for indoor multipath propagation,","cited_arxiv_id":null,"evidence_quote":"Supplies the Saleh-Valenzuela cluster model used to fit measured power delay profiles and extract cluster and ray decay factors."},{"cited_title":"Efficient downlink channel reconstruction for FDD multi-antenna Systems,","cited_arxiv_id":null,"evidence_quote":"Provides the in-band and out-of-band channel reconstruction concept that motivates the sub-band CSI consistency analysis."},{"cited_title":"Efficient multiband channel reconstruction and tracking for hybrid mmWave MIMO systems,","cited_arxiv_id":null,"evidence_quote":"Extends the multiband reconstruction idea used to argue that similar spatial CSI across sub-bands enables distributed channel estimation."}],"review_version":1}