{"id":"ceedf1dc-1f35-4b9b-be02-9e2d813df385","arxiv_id":"2505.23394","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A ray-shaped antenna array made of directly connected subarrays, switched without phase shifters, can match or beat conventional hybrid-beamforming arrays in mmWave simulations at a fraction of the hardware cost.","lead":"This paper proposes a new antenna layout for high-frequency wireless: many cheap antenna elements arranged in straight lines, or rays, each pointing in a different direction, with switches selecting which rays connect to the few radio chains. No phase shifters are needed, which could sharply cut hardware cost while keeping or improving beam performance for future millimeter-wave and terahertz networks.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The RAA-vs-ULA comparison fixes the per-sULA array factor M while letting RAA use N×M elements and a larger 2D aperture; the claimed gains and cost savings therefore depend on assuming the extra antennas and footprint are essentially free, an assumption the paper states but does not justify.","rationale":"The reader's weakest assumption identifies the same load-bearing issue: the comparison normalizes by the array gain factor M rather than by total antenna count or aperture, and the cost/performance conclusions depend on the assumption that extra antenna elements and physical size are negligible. My stress-test agrees with that reading. The paper is mathematically coherent within its chosen normalization, and the sULA beam-pattern analysis, orientation design, and uplink greedy algorithm appear sound, so I do not see a reason to move the verdict from CONDITIONAL to REJECT. The concern is material enough to require a concrete sensitivity check before accepting the central claim at face value; the proposed same-element-budget ULA baseline would settle whether the performance gain is an artifact of unequal resources. The downlink exhaustive-search issue noted by the reader is real but secondary, since Algorithm 2 is explicitly described as exhaustive over S in Section V and the efficiency claim is mainly about the greedy uplink Algorithm 1. The practical feasibility of low-cost high-directivity elements and the larger physical footprint also remain unverified, which supports keeping the verdict CONDITIONAL rather than ACCEPT.","tokens_in":20347,"tokens_out":15856,"duration_ms":201926,"concrete_test":"Re-run the Section VI simulations (Figs. 8, 9, 11) with an additional ULA-HBF baseline using the same total number of antenna elements as the RAA (N×M, e.g., 25,728 for M=128, N=201) while keeping N_RF, the UMa NLoS channel model, and the DFT-codebook beamformer fixed. Compute single-user SNR, multi-user sum rate, and max-min SINR versus transmit SNR for this baseline and for the RAA. If the longer ULA matches or exceeds RAA on the performance metrics, the claimed performance advantage is an artifact of unequal element/aperture budgets rather than an architectural advantage; the authors should then also report a cost-constrained comparison (e.g., equal total hardware cost) to separate cost savings from performance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central comparison in Section III-C and the simulations in Section VI normalize by 'the same array gain factor M' (e.g., M=128 for both RAA and ULA-HBF), not by total antenna count, aperture, physical footprint, or total hardware cost. For the RAA, each active beam uses M elements, but the full system deploys N×M elements (N=201, M=128 in the example, i.e., 25,728 antennas) arranged as N sULAs in a 2D fan, whereas the ULA baseline has only M=128 elements on a line. Equations (19)-(21) and (31) then allow the RAA element pattern to have phi_3dB=0.3pi and peak gain 5.13 dB while forcing the ULA element to phi_3dB=pi and 0 dB peak, so the ~5 dB single-user SNR advantage in Fig. 8 is largely the assumed element-pattern difference, not a derived property of the architecture. The paper's cost comparison (cost_RAA vs cost_ULA) relies on p_ant=0.01 dollars and ignores the cost and physical impact of the additional N×M antennas, switches, interconnects, and the larger footprint that the authors themselves concede in Section VII. Thus the load-bearing assumption is that extra antenna elements and array size are economically and physically negligible. If a ULA-HBF were allowed the same total element budget or the same aperture, its beamwidth and gain would improve substantially; if antenna cost or footprint were non-negligible, the RAA cost and practicality advantages would shrink. This is not an internal contradiction, but it is a resource-budget asymmetry that the paper never quantifies or tests.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a new multi-antenna architecture, the ray antenna array (RAA), in which N directly-connected uniform linear subarrays (sULAs) are arranged with carefully chosen physical orientations. Steering is performed by selecting sULAs through a switch network rather than by phase shifters, which the authors argue is much cheaper at mmWave/THz frequencies. The paper derives the sULA beam pattern, the orientation spacing and array-count formulas, and sufficient conditions on the element pattern (Theorems 1–3). It then formulates uplink and downlink joint beamforming and ray-selection problems, proposes a greedy uplink algorithm and an alternating downlink algorithm, and reports simulations showing a roughly 5 dB single-user SNR gain over a conventional ULA with hybrid beamforming, as well as large hardware-cost reductions.","tokens_in":20669,"tokens_out":11402,"duration_ms":122811,"significance":"The architectural idea is genuinely interesting: replacing phase shifters with physically oriented subarrays and switches could be a useful low-cost option for high-frequency arrays, and the core sULA beam-pattern mathematics (Dirichlet kernel analysis, beamwidth, orientation spacing) is correct and clearly presented. The greedy uplink algorithm is shown to match exhaustive search in a small example, which is a solid algorithmic contribution. However, the central comparative claims are currently weakened by a resource-budget asymmetry (RAA uses N×M elements versus M for the ULA), by an element-pattern assumption whose physical cost and realizability are not analyzed, and by a downlink algorithm that relies on exhaustive search. If these issues are rigorously addressed, the paper could be a valuable contribution to the mmWave/THz architecture literature.","major_comments":[{"comment":"The performance comparison normalizes by the array gain factor M rather than by total element count or aperture. In the large example (M=128, N=201), the RAA deploys 25,728 antenna elements in a 2D fan, while the ULA baseline has 128 elements on a line. If the ULA were allowed the same total element budget (or the same aperture), its beamwidth would be far narrower than 2 arcsin(2/M) and its array gain correspondingly larger, so the claimed advantages in angular resolution and beamforming gain would not hold in an equal-resource comparison. The paper should add such a comparison, or explicitly justify why equal-array-gain-factor normalization is the appropriate fairness criterion for the claims in the abstract.","section":"Section III-C, Eqs. (12), (28), and Section VI"},{"comment":"The roughly 5 dB advantage in Fig. 8 is largely an input choice rather than a derived architectural property: the simulations set the RAA element 3 dB beamwidth to phi_3dB=0.3pi (peak gain 5.13 dB) and the ULA element beamwidth to pi (0 dB), and Eq. (21) then converts that beamwidth difference into the gain difference under equal G_sum. The paper does not establish that high-directivity RAA elements with phi_3dB around 0.3pi can be realized at the assumed $0.01 unit cost and without increasing element size, mutual coupling, or footprint, nor does it quantify the larger size admitted in Section VII. A feasibility analysis or a sensitivity study over phi_3dB and element cost should be included.","section":"Section III-C, Eqs. (19), (21), (31), and Section VI, Fig. 8"},{"comment":"The downlink alternating optimization algorithm uses exhaustive search over all N choose N_RF sULA subsets in every iteration of Problem (47). Its complexity is combinatorial in N_RF and N, so it is not efficient for the large configuration (N=201, N_RF=8) used elsewhere in Section VI. The downlink simulations consequently use only M=6 and N_RF=3, where the search is feasible. The abstract's claim of efficient algorithms for multi-user RAA communications is therefore not supported for the downlink case; a polynomial-complexity ray-selection method (e.g., a greedy scheme analogous to Algorithm 1) is needed, or the complexity and scalability limitations should be clearly stated.","section":"Section V, Algorithm 2, Problem (47)"}],"minor_comments":[{"comment":"The statement that the function sqrt(G(phi-eta_n*)) |H_M(sin(phi-eta_n*))| is 'a periodic function' is inaccurate; the subsequent argument works because every angular segment has the same functional form up to a shift. Please rephrase.","section":"Appendix A"},{"comment":"The caption states that G(zeta)=1 for the plotted beam patterns, but this is an idealized element pattern; the later RAA design uses the 3GPP pattern in Eq. (20). Clarify in the caption that Fig. 4 isolates the array-factor behavior.","section":"Section III-A, Fig. 4"},{"comment":"The floor operation in Eq. (13) gives a slightly conservative number of sULAs; stating the resulting coverage radius explicitly (for example, as (floor(eta_max/s)+1)*s) would make the coverage argument easier to follow.","section":"Section III-B, Eqs. (13), (14)"},{"comment":"The simulations use only 50 channel realizations; adding error bars or confidence intervals, especially in the multi-user sum-rate and max-min SINR plots, would help assess the statistical reliability of the observed gains.","section":"Section VI"},{"comment":"The cost model counts N_RF*N switches for the RSN and M*N_RF phase shifters for the ULA, but the physical topology of the RSN (e.g., whether each RF chain needs a full N-way selector or a smaller network) is not described in detail; a sentence describing the switching topology and its control overhead would clarify the cost comparison.","section":"Section III-C, cost comparison"},{"comment":"The phrase 'Noted that' should be 'Note that', and the statement about 'prohibitive complexity' of exhaustive search in Section V could be quantified by giving the exact order of the search space.","section":"Section IV, first paragraph"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the core sULA mathematics is sound, but the main comparative claims rest on a resource-budget asymmetry and an unexamined element-directivity assumption. I would ask the authors to (i) provide a same-element-count or same-aperture comparison, (ii) add a feasibility/sensitivity analysis for the directional element pattern and its cost, and (iii) replace or clearly scale the exhaustive-search downlink step. The VTC-2025 overlap in [1] and the reproducibility of the Qorvo price data from [30] may also deserve editorial attention."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take on arXiv:2505.23394. It proposes a ray antenna array: many small uniform linear arrays, each pointed at a fixed angle, all elements directly connected inside a sULA, and a switch network selects which sULAs feed the RF chains. No phase shifters anywhere. The math is mostly sound: the sULA beam pattern derivation, the null-to-null beamwidth, the orientation spacing rule, and the greedy uplink selection algorithm all check out. The layout and the joint ray-selection formulation are new relative to the cited lens-array and switch-based HBF work.\n\nThe real soft spot is the comparison, not the math. The paper normalizes by the per-sULA array factor M, so the RAA gets to use N×M antennas (N=201, M=128 in the example) while the ULA baseline only gets M=128. The ~5 dB single-user gain in Fig. 8 follows directly from that asymmetry plus the hand-set element beamwidths: 0.3π for RAA elements, π for the ULA, with equal total power gain, and Eq. (21) then gives the RAA element a roughly 5 dB higher peak gain. So the headline performance advantage is largely an input assumption rather than a derived consequence of the architecture. The cost comparison does the same thing: it charges the ULA for thousands of phase shifters but ignores the extra 25,000+ antennas, the switch network, and the larger physical aperture. The authors do admit in Section VII that the RAA requires larger size and list antenna blockage as future work, so this is not hidden, but they never test how the comparison changes if the ULA gets the same element budget or if the extra antennas and footprint have any cost.\n\nA second issue: the downlink algorithm (Algorithm 2) uses exhaustive search over ray selections inside each alternating step, so calling it 'efficient' is an overstatement for realistic sizes. The uplink greedy algorithm is genuinely efficient and near-optimal in the small verified example. Also, no code, data, or hardware prototype, so the practical claim that high-directivity elements can be built at $0.01 each remains unverified.\n\nNone of this is fatal. The architecture is plausible, the derivations are careful, and the orientation-spacing design rule is a useful contribution. The paper deserves a serious referee. I would want the comparison re-normalized, a sensitivity check on element beamwidths, and the downlink complexity addressed before publication. I would bring it to reading group and would cite it if I worked on low-cost mmWave arrays.","headline":"Promising phase-shifter-free array architecture with sound math, but the headline gains rest on an unequal antenna budget and hand-picked element patterns.","tokens_in":21269,"tokens_out":3199,"would_cite":true,"duration_ms":31235,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A ray antenna array made of directly connected uniform linear subarrays can steer beams with switches instead of phase shifters, and the paper argues this gives finer uniform angular resolution, higher gain, and lower cost than a…","keywords":["ray antenna array","phase-shifter-free beamforming","hybrid beamforming","mmWave MIMO","angular resolution","ray selection network","cost-effective MIMO","terahertz communications"],"falsifier":"A concrete test: simulate or build the 38 GHz example with a ULA using the same total element budget of N×M=25,728 antennas (or the same physical aperture) under the same channel, and compare achievable rates and hardware cost; if that equally financed ULA matches or beats the RAA, the claimed advantage over fairly compared alternatives fails. Alternatively, measure the realized antenna element pattern of a fabricated RAA ray at $\\phi_{3\\mathrm{dB}}=0.3\\pi$ and check whether the assumed directivity, and hence the 5 dB SNR gain, is physically obtained.","tokens_in":20041,"feed_emoji":"📡","tokens_out":6143,"duration_ms":52487,"temperature":0.7,"pith_summary":"This paper proposes a multi-antenna architecture called the ray antenna array (RAA) that does beamforming without any phase shifters. The array is built from many small uniform linear arrays (sULAs), each with its antennas wired directly together so that it naturally forms a beam along its physical orientation; a switch network then connects the best rays to the radio-frequency chains. The paper's central claim is that, compared with a conventional uniform linear array using hybrid analog/digital beamforming, the RAA has three advantages at the same array gain factor: uniform and finer angular resolution across the coverage range, higher beamforming gain because each ray can use more directional antenna elements, and drastically lower hardware cost. The argument matters because phase shifters are expensive and hard to build at millimeter-wave and terahertz frequencies, so a phase-shifter-free architecture could make large antenna arrays practical in 6G base stations. Simulations with a 47.2 GHz channel model show the RAA beating the ULA baseline in single-user and multi-user uplink and downlink scenarios.","feed_headline":"Ray antenna array beams without a single phase shifter","feed_subtitle":"The switch-based design claims sharper uniform beams and about one-sixth the hardware cost of a hybrid ULA at mmWave.","key_machinery":"The load-bearing object is the sULA: M antennas spaced half a wavelength apart and wired directly to one port, so their signals add coherently only for a wave arriving along the line of the array. Its array response is $f(\\phi,\\eta_n)=M H_M(\\sin(\\phi-\\eta_n))\\,b(\\phi-\\eta_n)$, with $H_M$ the Dirichlet kernel; this identity carries the argument because it shows the beam direction is set by geometry, not by tunable phases. The orientation rule $\\eta_n=n\\arcsin(2/M)$ and the count $N=2\\lfloor \\eta_{\\max}/\\arcsin(2/M)\\rfloor+1$ tile the coverage range with equal beamwidths, and the radiation-pattern condition in Theorem 1, $\\phi_{3\\mathrm{dB}}\\ge\\arcsin(2/M)$ with $G(0.5\\phi_{3\\mathrm{dB}})\\ge(\\varepsilon/|H_M(\\sin(0.5\\arcsin(2/M)))|)^2$, is what lets the array use directional elements. The ray selection network, a binary matrix $S\\in\\{0,1\\}^{N_{\\mathrm{RF}}\\times N}$ that chooses which sULAs reach the RF chains, converts the geometric beams into flexible MIMO transmission.","core_discovery":"The paper's central claim is that N simple uniform linear arrays, each with M directly connected antennas and oriented to cover a different slice of the angular range, can replace an M-element ULA with hybrid beamforming while improving performance. For a ray with orientation $\\eta_n$, the sULA output for a path at angle $\\phi$ is $f(\\phi,\\eta_n)=M H_M(\\sin(\\phi-\\eta_n))\\,b(\\phi-\\eta_n)$, where $H_M$ is the Dirichlet kernel; the beam peaks exactly when $\\phi=\\eta_n$, with no phase shifters. The design sets adjacent ray orientations $|\\eta_n-\\eta_{n-1}|=\\arcsin(2/M)$ so the null of one ray aligns with the peak of its neighbor, giving every ray the same null-to-null beamwidth $\\phi_{\\mathrm{BW}}=2\\arcsin(2/M)$. Because each ray covers only a fraction of the total angle range, the antenna elements can have 3dB beamwidth $\\phi_{3\\mathrm{dB}} \\ge \\arcsin(2/M)$ instead of the ULA's $\\phi^{\\mathrm{ULA}}_{3\\mathrm{dB}}\\ge 2\\phi_{\\max}$, and this narrower element pattern raises the peak gain through the conservation relation $G(0)\\approx G_{\\mathrm{sum}}/(1.066\\,\\phi_{3\\mathrm{dB}})$. The paper proves the resulting beamwidth is never wider than the ULA's DFT-codebook beamwidth and is strictly narrower away from boresight, and it gives a cost comparison showing RAA at roughly 17% of the ULA-HBF hardware cost in its example.","pith_inferences":["Editorial inference: the comparison is normalized to the same array gain factor M, not the same total element count; if the ULA were allowed N×M elements or the same physical aperture, the resolution and gain advantages would shrink, so the practical merit depends on the cost and size of the extra antennas being as negligible as assumed.","Editorial inference: the architecture's beam directions are discrete, set by the ray orientations, so angular alignment error relative to a user's true direction is bounded by half the ray spacing; a testable extension is to quantify how this quantization loss behaves when N is small or when M is modest.","Editorial inference: the same phase-shifter-free idea could apply to sensing and localization at mmWave and terahertz bands, since uniform angular resolution is valuable for angle-of-arrival estimation; the paper only mentions sensing as future 3D-RAA work.","Editorial inference: the directivity gain at $\\phi_{3\\mathrm{dB}}=0.3\\pi$ is assumed achievable at no extra fabrication cost; a hardware prototype measuring the realized element pattern would settle whether the SNR gains survive in practice."],"forward_implications":["A 128-element-class base station with $N_{\\mathrm{RF}}=16$ chains can drop from 2048 phase shifters to a switch network, cutting the quoted hardware cost from about 268,700 USD to about 46,300 USD in the paper's 38 GHz example.","Angular resolution no longer degrades for users at the edge of the cell: every ray has the same beamwidth, whereas a ULA's DFT beam widens away from boresight.","Because each sULA covers only a small angular sector, antenna elements can be made more directional, which the simulations turn into roughly 5 dB higher SNR for single-user uplink.","The greedy ray selection for uplink sum rate and the alternating optimization for downlink max-min SINR reach near-optimal performance in the paper's tests, so the architecture's gains do not require exhaustive search.","Isotropic-element RAA still matches or edges out ULA at high SNR in downlink, showing the resolution advantage alone carries value even without directional elements."],"supporting_citations":[{"why":"supplies the Qorvo unit prices for phase shifter, RF switch, and antenna element used in the hardware cost comparison.","marker":"[30]"},{"why":"supplies the 3GPP UMa NLoS channel model used in all simulations at 47.2 GHz.","marker":"[28]"},{"why":"supplies the DFT codebook and array signal processing baseline for the ULA comparison.","marker":"[29]"},{"why":"defines the fully-connected hybrid beamforming architecture whose phase-shifter count the RAA is compared against.","marker":"[10]"},{"why":"establishes switch-based hybrid beamforming as the prior alternative that motivates replacing phase shifters with switches.","marker":"[12]"},{"why":"provides the lens antenna array as a competing cost-reduction architecture for mmWave MIMO.","marker":"[20]"},{"why":"supplies the partially-connected HBF structure that reduces phase shifters at the cost of aperture, a baseline context for the RAA trade-offs.","marker":"[13]"}],"fun_headline_variants":["No phase shifters? Ray array beams at a fraction of the cost","Switch-only ray array delivers uniform beams at low cost","Ray antenna array: cheaper mmWave beams without phase shifters","Sharper beams, lower hardware cost: the ray array approach","Ray array achieves uniform angular resolution with simple switches"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison gives the RAA N×M antenna elements while the ULA gets only M, and it assumes the extra antennas cost about a cent each and take negligible space, so the cost and resolution advantages rest on antennas being far cheaper and physically smaller than phase shifters.","fun_headline_variants_meta":{"raw":{"variants":["No phase shifters? Ray array beams at a fraction of the cost","Switch-only ray array delivers uniform beams at low cost","Ray antenna array: cheaper mmWave beams without phase shifters","Sharper beams, lower hardware cost: the ray array approach","Ray array achieves uniform angular resolution with simple switches"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000379,"raw_usage":{"total_tokens":2152,"prompt_tokens":1217,"completion_tokens":935,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":833,"completion_tokens_details":{"reasoning_tokens":852}},"tokens_in":833,"tokens_out":935,"duration_ms":9168,"temperature":1.0,"reasoning_tokens":852,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:48:28.040876+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: simulate or build the 38 GHz example with a ULA using the same total element budget of N×M=25,728 antennas (or the same physical aperture) under the same channel, and compare achievable rates and hardware cost; if that equally financed ULA matches or beats the RAA, the claimed advantage over fairly compared alternatives fails. Alternatively, measure the realized antenna element pattern of a fabricated RAA ray at $\\phi_{3\\mathrm{dB}}=0.3\\pi$ and check whether the assumed directivity, and hence the 5 dB SNR gain, is physically obtained.","supporting_citations":[{"cited_title":"Qorvo official website,","cited_arxiv_id":null,"evidence_quote":"supplies the Qorvo unit prices for phase shifter, RF switch, and antenna element used in the hardware cost comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the DFT codebook and array signal processing baseline for the ULA comparison."},{"cited_title":"Spatially sparse precoding in millimeter wave MIMO systems,","cited_arxiv_id":null,"evidence_quote":"defines the fully-connected hybrid beamforming architecture whose phase-shifter count the RAA is compared against."},{"cited_title":"Millimeter wave MIMO with lens antenna array: A new path division multiplexing paradigm,","cited_arxiv_id":null,"evidence_quote":"provides the lens antenna array as a competing cost-reduction architecture for mmWave MIMO."},{"cited_title":"Energy-efficient hybrid analog and digital precoding for mmWave MIMO systems with large antenna arrays,","cited_arxiv_id":null,"evidence_quote":"supplies the partially-connected HBF structure that reduces phase shifters at the cost of aperture, a baseline context for the RAA trade-offs."}],"review_version":1}