{"id":"7218ebe8-2130-4b5f-a3cf-eda49b7f34c7","arxiv_id":"2502.00139","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A 28 GHz prototype shows a single RF chain with per-antenna time delays can form frequency-dependent beams and serve four users simultaneously, with simulations projecting near-doubled uplink coverage.","lead":"A prototype antenna that uses tiny time delays to bend different frequencies in different directions let a single radio chain serve four users at once in the millimeter-wave band. Simulations suggest this 'joint phase-time array' design could roughly double uplink coverage compared with standard analog beamforming.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prototype in Sec. III uses 16 per-antenna downconversion mixers before IF delay and combine, so the abstract's 'single TRX RF chain' claim is unsupported; the experiment is a multi-chain IF beamformer, not the JPTA architecture of Fig. 1.","rationale":"The reader's weakest_assumption targets the system-level simulation's scheduling and channel model. That concern is legitimate but secondary, because the simulation is explicitly an idealized model and the qualitative N-fold gain follows from the TDM-vs-FDM comparison, not from a hidden error. The paper's novel contribution and headline result is the prototype demonstration of a single TRX serving four users. The prototype description in Section III, however, describes 16 receiver channels, each with its own downconversion mixer, followed by IF delay and combine. This does not match the JPTA architecture in Fig. 1, which has one TRX after the delay/phase network. The experiment thus demonstrates a different, more hardware-intensive architecture. Without a single-chain prototype or a clear statement that IF delays emulate the RF delays in a multi-chain testbed, the abstract overclaims. The missing EVM figures and delay-line specifications are additional gaps, but the architecture mismatch is more fundamental: even complete measurement data would not turn a 16-mixer receiver into a single-TRX demonstration. Therefore, the verdict should be REJECT for the current version, with a path to revision by clarifying the emulation nature and adjusting claims.","tokens_in":123,"tokens_out":19535,"duration_ms":453441,"concrete_test":"Request the prototype's detailed block diagram and bill of materials. Count the number of ADMV1018 mixers and the number of ADC/spectrum-analyzer inputs; if the delay lines are placed after per-element mixers and combined into one digitizer, the setup is a multi-chain IF-delay beamformer. Verify whether any delay element operates at 28 GHz RF before a single combiner; if not, the abstract's 'single TRX RF chain' claim is unsupported.","verdict_should_be":"REJECT","load_bearing_attack":"The central claim in the abstract—'With newly developed delay elements, we demonstrate that a single TRX RF chain can serve four users in four different directions'—is not supported by the prototype described in Section III. The text states that on the gNB RX side, '16 RF channels are integrated onboard,' each connected to a 1x4 patch antenna and 'down-converted to the 0.5 GHz intermediate frequency (IF) through the mixer (ADMV1018).' The IF signals are then 'delayed by 16 delay lines separately and then combined' before a spectrum analyzer and digital scope. This is a per-element downconversion architecture with 16 independent RF/IF chains; it is not the single-TRX JPTA shown in Fig. 1 and Eq. (1), where all antenna branches share one TRX after the delay and phase-shift network. The delays are at IF, not RF true-time-delay elements, so the 'newly developed delay elements' are not demonstrated to be the RF delay elements required by the architecture. The experiment therefore does not verify the claimed single-RF-chain feasibility; it verifies a multi-chain IF-delay beamformer that resembles hybrid beamforming with per-element downconversion.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes joint phase-time arrays (JPTA), an analog beamforming architecture in which each antenna branch contains a tunable time-delay element and a phase shifter, all sharing a single RF chain. The authors describe Type-1 (discrete multi-beam) and Type-2 (rainbow) beam designs, report a 28 GHz prototype with 16 receive channels and IF delay lines that serves four UEs in four directions with measured EVM degradation of 0.5–3 dB, and present system-level simulations using a large-scale-fading-only model. The simulations claim that JPTA extends uplink throughput coverage by 100% and gives up to 830% throughput gain over phased-array beamforming at cell edge. The abstract states that the prototype demonstrates that a single TRX RF chain can serve four users in four different directions in the mmWave band.","tokens_in":8321,"tokens_out":3517,"duration_ms":39707,"significance":"If the single-chain JPTA claim is correct, the architecture is an attractive low-cost alternative to multi-RF-chain hybrid beamforming, and the system-level gains would be relevant for 6G mmWave design. The paper has strengths: a physical prototype with measured EVM results, no parameters fitted to the measurements, and explicit statement of simulation assumptions. The measured 0.5–3 dB EVM degradation matching the predicted 1–3 dB beamforming-gain loss is a useful checkpoint. However, the prototype does not implement the single-TRX architecture claimed in the abstract, and the headline simulation gains are largely forced by the scheduling model rather than by the JPTA beamforming itself. The paper is therefore a useful progress report, but its central demonstration claim needs either stronger evidence or careful reformulation.","major_comments":[{"comment":"The prototype described on the gNB RX side does not implement the single-TRX JPTA architecture of Fig. 1 and Eq. (1). The text states that 16 RF channels are integrated onboard, each connected to a 1x4 patch antenna and down-converted to 0.5 GHz IF through its own mixer, and that the IF signals are delayed by 16 delay lines separately and then combined. This is a multi-chain beamformer with per-element downconversion and IF-domain delays, not a single RF chain with RF true-time-delay elements preceding one TRX. Consequently, the abstract's claim that the prototype demonstrates 'a single TRX RF chain can serve four users in four different directions' is unsupported by the described experiment. The authors should either revise the architecture claim to describe an IF-delay multi-chain beamformer that validates the frequency-dependent multi-beam concept, or they must show a single-TRX RF-delay implementation with measured results.","section":"Section III"},{"comment":"The headline throughput and coverage gains in Fig. 9 are largely a direct consequence of the resource-allocation rule rather than a property of JPTA beamforming. In the cell-edge regime the text states that both schemes operate at MCS 0 with 4 RBs, but PAA users transmit only once every NUE slots while JPTA users transmit in every slot. This allocation structure forces the NUE-fold throughput ratio and, with path-loss exponent beta=3, the coverage ratio NUE^(1/3) (e.g., 2 for 8 UEs). The authors acknowledge the scheduling premise in words, but the abstract and conclusion present the 100% coverage increase and 830% gain as JPTA system-level benefits without this qualification. The paper should state clearly that these numbers are upper bounds under an idealized scheduler that guarantees every user 4 RBs in every slot, and should show how the gains depend on the scheduler, the minimum-RB constraint, and the path-loss exponent.","section":"Section IV-B"},{"comment":"The simulation results in Fig. 9 are deterministic: only one ring of UEs is active at a time, only large-scale fading is included, and no repeated trials or error bars are reported. Since the central quantitative claims are 100% coverage extension and 830% throughput gain, the absence of any statistical variation or sensitivity analysis makes it difficult to assess whether those numbers are robust. The authors should add sensitivity studies for the path-loss exponent beta, the UE angular distribution, and the scheduling parameters, or at least provide confidence intervals from multiple UE drops.","section":"Section IV-A"}],"minor_comments":[{"comment":"The three figures showing the setup and measured EVM results are all referenced as 'Fig. III' in the text, which appears to be a LaTeX placeholder error; the figure numbers need to be corrected.","section":"Section III and Figures 3–5"},{"comment":"Equation (1) defines the downlink transmitted signal, while the prototype and simulation are for uplink reception. The authors should clarify whether the same JPTA model applies to uplink combining and, if so, state the reciprocity or transpose relationship explicitly.","section":"Eq. (1) and Section IV"},{"comment":"The claim that emulating Type-1 JPTA beam patterns requires at least as many TRXs as subbands is stated without a proof or a precise reference. A short derivation or a citation to a specific equation in [4] would make the comparison with hybrid precoding easier to verify.","section":"Section II-C"},{"comment":"The sentence 'having N1 UL time slots with N2/k RBs ... is always superior to having N1/k UL time slots with N2 RBs' uses undefined variables N1, N2, and k. These should be defined, and the statement should be stated as an inequality with the appropriate SNR model, since it underlies the claimed gain.","section":"Section IV-B"},{"comment":"There is a capitalization typo in the final paragraph: 'Therefore, We believe' should be 'Therefore, we believe'. Also, the axes labels in Fig. 7 appear to show 'Frequency (Hz) 108' with a missing superscript, and the angle axis label is unclear; the figure should be reformatted for readability.","section":"Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The central architectural mismatch between the abstract's 'single TRX RF chain' claim and the actual multi-chain IF-delay prototype is the main reason for major revision. The simulation results are internally consistent but are a product of the scheduling model; the authors should be encouraged to reframe the claims as an idealized comparison and to add sensitivity analysis. The paper is within the scope of the journal and the underlying JPTA concept is interesting, but the presented evidence does not yet support the abstract's strongest statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, the physical prototype does not demonstrate what the abstract says. The gNB receiver uses 16 RF chains, each with its own mixer downconverting to 0.5 GHz IF; the delays are applied at IF and then combined. That is a multi-chain IF beamformer, not the single-TRX architecture of Fig. 1. So the claim 'a single TRX RF chain can serve four users in four different directions' is unsupported by the experiment. The delays are also not the 'newly developed delay elements' in the RF path. This is a real discrepancy, not a nitpick.\n\nSecond, the system-level simulation results are largely built into the scheduling model. At the cell edge, every JPTA user gets a minimum of 4 RBs in every slot, while a PAA user gets 4 RBs only once per N slots. That allocation rule, combined with the path-loss exponent, directly forces the N-fold throughput gain and the N^(1/3) coverage ratio. The paper is honest about this—it says the gain increases to N×100% at the cell edge—but it means the 830% gain and 100% coverage extension are consequences of the model, not independent findings.\n\nWhat the paper does well: it gives a clear description of JPTA beam design, shows Type-1 and Type-2 beams with concrete delay/phase solutions, and includes a real testbed with measured EVM. The measured EVM degradation of 0.5–3 dB roughly matches the predicted beamforming-gain loss, which is a useful sanity check. The simulation study for 2, 4, 8, and 16 UEs is also a reasonable extension of prior work, even if the architecture and coverage analysis came from the authors' earlier papers.\n\nSoft spots beyond the headline mismatch: the prototype used only 8 of 16 receive channels, a 1.2 m link, and backed-off TX power, with no repeated trials or error bars. That is acceptable for a feasibility demo but not for a quantitative claim. Also, the paper does not provide code or data, so the simulations are not independently checkable.\n\nWho is this for? Reading-group material for people working on TTD/mmWave front-ends, and a signal to the authors that the prototype write-up needs to be aligned with what was actually built. I would send it to peer review—the mismatch can be fixed by rewriting the abstract and clarifying the prototype architecture—but in its current form the central claim should not be accepted. I would not cite it for the single-RF-chain feasibility; I might cite it for the measured IF-delay beamforming results if I worked in that niche.","headline":"The prototype is a 16-channel IF-delay beamformer, not the single-RF-chain JPTA the abstract promises; treat the headline claim as unproven, though the beamforming concept and simulations have merit.","tokens_in":8888,"tokens_out":1552,"would_cite":false,"duration_ms":18020,"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":"This paper claims that adding true-time-delay elements to an analog beamformer lets a single RF chain create frequency-dependent beams, so one mmWave base station can serve multiple users in different directions simultaneously; a 28 GHz…","keywords":["joint phase-time array","true time delay","mmWave beamforming","analog beamforming","frequency division multiplexing","multi-user MIMO","28 GHz prototype","6G radio front end"],"falsifier":"Run the same uplink simulation with a standardized proportional-fair or max-throughput scheduler that does not reserve a minimum of four resource blocks per user per slot, and compare the cell-edge throughput ratio between JPTA and phased array; if the ratio falls materially below the number of multiplexed users, the claimed N-fold cell-edge gain and the $N^{1/3}$ coverage ratio are artifacts of the paper's allocation rule rather than properties of the hardware.","tokens_in":7862,"feed_emoji":"📡","tokens_out":5701,"duration_ms":54577,"temperature":0.7,"pith_summary":"This paper argues that a millimeter-wave base station can multiplex several users in different directions using a single radio-frequency chain, by replacing ordinary phase-shifter beamforming with a joint phase-time array (JPTA). Each antenna branch of a JPTA adds a programmable true-time-delay element next to a phase shifter, so the array's beam direction varies with frequency. That frequency dependence lets the base station split its bandwidth among users located in different angles, which a conventional analog beamformer cannot do. The paper backs the idea with a 28 GHz prototype that receives four simultaneous user signals with one chain, and with system simulations that put the cell-edge uplink throughput of a JPTA system N times above a phased-array baseline and extend coverage by roughly a factor of two. A sympathetic reader would care because this gain comes without extra power-hungry RF chains, the main cost of hybrid and digital beamforming.","feed_headline":"Single RF chain serves four mmWave users at once","feed_subtitle":"Time-delay elements make analog beams frequency-dependent, doubling simulated uplink cell-edge coverage.","key_machinery":"The central object is the joint phase-time array (JPTA): an antenna array in which each element is preceded by a programmable true-time-delay element and a phase shifter, all fed by one RF chain. The delay unit contributes a phase $2\\pi f_k \\tau_m$ that depends linearly on sub-carrier frequency $f_k$, so the composite weight $p_k$ can be a different steering vector on every sub-carrier. The paper uses two beam designs: Type-1, where the band is split and each subband points at a discrete direction, solved by the optimization in equation (4) using the algorithm of [6] or the single-shot method of [12]; and Type-2, where a closed-form delay profile $\\tau_m = (m-1)/W \\sin(\\Delta\\theta/2)$ produces a continuous 'rainbow' sweep across an angular range. This frequency-to-angle mapping is what lets one chain schedule different users on different resource blocks, and it is also what demands high-resolution delay elements with a wide delay range.","core_discovery":"The central claim is that the extra degree of freedom introduced by time delays turns an otherwise frequency-flat analog beam into a frequency-dependent one, and that this is enough to make a single RF chain serve multiple users in separate directions by frequency-division multiplexing. The paper derives the array response as $p_k = \\frac{1}{\\sqrt M}[e^{j\\phi_1+2\\pi f_k\\tau_1}, \\dots, e^{j\\phi_M+2\\pi f_k\\tau_M}]^T$, where the phase-shifter terms $\\phi_m$ set a common angle and the delay terms $\\tau_m$ make the beam sweep with frequency. With two beam classes, discrete-angle (Type-1) and rainbow (Type-2), it designs the delays by optimizing the fit between the achieved and desired steering vectors. The demonstration at 28 GHz with a 400 MHz band assigns 100 MHz to each of four UEs and reports EVM values 0.5–3 dB worse than single-beam analog reception, matching the predicted beamforming-gain loss. The system-level simulation then reports that at the cell edge a JPTA user gets $N$ times the throughput of a phased-array user because it transmits every slot with its share of resource blocks, while the phased-array user transmits only once per $N$ slots, and the coverage distance grows by $N^{1/3}$ for path-loss exponent 3.","pith_inferences":["If the scheduler does not guarantee every user a minimum number of resource blocks in every slot, the N-fold cell-edge ratio and the $N^{1/3}$ coverage ratio are upper bounds rather than typical gains; a proportional-fair scheduler that lets best-effort users burst could shrink the gap between JPTA and PAA.","The 1–3 dB per-beam gain loss is incurred because each subband steers the full array with a compromised weight; the paper notes this loss becomes less critical with larger arrays, so the architecture's advantage should improve as antenna panels scale up.","The same frequency-to-angle mapping could be used in reverse at the UE side: a JPTA receiver could distinguish multiple simultaneous transmitters by their subband, which suggests extensions to uplink multi-user detection without extra chains beyond what the paper demonstrates."],"forward_implications":["A JPTA base station can schedule users in different directions in the same time slot by assigning them non-overlapping subbands, eliminating the analog-beamforming constraint of one beam per slot.","At cell edge, where users are limited to the lowest modulation and a minimum of four resource blocks, per-user throughput becomes $N$ times the phased-array baseline because JPTA users transmit every slot instead of once per $N$ slots.","With path-loss exponent 3, uplink throughput coverage distance extends by the cube root of the number of multiplexed users, about $2\\times$ for $N=8$ users.","The maximum delay a JPTA codebook needs grows with the angular spread and number of users (2.5 ns for 2 users up to 35 ns for 16 users in the paper's examples), so the required delay range is a concrete design parameter.","As the number of scheduled users grows, the Type-1 discrete-angle beam converges to the Type-2 rainbow beam, so the same hardware can serve both scheduled multi-user traffic and fast beam sweeping."],"supporting_citations":[{"why":"Introduces the JPTA architecture and the iterative beam-design algorithm that Section II-B uses to solve the Type-1 optimization.","marker":"[6]"},{"why":"Provides the coverage-ratio analysis ($N^{1/3}$ for path-loss exponent 3) and the EESM-to-BLER throughput mapping used in Section IV.","marker":"[7]"},{"why":"Supplies the heuristic single-shot solution for Type-1 JPTA weights referenced in Section II-B.","marker":"[12]"},{"why":"Defines the rainbow-beam closed-form delay profile for Type-2 JPTA beams used in Section II-B.","marker":"[15]"},{"why":"Gives the urban-micro propagation measurements whose path-loss exponents justify the default $\\beta=3$ in the simulations.","marker":"[17]"},{"why":"Frames the comparison with hybrid beamforming in Section II-C, where emulating JPTA patterns is said to require many RF chains.","marker":"[4]"}],"fun_headline_variants":["Time-delay beams let one chain serve four users","Single RF chain, four mmWave beams from JPTA","Delay elements make one chain beam four ways","One chain, four users: JPTA at 28 GHz","Frequency-dependent beams double cell-edge uplink"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline coverage and throughput numbers rest on a scheduling rule in which every JPTA user receives at least four resource blocks in every uplink slot, while a phased-array user on the same cell edge gets four blocks only once per sweep round, together with a fixed path-loss exponent of 3.","fun_headline_variants_meta":{"raw":{"variants":["Time-delay beams let one chain serve four users","Single RF chain, four mmWave beams from JPTA","Delay elements make one chain beam four ways","One chain, four users: JPTA at 28 GHz","Frequency-dependent beams double cell-edge uplink"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000221,"raw_usage":{"total_tokens":1483,"prompt_tokens":1008,"completion_tokens":475,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":398}},"tokens_in":624,"tokens_out":475,"duration_ms":5692,"temperature":1.0,"reasoning_tokens":398,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T20:02:34.929518+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same uplink simulation with a standardized proportional-fair or max-throughput scheduler that does not reserve a minimum of four resource blocks per user per slot, and compare the cell-edge throughput ratio between JPTA and phased array; if the ratio falls materially below the number of multiplexed users, the claimed N-fold cell-edge gain and the $N^{1/3}$ coverage ratio are artifacts of the paper's allocation rule rather than properties of the hardware.","supporting_citations":[{"cited_title":"Joint phase-time arrays: A paradigm for frequency-dependent analog beamforming in 6G,","cited_arxiv_id":null,"evidence_quote":"Introduces the JPTA architecture and the iterative beam-design algorithm that Section II-B uses to solve the Type-1 optimization."},{"cited_title":"Extending uplink coverage of mmwave and terahertz systems through joint phase-time arrays,","cited_arxiv_id":null,"evidence_quote":"Provides the coverage-ratio analysis ($N^{1/3}$ for path-loss exponent 3) and the EESM-to-BLER throughput mapping used in Section IV."},{"cited_title":"mmFlexible: Flexible directional frequency multiplexing for multi-user mmWave networks,","cited_arxiv_id":null,"evidence_quote":"Supplies the heuristic single-shot solution for Type-1 JPTA weights referenced in Section II-B."},{"cited_title":"Rainbow-link: Beam-alignment-free and grant-free mmW multiple access using true-time-delay array,","cited_arxiv_id":null,"evidence_quote":"Defines the rainbow-beam closed-form delay profile for Type-2 JPTA beams used in Section II-B."},{"cited_title":"Millimeter wave and terahertz urban mi- crocell propagation measurements and models,","cited_arxiv_id":null,"evidence_quote":"Gives the urban-micro propagation measurements whose path-loss exponents justify the default $\\beta=3$ in the simulations."},{"cited_title":"An overview of signal processing techniques for millimeter wave MIMO systems,","cited_arxiv_id":null,"evidence_quote":"Frames the comparison with hybrid beamforming in Section II-C, where emulating JPTA patterns is said to require many RF chains."}],"review_version":1}