{"id":"bce4ac13-a869-401d-b5d2-afdd61e10448","arxiv_id":"2505.00297","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A custom low-noise, low-drift dual-channel DC source was built and validated on a 66-qubit processor, maintaining qubit coherence and frequency stability over 12 hours.","lead":"QPower is a compact two-channel DC voltage source built for biasing superconducting qubits and their amplifiers, with reported noise below 20 nV/√Hz and 12-hour drift under 5 µV. When used on a 66-qubit processor, the qubit frequency stayed within ±40 kHz over 12 hours, suggesting a practical alternative to bulky commercial bias supplies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reported ripple (<500 µVpp) and RMS noise (335 µVrms) are mathematically incompatible: any AC signal satisfies Vpp ≥ 2·Vrms, so 335 µVrms forces Vpp ≥ 670 µV; the central noise benchmarks cannot both be correct.","rationale":"The reader's weakest assumption correctly identifies that the headline specs are close to instrument floors, making the measurement protocol important. My concern is stronger and independent of calibration: the reported ripple and RMS values violate a bound that holds for every AC signal, so they cannot both be true under the stated measurement conditions. This directly threatens the central claim of low-noise, stable DC output. The proposed check—computing RMS and peak-to-peak from the same raw record—would settle whether the discrepancy is a unit error, a bandwidth difference, or a misstatement of one of the specs. I do not change the overall CONDITIONAL verdict, but the acceptance condition should now include resolving this ripple/RMS inconsistency, not merely re-measuring near the instrument floor.","tokens_in":9810,"tokens_out":11920,"duration_ms":126615,"concrete_test":"Extract the time-domain record behind Fig. 3(a) and Fig. 3(b) at the same output voltage (especially 7 V) with the same 20 MHz bandwidth, and compute both peak-to-peak and RMS from the same waveform. If RMS ≈ 335 µVrms, the peak-to-peak must be ≥ 670 µV, so a <500 µVpp result cannot be reproduced. Alternatively, the authors should release the raw traces or explicitly state the different output voltages, bandwidths, or definitions of 'ripple' used for the two headline numbers.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing issue is internal: the ripple and RMS noise figures reported in Section II and Fig. 3 cannot both describe the same output. Fig. 3(a) reports <470 µVpp ripple over DC–20 MHz; Fig. 3(b) reports RMS noise up to 335 µVrms at 7 V output, and the text says the RMS measurement used 'identical configurations'. For any zero-mean AC signal, RMS ≤ Vpp/2, because every sample is bounded by ±Vpp/2. Hence 335 µVrms forces Vpp ≥ 670 µV. Even subtracting the stated 80 µVrms oscilloscope floor gives a DUT RMS of about 325 µVrms, still forcing Vpp ≥ 650 µV. The claim '<500 µVpp' is therefore mathematically incompatible with the reported RMS value. If 'ripple' is meant to exclude random noise, that distinction must be stated explicitly; the current text defines ripple as the residual AC voltage and the abstract presents '<500 µVpp' as an unconditional headline spec. This is not a calibration subtlety; it is a logical inconsistency in the central electronics benchmarks.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the design and characterization of QPower, a custom dual-channel DC source module intended for biasing superconducting quantum processors and cryogenic amplifiers. The claimed key specifications are a ±7 V, 200 mA output with noise spectral density below 20 nV/√Hz at 10 kHz, output ripple below 500 µVpp over a 20 MHz bandwidth, long-term drift below 5 µVpp over 12 hours, and channel-to-channel crosstalk below 0.3 ppm. The authors also integrate three QPower modules into the control stack of a 66-qubit processor and report T1 = 87.6 µs, Ramsey T2 = 5.1 µs, and qubit-frequency drift within ±40 kHz over 12 hours, which they interpret as evidence that the source enables state-of-the-art qubit coherence and stability.","tokens_in":10046,"tokens_out":3295,"duration_ms":36717,"significance":"If the reported electronics specifications are correct, QPower would be a useful, compact, low-power alternative to commercial DC sources for intermediate-scale superconducting quantum systems, and the qubit-level integration provides a relevant end-to-end demonstration. The strengths of the paper are that the performance numbers are direct measurements rather than fitted parameters, and the qubit benchmark is an honest integration test. However, the central electronics claims contain an internal mathematical inconsistency and several headline numbers sit close to the instrument floors, so the significance of the work cannot be assessed until the measurement protocol and the reported values are reconciled.","major_comments":[{"comment":"The reported ripple and RMS noise figures cannot both describe the same output signal. For any AC signal, the peak-to-peak amplitude Vpp and RMS value satisfy Vrms ≤ Vpp/2, because every sample is bounded by ±Vpp/2. Fig. 3(b) reports a maximum RMS noise of 335 µVrms at 7 V output, which therefore requires Vpp ≥ 670 µV. This is incompatible with the ripple claim of <500 µVpp in Fig. 3(a), even after subtracting the stated 80 µVrms oscilloscope floor, which would still force Vpp ≥ 650 µV. If 'ripple' is intended to exclude random noise or to refer only to deterministic periodic components, that distinction must be stated explicitly in Section II and in the abstract; as written, the text defines ripple as 'residual AC voltage superimposed on DC output' and presents '<500 µVpp' as an unconditional headline specification. This is a load-bearing inconsistency in the central electronics benchmark and must be resolved.","section":"Section II, Fig. 3(a)-(b)"},{"comment":"The claimed performance margins over the measurement-instrument floors are small and the measurement protocol is not sufficiently specified. The ripple claim of <500 µVpp is only a factor of five above the stated oscilloscope floor of ~100 µVpp, and the noise spectral density of 20 nV/√Hz is only a factor of two above the analyzer floor of ~10 nV/√Hz. The manuscript does not report how the floor was subtracted (or whether it was subtracted), the number of acquisitions, the resolution bandwidth, averaging settings, or repeated measurements on more than one unit. Without this information, the reader cannot distinguish a genuine specification from instrumentation-limited bounds. Please provide a complete metrology description for each panel of Fig. 3, including same-settings floor measurements at the same bandwidths and output conditions, and state the uncertainty or reproducibility of each headline number.","section":"Section II, Fig. 3 captions and Section II 'Electronics performance'"},{"comment":"The causal claim that QPower 'directly correlates DC source noise performance with quantum coherence preservation' is not supported by the data as presented. The qubit coherence times and frequency drift are measured at a single flux-sensitive operating point using QPower, but there is no control measurement with a commercial reference source under otherwise identical conditions, and no quantitative model that connects the measured DC noise or drift to the observed ±40 kHz frequency fluctuation and T2 values. The integration test demonstrates that QPower does not visibly degrade the processor, which is valuable, but the stronger causal wording in the Conclusion should be softened or supported by a comparative benchmark.","section":"Section III and Conclusion"}],"minor_comments":[{"comment":"Use consistent notation for 'QPower' and 'Qpower', and consistently write peak-to-peak quantities as 'µVpp' in Table I and the abstract; the table currently gives '<500 µV (20 MHz bandwidth)' without the 'pp' subscript.","section":"Throughout"},{"comment":"There is a typo in the Fig. 5(d) label 'Ramsy' (should be 'Ramsey'), and in the text '87.6±3 µs s' contains a stray 's'.","section":"Section III, Fig. 5"},{"comment":"Please specify the load conditions (output current, cable/termination, and integration time) for the 12-hour drift measurement in Fig. 3(e) and the crosstalk measurement in Fig. 3(f), since DC source drift and crosstalk can depend strongly on load.","section":"Section II, long-term stability and crosstalk measurements"},{"comment":"The DAC reference [33] is an unrelated conference paper; the AD5791 and AD8676 datasheets should be cited directly, and the reference list should be checked for part-number and citation consistency.","section":"References and part numbering"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern raised by the reader lands: the Vpp-versus-RMS inconsistency in Section II is real and central. The paper is otherwise a straightforward hardware paper, and the qubit integration is a positive feature, but the electronics metrology needs to be made self-consistent and the measurement floors need to be handled much more carefully before the claimed specifications can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's the quick read on arXiv:2505.00297. The authors built a custom dual-channel DC source for superconducting qubit control, integrated it into a 66-qubit setup, and report low noise and drift. The integration work is real and useful: small footprint, low power, and qubit coherence numbers T1 = 87.6 µs, T2 = 5.1 µs, with ±40 kHz drift over 12 hours. But there is a load-bearing internal contradiction in the electronic specifications.\n\nThe paper claims ripple <500 µVpp (Fig. 3a) and simultaneously reports RMS noise up to 335 µVrms at 7 V output (Fig. 3b), explicitly saying the RMS measurement used identical configurations. For any AC signal, Vpp ≥ 2·Vrms. 335 µVrms forces Vpp ≥ 670 µV. Even subtracting the stated 80 µVrms instrument floor leaves roughly 325 µVrms, which still forces Vpp ≥ 650 µV. So the headline ripple spec cannot be correct. This is not a subtle calibration issue; the two headline numbers are mutually exclusive.\n\nThe other specs also sit close to the measurement floors: 500 µVpp versus a 100 µVpp oscilloscope floor, and 20 nV/√Hz versus a 10 nV/√Hz analyzer floor. That would warrant more careful protocol detail, but the internal inconsistency is decisive.\n\nOn the positive side, the design is conventional but sensibly assembled: LTZ1000 reference, AD5791 DAC, LDOs, amplifier array with phase compensation. The 66-qubit integration is a genuine step beyond bench characterization, even if it lacks a controlled A/B baseline against a commercial source. The conclusion's language that qubit stability 'directly correlates' with QPower noise is stronger than the evidence supports without such a baseline.\n\nRecommendation: send to peer review, but with the expectation of major revision. The authors need to correct or clarify the ripple and RMS figures, specify measurement bandwidths and configurations, or explicitly distinguish periodic ripple from broadband noise. As written, the central specs are not believable, and a referee should not have to unwind this much. If fixed, the paper would be a useful contribution to quantum control electronics. As is, I would not cite it in my own work.","headline":"Useful 66-qubit integration work, but the headline ripple and RMS noise numbers contradict each other mathematically, so the central specs are not credible as written.","tokens_in":10637,"tokens_out":3698,"would_cite":false,"duration_ms":35791,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Lx"],"model":"deepseek-v4-flash","headline":"A compact two-channel DC source achieves 20 nV/√Hz noise and keeps a 66-qubit processor's qubit frequencies stable to ±40 kHz for 12 hours.","keywords":["superconducting quantum circuits","superconducting qubit","low-noise DC source","QPower","flux bias","qubit coherence","cryogenic amplifier bias","quantum control electronics"],"falsifier":"Take a QPower channel set to 7 V and measure its output noise with an independently calibrated, battery-powered low-noise preamplifier and digitizer whose input noise is below 5 nV/√Hz at 10 kHz and below 1 µVrms; if the measured noise and ripple do not remain at or below the claimed levels after subtracting this front-end's noise, then the published numbers include instrument floor.","tokens_in":9627,"feed_emoji":"⚡","tokens_out":7916,"duration_ms":78037,"temperature":0.7,"pith_summary":"The paper reports a custom dual-channel DC voltage source, QPower, built for the biasing needs of superconducting quantum processors: flux bias for tunable qubits and couplers, and supply for cryogenic and room-temperature amplifiers. It claims that one compact 2U module can deliver ±7 V at up to 200 mA per channel with a noise spectral density of 20 nV/√Hz at 10 kHz, ripple below 500 µVpp over 20 MHz bandwidth, and drift below 5 µVpp over 12 hours. The point of these specifications is that DC biasing, not just fast microwave control, becomes a bottleneck as qubit counts grow, and a small low-power module that matches commercial instruments removes that bottleneck. Integration into a 66-qubit processor is reported to preserve T1 = 87.6 µs, Ramsey T2 = 5.1 µs, and qubit frequency drift within ±40 kHz over 12 hours.","feed_headline":"DC source keeps a 66-qubit chip coherent for 12 hours","feed_subtitle":"Two-channel ±7 V module: 20 nV/√Hz noise, 5 µV drift, and T1 = 87.6 µs in a superconducting processor.","key_machinery":"The load-bearing hardware chain is the voltage-reference-to-output cascade: an LTZ1000ACH reference (1.2 µV/√Hz noise, 0.05 ppm/°C) feeding a 20-bit AD5791 DAC, followed by a buffered amplifier array with a phase-compensation network. The compensation network is the element that makes the high-current output stable: it raises the feedback phase margin from 18.1° to 64.8° and the unity-gain bandwidth from 5.95 MHz to 8.12 MHz, eliminating self-oscillation and reducing output ripple from 5 mVpp to 500 µVpp. The three-stage power-conditioning chain (isolating DC-DC converters, LT3045/LT3094 LDOs, and a π-type low-pass filter) is what keeps the output noise below 20 nV/√Hz at 10 kHz and the long-term drift below 5 µVpp.","core_discovery":"The central claim is that QPower, a distributed-reference DC source using an LTZ1000ACH reference, a 20-bit DAC, LDO regulation, and a phase-compensated amplifier array, can deliver precision DC bias with performance comparable to premium commercial instruments while occupying a two-channel board and drawing 15 W per channel. The paper argues this is achieved by a three-stage noise-suppression architecture—DC-DC isolation, low-dropout regulation, and a π-type 100 Hz low-pass filter—and by a compensation network that raises the output stage's phase margin from 18.1° to 64.8°, cutting ripple from 5 mVpp to below 500 µVpp. As validation, QPower modules bias a Josephson parametric amplifier, a cryogenic HEMT and room-temperature LNA, and a qubit/coupler on a 66-qubit processor, yielding T1 = 87.6 ± 2.0 µs, T2,ramsey = 5.1 ± 0.4 µs, T2,echo = 23.5 ± 1.2 µs, and 12-hour qubit frequency stability within ±40 kHz. Underlying the significance is the scalability argument: the module replaces racks of commercial DC sources with small, low-power units.","pith_inferences":["If the headline noise and ripple figures are confirmed with a lower-floor measurement chain, the same reference-and-LDO topology could be adapted to bipolar current sources or programmable bias ramps, since the module already has a digital interface.","The reported ±40 kHz qubit frequency drift over 12 hours implies about 16 kHz per µV of voltage drift if source drift were the only cause; comparing that slope with independent flux-noise spectroscopy would separate source drift from intrinsic qubit flux noise.","A natural next test is to measure crosstalk across multiple QPower modules in a rack, since the reported 0.3 ppm figure is within a single board and inter-module isolation is the quantity that matters for thousand-qubit systems.","The high-frequency spurious spec below −95 dBm suggests the module could also bias sensitive RF amplifiers in precision metrology outside quantum computing."],"forward_implications":["A single QPower module can simultaneously bias a qubit flux line, a tunable coupler, a Josephson parametric amplifier, and cryogenic and room-temperature LNAs, replacing several bench instruments.","At 15 W per channel and two channels per board, scaling to hundreds of qubits requires far less power and rack space than commercial DC sources.","The reported 12-hour drift below 5 µVpp and ±40 kHz qubit frequency stability mean long calibration and data-acquisition runs need not be interrupted by DC drift.","The distributed-reference, modular design is presented as extensible to other quantum platforms such as silicon spin qubits and to precision measurement applications.","With 200 mA maximum output per channel, the module can power active amplifier chains, not just provide bias voltages."],"supporting_citations":[{"why":"Supplies the prior DC-bias source design whose distributed-reference topology and LTZ1000ACH choice QPower adopts and extends.","marker":"[16]"},{"why":"The LTZ1000 datasheet that specifies the reference's 1.2 µV/√Hz noise and 0.05 ppm/°C temperature coefficient, which the design relies on for low-noise, low-drift output.","marker":"[31]"},{"why":"Describes the 66-qubit processor used for the qubit coherence and frequency-stability benchmark.","marker":"[32]"},{"why":"The microwave measurement and control system QPower is integrated into; it supplies the control and readout chain for the qubit benchmarking.","marker":"[8]"},{"why":"Provides the DAC transfer-function equation used to map the 20-bit digital code to the output voltage.","marker":"[33]"}],"fun_headline_variants":["QPower: 20 nV/√Hz noise, 5 µV drift for 66-qubit system","Low-noise DC source: 20 nV/√Hz, 12-hour stability on 66 qubits","Dual-channel DC: 20 nV/√Hz noise, T1 87.6 µs on 66-qubit chip","Custom DC module: 20 nV/√Hz noise, 5 µV drift for qubits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline noise and ripple figures assume the measurement instruments' own noise floors were cleanly separated from the module's output, since the claimed ripple is only about five times the oscilloscope floor and the claimed noise only about twice the spectrum analyzer floor.","fun_headline_variants_meta":{"raw":{"variants":["QPower: 20 nV/√Hz noise, 5 µV drift for 66-qubit system","Low-noise DC source: 20 nV/√Hz, 12-hour stability on 66 qubits","Dual-channel DC: 20 nV/√Hz noise, T1 87.6 µs on 66-qubit chip","Custom DC module: 20 nV/√Hz noise, 5 µV drift for qubits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001035,"raw_usage":{"total_tokens":4420,"prompt_tokens":1073,"completion_tokens":3347,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":689,"completion_tokens_details":{"reasoning_tokens":3230}},"tokens_in":689,"tokens_out":3347,"duration_ms":25318,"temperature":1.0,"reasoning_tokens":3230,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:45:37.659201+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a QPower channel set to 7 V and measure its output noise with an independently calibrated, battery-powered low-noise preamplifier and digitizer whose input noise is below 5 nV/√Hz at 10 kHz and below 1 µVrms; if the measured noise and ripple do not remain at or below the claimed levels after subtracting this front-end's noise, then the published numbers include instrument floor.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the prior DC-bias source design whose distributed-reference topology and LTZ1000ACH choice QPower adopts and extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The LTZ1000 datasheet that specifies the reference's 1.2 µV/√Hz noise and 0.05 ppm/°C temperature coefficient, which the design relies on for low-noise, low-drift output."},{"cited_title":"Zhang, X","cited_arxiv_id":null,"evidence_quote":"The microwave measurement and control system QPower is integrated into; it supplies the control and readout chain for the qubit benchmarking."},{"cited_title":"Wang and X","cited_arxiv_id":null,"evidence_quote":"Provides the DAC transfer-function equation used to map the 20-bit digital code to the output voltage."}],"review_version":1}