{"id":"f9bdb5c0-c074-407d-bcd3-13dc15095e8e","arxiv_id":"2607.20355","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A prototype MIT-LL CCD read with the Stanford MCRC ASIC achieves 5 Mpixel/s at 3.85 e- read noise, and a trap-based noise model offers an explanation for a temperature-dependent noise resonance.","lead":"X-ray detector engineers report that a small prototype CCD, paired with a custom ASIC and optimized clock and bias settings, reads out at up to 5 million pixels per second with less than 4 electrons of read noise. The work is a step toward the fast, low-noise imaging spectrometers that future large X-ray observatories need.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central mission-level claim is supported only by measurements on the small-format CCID-93 prototype; the paper acknowledges CCID-100 (the actual mission-scale device) is not yet characterized, so the transfer of speed/noise performance to large-format, multi-channel operation is untested.","rationale":"The reader identified the most load-bearing assumption: the CCID-93-to-CCID-100 transfer. The manuscript itself states that CCID-100 characterization is future work, so the strongest claim overreaches. The measured numbers on the prototype are credible and independently corroborated by the FWHM, so no issue with the measurement itself. The remaining concerns (selection of 'best' noise, tuned noise model, lack of code/data) are real but secondary; they would not change the verdict. Therefore I recommend the verdict remain CONDITIONAL.","tokens_in":13148,"tokens_out":8223,"duration_ms":66925,"concrete_test":"Test a CCID-100 device (or an equivalent 16-channel large-format device) with the same MCRC readout at 173 K, operating all 16 channels simultaneously, at serial transfer speeds of 2, 4, and 5 Mpixel/s per channel. Measure per-channel read noise (overscan method) and 5.9 keV single-pixel-event FWHM. Compare the channel-mean and channel-to-channel spread against Table 1. If any channel's read noise at 4 Mpixel/s exceeds ~4 e- rms, or if the array-averaged FWHM at any speed degrades by more than ~2 eV relative to the CCID-93 values, the 'can deliver' claim for a full strategic mission is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim in Section 8 ('Our results demonstrate clearly that the MIT-LL CCD fabrication process and state-of-the-art electronics can deliver the speed and noise performance required of future strategic X-ray missions') is not yet supported by data on the device that would actually fly. All headline numbers in Table 1 come from a single small-format CCID-93 (512x512, one readout channel). The very next sentence in Section 8 says the group is 'now focusing efforts to characterize and optimize larger-format, 16-channel 1440x1440 pixel CCID-100 devices, which have the same output stage configuration as the CCID-93.' This is an explicit admission that the large-format device has not been characterized. 'Same output stage configuration' does not guarantee identical performance: the CCID-100 has 16 parallel output stages; multi-channel operation can introduce crosstalk, ground/power coupling, and thermal gradients; the longer serial register and larger array can increase clock loading and charge transfer inefficiency at high speed; and the MCRC ASIC must drive 16 channels, possibly with multiple chips. Any of these could raise read noise above the sub-4-e- values shown in Table 1 or degrade the FWHM. Since the paper's central claim is about delivering mission-required performance, the missing demonstration on a mission-scale device is the key unverified step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports development and testing of a fast, low-noise CCD readout chain for future X-ray missions, combining the MIT-LL CCID-93 prototype CCD with the Stanford Multi-Channel Readout Chip (MCRC) ASIC. Contributions include locally buffered summing-well and reset-gate clock drivers, an automated four-parameter bias-scan procedure, and measured read-noise and FWHM values at serial speeds of 2–5 Mpixel/s. At 173 K the authors report read noise of 2.18–3.85 e− and Fe-55 5.9 keV FWHM of 124.9–130.8 eV. They also present waveform and PSD analyses and a trap-based model intended to explain a noise resonance near 240–260 K. The central claim, stated most strongly in Section 8, is that the MIT-LL process and the MCRC electronics 'can deliver the speed and noise performance required of future strategic X-ray missions.'","tokens_in":13594,"tokens_out":3315,"duration_ms":31907,"significance":"If the measured performance transfers to mission-scale detectors, the work is significant: it demonstrates that CCD-based X-ray imagers can reach frame rates an order of magnitude above legacy systems while maintaining sub-4-e− read noise and good spectral resolution. The paper's strengths are the concrete, speed-resolved measurements in Table 1, the practical engineering of onboard clock drivers that increase usable waveform samples, the systematic bias-scan methodology, and the use of PSD/waveform analyses to localize noise sources. These are useful contributions for detector development. However, the central mission-level claim is not yet supported because all quantitative results come from a single small-format, single-channel CCID-93; the mission-scale CCID-100 is explicitly stated as not yet characterized. The bias-scan outlier filtering and the noise-model parameters also require stronger justification before the headline numbers and physical interpretation can be taken at face value.","major_comments":[{"comment":"The statement 'Our results demonstrate clearly that the MIT-LL CCD fabrication process and state-of-the-art electronics can deliver the speed and noise performance required of future strategic X-ray missions' is stronger than the evidence. All Table 1 values are from the 512×512, single-output CCID-93, while the mission-relevant CCID-100 (1440×1440, 16 channels) is acknowledged as not yet characterized. 'Same output stage configuration' does not guarantee identical performance: multi-channel crosstalk, clock loading, thermal gradients, and ASIC fan-out can all affect noise and speed. Please either present CCID-100 data or revise the conclusion to a clearly labeled projection, distinct from demonstrated performance.","section":"Section 8, Table 1"},{"comment":"The bias-scan procedure filters out 'low noise outliers' because 'they reflect a collapse in the gain.' No quantitative criterion is given for identifying such outliers, and no validation is shown that the excluded points indeed have collapsed gain rather than genuinely improved noise. Since the Table 1 optimum is found from this scan, an arbitrary or unstable filtering threshold could bias the headline noise value. Please specify the outlier rule, report how many points were excluded, and demonstrate with gain measurements that the excluded points are non-functional.","section":"Section 5"},{"comment":"The output-stage noise model is presented as explaining the noise resonance, but the key parameters (N_trap, ΔE, σ) are freely chosen, and no quantitative fit or uncertainty is given. The model 'reproduces' the resonance by construction over the plotted parameter ranges, so it does not independently validate the VO-trap interpretation. Please reframe the model as illustrative, or provide an independent measurement or likelihood fit that constrains the parameters and demonstrates predictive power.","section":"Section 7.2, Eq. (3), Figs. 12–13"},{"comment":"The phrase 'required of future strategic X-ray missions' is not operationalized. The paper never lists the specific noise, FWHM, frame-rate, or power requirements of GOMaP, AXIS, or any other stated mission, nor does Table 1 compare the measured values to those requirements. Without explicit requirements and a compliance table, the central claim is not quantitatively supported. Please add the requirements and show how the measurements satisfy them, or temper the claim accordingly.","section":"Sections 1 and 8"}],"minor_comments":[{"comment":"The uncertainties on noise and FWHM (e.g., 2.18±0.01 e−) are reported to two decimals, but the statistical method used to derive them is not described. Please state how many frames/pixels were used and how the error bars were computed.","section":"Table 1"},{"comment":"The expanding-baseline analysis is described qualitatively. It would help to show the expected sqrt(N) white-noise curve on each panel of Figure 10, or to state explicitly how the standard deviation is normalized, so the deviation can be assessed by eye.","section":"Section 7.1"},{"comment":"The thermal-noise formula assumes identical noise from both stages and a fixed 60 MHz bandwidth. These assumptions should be justified or their sensitivity explored, since the quoted 'total noise' depends on them.","section":"Section 7.2.1, Eq. (2)"},{"comment":"Equation (3) uses a single capture cross-section and equal trapping/detrapping probabilities per time step. This is a strong simplification for interface traps; please note the limitation or use distinct capture and emission time constants if data require it.","section":"Section 7.2.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid instrumentation progress report with useful measurements and engineering. The main issue is the gap between the demonstrated small-format CCID-93 results and the mission-level conclusion. This is fixable by either adding CCID-100 data or carefully limiting the claims. The noise-model section would benefit from being framed as a hypothesis rather than a demonstrated physical explanation. No concerns about academic integrity or reference handling."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things. First, the measured readout performance here is credible and worth taking seriously: 2.18 e- at 2 Mpixel/s, 3.07 e- at 4 Mpixel/s, 3.85 e- at 5 Mpixel/s, with 5.9 keV FWHM between 124.9 and 130.8 eV. These numbers come from a real CCID-93 with the MCRC ASIC, and the error bars are small. Second, the paper's central claim that this demonstrates clearly the performance needed for future strategic X-ray missions is not supported by the data in front of you. The stress-test note is right: the headline numbers come from a single small-format, one-channel device. The paper itself says the CCID-100, the actual 16-channel mission-scale device, has not yet been characterized. Same output stage configuration is not the same as measured performance; crosstalk, clock loading, thermal gradients, and multi-channel ASIC drive can all degrade noise. That is a load-bearing gap, and it is the main reason I would not sign off on the strongest wording. What is genuinely new and good: the bias-scan method for co-optimizing RG, OG, RD across temperature is practical and clearly described; the onboard SW/RG clock drivers with snubber termination are a real engineering contribution; the PSD/expanding-baseline analysis is a sensible way to localize 1/f noise to the output stage. The RTN-based trap model is thought-provoking but not a prediction, because its parameters are chosen to reproduce the resonance the model is meant to explain. The paper does acknowledge this indirectly by exploring a range of energies and cross-sections, but it should be framed as an illustrative hypothesis, not a validated model. The post-hoc filtering of low-noise outliers in the bias scan is a minor concern; as long as the final operating point is stable and repeatable, it is acceptable, though the method would be stronger with a stated criterion. Who gets value from this: anyone building or planning fast X-ray CCD readout systems, and the GOMaP-era mission studies that need to know whether CCDs can reach the required frame rates. It deserves a serious peer review, but the referee should ask the authors to moderate the conclusion, say promising path toward instead of clearly deliver, and to clarify the model's status. If they later show CCID-100 data, this paper becomes much more important. I would bring it to a reading group only if people care specifically about detector readout electronics; otherwise, the main takeaway is the measured noise numbers and the bias-scan technique.","headline":"Solid engineering report with believable speed/noise numbers; the mission-level claim runs ahead of the data.","tokens_in":740,"tokens_out":885,"would_cite":true,"duration_ms":20935,"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":"A prototype CCD and custom ASIC achieve sub-4-electron read noise at 5 Mpixel/s, the speed and noise combination future strategic X-ray missions require.","keywords":["X-ray CCD","low-noise readout","ASIC","MCRC","bias optimization","1/f noise","output stage","fast readout"],"falsifier":"Measure read noise on a large-format CCID-100 detector at 173 K at 5 Mpixel/s per channel using the same ASIC and bias-scanning procedure; if the read noise exceeds 4 e- or the 5.9 keV line width exceeds about 135 eV, the central claim that the fabrication and electronics meet future mission requirements is not supported.","tokens_in":13099,"feed_emoji":"🛰️","tokens_out":10150,"duration_ms":64715,"temperature":0.7,"pith_summary":"This paper demonstrates that a prototype CCD (CCID-93) read out by a custom ASIC (MCRC) achieves read noise of 2.18 electrons at 2 Mpixel/s and 3.85 electrons at 5 Mpixel/s, with 5.9 keV X-ray line widths of 125-131 eV. That combination of speed and noise is about an order of magnitude faster frame rate than legacy X-ray CCDs while keeping noise below 4 electrons. The paper also introduces an automated scan of four bias voltages that finds minimum-noise operating points, and a physical model of charge trapping in the output stage that explains a temperature-dependent noise resonance. If these results carry over to larger-format detectors, future X-ray missions can have fast, low-noise imaging spectrometers.","feed_headline":"CCD readout hits sub-4-electron noise at 5 Mpixel/s","feed_subtitle":"Prototype CCD and ASIC read at 5 Mpixel/s while keeping 5.9 keV line width below 131 eV.","key_machinery":"The key elements are the two-stage output stage of the CCID-93 (a p-JFET source follower followed by an n-MOSFET), the MCRC ASIC providing eight parallel low-noise readout channels, locally buffered summing-well and reset-gate clock drivers that recover roughly 20% more waveform samples, and a four-dimensional bias scan over reset-gate high/low, output-gate, and reset-drain voltages that finds minimum-noise operating points. The noise model is a random-telegraph-noise simulation with thermally activated traps, which reproduces the resonance in noise versus temperature and readout speed.","core_discovery":"The central claim is that the CCID-93 detector, the MCRC readout ASIC, onboard fast clock drivers, and the bias-scanning procedure together deliver the speed and noise performance required of future strategic X-ray missions. Concretely, at 173 K the read noise is 2.18±0.01 e- at 2 Mpixel/s, 3.07±0.01 e- at 4 Mpixel/s, and 3.85±0.02 e- at 5 Mpixel/s, while the single-pixel 5.9 keV Fe-55 line width is 124.9±1.0 to 130.8±0.9 eV. The paper further posits that the observed 1/f noise and the noise resonance near 240-260 K arise from trapping and detrapping of charge carriers in the output stage, with trap parameters consistent with shallow, weakly interacting defects such as vacancy-oxygen complex","pith_inferences":["If the small-format results transfer to the 16-channel CCID-100, a full 1440x1440 frame could be read in under a second, enabling rapid time-domain X-ray observations that current missions cannot do.","The trap model implies that detectors with different fabrication histories could have noise resonances at different temperatures, so flight detectors should be individually characterized before the operating point is fixed.","The bias-scanning and PSD-analysis methods could be adopted as standard calibration practices for any CCD-based X-ray instrument, potentially shortening ground-testing time."],"forward_implications":["Future X-ray missions can use CCDs at frame rates an order of magnitude faster than legacy observatories while keeping read noise below 4 electrons.","The automated bias scan efficiently finds optimal operating points for detectors, which should streamline the tuning of large-format, multi-channel devices.","The onboard clock architecture reduces electromagnetic interference and enables serial transfer speeds up to 5 Mpixel/s.","The trap model predicts that the noise resonance shifts to higher temperatures at higher readout speeds, guiding selection of operating temperature."],"fun_headline_variants":["CCD noise down to 2.18 e- at 2 Mpixel/s with bias scan","Sub-4 e- read noise at 5 Mpixel/s on X-ray CCD","Bias-scan method enables fast, low-noise CCD readout","X-ray CCD readout: 2.18 e- noise at 2 Mpixel/s","Fast, low-noise CCD readout: 3.85 e- at 5 Mpixel/s"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The paper assumes that performance measured on the small-format CCID-93 prototype transfers to the large-format, 16-channel CCID-100 detectors that share the output stage but differ in layout, power, thermal, and cross-talk conditions; this transfer has not yet been tested.","fun_headline_variants_meta":{"raw":{"variants":["CCD noise down to 2.18 e- at 2 Mpixel/s with bias scan","Sub-4 e- read noise at 5 Mpixel/s on X-ray CCD","Bias-scan method enables fast, low-noise CCD readout","X-ray CCD readout: 2.18 e- noise at 2 Mpixel/s","Fast, low-noise CCD readout: 3.85 e- at 5 Mpixel/s"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000535,"raw_usage":{"total_tokens":2434,"prompt_tokens":794,"completion_tokens":1640,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":1525}},"tokens_in":538,"tokens_out":1640,"duration_ms":22738,"temperature":1.0,"reasoning_tokens":1525,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:03:10.328991+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure read noise on a large-format CCID-100 detector at 173 K at 5 Mpixel/s per channel using the same ASIC and bias-scanning procedure; if the read noise exceeds 4 e- or the 5.9 keV line width exceeds about 135 eV, the central claim that the fabrication and electronics meet future mission requirements is not supported.","supporting_citations":[],"review_version":1}