{"id":"84046697-79ad-48cd-9702-21bf2998ff97","arxiv_id":"2501.03283","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Ground tests with a CPU-unlimited parallel processor characterize three high-count-rate effects in the XRISM/Resolve microcalorimeter and produce empirical models for correcting them.","lead":"This paper measures how the XRISM/Resolve X-ray microcalorimeter loses performance at high count rates, using ground test data with a special setup that bypasses the onboard CPU limit. It models three effects, CPU overload, pulse pile-up, and electrical cross talk, so that bright-source observations can be planned and corrected.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (2) mixes PSP and SCDP processing paths, so the fitted cross-talk model may absorb CPU-loss and pile-up residuals rather than isolating cross-talk.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing issue: Eq. (2) mixes the PSP and SCDP processing paths, so the derived FWHM_excess and the subsequent Eq. (3) cross-talk model are not isolated from CPU-loss and dead-time effects. This is the central quantitative claim because it drives the 206 km/s versus 200 km/s astrophysical illustration and the guidance on when to apply cross-talk cuts. The paper is otherwise a careful, honest ground-calibration study with real measurements, a useful CPU-load model, and a reasonable pile-up exposure correction; those components are not in question. The concern is specific and testable: rederive Eq. (3) from SCDP-only comparisons or from PSP data below the CPU limit. If the relation is robust, the paper's main phenomenological conclusion survives; if not, the cross-talk model and its astrophysical application need revision. This does not move the overall verdict, which remains CONDITIONAL pending that check.","tokens_in":20768,"tokens_out":3502,"duration_ms":36892,"concrete_test":"Re-derive Eq. (3) using only SCDP data for both terms: set FWHM_withXtalk to the SCDP blue points (no cross-talk cut) and FWHM_noXtalk to the SCDP orange points (first-neighbor cut) in the right panel of Fig. 16, both processed without CPU limitation and with the same pile-up screening. Refit the quadratic relation of Eq. (3). If the coefficients change substantially from 0.125 and 0.054, or if the relation becomes inconsistent with the published fit, the cross-talk model is contaminated by CPU-loss and pile-up differences. As a second check, refit using only PSP data below the CPU limit (rate < 4.2 s^-1 pixel^-1), where PSP and SCDP should agree, and compare the extrapolation to high rates.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central cross-talk model, Eq. (3), is trained on FWHM_excess defined in Eq. (2) as a quadrature difference between FWHM_withXtalk from the PSP path without a cross-talk cut and FWHM_noXtalk from the SCDP path with a cross-talk cut (Fig. 16, left blue vs. right orange). This assumes the only relevant difference between the two paths is cross talk, but the paths also differ in CPU behavior and pile-up processing. The PSP path suffers event loss and dead time beyond the CPU limit, and the paper itself notes that the cross-talk cut is incomplete in PSP because cross-talk parents can fall inside dead-time intervals. The SCDP path has no CPU limit and uses a smaller pile-up rejection threshold (dtthres ~0.8 ms versus ~2 ms in PSP, §3.3.2). Thus FWHM_excess as constructed can contain CPU-loss-induced degradation and residual pile-up differences, not just untriggered electrical cross talk. Since Eq. (3) with coefficients 0.125 and 0.054 is the quantitative engine behind the astrophysical conclusion (vturb = 206 km/s instead of 200 km/s), a contamination in FWHM_excess propagates directly into the recommended cross-talk correction and the simulated impact on GX 13+1. The concern is not that cross talk is absent; it is that the separable decomposition claimed in the paper has not been cleanly demonstrated at the point where it matters most.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents ground-test measurements of the XRISM/Resolve microcalorimeter at high count rates, exploiting a unique setup in which the flight PSP processing is run in parallel with an SCDP path that is free of CPU limits. The authors model three effects that degrade performance at high count rates: the CPU limit (a linear grade-dependent CPU-load model fit to Ni data and validated on KBr data), pulse pile-up (an effective exposure-time model based on a Poisson survival fraction), and untriggered electrical cross talk (an empirical relation between an excess width FWHM_excess and the cross-talk-contaminated fraction beta_XTalk). These models are then applied to a simulated observation of GX 13+1, where the predicted cross-talk broadening of 1.19 eV inflates the recovered turbulent velocity from 200 to 206 km/s if left uncorrected.","tokens_in":21208,"tokens_out":7517,"duration_ms":68271,"significance":"The paper addresses a practical and important problem for high-resolution microcalorimeter spectroscopy at high count rates. Its principal strengths are the unique parallel PSP/SCDP ground-test dataset, the successful use of lost-event telemetry to correct bad time intervals to within 4%, and the independent validation of the CPU-load model on KBr data. The resulting phenomenological tools (Eqs. 1, 3, 4, 5) are directly useful for observation planning, and the warning about artificially flux-dependent turbulent velocities is scientifically important. The main weakness is the construction of FWHM_excess, which mixes the PSP and SCDP processing paths and therefore may not isolate cross talk as cleanly as claimed; this affects the quantitative astrophysical prediction.","major_comments":[{"comment":"The excess broadening FWHM_excess is computed as (FWHM_withXtalk^2 - FWHM_noXtalk^2)^{1/2}, where FWHM_withXtalk is the PSP-processed spectrum without a cross-talk cut (blue bins, left panel of Fig. 16) and FWHM_noXtalk is the SCDP-processed spectrum with a first-neighbor cross-talk cut (orange bins, right panel of Fig. 16). These two paths differ in more than the cross-talk cut: the PSP path is subject to the CPU limit and associated event loss, while the SCDP path is not, and the pile-up rejection thresholds differ (approximately 2 ms in PSP versus 0.8 ms in SCDP; see §3.3.2). The paper itself notes in §3.3.3 that beyond the CPU limit some cross-talk parents fall inside dead-time intervals and cannot be identified. Consequently, FWHM_excess — and therefore the fitted coefficients in Eq. (3) — can absorb CPU-loss broadening and residual pile-up differences rather than isolating untriggered electrical cross talk. Because Eq. (3) is the engine behind the simulated GX 13+1 excess broadening of 1.19 eV and the vturb = 206 km/s result, the separable decomposition that the paper claims is not cleanly demonstrated at this load-bearing point. I recommend recomputing FWHM_excess from the SCDP no-cut and SCDP first-neighbor-cut spectra of the right panel of Fig. 16, where only the cut differs, or providing a quantitative demonstration that the PSP and SCDP no-cut FWHM values agree at all count rates used in the fit.","section":"§4.1.3, Eq. (2)"},{"comment":"The hardness correction applied to FWHM_excess for GX 13+1 is a linear scaling by E_GX13+1/E_Ni-K = 4.14/7.47 = 0.58, with no empirical or simulated calibration. The paper states in §4.1.3 that the degradation is highly dependent on spectral hardness, but the assumed linear dependence on count-weighted average photon energy is an ad hoc ansatz that directly sets the predicted 1.19 eV broadening and the vturb inflation. A validation of this scaling, or at least an error bar that propagates the uncertainty in the hardness scaling, is needed before the astrophysical conclusion can be quantified.","section":"§4.2.2, hardness scaling"},{"comment":"The pile-up model alpha_noPileUp is fitted to a deterministic curve generated by assuming the Poisson survival fraction exp(-2*nu*Delta_t_thres) with Delta_t_thres = 2 ms, which is the same formula introduced in §3.3.2. The fit therefore provides no independent validation of the pile-up model; only the 4% shortfall in Steps 2 and 19 of Table 3 is direct empirical evidence, and that evidence is consistent but not strongly constraining. The text should state explicitly that Fig. 19 is a reparameterization of the assumed analytic pile-up formula rather than a measurement, so that readers do not interpret the good fit in Fig. 19 as empirical validation.","section":"§4.1.2, Fig. 19"}],"minor_comments":[{"comment":"In the sentence about the cross-talk cut working only partially, \"a some events\" should be \"some events\".","section":"§3.3.3"},{"comment":"In the phrase \"the Mn K α enery band\", \"enery\" should be \"energy\".","section":"§3.2.2"},{"comment":"The pointer \"Figures 25 and 4.2.2\" should be \"Figures 25 and 26\".","section":"§4.2.2"},{"comment":"The phrase \"the most right-handed data bins\" should be \"the rightmost data bins\".","section":"§3.3.3"},{"comment":"Consider adding a sentence in the caption clarifying that columns (1)-(7) are counts rates in s^-1 pixel^-1 and that the 4% shortfall in Steps 2 and 19 is the pile-up effect discussed in the text.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a technically solid instrument-calibration contribution, and the parallel PSP/SCDP ground test is a valuable dataset. The main technical concern (Eq. 2) is fixable with data that already appear in Fig. 16, so this is within the scope of a major revision rather than a rejection. The hardness scaling is also ad hoc and should be flagged as an extrapolation with a systematic range. The paper is well suited to an instrument/calibration journal; the astrophysical application is illustrative."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is worth taking seriously. Its main contribution is a ground test that runs PSP-equivalent processing without CPU limits in parallel through SCDP, giving a unique handle on CPU-induced event loss that cannot be obtained in orbit. The CPU-load model (Eq. 1) is fitted on Ni data and validated on independent KBr data—good practice. The bad-time-interval correction recovers count rates to within 4% at the highest rates, which is a concrete, useful result. The empirical models for pile-up and cross-talk exposure time are clearly intended for observation planning, and the GX 13+1 simulation illustrates the risk of spurious turbulent-velocity trends if cross talk is ignored.\n\nThe main soft spot is exactly what the stress-test note flags: Eq. (2) defines FWHM_excess as the quadrature difference between PSP data without a cross-talk cut and SCDP data with a cross-talk cut. The two paths differ not only in the cut but also in CPU event loss, dead time, and pile-up threshold (SCDP uses ~0.8 ms vs ~2 ms). The paper itself notes that the PSP cross-talk cut is incomplete because parents can fall inside dead-time intervals. So the fitted Eq. (3) may absorb CPU-loss and pile-up residuals rather than isolating untriggered electrical cross talk. This does not make the cross-talk effect disappear—the qualitative trend is clear and the recovery with the cut is real—but it means the clean separation into three effects is not demonstrated at exactly the point that feeds the astrophysical conclusion.\n\nMinor issues: fit uncertainties for Eq. (3) are not reported, and the spectral-hardness scaling (0.58 ratio of mean energies) is ad hoc and unvalidated. The models are openly phenomenological, which is fine for planning tools, but the GX 13+1 numbers should be treated as illustrative. The data are not public, though the methods are detailed enough for the instrument team to re-implement.\n\nThe paper is honest, gives clear guidance for observers, and ships a genuinely new dataset plus two independently validated pieces (CPU model, bad-time correction). I would send this to peer review, asking the authors to quantify the cross-talk model uncertainty and to discuss the PSP/SCDP confound explicitly.","headline":"Solid instrument-calibration study with a unique parallel-processing ground test; the cross-talk model rests on a somewhat mixed comparison, but the paper deserves serious refereeing.","tokens_in":21793,"tokens_out":1592,"would_cite":true,"duration_ms":16952,"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":"Ground tests with a parallel CPU-unlimited processor separate the three high-count-rate degradation mechanisms of XRISM/Resolve and quantify the cross-talk contribution with a quadratic relation.","keywords":["XRISM","Resolve","X-ray microcalorimeter","high count rate","cross talk","pile-up","CPU limit","spectral resolution"],"falsifier":"Measure the $\\mathrm{FWHM_{excess}}$ versus $\\beta_{\\rm XTalk}$ relation using only ground-test steps where the count rate stays below the CPU limit (no PSP overflow), so the PSP and SCDP paths differ only by the cross-talk cut; if the best-fit quadratic differs from Eq. (3), CPU-loss degradation contaminated the calibration. An in-orbit counterpart would be a bright source with an independently known intrinsic line width, checking whether the uncorrected spectrum shows exactly the predicted extra broadening.","tokens_in":20604,"feed_emoji":"🔭","tokens_out":6950,"duration_ms":66647,"temperature":0.7,"pith_summary":"XRISM/Resolve is an X-ray microcalorimeter whose high energy resolution degrades when the observed source is bright. This paper exploits a ground test in which the flight hardware recorded events through the normal onboard processor while, in parallel, an unlimited software processor handled the same data stream, exposing the losses caused by the onboard CPU. It separates the degradation into three effects — CPU overflow, pulse pile-up, and untriggered electrical cross talk — and supplies phenomenological models for each, including a quadratic relation between the cross-talk contamination fraction and the excess line broadening ($0.125\\, \\mathrm{FWHM_{excess}}^2 + 0.054\\, \\mathrm{FWHM_{excess}} = \\beta_{\\rm XTalk}$). If the models hold, observers can recover true exposure times to within 4% during CPU overflow and can predict when a cross-talk cut will restore resolution, preventing spurious astrophysical conclusions such as a false increase of turbulent velocity with source brightness.","feed_headline":"Ground tests quantify how high rates widen X-ray lines","feed_subtitle":"XRISM/Resolve separates CPU loss, pile-up, and cross talk so observers can correct width and exposure.","key_machinery":"The central machinery is a set of three phenomenological models: a linear CPU-consumption model in which each event grade contributes a fixed load per quadrant, a pile-up live-time factor $\\exp(-2\\nu\\Delta t_{\\rm thres})$, and the empirical cross-talk equation relating $\\mathrm{FWHM_{excess}}$ to $\\beta_{\\rm XTalk}$. The decisive experimental mechanism is the parallel data path: a data repeater inserted between the analog electronics and the flight Pulse Shape Processor feeds the same raw stream to a software processor with essentially no CPU limits, so the true incoming rate, the contamination fraction, and the bad-time intervals can be measured directly. The cross-talk model is calibrated on Ni K-$\\alpha$ lines and then extrapolated to astrophysical spectra by scaling the excess broadening by the ratio of the count-weighted mean photon energies.","core_discovery":"On its own terms, the paper establishes that high-count-rate degradation of Resolve can be decomposed into several mechanisms, with untriggered electrical cross talk being the main cause of energy-resolution loss at very high rates. The cross-talk contribution is quantified by the empirical equation $0.125\\, \\mathrm{FWHM_{excess}}^2 + 0.054\\, \\mathrm{FWHM_{excess}} = \\beta_{\\rm XTalk}$, where $\\beta_{\\rm XTalk}$ is the fraction of events contaminated by cross-talk children from neighboring pixels and $\\mathrm{FWHM_{excess}}$ is the quadrature excess broadening obtained by comparing spectra with and without the cross-talk cut. Because $\\beta_{\\rm XTalk}$ can be computed from the count-rate map of any planned observation, the relation turns a detector artifact into a planning tool. The paper demonstrates the consequences with a simulated GX 13+1 observation: ignoring the effect inflates the measured turbulent velocity from 200 to 206 km/s, while applying the cross-talk cut restores the resolution at the cost of live time; doubling the source flux pushes one CPU past its limit, so the cut can no longer be fully applied.","pith_inferences":["If Eq. (3) is calibrated with a hard Ni line, its extrapolation to softer astrophysical spectra through an energy-scaling factor has not been validated with data; a dedicated soft-line ground calibration or an in-orbit cross-check would test that scaling.","The cross-talk model is built on uniform-illumination Ni data with counts averaged across pixels; for real point sources with steep count-rate gradients, per-pixel $\\beta_{\\rm XTalk}$ and neighbor rates should be used, and the paper's GX 13+1 map indeed shows the effect peaking in the first neighbor pixel rather than the source pixel.","The same parallel-processing methodology could be applied to other microcalorimeter instruments to separate onboard CPU losses from detector physics before launch.","When the CPU overflows, cross-talk parents inside bad-time intervals cannot be identified, so the cross-talk cut is incomplete; in that regime it may be necessary to model the residual broadening rather than rely on the cut."],"forward_implications":["Observers can correct the true exposure time of each pixel using the bad-time telemetry; in the ground test this recovers the incoming rate to within 4% even during PSP overflow, with the remaining shortfall attributed to pile-up.","The CPU-consumption model predicts whether a planned observation will overflow: for a 270 mCrab GX 13+1 with the 1/4 ND filter the load is 0.667 on the busiest quadrant, so no event loss is expected, but doubling the flux pushes one CPU to 1.21 and causes loss.","The cross-talk model lets planners choose: without the cut a 100 mCrab irradiated point source suffers about 2.9 eV of line distortion, while the cut restores resolution but leaves 83.5% live time.","Uncorrected cross talk biases astrophysical parameters: a true turbulent velocity of 200 km/s is measured as 206 km/s, and the bias grows with flux (211 and 214 km/s at double and triple flux), creating a spurious flux dependence.","Pile-up screening based on RISE TIME versus energy and SLOPE DIFFER reduces the pile-up bump above 10 keV to 32.8% of its original count, and the residual dead time is described by $\\exp(-2\\nu\\Delta t_{\\rm thres})$ with $\\Delta t_{\\rm thres}\\simeq 2$ ms for the PSP."],"supporting_citations":[{"why":"Supplies the PSP event-processing design, grade definitions, and buffer-loss mechanism on which the CPU-consumption model is built.","marker":"[15]"},{"why":"Provides the calorimeter array design, thermal conductance, and pixel wiring layout that determine the cross-talk geometry.","marker":"[12]"},{"why":"Describes the signal chain electronics, including the high-impedance cross-talk path before the JFET amplifiers.","marker":"[14]"},{"why":"Documents the ground-calibration rotating target source used for the uniform illumination test.","marker":"[22]"},{"why":"Defines the RISE TIME versus energy and SLOPE DIFFER screening criteria used to identify pile-up events.","marker":"[26]"},{"why":"Provides the orbital ephemeris and flux level of GX 13+1 used in the simulated astrophysical observation.","marker":"[28]"},{"why":"Supplies the ionized-absorption model and wind parameters used as the input spectrum for the GX 13+1 simulation.","marker":"[30]"}],"fun_headline_variants":["XRISM pins line blur at high rates on cross talk","Ground test untangles CPU, pile-up, cross talk for XRISM","Cross talk dominates XRISM's high-rate resolution loss","New correction restores XRISM spectra at high count rates","XRISM ground test turns detector artifact into planning tool"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The cross-talk model assumes that the only difference between the two processing paths used to measure excess broadening is the presence or absence of the cross-talk cut, although the PSP path also suffers CPU-loss and dead-time effects that get folded into the fitted relation.","fun_headline_variants_meta":{"raw":{"variants":["XRISM pins line blur at high rates on cross talk","Ground test untangles CPU, pile-up, cross talk for XRISM","Cross talk dominates XRISM's high-rate resolution loss","New correction restores XRISM spectra at high count rates","XRISM ground test turns detector artifact into planning tool"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1506,"prompt_tokens":984,"completion_tokens":522,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":437}},"tokens_in":600,"tokens_out":522,"duration_ms":5298,"temperature":1.0,"reasoning_tokens":437,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:06:44.907335+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the $\\mathrm{FWHM_{excess}}$ versus $\\beta_{\\rm XTalk}$ relation using only ground-test steps where the count rate stays below the CPU limit (no PSP overflow), so the PSP and SCDP paths differ only by the cross-talk cut; if the best-fit quadratic differs from Eq. (3), CPU-loss degradation contaminated the calibration. An in-orbit counterpart would be a bright source with an independently known intrinsic line width, checking whether the uncorrected spectrum shows exactly the predicted extra broadening.","supporting_citations":[{"cited_title":"Ishisaki , S","cited_arxiv_id":null,"evidence_quote":"Supplies the PSP event-processing design, grade definitions, and buffer-loss mechanism on which the CPU-consumption model is built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the calorimeter array design, thermal conductance, and pixel wiring layout that determine the cross-talk geometry."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the signal chain electronics, including the high-impedance cross-talk path before the JFET amplifiers."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the ground-calibration rotating target source used for the uniform illumination test."},{"cited_title":"Mochizuki, M","cited_arxiv_id":null,"evidence_quote":"Defines the RISE TIME versus energy and SLOPE DIFFER screening criteria used to identify pile-up events."},{"cited_title":"Iaria , T","cited_arxiv_id":null,"evidence_quote":"Provides the orbital ephemeris and flux level of GX 13+1 used in the simulated astrophysical observation."},{"cited_title":"Tomaru , C","cited_arxiv_id":null,"evidence_quote":"Supplies the ionized-absorption model and wind parameters used as the input spectrum for the GX 13+1 simulation."}],"review_version":1}