{"id":"be429cec-2c2f-4050-9185-f5c1f678a1ac","arxiv_id":"1907.07648","paper_version":1,"verdict":"UNVERDICTED","confidence":"UNKNOWN","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Pith review generated a malformed one-line summary.","lead":"Pith review generated a malformed lead; see the summary below.","discovery_kind":"unclear","skeptic_critique":{"model":"claude-opus-4-7","headline":"Deconvolution-free imaging at the claimed dynamic range (10^6–10^7) is the architectural keystone, and the phase/beam-stability budget needed to achieve it across 15 km baselines at L-band is far tighter than the 1° figure quoted in §3.","rationale":"This is an Astro2020 APC (project) white paper, not a results paper, so \"correctness\" maps onto whether the proposed instrument can plausibly meet its stated specs. The reader's payload was malformed and gave no usable assessment, so I am not building on it.\n\nReading the document in good faith, the science case (FRBs, PTA, GW follow-up, sky surveys) is straightforward and well motivated; the cost and schedule have ordinary project-management uncertainties but are not load-bearing for the scientific claim. The single distinctive technical assertion that makes DSA-2000 different from \"another L-band array\" is the deconvolution-free, image-only data architecture enabled by a dense optimized configuration. Every downstream claim — real-time imaging, no public visibility archive, the modest digital-backend cost, the operations FTE count, and the \"first true radio camera\" framing — follows from that.\n\nThe paper's own phase-error budget (§3) is consistent at D = 10^5 but their target dynamic range is stated as 10^6–10^7. The corresponding tolerances on residual antenna phase, direction-dependent beam errors, and ionospheric reconstruction at L-band over 15 km baselines are at the 0.01–0.1° level per antenna and are not demonstrated. They are flagged as items to be prototyped, which is appropriate for an APC white paper, but it means the central architectural claim is a contingent one. The Fig. 7 demonstration uses a 10-second snapshot on the SKA Data Challenge field (not the real source-count distribution) and reaches dynamic range 5×10^5 after image-plane deconvolution — short of the 10^6–10^7 target, and on a more benign sky than the actual survey will see.\n\nA CONDITIONAL adjustment from UNVERDICTED is warranted: the proposal is coherent and the science compelling, but the load-bearing technical claim has a quantitative gap between the cited phase-budget calculation and the dynamic range required for the \"no-visibility, image-only\" data strategy. The concrete test above — a realistic end-to-end simulation including ionosphere and the actual L-band source distribution — would settle whether the architecture survives contact with the real sky or whether a fallback (visibility retention, conventional CLEAN) must be designed in, with attendant cost/data-volume implications. None of this should be read as ad hominem; the team has a credible hardware track record from DSA-10/110.","tokens_in":3099,"tokens_out":3291,"duration_ms":58853,"concrete_test":"Run an end-to-end simulation of a single 15-min DSA-2000 pointing on a realistic L-band sky drawn from VLASS/NVSS source counts (not just the SKA Data Challenge cut-out used in Fig. 7), including: (a) Kolmogorov ionospheric phase screens with TEC variance typical of a mid-latitude site at 1.4 GHz on baselines up to 15 km, (b) per-dish primary-beam errors at the level expected from the proposed 5 m dish surface and pointing spec, and (c) all sources >10 mJy in the 10.6 deg² field, not just the global top 40. Apply only direction-dependent calibration plus image-plane deconvolution (no visibility-domain CLEAN). If the residual rms exceeds 2 μJy/beam in single epochs, or if stacks of 16 epochs fail to reach 500 nJy/beam in fields containing any source brighter than ~1 Jy, the deconvolution-free architecture (and the data-archive plan built on it) needs revision.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The white paper's distinctive design choice — discarding visibilities, doing real-time GPU gridding/imaging, and serving images rather than raw data — rests on the claim that the dense 2000-element configuration produces sidelobes low enough that \"image-plane deconvolution\" suffices for all but the ~40 brightest sources. The cost model, archive plan, and \"true radio camera\" framing all flow from this.\n\nThe phase-error requirement quoted in §3 is φ ≈ N/(√2 D) ≈ 1° for D > 10^5 with N = 2000. That is internally consistent for D = 10^5, but the array's quoted target dynamic range is 10^6–10^7 (§2.2 and Fig. 7 caption). Plugging in D = 10^6 gives φ ≈ 0.08°; D = 10^7 gives ≈ 0.008° per antenna. Those tolerances must be met simultaneously by (i) residual antenna gain/phase calibration, (ii) direction-dependent primary-beam errors across a 10.6 deg² field, (iii) pointing jitter on motorized 5 m dishes, and (iv) ionospheric phase screen reconstruction across ~15 km baselines at 0.7–2 GHz, where TEC fluctuations routinely exceed several degrees of phase. The paper lists each of these as a \"technology driver\" but does not show that they can be jointly achieved at the 0.01°-class level.\n\nA related concern: the \"only 40 sources brighter than 10 Jy\" framing understates the sidelobe-confusion problem. At 1.4 GHz the source count gives thousands of sources >100 mJy in any 10 deg² field. With a 2 μJy/beam single-epoch noise floor and a 500 nJy/beam stack, even 10^4-level residual sidelobes from 100 mJy sources are at the noise floor; this dominates the stack in any field with even one moderately bright source, not just the 40 strongest in the sky.\n\nIf image-plane deconvolution alone cannot reach 10^6–10^7 over the survey, the project still produces a strong instrument, but the \"no visibility archive / real-time imaging\" architecture — and the cost/operations envelope it implies — would need revision.","agreement_with_reader":"partial"},"referee_report":{"model":"claude-opus-4-7","summary":"The manuscript is an Astro2020 APC white paper presenting the DSA-2000 concept: a 2000-element array of 5 m dishes operating at 0.7–2 GHz, designed as a dedicated radio survey \"camera\" delivering science-ready images rather than visibilities. Headline deliverables are a cadenced all-sky survey (3π sr to 500 nJy/beam stacked rms over 16 epochs), deep drilling fields at ~100 nJy/beam, NANOGrav pulsar timing, GW follow-up, and ~10^4 FRB localizations/year. The architectural keystone is real-time GPU-based gridding and imaging in the absence of visibility-domain deconvolution, enabled by a dense configuration with low sidelobes in 15-min tracks plus image-plane deconvolution for a small number of bright sources. The paper builds on the operational DSA-10 and the funded DSA-110 prototypes, and provides a $96 M construction cost estimate and a schedule targeting first light in 2026.","tokens_in":3770,"tokens_out":3521,"duration_ms":62753,"significance":"If the deconvolution-free imaging architecture works at the claimed dynamic range and the SEFD/efficiency targets are met, the science case is genuinely transformational: an order-of-magnitude survey-speed gain over SKA1-mid at L-band, a billion-source cadenced sky map complementary to LSST/SPHEREx/eROSITA, the principal NANOGrav timing engine, and a unique FRB localization machine. The proposal is anchored by demonstrated hardware heritage (DSA-10 on sky, DSA-110 funded with PDR passed, custom 6 K LNA prototypes, FRB localization to host galaxy in Ravi et al. 2019), an explicit cost breakdown with contingencies, a concrete schedule, and a falsifiable design specification (Table 1). These are real strengths for an APC-format document. The chief risk is technical, not scientific: the entire data-reduction philosophy and cost model rest on a calibration/imaging budget that is asserted rather than demonstrated.","major_comments":[{"comment":"The phase-error budget is internally inconsistent with the quoted dynamic-range target. §3 derives φ ≈ N/(√2 D) ≈ 1° per antenna for D > 10^5, but §2.2 and the Fig. 7 caption claim D > 10^6 (likely > 10^7 in final configuration). Substituting these targets gives φ ≈ 0.08° (D=10^6) and ≈ 0.008° (D=10^7) per antenna — to be met jointly by residual gain/phase calibration, direction-dependent primary-beam errors over 10.6 deg², pointing jitter on motorized 5 m dishes, and ionospheric phase reconstruction across 15 km baselines at 0.7–2 GHz, where TEC-driven phase fluctuations routinely exceed several degrees. The paper should either (a) reconcile the quoted φ with the operative D, (b) state which D the cost/data-product model actually requires, or (c) present a budget showing how the four error contributions add to meet the stricter tolerance. As written this is the load-bearing assumption f","section":"§3 (technology drivers) and §2.2"},{"comment":"The 'only 40 sources >10 Jy' framing understates the residual-sidelobe confusion problem. Standard 1.4 GHz source counts give of order 10^3–10^4 sources >100 mJy in any 10.6 deg² field; with a single-epoch 2 μJy/beam noise floor and a 500 nJy/beam stack target, even 10^-4-level residual sidelobes from the bright population can dominate the noise budget across much of the survey area. Fig. 7 demonstrates image-plane deconvolution on a single ~320 mJy source from a 10 s snapshot. The paper should quantify (e.g., via simulation with the realistic source count, not just one bright source) the fraction of pixels in a typical 15-min image whose noise is dominated by the aggregated sidelobe response of the full source population, and demonstrate that image-plane deconvolution remains tractable in that regime.","section":"§2.2, Fig. 7"},{"comment":"The claim that calibration/imaging is 'completely deterministic' once non-linear deconvolution is removed glosses over direction-dependent calibration, which §3 then identifies as requiring a GPU realization of the RIME including ionospheric phase-screen reconstruction. Pointing self-calibration (Bhatnagar & Cornwell 2017) is iterative and non-linear in the gain solutions. The paper should clarify whether the real-time pipeline assumes pre-existing solutions (from a prior pass or external catalog) or whether on-line iterative DD calibration is required, and quote the GPU compute load for the latter case — this is relevant to the $25 M digital backend and 50 kW GPU power line items in §5.","section":"§2.2, §3"},{"comment":"The SEFD = 2.5 Jy figure depends on Tsys = 25 K and 70% aperture efficiency over 0.7–2 GHz, with an LNA average of 6 K. §3 acknowledges the 6 K target has only been demonstrated within the DSA-110 sub-band (1.28–1.53 GHz). Because survey speed and all sensitivity-derived science cases scale as SEFD^-2, the manuscript should either state how the budget degrades if Tsys is realized at, say, 35 K averaged over the full octave (a plausible outcome for an ambient-temperature wideband QRFH+LNA), or carry that as an explicit risk in the cost/scope discussion.","section":"§2.2 / Table 1"},{"comment":"Discarding visibilities in favor of images is presented as a feature, but it forecloses post-hoc reprocessing — re-calibration with improved sky models, alternative weighting schemes, peeling of newly identified bright sources, and the kind of archival re-imaging that drove much of NVSS/FIRST's long-term scientific value. The paper should either describe a visibility-retention policy (even short-baseline-averaged) or argue why the science case is robust to this loss. As stated, the 70 PB data-product budget contains no visibility tier.","section":"§2.2, last paragraph; §5 (data management)"}],"minor_comments":[{"comment":"'> 1 unique billion radio sources' — typo; should read '> 1 billion unique radio sources' as in the body.","section":"Abstract"},{"comment":"'Maglicchetti et al. 2018' in text vs. 'Magliocchetti, M.' in the reference list — fix spelling.","section":"§1"},{"comment":"Specify whether the SEFD comparison uses on-axis or beam-averaged sensitivity, and whether plotted survey speeds include realistic time overheads (the 20% in Table 1) or are theoretical maxima.","section":"Fig. 1 caption"},{"comment":"'2µJy/beam' single-epoch rms over 6000 15-min pointings on 3π sr should be made consistent with Table 1's 'All-Sky Survey (per epoch) 30,000 deg² @ 2 μJy/bm'; 3π sr is ~30,940 deg², which is fine, but the pointing count and assumed FoV (10.6 deg²) imply ~63,600 deg² of footprint — readers will want a one-line explanation of the overlap/mosaicking factor.","section":"§1, Key Objective 1"},{"comment":"Define the symbols in φ ≈ N/(√2 D) (especially what 'N' represents — number of antennas, here 2000 — and whether D is per-pixel or per-image dynamic range), and cite the original derivation.","section":"§3, phase-error formula"},{"comment":"State explicitly which dish/beam model and ionospheric assumptions were used for panel D/E/F simulations, since these set the credibility of the deconvolution-free demonstration.","section":"Fig. 7 caption"},{"comment":"Operations cost is quoted in $ M/yr but the lifetime operations envelope (e.g., 5-yr prime phase × $6.6 M ≈ $33 M) is not summed; worth stating for comparison with the $96 M construction cost.","section":"§5, Table 2"},{"comment":"The 'NSF Mid-scale Research Infrastructure-2 program' cap is stated as $70 M but Table 2 totals $96.25 M; the sentence noting partner contributions for the gap is good but could quantify the expected partner share.","section":"§4"},{"comment":"Several references are arXiv-only (Hotokezaka 2019, Mezcua 2019, Ravi 2019, Kocz 2019); update to journal versions where available.","section":"References"}],"recommendation":"minor_revision","confidential_remarks":"This is an Astro2020 APC white paper, i.e., a project concept document for the decadal process and an MSRI-2 proposal — not a standard research article. The appropriate referee bar is whether the concept is well-posed, the heritage credible, and the risks identified. By that standard the document is solid. The substantive technical concern that should be flagged to the authors (and is appropriate to surface in the decadal review) is the inconsistency between the φ ≈ 1° phase budget written in §3 and the D > 10^6–10^7 dynamic range claimed in §2.2/Fig. 7 — the entire deconvolution-free architecture, and hence the cost model, hinges on which of those numbers is binding. I would not hold up publication of the white paper for this, but it should be addressed in the body before the document is used as a proposal anchor."},"author_rebuttal":{"model":"claude-opus-4-7","summary":"We thank the referee for a careful and constructive report. The five major comments cluster around a single legitimate concern: the manuscript asserts the calibration/imaging budget that underpins the deconvolution-free architecture rather than demonstrating it, and in places the text conflates distinct quantities (PSF sidelobe level vs. calibration-limited dynamic range; design targets vs. demonstrated hardware performance). We accept this critique and will revise accordingly. In particular, we will (i) rewrite the dynamic-range discussion in §2.2/§3 to separate uv-plane PSF sidelobe levels from image-plane calibration dynamic range and add an explicit per-antenna phase-error budget across the four contributing terms; (ii) replace the single-bright-source Fig. 7 demonstration with an end-to-end simulation using a realistic 1.4 GHz source count populated over the full 10.6 deg² primary beam, and quantify aggregated residual-sidelobe confusion; (iii) soften 'completely deterministic' to 'bounded and convergent' and document the on-line DD calibration iteration count assumed by the $25 M / 50 kW digital-backend line; (iv) add an explicit Tsys risk row showing how survey speed and the science tiers degrade if the wideband LNA realizes ~35 K rather than 25 K averaged across the octave; and (v) add an explicit visibility-retention policy to §2.2/§5, since the submitted draft did not articulate one. Several of these items — most importantly the forward-modeling validation of the ph","responses":[{"response":"The referee is correct that the text conflates two different dynamic-range numbers. The relevant figure for the science data products is the per-15-min-image dynamic range required so that residual sidelobes from the brightest in-beam source do not dominate the 2 μJy/beam single-epoch noise. For typical fields this is D ≈ a few × 10^5, which is what the φ ≈ 1°/antenna budget in §3 was derived for. The D > 10^6–10^7 numbers cited in §2.2/Fig. 7 refer to the *PSF sidelobe level* of the array configuration itself (a geometric property of the uv coverage in a 15-min track), not to the calibration-limited image dynamic range. We will rewrite §2.2 and the Fig. 7 caption to separate these two quantities cleanly, and add an explicit error budget table partitioning the 1° per-antenna allocation among: (i) residual direction-independent gain/phase, (ii) DD primary-beam errors over the 10.6 deg² field, (iii) pointing jitter (with the dish stiffness/thermal spec set by this allocation), and (iv) residual ionospheric phase after screen reconstruction across 15 km baselines. We agree this is the load-bearing assumption of the architecture and that the current text does not adequately defend it; quantitative substantiation will be a deliverable of the design phase prototype and forward-modeling work scheduled in 2020–2021.","revision_made":"yes","referee_comment":"Phase-error budget is internally inconsistent with the dynamic-range target: §3's φ ≈ N/(√2 D) ≈ 1° is for D > 10^5, but §2.2/Fig. 7 quote D > 10^6 (and possibly 10^7). The stricter tolerance must be met jointly by gain calibration, DD primary-beam errors, pointing jitter, and ionospheric phase, none of which are budgeted."},{"response":"Agreed. The Fig. 7 demonstration was illustrative of the image-plane deconvolution algorithm on a single bright source and was not intended as a confusion-budget proof. The quoted '40 sources' figure refers only to the population for which we would consider visibility-domain model subtraction prior to gridding; image-plane deconvolution is intended to handle the much larger ~mJy–Jy population. We will replace the single-source Fig. 7 demonstration (or supplement it) with an end-to-end simulation drawn from a realistic 1.4 GHz source-count realization (e.g., Wilman et al. 2008 / SKA Data Challenge field) populated across the full 10.6 deg² primary beam, and report (a) the fraction of pixels in a typical 15-min image whose noise budget is dominated by aggregated residual sidelobes, and (b) the per-image image-plane deconvolution compute cost in that regime. This simulation is part of the array forward-modeling work item already scheduled in §4. We acknowledge the current manuscript does not establish this, and we will not claim the architecture is validated against aggregated-sidelobe confusion until the simulation is in hand.","revision_made":"yes","referee_comment":"The '40 sources > 10 Jy' framing understates residual-sidelobe confusion: 10^3–10^4 sources > 100 mJy lie in each 10.6 deg² field, and at the 500 nJy/beam stacked target even 10^-4-level sidelobes aggregate. Fig. 7 only demonstrates one 320 mJy source. Quantify via realistic source-count simulation."},{"response":"The referee is right that 'completely deterministic' is too strong. What we mean is that *blind, visibility-domain CLEAN-style deconvolution* — the dominant computational and algorithmic uncertainty in conventional radio imaging — is not in the real-time path. DD calibration (gain screens, pointing self-cal, ionospheric phase-screen reconstruction) is iterative, but operates on a small, bounded number of parameters with strong priors from external catalogs (NVSS, VLASS) and pre-measured antenna beam patterns, and is expected to converge in a fixed, scheduleable number of iterations per 15-min block. We will rewrite the offending sentence in §2.2 to say 'bounded and convergent' rather than 'completely deterministic.' For the compute budget: the $25 M / 50 kW GPU line in §5 was sized assuming on-line iterative DD calibration via a GPU-RIME implementation, with the iteration count set by the worst-case ionosphere; we will state this assumption explicitly and add the per-15-min iteration budget to §3. A more detailed compute-load justification, including comparison to existing GPU gridding benchmarks (van der Tol et al. 2018), will appear in the PDR documentation and will be summarized in the next revision.","revision_made":"yes","referee_comment":"The 'completely deterministic' framing glosses over DD calibration, which is iterative (e.g., pointing self-cal, Bhatnagar & Cornwell 2017). Clarify whether on-line iterative DD calibration is required and quote GPU compute load — relevant to the $25 M digital backend and 50 kW GPU line."},{"response":"Accepted. The 25 K system temperature is a design target predicated on extending the demonstrated 6 K LNA performance across the full octave with a wideband QRFH feed, and the manuscript should not present this target as if it were already achieved. We will add an explicit risk/sensitivity statement to §2.2 and §3 quantifying the degradation: at Tsys = 35 K (SEFD ≈ 3.5 Jy) survey speed degrades by a factor (35/25)² ≈ 2; the all-sky stacked rms rises from 500 to ~700 nJy/beam, and the per-epoch rms from 2 to ~2.8 μJy/beam. The science cases anchored on the stacked map (billion-source catalog, AGN/SF census to z ≳ 1) remain qualitatively intact at this degraded sensitivity, while the deep-drilling and GW-follow-up tiers would scale proportionally in integration time. We will add this as an explicit risk row and treat the wideband LNA/feed development as the principal sensitivity risk in the cost/scope discussion.","revision_made":"yes","referee_comment":"SEFD = 2.5 Jy assumes Tsys = 25 K and 70% efficiency across 0.7–2 GHz, but the 6 K LNA has only been demonstrated in the DSA-110 sub-band. Survey speed scales as SEFD^-2 — state how the budget degrades at, e.g., Tsys ≈ 35 K, or carry the risk explicitly."},{"response":"This is a fair criticism and one we have wrestled with internally. The full visibility data rate of the DSA-2000 is prohibitive for indefinite retention at the scale of the project's data-management budget, which is why the current 70 PB envelope is image- and cube-dominated. However, we agree the manuscript should articulate an explicit visibility policy rather than leave it as an omission. Options under active consideration, all of which preserve substantial reprocessing capability, include: (i) retention of baseline-averaged and time-averaged 'core-only' (short-baseline) visibilities for the full survey, supporting re-weighting and large-scale re-imaging; (ii) on-demand retention of full-resolution visibilities for limited time/sky windows (e.g., GW triggers, FRB fields, deep-drilling fields, transient candidates) where post-hoc peeling and re-calibration are most scientifically valuable; (iii) retention of calibration solutions and residual visibilities (post sky-model subtraction) at reduced volume. We will add a paragraph to §2.2 / §5 describing this policy and an associated additional storage tier in the data-management cost line. We acknowledge this was a gap in the submitted manuscript.","revision_made":"yes","referee_comment":"Discarding visibilities forecloses post-hoc reprocessing (re-calibration with improved sky models, alternative weighting, peeling of new bright sources, archival re-imaging) — much of NVSS/FIRST's long-term value came from this. Describe a visibility-retention policy or argue robustness to its loss. The 70 PB budget contains no visibility tier."}],"tokens_in":4214,"tokens_out":3641,"duration_ms":58839,"standing_objections":["End-to-end empirical validation of the joint phase-error budget (gain + DD beam + pointing + ionosphere) against the operative image dynamic range cannot be provided within this APC white paper; it is a design-phase deliverable. We will state this explicitly rather than claim the budget is closed."]},"desk_editor":{"model":"claude-opus-4-7","letter":"This is an Astro2020 APC white paper for the DSA-2000 — 2000 5-m dishes, 0.7–2 GHz, designed as a survey camera that outputs images rather than visibilities. Read it as a project pitch, not a refereed result.\n\nWhat's genuinely new and worth taking seriously: the design philosophy. A dense, 15-km, 2000-element configuration optimized to give low sidelobes in 15-min tracks, with real-time GPU gridding/imaging and no visibility archive. If it works, it changes what a radio survey instrument looks like and how its data are used. The team has actual hardware heritage (DSA-10 on sky, DSA-110 funded), the cost model scales from real per-antenna unit costs rather than wishful numbers, and the science case — cadenced all-sky to 500 nJy/beam stacked, NANOGrav timing, FRB localization at ~10^4/yr, GW follow-up — is well-motivated and largely independent of any single subsystem.\n\nThe soft spot the stress-test note flags is real and is the right one to flag. The φ ≈ N/√(2D) figure of \"≈1°\" in §3 is computed at D=10^5, but the paper elsewhere (§2.2, Fig. 7) needs D > 10^6, plausibly 10^7. That pushes the per-antenna phase budget to ~0.08° or ~0.008°, and that budget has to absorb residual gain/phase calibration, direction-dependent primary-beam errors over a 10.6 deg² field, pointing jitter, and ionospheric phase across 15 km at L-band simultaneously. The paper lists these as \"technology drivers\" without showing they close jointly. The \"only 40 sources >10 Jy\" framing also undersells the problem: at 1.4 GHz, hundreds to thousands of >100 mJy sources will sit in any 10 deg² field, and 10^-4–10^-5 residual sidelobes off those will set the stack floor long before you reach 500 nJy/beam. If image-plane deconvolution alone can't deliver, you still get a strong instrument but the no-visibility-archive architecture, and the cost envelope built on it, would have to give.\n\nNone of this is fatal to the concept. It is the question a serious referee should press the team on, ideally with an end-to-end simulation that includes ionosphere and beam errors at survey-realistic source densities.\n\nWorth reading if you care about radio surveys or PTA/FRB infrastructure. I'd bring it to a reading group as a design-philosophy discussion. As a white paper it absolutely warrants serious engagement.","headline":"A serious project pitch with a real new architectural idea, but the dynamic-range arithmetic in §3 doesn't match the >10^6 dynamic range the rest of the paper depends on.","tokens_in":4636,"tokens_out":2107,"would_cite":true,"duration_ms":35005,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"claude-opus-4-7","evidence":[{"relation":"unclear","rs_module":"N/A — engineering specifications","rs_theorem":null,"paper_passage":"2000× 5 m dishes ... system temperature of 25 K, with aperture efficiency of 70% (SEFD 2.5 Jy)"},{"relation":"unclear","rs_module":"Foundation/Inequalities (general AM-GM-style bound, but used here for instrument calibration, not RS cost forcing)","rs_theorem":null,"paper_passage":"the average residual phase error per antenna, φ, required for an image of dynamic range D is φ ≈ N/√(2D)"}],"headline":"Radio survey instrument white paper; engineering/observational scope with no contact with RS forcing chain.","alignment":"orthogonal","rationale":"This is an APC (project) white paper proposing the DSA-2000 radio survey camera: 2000×5m dishes, 0.7–2 GHz, optimized for survey speed, FRB localization, pulsar timing arrays, and GW follow-up. The content is entirely about telescope architecture, calibration tolerances, cost estimates, and survey science cases (AGN census, star formation, HI cubes, FRBs as cosmological probes). There is no cost-functional reasoning, no ratio-symmetric structure, no φ-ladder, no 8-tick periodicity, no parameter-free derivation of constants — the paper has parameters everywhere (system temperature, aperture efficiency, dish count, dynamic range targets, etc.) and these are engineering choices, not derived. RS has no opinion on radio telescope design or survey sensitivity. The skeptic's concern about phase-stability budgets is a legitimate engineering critique but is also orthogonal to RS. None of the RS modules in the corpus (Cost.Jcost, PhiForcing, DimensionForcing, EightTick, etc.) bear on the paper's claims.","tokens_in":16075,"confidence":"high","tokens_out":772,"duration_ms":17656,"cache_read_input_tokens":415308,"cache_creation_input_tokens":0},"lean_confirmation":{"model":"claude-opus-4-7","status":"out_of_scope","citations":[],"rationale":"This is a ground-based instrument proposal/white paper. Its load-bearing claims are empirical (sensitivity calculations from radiometer equation applied to specific hardware, FRB rate estimates, cost projections). There is no mathematical identity or structural theorem at the core that shape-of-logic could machine-check. The reader payload is also marked malformed/unverdicted.","tokens_in":15159,"confidence":"high","tokens_out":596,"duration_ms":11987,"inferential_bridge":"The paper is an instrument white paper proposing the DSA-2000 radio survey telescope. Its central claims are engineering/observational: predicted sensitivity (500 nJy/beam), survey speed, FRB detection rates, cost estimates, and schedule. None of these rest on a mathematical/structural identity that Lean could prove. Lean would have to prove an empirical instrument performance claim, which is categorically out of scope for a shape-of-logic formal library.","load_bearing_premise":"The DSA-2000 array, as specified (2000×5m dishes, 25K system temperature, 70% aperture efficiency, 0.7-2 GHz band), will achieve 500 nJy/beam rms noise in the combined all-sky survey and detect ~10^9 radio sources and ~10^4 FRBs/year. This is an empirical engineering and astronomical performance prediction.","cache_read_input_tokens":415178,"cache_creation_input_tokens":0},"pith_extraction":{"msc":[],"pacs":[],"model":"claude-opus-4-7","headline":"A proposed 2000-dish radio array would survey the sky ten times faster than the next flagship facility by outputting calibrated images, not visibilities.","keywords":["radio interferometry","all-sky survey","fast radio bursts","pulsar timing array","gravitational-wave follow-up","GPU imaging pipeline","survey telescope","multi-messenger astronomy"],"falsifier":"Build the planned six-antenna prototype and full LNA-plus-feed chain and verify on sky that a 15-minute snapshot achieves a synthesized-beam dynamic range above 10^5 with residual per-antenna phase errors near 1 degree and system temperature near 25 K across 0.7–2 GHz. If the achieved dynamic range, sidelobe level, or system temperature falls materially short, the deconvolution-free imaging premise — and with it the survey-speed advantage — fails.","tokens_in":151,"feed_emoji":"📡","tokens_out":2198,"duration_ms":50227,"temperature":0.7,"pith_summary":"The paper lays out the case for a radio survey instrument built around 2000 small, cheap, fully steerable 5 m dishes spread across roughly 15 km, covering 0.7–2 GHz. Because the array is dense and the configuration is optimized for low sidelobes in a 15-minute track, the authors argue that they can skip visibility-domain deconvolution entirely and run flagging, calibration, gridding and imaging on a real-time GPU backend, delivering science-ready images directly to users. The headline survey would map the sky north of −30° every four months for five years, stacking to a 500 nJy/beam, roughly 10^9-source full-Stokes map, while time is also carved out for daily deep-drilling fields, weekly-to-monthly pulsar timing of millisecond pulsars, hour-per-day gravitational-wave afterglow follow-up, and a commensal fast-radio-burst search projected to localize about 10^4 bursts per year. The argument is essentially that survey speed at this band is bought more cheaply by multiplying small dishes with good receivers than by building fewer large ones, and that doing the imaging on the fly is what makes such a numerous array operationally tractable.","feed_headline":"A 2000-dish radio camera built to skip deconvolution","feed_subtitle":"Dense small-dish array aims for billion-source all-sky maps and about 10,000 fast radio bursts per year.","key_machinery":"A dense 2000×5 m aperture whose configuration is engineered so that a single 15-minute snapshot already has sidelobe levels low enough (dynamic range >10^5, projected >10^7 in the final layout) that imaging becomes a deterministic, GPU-friendly pipeline: flagging, direction-dependent calibration via a GPU realization of the radio interferometric measurement equation, gridding and imaging, with cheap image-plane deconvolution handling rare bright-source residuals.","core_discovery":"The authors propose a 2000-element array of low-cost 5 m dishes whose value comes not from large collecting area per dish but from being a dense, deliberately configured aperture that delivers point-source sensitivity comparable to a flagship next-generation radio facility while surveying the sky ten times faster. The configuration is chosen so that a 15-minute snapshot already has low enough sidelobes (dynamic range above 10^5) that the standard, expensive step of visibility-domain deconvolution can be skipped, and calibration plus imaging can run deterministically in real time on a GPU pipeline. The output is therefore a science-ready image stream — a true \"radio camera\" — rather than raw","pith_inferences":["The economic argument — that many small dishes plus modern receivers and GPUs beat fewer large dishes for survey work in this band — would, if borne out, reshape how mid-frequency radio facilities are costed and proposed.","Skipping blind deconvolution makes the data products vastly easier for non-radio astronomers to use, which is itself a sociological shift: the user base would resemble that of optical surveys rather than that of traditional interferometry.","The 1-degree residual phase budget at dynamic range 10^5 is tight given ionospheric structure across a 15 km aperture; achievability of that budget will likely be the practical bottleneck rather than receiver hardware.","The dense-core layout that enables low snapshot sidelobes also limits angular resolution to 3.5 arcsec, so this instrument is structurally complementary to, not competitive with, longer-baseline arrays for source morphology."],"forward_implications":["A combined 500 nJy/beam, full-Stokes sky map north of −30° containing about a billion sources becomes the radio counterpart to optical, infrared and X-ray all-sky surveys.","Sixteen four-month epochs make the radio sky genuinely synoptic, opening a slow-transient volume roughly 1700 times larger than the current VLA sky survey.","Roughly 10^4 fast radio bursts localized per year would convert FRBs from a curiosity into a usable cosmological probe of the diffuse baryon census.","Hour-per-day cadence permits radio afterglow discovery for gravitational-wave compact-object mergers without requiring an electromagnetic precursor at other wavelengths.","Real-time GPU imaging that bypasses visibility-domain deconvolution would shift the operational model of radio interferometers from delivering visibilities to delivering images."],"fun_headline_variants":[],"cache_read_input_tokens":15536,"weakest_assumption_plain":"The whole survey-speed advantage rests on the bet that a carefully optimized dense layout plus on-the-fly calibration really does keep image sidelobes low enough that the array can skip the expensive visibility-domain deconvolution step that radio interferometry has historically required.","fun_headline_variants_meta":null,"cache_creation_input_tokens":6245},"created_at":"2026-05-08T23:10:43.512969+00:00","model_set":{"reader":"claude-opus-4-7"},"falsifier":"Build the planned six-antenna prototype and full LNA-plus-feed chain and verify on sky that a 15-minute snapshot achieves a synthesized-beam dynamic range above 10^5 with residual per-antenna phase errors near 1 degree and system temperature near 25 K across 0.7–2 GHz. If the achieved dynamic range, sidelobe level, or system temperature falls materially short, the deconvolution-free imaging premise — and with it the survey-speed advantage — fails.","supporting_citations":[],"review_version":1}