{"id":"e122169f-584b-4799-b21a-9968ceb042a4","arxiv_id":"2412.00831","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A luminance formula for retinal scan AR glasses that depends on pupil size is derived, and a 20-person study finds perceived resolution around 7 cycles per degree.","lead":"This paper derives a formula for the perceived brightness of retinal-scan augmented reality glasses and tests how finely participants could resolve laser-projected line patterns. The result is a human-centered way to compare these glasses with ordinary screens and a measured resolution limit around 7 cycles per degree.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pupil clipping can break the claimed inverse-pupil-area scaling; the paper's own small-eyebox limitation makes the 'beneficial' bright-light luminance increase conditional on perfect, sustained alignment.","rationale":"I read the paper in good faith. The photometric derivation (Eqs. 3–14) is internally consistent under the stated assumption that the entire RS beam passes through the pupil; the algebra cancels f_e and produces a dimensionally sound equivalent luminance. The perceived-resolution staircase is plausible, though the comparison to the IEC 30% cutoff is not fully explained. The load-bearing issue I find is the interaction between the proposed pupil-dependent luminance model and the hardware's own stated small-eyebox limitation. The paper claims the pupil-diameter dependence is 'beneficial' because it assimilates virtual image luminance to ambient luminance in bright environments. But this benefit presumes the total flux in the eye is insensitive to pupil size. In bright environments, pupils are smallest and the eyebox tolerance is tightest; the paper's Section 4 limitation means that any loss of alignment or beam clipping changes the flux, potentially wiping out the predicted fivefold increase. This is not an internal contradiction — the derivation is correct for the ideal case — but it means the central practical claim is overgeneralized without a tolerance analysis. The reader identified this exact assumption as weakest; I agree. Because the paper already hinges on this condition and the reader's verdict is CONDITIONAL, my stress test does not change the verdict. I would still accept the paper if revised to include a flux-versus-pupil/offset measurement or an explicit statement that the 'beneficial' behavior holds only for perfectly aligned, full-beam conditions and to quantify the tolerance.","tokens_in":10875,"tokens_out":12379,"duration_ms":114795,"concrete_test":"Mount the BML500P (or an equivalent RS engine) on an optical bench with an artificial eye whose entrance aperture simulates pupil diameters from 2.0 to 6.0 mm and whose lateral position can be offset in 0.1 mm steps relative to the scanning beam. At each (diameter, offset) pair, integrate the transmitted luminous flux over a full raster frame (e.g., with an integrating sphere behind the aperture) and record the ratio to the on-axis, large-pupil value. If the flux remains within 5% of the unclipped value for all offsets up to ±1 mm at 2.5 mm pupil diameter, the full-beam assumption is robust. If flux drops by more than, say, 20% for offsets in the range of normal microsaccades and display slip (≈0.5–1 mm), the claimed inverse-pupil-area benefit is not robust in practice and the paper should be revised to state the alignment tolerance as a condition.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central luminance claim, Eq. (12), L_RS = K_m ∑ V Φ / (4 A_P tan(α/2) tan(β/2)), is derived by equating retinal illuminances and assumes the entire RS beam flux ΦV_Scan enters the eye independently of pupil size (Fig. 2, Section 2.1.1). The paper's headline insight — that pupil constriction in bright environments raises equivalent luminance and thus maintains contrast — depends entirely on this flux being pupil-independent. However, Section 4 explicitly lists as a limitation the 'small exit pupil of the single eyebox prototype' and states that 'the RS display is only visible if the narrow light beam fully passes through the pupil.' Under real usage, head/eye movements and pupil constriction in bright light (to 2–4 mm) reduce the alignment tolerance, so partial beam clipping becomes likely. If clipping occurs, ΦV_Scan becomes a function of A_P and of lateral offset, the inverse-scaling in Eq. (12) breaks down, and the claimed benefit can be reduced or reversed. The paper provides no tolerance analysis, no eye-tracking data, and no bench measurement of flux versus pupil position; the psychophysical study uses a chinrest and a 5-minute fitting that artificially sustain the full-beam condition. Therefore the practical central claim is not established for authentic wearable use, only for the ideal static-alignment regime.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper addresses brightness and resolution metrology for retinal scan (RS) augmented-reality glasses. The authors derive an equivalent luminance L_RS for an RS display by equating its retinal illuminance with that of a Lambertian display, obtaining L_RS = K_m Σ V(λ)Φ / (4 A_P tan(α/2)tan(β/2)) (Eq. 12), where A_P is the pupil area. They then use a Stanley–Davies pupil model to express L_RS and an ambient contrast ratio ACR_RS as functions of ambient luminance, and argue that pupil constriction in bright environments increases the equivalent luminance—a property they call beneficial for AR. In the experimental part, 20 participants performed a 3-down-1-up staircase discrimination task on a holographic RS glasses prototype, yielding perceived resolution thresholds of about 7 cyc/deg across different line orientations, colors, brightnesses, and backgrounds. The study also reports an IEC 61947-2 optical resolution of 3.4 cyc/deg for green and discusses why the perceived threshold exceeds it.","tokens_in":11109,"tokens_out":17590,"duration_ms":155312,"significance":"The derivation in Eqs. (3)–(14) is a genuinely parameter-free analytic result: the only inputs are measured optical powers, wavelengths, field of view, and pupil diameter, and the eye's focal length fe cancels. This gives RS displays a falsifiable, human-centric brightness metric that is directly comparable with conventional displays, and the predicted inverse scaling of L_RS with pupil area is a clear quantitative claim that can be tested. The psychophysical design is a reasonable and clearly described first step for bringing the human visual system into AR display metrology. The main caveats are that the practical brightness benefit depends on the full-beam pupil transmission assumption, which is not validated for realistic eyebox tolerances, and that the resolution comparisons are reported without statistical uncertainty.","major_comments":[{"comment":"The derivation of Eq. (12) is internally consistent, but the claim that pupil constriction is beneficial for AR use cases rests on the assumption that the entire RS beam flux ΦVScan enters the eye independently of pupil area (§2.1.1, Fig. 2). The paper's own Discussion states that the prototype has a small exit pupil and that the RS display is only visible if the narrow beam fully passes through the pupil. In realistic wearable use, head and eye movements and the smaller pupil diameters typical of bright environments (2–4 mm) make partial clipping likely; once clipping occurs, ΦVScan depends on A_P and on lateral alignment, and the inverse-pupil scaling in Eq. (12) breaks down. No tolerance analysis, eye-tracking data, or bench measurement of transmitted flux versus pupil position is provided, and the psychophysical study used a chinrest with a 5-minute fitting procedure that maintains the full-beam condition. The paper should either add a clipping factor (or measured tolerance curve) to Eq. (12), or restrict the 'beneficial pupil dependence' claim to the ideal static-alignment regime.","section":"§2.1.1, Eq. (12); §4 (Discussion)"},{"comment":"The perceived-resolution results are reported as point estimates (7.1 cyc/deg horizontal, 7.4 vertical; 6.7 red, 7.4 green, 7.3 blue; etc.) without confidence intervals or any inferential test. Because the staircase data are dichotomous pass/fail counts per level, the statements that background conditions 'did not have a notable influence' and that red stimuli were 'slightly worse' are not statistically supported. A mixed-effects logistic regression or a bootstrap over participants would quantify the uncertainty and should be added before the cross-condition comparisons are claimed.","section":"§3.2, Table 1, Figs. 6–9"},{"comment":"The comparison between the perceived resolution threshold (7.4 cyc/deg for green) and the IEC 61947-2 optical resolution (3.4 cyc/deg) is used to conclude that the optical resolution of the glasses, not human visual perception, was the limiting factor. This inference is not immediate, because the IEC cutoff is defined at 30% modulation rather than at the contrast detection threshold; a display can convey perceptible modulation beyond the 30% point. Without reporting the display's modulation transfer at the perceived threshold (or the spatial frequency at which modulation reaches the human contrast threshold), the data do not establish that the optical resolution is the limiting factor. The conclusion should be softened or supported by additional MTF/contrast measurements.","section":"§4 (Discussion), IEC 61947-2 comparison"}],"minor_comments":[{"comment":"The intermediate expression in Eq. (5), 'fe^2 tan(α/2) fe^2 tan(β/2)', is not the correct area product; it should read (2 fe tan(α/2))(2 fe tan(β/2)). The final equality and the subsequent formulas are correct, but the written line is confusing.","section":"Eq. (5)"},{"comment":"Equations (15) and (16) contain garbled notation for the Stanley–Davies model ('LA 3600 846' without explicit operators); please typeset the formula with clear parentheses and state the units of LA (cd/m²).","section":"Eqs. (15)–(16)"},{"comment":"The anti-guessing rule 'a line pattern was always presented after two correct responses' is not fully explained; please clarify how this interacts with the 3-down-1-up staircase.","section":"§2.2"},{"comment":"The reference 'Dressier and Rassotv (1981)' appears to contain misspelled author names (likely Dressler and Rassow); please correct.","section":"References"},{"comment":"The manuscript states that the study had a 'gamified design', but no gamification is described in Methods; please either describe the gamification or remove the term.","section":"§2.2 and §4"}],"recommendation":"major_revision","confidential_remarks":"The derivation is clean and the experimental design is a reasonable first step, but the practical generalization from the single-eyebox prototype is the main risk. The lack of inferential statistics in the resolution study and the unclear IEC comparison should be addressed. I do not see a circularity problem: the Stanley–Davies coefficients are imported inputs, not fitted to this paper's data. The manuscript is within the scope of the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The useful core is a clean derivation of an equivalent luminance for retinal scan displays, L_RS = K_m Σ V(λ) Φ / (4 A_P tan(α/2) tan(β/2)), with pupil area in the denominator. That is genuinely useful photometry for a display class that luminance cameras misreport, and it makes the ambient contrast ratio definition for RS displays explicit. The perceived-resolution experiment on a functional holographic RS prototype is a real addition; the field has very little human-in-the-loop data on this architecture. The staircase design is standard, and thresholds clustering around 7 c/d across fifteen conditions is a creditable measurement.\n\nThe soft spots are real but not fatal. First, the headline 'brighter in bright environments' depends on the beam fully passing through the pupil. The formula assumes RS flux is pupil-independent; the paper's own Discussion concedes the single-eyebox prototype loses visibility if the beam is clipped, and the psychophysical study used a chinrest plus a five-minute fitting. Without a tolerance analysis or eye-tracking data, the practical claim for wearable use is not established. The derivation itself is fine in the ideal static-alignment regime.\n\nSecond, the perceived-resolution thresholds come without confidence intervals or statistical tests. The comparison between the IEC optical resolution (3.4 c/d) and the perceived 7.4 c/d is left unexplained; the paper mentions the 30% modulation cutoff but does not quantify how it maps to perceptibility. That undercuts the conclusion that perceived resolution is limited by the optics.\n\nThird, minor: Eq. (5) has a typographical slip in the written area expression, though the final formula is correct.\n\nThe citation pattern is honest; prior ACR definitions are discussed and the standard-vs-additive ACR distinction is sensible. No circularity: the Stanley-Davies pupil model is an input, not a fitted parameter.\n\nThis paper deserves a serious referee. It is a solid, useful contribution to near-eye display metrology, not a paradigm shift. A referee should ask for error bars and a direct treatment of the clipping condition; the core derivation and the dataset are worth publishing after revision.","headline":"Useful derivation and dataset for retinal scan display metrology, but the practical brightness benefit is conditional on perfect beam-pupil alignment and the resolution thresholds lack error bars.","tokens_in":11636,"tokens_out":3007,"would_cite":true,"duration_ms":25245,"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":"The paper derives an equivalent luminance for retinal-scan AR glasses that scales inversely with pupil area, and supports it with a perceived-resolution study.","keywords":["retinal scan display","holographic optical element","augmented reality","equivalent luminance","ambient contrast ratio","perceived resolution","pupil diameter","psychophysics"],"falsifier":"Measure the equivalent luminance of a retinal-scan display while controlling pupil size independently, for example with artificial apertures of different diameters placed at the eye or with pharmacologically dilated pupils, and compare the result with Equation (14): if the luminance that matches a Lambertian reference does not scale as \\( 1/$d_P^{2}$ \\), the derivation is wrong. A simpler check is that, in a dark room with a dilating pupil, a retinal-scan display should appear dimmer rather than brighter, opposite to a Lambertian display.","tokens_in":10660,"feed_emoji":"🕶️","tokens_out":6671,"duration_ms":57381,"temperature":0.7,"pith_summary":"This paper tries to establish that the brightness of laser-scanning augmented reality glasses can be quantified by an equivalent luminance, defined as the luminance of a standard Lambertian display that would produce the same retinal illuminance. The formula depends on laser powers, wavelengths, field of view, and pupil diameter, and it implies that as the pupil shrinks in bright light, the virtual image gets relatively brighter. If true, retinal-scan displays have a self-compensating brightness advantage over conventional displays, which matters for designing AR glasses used outdoors. The paper also measures perceived resolution of a working prototype and finds thresholds around 7 cycles per degree, limited by the display optics rather than the eye.","feed_headline":"Retinal-scan AR glasses shine brighter as pupils shrink","feed_subtitle":"A new formula puts laser AR displays on the same luminance scale as ordinary screens, with a built-in outdoor boost.","key_machinery":"The load-bearing object is the equivalence \\( L_{\\mathrm{RS}} \\): the luminance of a hypothetical Lambertian display that would create the same retinal illuminance as the scanned laser beam. Only emitted flux, field-of-view angles, and pupil area enter; the eye's focal length cancels, so no assumptions about eye geometry are needed beyond pupil size. The companion quantity \\( \\mathrm{ACR}_{\\mathrm{RS}} = L_{\\mathrm{RS}}/(\\tau_C L_A) \\) then lets virtual-image contrast with the real world be computed in the same photometric units as ordinary displays.","core_discovery":"The central claim is that a retinal scan display's brightness can be written as \\( L_{\\mathrm{RS}} = \\dfrac{K_m \\sum V(\\$\\lambda$) \\Phi}{4 A_P \\tan(\\$\\alpha$/2) \\tan(\\$\\beta$/2)} \\), with \\( K_m \\) the luminous efficacy constant, \\( V(\\$\\lambda$) \\) the luminosity function, \\( \\Phi \\) the radiant powers of the RGB lasers, \\( A_P \\) the pupil area, and \\( \\$\\alpha$, \\$\\beta$ \\) the field-of-view angles. This is derived by equating the retinal illuminance from the scanned beam with the retinal illuminance from a Lambertian display and noting that the eye's focal length cancels. Because \\( A_P = \\pi (0.001\\, d_P)^2 / 4 \\) appears in the denominator, the same display looks brighter when the pupil is smaller. Combined with a published pupil-diameter-versus-ambient-luminance model, \\( L_{\\mathrm{RS}} \\) rises by a factor of about five from mesopic to bright photopic surroundings. The paper further defines \\( \\mathrm{ACR}_{\\mathrm{RS}} = L_{\\mathrm{RS}}/(\\tau_C L_A) \\) and argues that standard ambient contrast ratio, rather than one including the ambient transmission term in the numerator, is the right way to characterize these displays. In a 20-participant staircase study with a fully functional prototype, perceived resolution thresholds clustered around 7.1–7.4 cycles per degree, independent of background conditions, and well above the 3.4 cycles per degree measured by an IEC optical-resolution standard, indicating that the display optics limit what users can resolve.","pith_inferences":["Editorial inference: If \\( L_{\\mathrm{RS}} \\) is accepted as a metric, a direct brightness-matching experiment between a retinal-scan display and a calibrated Lambertian display should show that the retinal-scan display matches a luminance scaling as \\( 1/d_P^2 \\) when the pupil diameter is changed; the paper does not run such a test.","Editorial inference: The luminance-versus-ambient boost suggests an energy-budget design strategy where laser power is reduced in dark scenes while maintaining perceptually constant contrast outdoors, but the paper does not demonstrate such an adaptive power scheme.","Editorial inference: Because the human eye was not the resolution bottleneck up to about 7 cycles per degree, a retinal-scan display could potentially reduce its pixel density or holographic combiner area without a noticeable loss in perceived detail, until the optical resolution is improved.","Editorial inference: The proposed distinction between a standard ambient contrast ratio and an additive ambient contrast ratio could help unify comparisons across AR display types, but it would require industry consensus before becoming a standard measurement."],"forward_implications":["Retinal-scan displays can be specified with a luminance value that is directly comparable to ordinary Lambertian displays, provided the pupil diameter is known or modeled.","In bright ambient light, the pupil shrinks and the equivalent luminance rises, so virtual content stays visible outdoors without extra laser power.","The ambient contrast ratio \\( \\mathrm{ACR}_{\\mathrm{RS}} \\) can be used as a design target, since it stays well above the contrast level needed for reliable resolution perception even on bright backgrounds.","The perceived resolution threshold of roughly 7 cycles per degree is limited by the display optics, so future prototypes should first improve optical resolution and field of view before expecting users to see more detail.","IEC-style optical resolution measurements underestimate what users can actually resolve in retinal-scan glasses, so perceived-resolution testing is a useful addition to display metrology."],"supporting_citations":[{"why":"Supplies the Lambertian luminance and retinal-illuminance relations used in the equivalence derivation.","marker":"(Bass, Enoch and Lakshminarayanan, 2010)"},{"why":"Provides the luminous efficacy constant \\( K_m \\) and luminous efficiency function \\( V(\\lambda) \\) used to convert radiant flux to luminous flux.","marker":"(CIE JTC 2, 2019)"},{"why":"Provides the pupil-diameter versus ambient-luminance model used to make \\( L_{\\mathrm{RS}} \\) depend on environmental brightness.","marker":"(Stanley and Davies, 1995)"},{"why":"Alternative pupil-size model that the paper cites to show the qualitative relationship is not tied to one specific pupil formula.","marker":"(Watson and Yellott, 2012)"},{"why":"Describes the retinal-scan laser eyewear principle that the prototype and the derivation build on.","marker":"(Suzuki et al., 2018)"},{"why":"Supports the use of holographic optical elements as transparent combiners in near-eye displays.","marker":"(Maimone, Georgiou and Kollin, 2017)"},{"why":"Defines ambient contrast ratio and laser-beam-scanning benefits that the paper extends for retinal-scan displays.","marker":"(Fidler et al., 2021)"},{"why":"Gives the IEC 61947-2 optical resolution standard (3.4 cycles per degree) that the perceived-resolution results are compared against.","marker":"(International Electrotechnical Commission (IEC), 2001)"},{"why":"Explains the Troland convention for retinal illuminance, which the paper contrasts with its pupil-area-based \\( L_{\\mathrm{RS}} \\) formulation.","marker":"(Thibos, Lopez-Gil and Bradley, 2018)"}],"fun_headline_variants":["AR retinal scan brightness rises as pupils shrink","Smaller pupils boost perceived brightness of AR glasses","Retinal scan AR: luminance scales with pupil size","Pupil diameter controls AR display luminance","Study: pupil size is key to AR brightness"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes the narrow scanning beam (diameter under 0.5 mm) always passes completely through the pupil, so all emitted light enters the eye and the flux does not depend on pupil size; if a small or misaligned pupil clips the beam, the inverse-pupil-area scaling breaks down.","fun_headline_variants_meta":{"raw":{"variants":["AR retinal scan brightness rises as pupils shrink","Smaller pupils boost perceived brightness of AR glasses","Retinal scan AR: luminance scales with pupil size","Pupil diameter controls AR display luminance","Study: pupil size is key to AR brightness"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000251,"raw_usage":{"total_tokens":1601,"prompt_tokens":1033,"completion_tokens":568,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":649,"completion_tokens_details":{"reasoning_tokens":498}},"tokens_in":649,"tokens_out":568,"duration_ms":5764,"temperature":1.0,"reasoning_tokens":498,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T04:57:23.441309+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the equivalent luminance of a retinal-scan display while controlling pupil size independently, for example with artificial apertures of different diameters placed at the eye or with pharmacologically dilated pupils, and compare the result with Equation (14): if the luminance that matches a Lambertian reference does not scale as \\( 1/$d_P^{2}$ \\), the derivation is wrong. A simpler check is that, in a dark room with a dilating pupil, a retinal-scan display should appear dimmer rather than brighter, opposite to a Lambertian display.","supporting_citations":[{"cited_title":"and Lakshminarayanan, V","cited_arxiv_id":null,"evidence_quote":"Supplies the Lambertian luminance and retinal-illuminance relations used in the equivalence derivation."}],"review_version":1}