Answers are grounded in this paper and cite their sources.
Model calculations (SIMULATED, NOT MEASURED) from the published datasheet noise budget of the Hamamatsu ORCA-Flash4.0 V3 (C13440-20CU) scientific CMOS camera. The dynamic-range route gives 14.19 bits per pixel, but the channel-capacity route at full well (30,000 electrons, 1.6 electrons rms read noise) gives 7.44 bits per pixel per frame, because shot noise contributes 99.99 percent of the noise variance at full well. At 60 W and 100 frames per second the electrical cost is 1.92e-8 joules per delivered bit (5.20e7 bits per joule), and the camera spends 4,914 incident photons per bit at 560 nm. The shot-noise floor for on-off optical signalling at an error probability of 1e-3 is 6.91 photons per bit, or 2.45e-18 joules per bit: the camera is a factor of 711 above the floor in photon terms and 7.84e9 above it in electrical terms, and setting the read noise to zero would add only 0.000061 bits per pixel. A comparison with a digital MEMS microphone gives 1.58e-9 joules per bit. All numbers are produced by the deposited compute.py script.
No takes yet. Share an insight, caveat, or question.
William Alexander Aldred (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: