Optimal Photon Counting with Fast, Low Noise Astronomical Imagers

Abstract

Measuring rapid astrophysical phenomena requires many short exposures, which suffer from repeated read noise. A variety of recent technologies are achieving sub-electron read noise, which enable rapid time-domain studies of faint sources. These data can be analyzed with traditional methods, but the ability to resolve individual electrons enables a new class of more precise, photon-counting techniques. We present a photon-counting method for sub-electron read noise detectors which delivers optimal statistical uncertainties. We focus on Complementary Metal-Oxide-Semiconductor (CMOS) detectors and demonstrate our method on optical data collected by CMOS instrument "proto-Lightspeed," recently commissioned at the Magellan Clay telescope at Las Campanas Observatory. The method can be applied to non-CMOS detectors as well, such as skipper CCDs and linear-mode avalanche photodiodes. The method substantially improves analysis of faint sources, increasing signal to noise ratios by up to $\sim 130\%$, which would require 5.3$\times$ more observing time to accomplish with data alone. As an example we show detections of four normal rotation-powered pulsars with confirmed optical pulses, achieved in under two hours of total exposure. We provide the lightspeedpy pipeline for reducing proto-Lightspeed data with both photon counting and standard techniques.