A theory for image persistence in HgCdTe photodiodes

Image persistence can produce systematic errors, which remain significant in some applications even when buried in noise. Ideally the image persistence amplitude, linearity and decay over time could be calibrated independently for each pixel to levels well below the noise floor, however averaging multiple measurements to characterize persistence to this accuracy is impractical due to the long time scales for the decay and the need to emulate the exposure and readout timing of the observations to be calibrated. We examine a compromise where the initial persistence response is characterized independently for each pixel but the latter parts of the decay are assumed to follow the mean decay curve. When averaged spatially, persistence increases monotonically with stimulus amplitude until the photodiodes approach forward bias. For several Teledyne 1.7 μm cutoff HgCdTe detectors tested, persistence is linear over most of the normal signal range. We characterize the temporal response, and examine the dependence of charge emission time constants on total stimulus duration. We describe the suppression of persistence by signal in the current frame and begin to examine the superposition of the decay curves from multiple stimuli.