IRAC High Precision Photometry
Dark current in all IRAC frames is measured with a shutterless system designed by the IRAC instrument team. To remove the dark level as accurately as possible, weekly observations are made of a low background field at the north ecliptic pole. These images are then vetted for obvious persistent images (latents), and median combined to make a weekly dark. For each data frame, the nearest-in-time dark is used to correct that frame. The advantage of this weekly dark is that it removes any low level persistent images which may be present in the science frames.
This means that A) different darks are used for different observations of the gain map since that dataset was observed over many years, and B) for datasets which use the gain map, different dark frames are subtracted from the gain map than are subtracted from target observations.
We test the use of different darks in the gain map dataset by first generating a superdark frame and then applying it to all data frames of the same frame time. The superdark is generated by median combining all dark images taken during the warm mission of a single frame time. The median combine will reject all sources since this data is taken on-sky. The background level of the dark frames taken at the north ecliptic pole will vary as a function of time, but since most users here are doing aperture photometry, and not absolute photometry, we do not care that the mean level of the dark is incorrect by the amount that the zodiacal light fluctuates over a year baseline. The superdark is applied to each data frame by first backing out the calibrations already applied to the BCD including the weekly dark, and then calibrating those image with our superdark.
We test the superdark versus weekly dark in the gain map calibration star dataset by testing both versions in a set of ~30 different AORs taken on a different calibration star. We see lower scatter in those observations when using the weekly dark, therefore we do not use the superdarks in our current gain map dataset. The cause for this is likely to be local in time persistent images in the science frameswhich the local darks are able to partially mitigate whereas the superdarks are not.