| | 5 | Further investigation of the calibration errors point to an issue with images that have been excluded and the internal relphot calculation that determines the transparency/zeropoint offsets between images. Sources on all images are passed to relphot, but only images that are accepted into the stack have the magnitude offsets applied back to their sources. This creates ~4 magnitude offsets for the images that are intended to be masked from the stacking. However, this masking does not seem to be working correctly, so we occasionally select these uncalibrated sources for the PSF target determination. This explains why disabling this correction seemed to solve the problem: all exposures only differed by their transparency values, which are generally small. |
