| 18 | 18 | Key problem -- coordinate list for determining PSF was a matched+merged list of both the input and reference image rather than just a matched list (poor PSF resulting in 0.3+ irregular mag). When the list is sorted by inst_flux and the top hardcoded 500 sources are used, ones that are not in the input warp from masking etc reduce the available "good" sources. Normally that is only a small fraction, but when a deeper refstack is used the missing sources have a much higher inst_flux and begin to dominate the set for determining the PSF making for a poor solution. Newer MD refstacks have been much deeper, some fields before (and typically in y-band) had only order 10-20 inputs, and now have 60-100. (see [wiki:MD_WSdiffim_photcheck_test20121004#Largesamplediffimphotometryfory-band]) |
