In the attached figure I show the coadds without (left) and with (right) skycorrection.
I am wandering if I missed option to use for the skycorrection step: This is what I am using for skycorrection: skyCorrection.py /sps/lsst/HSC/Manal/hsc_output/ --calib /sps/lsst/HSC/Manal/HSC_Z_R_Band/CALIB/ --rerun rerun --id visit=1164^1166^1168^1172^1174^1180^1182^1188^1176^1184^1178^1170^1186^1190^1192 --batch-type=smp --cores=8
Thank you in advance
Hi Manal, Your skyCorrection.py call looks fine, assuming your sky frames in /sps/lsst/HSC/Manal/HSC_Z_R_Band/CALIB also look fine. We don’t use any special configs.
The images are completely in line with what we see as well and is a known limitation of the skyCorrection algorithm for coaddition which was reported in dmtn-080.lsst.io.
The coaddition algorithm assumes that the visits have been background subtracted before stacking and have a mean background of zero. This assumption is not true when doApplySkyCorr=True, which doesn’t remove the ripples at the edge of the focal plane (give me a minute while I dig up examples)
We’re working on new ways to mitigate this (both via background matching during coaddition and a replacement full focal plane background model during single frame processing), but in the meantime, I recommend using doApplySkyCorr=False if you’re not specifically interested in the wings of bright large galaxies and especially when you’re looking at tracts like 9812 with minimal dithers and patches at the edges of the fov.
Sky correction (RIGHT) does somethings better: Less oversubtraction around bright stars and galaxies, some of the instrumental amp offsets removed. But some features are worse: noticeable roll off at the edge of the focal plane and smaller scale ghosts are not removed, all which contribute to what I called “ripples” above, which when stacked with minimal dithers are causing the feature you’re seeing in your coadd.
