Fix Version/s: None
DM-6817 revealed a difference in the trace radii for PSF model stars between the LSST vs. HSC stack runs ( DM-6816 and DM-9028). DM-9535 uncovered that this is due to differences in the implementations of the Brighter-Fatter correction between the stacks (namely in how the second derivatives are computed).
To make a fair comparison between the LSST and HSC stacks, we should be using the same algorithm for the Brighter-Fatter correction. While an assessment of the optimal algorithm awaits
DM-9553, we concluded in DM-9535 that the HSC version seems to be doing a slightly better job, so we will adopt that on the LSST stack for this run.
In addition, I will set processCcd.charImage.measurePsf.reserveFraction=0.0 for this run (instead of the default of 0.2 for obs_subaru) and do the same in another run of the the RC dataset through the HSC 4.0.5 stack in order to minimize differences in calibrations due to PSF model star selection differences.
While the differences due to the BBox setting for adoptive moments were small (
DM-9411) I will also use consistent settings for both runs (here I will adopt the HSC stack to use LSST's setting).
This processing has been completed and the rerun is on tiger at Princeton: /tigress/HSC/HSC/rerun/lauren/LSST/
However, the issue of
DM-9862is present in this reduction, so the coadds should not be considered useful. Once a more suitable recent reduction of the RC dataset has been completed (certainly as part of DM-10043, but one may exist sooner).