PSF star selection: survey other surveys' methods (MU_MAX−MAG_AUTO, colour) #950

PSF stars are selected from the stellar locus in FWHM vs magnitude. Since #934 the box is the KDE peak ± 3σ̂. In poor seeing the locus broadens and the box widens with it: on two test exposures it kept 99% of locus stars instead of 96%, and non-stars in the box went from 1.7% to 2.7%. The star–galaxy separation has drawn criticism before, and #935 shows FWHM stops separating stars from galaxies beyond about 21 mag.

Survey how other surveys select PSF stars (Euclid, DES Y6, KiDS, HSC) and propose what to adopt. Two candidates came up at the telecon:

• MU_MAX − MAG_AUTO, a concentration measure that doesn't depend on size. Centre the PSF-star box on a KDE peak of the FWHM preselection #934 used it as the truth set. It needs only the r band, and it avoids selecting stars on the very quantity the PSF model fits.
• Colour. Hendrik has a multi-band star list from his own high-threshold SExtractor run.

Related: #935 (faint limit), #840 (star-catalogue depth and flux-radius bias).

Claude Opus 5.5 on behalf of Cail. From the 3x2pt telecon, 6 Oct 2026.