Repository navigation
Extend size of bright star masks [BUG] #662
Description
**Describe the bug**
Our masks for saturated stars are not big enough, we are getting object contamination in our catalog.{{ message }}
**Describe the bug**
Our masks for saturated stars are not big enough, we are getting object contamination in our catalog.
Activity
cailmdaley commentedon Jul 14, 2026
Possible major contributor found while debugging a first P3 test run on Nibi (#808):
mask.pyaverages the four GSC magnitudes withis not Noneguards that let NaN through, so any bright star with missing GSC bands getsmean = NaN, failsNaN < HALO_MAG_LIM, and silently receives no halo mask at all — and the brightest (saturated) stars are precisely the ones most likely to have missing bands. On one test CCD, fixing this took the halo-star count from 57 to 118, the recovered objects being the brightest in the field (mag 5.7–9.5).Fix is up in #824. If contamination persists after that, the remainder would point at halo radius scaling rather than missing masks — we plan to quantify (detections in halo annuli before/after) on a 10-tile P3 run.
— Claude (Fable) on behalf of Cail.
martinkilbinger commentedon Jul 15, 2026
After discussion with Mike and @aguinot it would be a good idea to have consistent and conservative masking within ShapePipe at exposure level and the structural masks / updated THELI setting that Hendrik is using.
cailmdaley commentedon Sep 25, 2026
Superseded by the external healsparse masks (#845); whether their star radii are large enough is tested empirically in #851.