Using NGMIX To Identify Junk Objects

There is a population of objects in ngmix that are somewhat starlike, but actually are junk. What are they and how can we use SExtractor to weed them out?


Plots of size signal-to-noise vs mag:

NGMIX009 : older, slower code, old psfs NGMIX010 : newer, faster code, new psfs

The difference is how the code handles these things which are actually LARGE size with LARGE uncertainty.

Correlation Functions

High end are correlated with Tycho stars (And other bright stars that are fainter than the Tycho catalog!)

This was made with a selection from ngmix010, but a similar plot can be made from ngmix009. These objects didn't appear...they're just handled differently (better?) which causes them to pop out.

Other clues

  • These have very low surface brightness, and high uncertainty.
  • They are often near bright stars or other defects.
  • They have strange shape measurements.
  • They are not isolated to the band at low positive S/N (I don't think.)
  • They have relatively large model radius (sextractor) vs flux radius; maybe distinct population.
  • They are relatively low surface brightness given their mag_auto, but not anomalously so.