With this test we evaluate the overall 'health' of the shapes by plotting a standard classifier against magnitude.


One 1/2 night worth of data in each of the passbands, in catalogs including CLASS_STAR evaluation.
Magnitude limit obtained in some manner.


Produce a CLASS_STAR vs MAG_AUTO plot using the catalog from single-epoch data of one night. There should be a clear two-pronged structure which merges at magnitude limit (check with noise, MAGERR, or mask, if it has been validated). The stellarity is computed with SExtractor and does not rely on PSF extraction or truth tables.


There should be less than 1% of objects with 0.1<CLASS_STAR<0.9 up to m_limit-0.5. This would just tell us the overall health of the images wrt to extracting shape information for star/galaxy separation.

Tools: could be a python script plotting this and calculating the corresponding pass/fail value. Currently exists as C++ script to be run in ROOT environment.


If there is a very large amount of noise clustering in the zone 0.1<CLASS_STAR<0.9 at bright magnitudes, check SPREAD_MODEL shape. If noisy too, then check visually known stars or galaxies for problems on image quality or tests in WL.