Details
-
Type:
Bug
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: cp_verify
-
Labels:None
-
Story Points:2
-
Epic Link:
-
Team:Data Release Production
-
Urgent?:No
Description
The bias and dark NOISE tests continue to fail for amplifiers with lower than nominal read noise. This ticket will fix that bug.
Attachments
Issue Links
- relates to
-
DM-38002 Determine if overscan-induced noise can be reduced
- To Do
The NOISE tests are failing in the comparison to the read noise. The issue seems to be that we're comparing the image statistics against the read noise derived from the serial overscan. This ignores that the serial read noise is just one sample of the read noise, and we should continue to compare against the nominal cameraGeom value (which should be updated regularly from analysis of a set of overcan residual noise).
This test also ignores the fact that the current image noise is actually something like the sqrt of {{ \sigma_
{read noise}^2 + \sigma_{noise added from row-by-row overscan correction}^2 + \sigma_{noise added from column-by-column overscan correction}^2 }}. Failing these tests because we haven't resolved the ~1 ADU additional overscan induced noise isn't useful, although it may be worth revisiting this after DM-38002 completes.