Difference image (imageDifference.py)
Coadded template (assembleCoadd.py)
Single Skymap Image (makeCoaddTempExp.py)
Calexp (processCcd.py) (independently warped against another calexp)
NOAO Instcal file
Science | Mask Plane | Variance Plane
This grid of images shows the workflow of taking an NOAO instcal and producing a difference image using a coadded template. In one NOAO instcal (bottom row), there exists a cosmic ray. This cosmic ray is properly flagged in the instcal. In processCcd (or potentially in the manual warping afterwards), additional flags are added, and the variance plane is widened compared to the wtmap from NOAO. The UNWARPED version of the calexp is then fed into makeDiscreteSkyMap.py and makeCoaddTempExp.py (–makePsfMatched=True) to create row three. In this row, the variance plane is greatly increased around the cosmic ray (potentially due to the Psf matching), and there is a very visible grid pattern in the “normal” parts of the image. The coadded template from assembleCoadd.py is shown in row two. Here, there also exists a strong grid pattern (not shown), but more concerning, the clipped mean filtering that is supposed to remove the high variance seems to fail. The cosmic ray is not visible in the coadded template science image, but the outline is visible in the coadded template variance image. These values are ~1e6. This could be because the clipped mean filtering only applies to the areas of high flux in the science image, meaning that the outline of the broadened variance plane does not get removed. NOTE: these artifacts exist all over the image, and are not likely unique to cosmic rays. Finally, when another visit uses the coadded template in imageDifference.py, very high variance relics remain due to the high variance in the coadd. These values are ~1e6-1e13.
- High variance relics remain in the coadd variance plane.
- Large-scale grid pattern exists in the noise of the coadd and tempExp variance plane.
- Variance around objects like cosmic rays becomes much broader in tempExp (perhaps due to psf matching).
- Mask plane in tempExp and coadd does not display properly in ds9.
Is this something people are familiar with? Is there a solution?