Anom-R3: Overscan correction is working w/o anomalies, including fixing jump in some CCDs that share a crate with the focus chips when the 2kx2k chips stop readout/clocking.
Raw images contain additional columns with prescan and overscan data. The median of these is usually subtracted row by row from the science part of the image in order to correct for temporal variations in the bias. For this, the variations need to be stable between the overscan / prescan and the science part of the image.
We need reduced (bias+overscan corrected, flat fielded, not yet background subtracted) science images from different nights of the commissioning period. The science images should be taken with short exposure times, so the sky background does not dominate potential (additive) overscan anomaly patterns too much.
What we're looking for are horizontal stripes of jumps in the background level of that are residuals even after overscan+bias correction. A way of finding these is generating Sextractor background maps with a mesh of 1pix height and full chip width and looking at them – periodic stripe patterns or jumps should immediately be visible.
GaryB suggestion: not sure that SExtractor will be too happy producing sky backgrounds at 1 pix resolution. Another way to do this would be just to measure the median value in each row of the bias+overscan-corrected images. Perhaps take the median of 1000 or so columns. Each exposure is then reduced to a single column of sky values. Can then examine these individually, or look at the median/mean of a bunch of exposures, looking for any jumps. Would need to repeat this for each CCD.
...or code that takes median along rows of an exposure.
If any stripe patterns or a jump in the CCDs that share a crate with the focus chips are visible in the background images described above, the test fails.
If the test fails, we need to check master bias images, master flat images, raw master, flat and science images for overscan patterns and their alignment between overscan and science region. This might trigger either changes to the DM pipeline or indicate actual electronics problems.
This appears to work fine. (PASSED)