Nasty Banding in Southern Coadds

T. Jarrett, IPAC
(981103)

Night of 980406s, scan "030", coadd "033" (also contains NGC 252)

Statistics Summary:

Note: sigma includes stars (so it is artificially high)

Using n-sigma rejection method, the statistics are as follows:


Identification of Banding Period

The cross-scan banding appears to be a sinusoidal remnant feature from constructive coaddition of the stepped frames. We can narrow down the period and amplitude of the feature by applying a fourier transform to the image data. Since the left and right sides of the coadd are affected to differing degrees, we compute the transform on each separately. The fourier transform is sensitive to real stellar objects, such as bright stars and large galaxies, both of which may have features with sizes on order of what we are looking for in the faint backgroud. Therefore, we perform median filtering to minimize confusion from large structures. The filter procedure is basically to collapse a "line" of data to one number using iterative median and n-sigma rejection. Finally, we apply the transform to the 2-D data (median line value versus line number). To bolster the statistics, we combine two lines in the median filter procedure; thus, one coadd collapses to 512 X line values X 2 (the "2" refers to the coadd split).

Scan 085 of 980416s contains a number of coadds affected by the cross-scan banding. The following gifs illustrate the banding, reduced line medians and resultant fourier transform.

The period of the banding feature is about 78" for this coadd example. The corresponding amplitude, ~0.15, is significant and proves that the banding feature present in this coadd is bad news for galaxy photometry.

The period seems to vary about +-10 degrees or so between coadds. The next set of gifs show the coadd "068" with a period closer to 90".

Notice that the banding feature is present along the left side of the coadd, although with an amplitude of about half that seen on the right side. The banding feature is present in the H and K images as well (at much fainter levels):


See also Gene Kopan's analysis of the J banding:
J Band Quadrant Electonics Noise Problem