# KPNO Mosaic chip-chip cross-talk coefficients

David W. Hogg
New York University
2005 May 9

To read this table, do the following: Find the column corresponding to the "target" chip (the chip you want to read). Look for significant amplitudes. For each "cross-talk" chip that has a significant amplitude in the column, subtract the amplitude times the cross-talk-chip image from the target-chip image. (In principle, you need to do some column and row flipping, but in practice you don't because the non-zero amplitudes are for chip pairs with identical orientations.)

chip12345678
1 -0.00000 -0.00003 +0.00001 -0.00001 -0.00001 -0.00000 -0.00000
2 +0.00137 +0.00000 -0.00001 -0.00003 +0.00001 -0.00001 -0.00002
3 +0.00001 -0.00001 +0.00002 -0.00002 +0.00000 +0.00001 -0.00001
4 -0.00000 -0.00001 +0.00290 -0.00005 -0.00001 -0.00001 -0.00003
5 -0.00002 -0.00002 +0.00001 +0.00000 +0.00166 -0.00002 -0.00000
6 -0.00001 +0.00000 -0.00000 -0.00001 +0.00003 +0.00000 -0.00002
7 +0.00001 -0.00002 -0.00003 -0.00002 -0.00002 -0.00003 +0.00083
8 +0.00000 +0.00001 +0.00001 +0.00001 +0.00000 +0.00000 +0.00000

These coefficients were found with mosaic_crosstalk and mosaic_crosstalk_analyze in the astrometry.net codebase.

Briefly, the method is as follows:

1. Read in pairs of chips, one being called the "target" and one being called the "cross-talking".
2. Find the highest-variance 256x512 (yes, vertical rectangle is a good idea) sub-image of the cross-talking chip.
3. Find all the pixels in the sub-image that are simultaneously far from median for the cross-talking image and close to median in the target image. This reduces false apparent cross-talk from accidentally overlapping sources.
4. Fit the identified pixels in the target sub-image as a linear combination of the form
a C + b + c x + d y
where a,b,c,d are free parameters, C is the cross-talking image pixel values, a is your single-image estimate of the cross-talk, b is the "sky" level, and c and d allow for gradients in the sky level.
5. Repeat the above for all chip pairs on all on-source images taken in your run.
6. The median a values (median taken across images) are your cross-talk coefficients, given above.
We believe that most other techniques (including that implemented by the "official" procedure "xtcoeff") are at least slightly biased relative to the technique described here.

The IDL code that performs these functions is

This code depends on the skymaps.info codebase idlutils.