Each pair of CCDs share the same ARCON controller box, with some crosstalk occuring between these chips. This is best seen as a "reflection" of heavily saturated stars, seen faintly in the "victim" chips. To correct for this, we must first determine the crosstalk coefficients between chips. The IRAF task 'xtcoeff' does a pretty good job of this, although it is necessary to revise the coeficients slightly for best results. The task 'xtalkcor' allows you to view an output image scaled with the adopted xtalk coefficients. Stars that are saturated on a "source" chip should be invisible on a "victim" chip. Since the cross-talk coeficients are simply written to a text file, they are easily modified until one finds good results.
The best data to use for this is one with very little sky background but lots of saturated stars on all the chips.
Below are the coeficents we found at Kitt Peak. They differ somewhat from
the standard values, but we found obvous residuals when we used the
the standard values. We remeasured it again for the February run, and
by visual inspection arrived at the same coeficients.
| Victim | Source | Old | Aug2000(Buell) | Sept2000(us) | Sept2002(us) | March2003(Buell) |
|---|---|---|---|---|---|---|
| im1 | im2 | 0.00116 | 0.00133 | 0.00152 | 0.00152 | 0.00162 |
| im2 | --- | --- | --- | --- | --- | |
| im3 | im4 | 0.00289 | 0.00267 | 0.00310 | 0.00310 | 0.00298 |
| im4 | --- | --- | --- | --- | --- | |
| im5 | --- | --- | --- | --- | --- | |
| im6 | im5 | 0.00152 | 0.00159 | 0.00168 | 0.00175 | 0.00175 |
| im7 | --- | --- | --- | --- | --- | |
| im8 | im7 | 0.00062 | 0.00085 | 0.00085 | 0.00085 | 0.00085 |
xtcoeff obj2089 xtalkobj2089.dat @vnoao8 @snoao8 PACKAGE = mscred TASK = xtcoeff input = obj2089 List of mosaic exposures output = xtalkobj2089.dat Output crosstalk file victim = @vnoao8 List of victim extensions source = @snoao8 List of source extensions (smin = 20000.) Minimum source pixel (smax = INDEF) Maximum source pixel (medfact= 0.5) Median factor (maxcoef= 0.01) Maximum coefficient value to consider (niterat= 3) Maximum number of rejection iterations (low = 3.) Low rejection sigma factor (high = 3.) High rejection sigma factor (interac= no) Examine and fit interactively? (verbose= yes) Output to terminal? clobber = no Clobber existing crosstalk file? (mode = ql)
Note that the newer versions (MSC V4.8) also want a good estimate of the background of both the source and victim chips, plus object masks, etc., etc. My recommendation is to start with some "standard" values (either mine or the revised March 2003 NOAO ones above) and fiddle around with them, testing with xtalkcor below, until you find the ones that work best on your data.
That said, the output file of xtcoeff will resemble the following:
# XTCOEFF: NOAO/IRAF V2.11.3EXPORT sholmes@tofu Thu 10:06:58 19-Jul-2001 # Images: obj2089 im1 im2 0.001144 (0.000005, 219.0) im2 im1 0.000070 (0.000005, 13.3) im3 im4 0.001253 (0.000004, 282.9) im4 im3 0.001457 (0.000011, 130.9) im5 im6 0.000160 (0.000009, 17.4) im6 im5 0.001045 (0.000005, 209.9) im7 im8 0.000597 (0.000010, 61.7) im8 im7 0.001198 (0.000004, 300.0)
The columns indicate the VICTIM chip (where you will see the problem), the SOURCE chip (where the problem comes from), the cross-talk coefficient, and, in parthesis, the uncertainity in the coefficient and the sigma uncertainity. In this case we see very significant coefficients for im1 being affected by im2 (200+ sigma), but more marinal results for im2 being affected by im1 (13 sigma). We can compare these numbers with the default cross-talk coefficients found by the Tololo staff at the time; they are in the database mscdb$noao/Mosaic2 (for the CTIO camera). There the coefficients applicable to our run (in August 2000) can be found in CAL0009/xtalkA0009:
# XTCOEFF: NOAO/IRAF V2.11.3EXPORT mosaic@ctio4m Fri 16:56:54 01-Sep-2000 # New coeffs determined after significant La Serena lab work in Aug 2000 im1 im2 0.001190 (0.000005, 217.0) im2 im1 im3 im4 0.001373 (0.000009, 148.2) im4 im3 0.001388 (0.000006, 239.7) im5 im6 im6 im5 0.001460 (0.000004, 328.7) im7 im8 im8 im7 0.001177 (0.000003, 356.4)
Thus we see that the same cases where we found relatively low significance coefficents were set to zero in the default database. A comparision of our numbers with theirs show ver similar results (.001144 vs .001190, say) except for im6/im5, where we found .001045 and they found .001460. Additionally we found a connection between im7 and im8 which they did not, at the 60 sigma significance level. It is relatively easy to see a difference of .0006--- a saturated star (30,000 ADU) would show up in the "victim" with a level of 18 ADU, which should be easy to see. Are these low significnace coefficents real? We might choose to delete these for now, and then carefully examine the images corrected using our new table:
# XTCOEFF: NOAO/IRAF V2.11.3EXPORT sholmes@tofu Thu 10:06:58 19-Jul-2001 # Images: obj2089 im1 im2 0.001144 (0.000005, 219.0) im2 im1 im3 im4 0.001253 (0.000004, 282.9) im4 im3 0.001457 (0.000011, 130.9) im5 im6 im6 im5 0.001045 (0.000005, 209.9) im7 im8 0.000597 (0.000010, 61.7) im8 im7 0.001198 (0.000004, 300.0)
First, we should look at some frames that contain strong saturated stars in im8 to see if we can see them in im7. Secondly, we will use our file with xtalkcor to see if we are happy with how good a job it does in getting rid of the ghosts of saturated stars.
Our conclusion was that we CAN see saturated stars from im8 in im7, although the coefficient is a little too big---we found that 0.000200 (rather than 0.000600) worked well. Furthermore, we find that im1 DOES act as a source for images on im2, although experimentation suggests that a coefficient of 0.00200 works well there as well. So, we adopt:
# XTCOEFF: NOAO/IRAF V2.11.3EXPORT sholmes@tofu Thu 10:06:58 19-Jul-2001 # Images: obj2089 im1 im2 0.001150 im2 im1 0.000200 im3 im4 0.001375 im4 im3 0.001375 im5 im6 im6 im5 0.001450 im7 im8 0.000200 im8 im7 0.001170
PACKAGE = mscred TASK = xtalkcor input = @Z List of input files output = p//@Z List of output files bpmasks = List of output bad pixel mask rootnames xtalkfil= xtalkobj2089.dat List of crosstalk files (bpmthre= -10.) Corrections threshold for bad pixels (fextn = fits) File extension (noproc = no) List operations only? (fd1 = ) (fd2 = ) (mode = ql)