Reset-subtraction test
  21 Dec 1999
  
  Test to determine if online processing done by PixCel (reset correction, 
  rotate, and assembly) is working correctly.  The test data were taken by 
  rgm/aef at Las Campanas on 15 Dec 1999 (the data were obtained by ftp to 
  rigel.lco.cl in /data1/19991215/ioa/irx*fits).
  
  Run 9137:
    Raw data with no on the fly processing
    One 1024x512 image per chip quadrant containing reset and data portions
    RRR mode
    assemble off
    rotate off
    subtract off
    read 000 contains the reset for read 001
    read 001 contains the reset for read 002
    etc.
  
  Run 9138:
    As above assembled in chips and with reset subraction
    RRR mode
    assemble on
    rotate off
    subtract on
  
  Run 9135:
    Same as above except that data has been also rotated on the fly so that 
    all datasets are coaligned NS.  This is our standard observing mode.
    RRR mode
    assemble ON
    rotate   ON
    subtract ON
  
  Test: 
  
  Assemble and reset-correct run 9137 and make sure this agrees with run
  9138.  Rotate the assembled/corrected run 9137 and make sure the output
  agrees with run 9135.  The rotations applied to 9137 were 270, 0, 90,
  and 180 degrees for chip1, chip2, chip3, and chip4 respectively (as
  specified in the pixcel.ini file).  Assembly was done by arranging the
  quadrants in the image like this:
  q4 q3 
  q1 q2
  
  Procedure:
  
  IDL> subtract
  IDL> assemble
  IDL> diff
  IDL> rotdiff
  see programs in /data/cass60d/sabbey/reset/*.pro
  
  Result:
  
  The difference images (9138-9137 and 9135-9137) are perfectly flat
  noise images for all loops (difference images and original data are
  available in /data/cass60d/sabbey/reset).  
  
  However, a few minor comments:
  
  1) the reset level (mean signal level in the reset portion of the frame)
     typcically varied between loop 0 and loop 1 but was constant between later 
     loops.  For example:
  
     cl> imstat irx_09137_c1q1*fits[1:512,*]
  
                 IMAGE                    MIDPT     MODE      STDDEV
     irx_09137_c1q1_000.fits[1:512,*]     4531.     4516.     260.9
     irx_09137_c1q1_001.fits[1:512,*]     4471.     4392.     371.6
     irx_09137_c1q1_002.fits[1:512,*]     4472.     4389.     370.9
     irx_09137_c1q1_003.fits[1:512,*]     4472.     4389.     369.4
     irx_09137_c1q1_004.fits[1:512,*]     4471.     4387.     369.4
  
  2) the difference images do not have a median signal level of 0, especially 
     for loop 1 and loop 2 which have +/- tens of counts.  with the reset
     level known to vary during the first couple of loops, this is possibly
     not surprising but should probably be looked into more.

     cl> imstat irx.9135-9137.c3.00?.fits
                 IMAGE             MIDPT      MODE     STDDEV
     irx.9135-9137.c3.001.fits     -19.20    -23.52     66.7
     irx.9135-9137.c3.002.fits     -29.68    -32.08     67.0
     irx.9135-9137.c3.003.fits      -2.85     -3.42     64.6
     irx.9135-9137.c3.004.fits      -7.42     -1.44     66.2

     cl> imstat irx.9138-9137.c3.00?.fits
                 IMAGE             MIDPT      MODE     STDDEV
     irx.9138-9137.c3.001.fits    -16.15     -20.62     66.58
     irx.9138-9137.c3.002.fits    -29.70     -29.61     66.38
     irx.9138-9137.c3.003.fits     -1.57      -7.22     65.76
     irx.9138-9137.c3.004.fits     -1.40       2.81     65.86
  
  3) the cross of bad columns/rows at the borders between quadrants sometimes 
     leaves a significant residual in the difference frames.
  
  4) it appears that a dust speck moved between observations leaving a 
     noticeable residual speck in one of the difference frames.


irx.9135-9137.c1.002.fits.gz:

irx.9138-9137.c1.002.fits.gz: