The ISP data from late 2006 / early 2007 makes a good demo case for building and testing detrend images.  

Some notes from the analysis:

 * the shutter correction can be measured and applied, but there are some limitations:
  * the input images with excessive saturated pixels should be excluded - this is especially an issue for the ISP since it seems to saturate with a severe non-linear regime (>60k).
  * the correction does not work below ~0.05 sec.  The correction image ranges from -4 msec to +45ms.  
  * the correction appears to have some variations as a function of time: 
the residual images include a term which appears to be the difference between the shutter correction at two small offsets.  
I conclude that the shutter, detector, and optics are not remaining co-aligned.  this effect appears at below the 1% level.  

 * flat-field residual images (verify mode)
  * g,r,i images are easy to explain
   * short exposures (<0.05) show the effect of the shutter correction hitting its limits
   * some images show large scale gradients, perhaps clouds or sky illumination pattern
   * some images show the effect of non-linearity / saturation

  * z band images have some peculiarities
   * the show all the expected features above (not peculiar)
   * images from 2007/1/18 seem to be very poor
   ** most of these are 240 sec exposures - bad dark correction??
   ** some of these have stars (ie, night, not twilight) - IR sky glow?

  * y band images show similar issues
   * problem from 2007/1/18

'''looking in detail, it is clear that the dark correction is poor for long exposures'''