The amp geometry in the current readouts is slightly wrong.

The physical geometry, as detailed in the Hamamatsu docs, is:

. 8 leadin pixels
. 512 image pixels

We currently request 32 pixels of overscan. For all amps, the images show:

. 9 leadin pixels
. 512 image pixels
. 31 overscan pixels

From the pixels levels, which vary slightly between amps, it appears that the 0th leadin pixel in the amp image is actually an overscan pixel, presumably the last one.

The current image headers do not have a geometry version number. I suggest that we adjust the stack ISR routines to drop 9 pixels and use 31 pixels of overscan for such files (geometry version 0, call them). Subsequent images from the acquisition software should correct this problem and set the version number to >= 1. When the ISR sees that, it should not correct.

If such a procedure would be too difficult, we can rewrite existing data: none of it has been created outside a lab, and there are only a few hundred real images.

If you want to confirm this, i'll suggest one of the lab flats, say 2015-12-22/PFSA00722192.fits. Yes I know that should be PFLA.