There are different recipes in use for stacks:

THREEPI_STACK - used for LAP and SAS
NIGHTLY_STACK - used for nightlyscience.

For a test, I processed an MD nightlyscience (md09.20110915, i band) using the THREEPI_STACK recipe, to compare to NIGHTLY_STACK. The difference between the 2 recipes is mainly that one (threepi) puts the output in some kind of logflux way, and nightlyscience does not.

I'm still processing and downloading, but I did do the following for skycell 20 (not a very clean skycell):

 * I looked at the catalogs mapped onto the images: psphot misses a lot for both, and detects fake things (for both)

 * I made a histogram plot, red=  threepi_stack, black = nightly_stack.  The red line looks very similar to the SAS as before, and the black looks similar in shape to skyprobe (and md). 

[[Image(stacks.png)]]

 * worth noting is that the values in wt.fits for threepi are sometimes negative (what does that mean?).  This does not happen in the wt.fits for the nightlystack recipe.

Questions: do we need to tweak psphot to work with these logflux output files?