Heather's stacking adventures

This wiki describes my adventures while trying to stack things in both skyprobe and in gpc1.

I am trying to process skyprobe images using ppStack. This has not been done before, but I have a wiki to describe some of my steps (see IspSasProcessing). The stacks look pretty crummy, which led to the following investigations:

1. why are there big chunks missing
2. why did it take 10k seconds?
3. why are the edges nan'd off (gene says convolution, so that's ok for now)

These are stacks of roughly 100-200 skyprobe warps (in a special skyprobe tessalation where the pixel size is the size of a skyprobe pixel, and the 'skycell' is the size of a projection cell. The first investigation (why are there big chunks missing) I answered by writing a tool (in ipptrunk/tools/heather/stack_inspector.pl) - give it a stack_id, it will make binned fits of the stack and the input warps. You can easily open up the stack and all the inputs, which I did with skycell 1315 (g band). What I found out is that everything is covered multiple times. The skycell should be completely filled with pixels.

At this point, I committed a sin: I downloaded IRAF to see if it could stack better. I did the dumbest stack possible (I fed it 200 images and said, go make a stack), and I stacked stack_id 33. IRAF complied and gave me a stack ~ 1 minute later, completely filled, very deep, very nice. Note that the astrometry is crappy for some of the warps, and it prints through on both ipp and iraf.

(left IRAF right IPP)

Next was to play with gpc1 images with IRAF for SAS. I downloaded a random g band skycell ( skycell.1495.014 ), grabbed the inputs, and stacked. I discovered 2 things:

1. Iraf doesn't like compressed files
2. Iraf doesn't like different bzero and bscale.

Ken played with IRAF, and got a better IRAF stack:

Apr  1 12:15: IMCOMBINE
  combine = median, scale = none, zero = none, weight = none
  reject = avsigclip, mclip = yes, nkeep = 1
  lsigma = 3., hsigma = 2.5
  lthreshold = -1000., hthreshold = 1000.
  blank = 0.
                Images 
  /home/heather/sas_v8_stacks/o5745g0409o.356779.wrp.453852.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5745g0421o.356794.wrp.453876.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5745g0432o.356807.wrp.453888.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5745g0442o.356814.wrp.453903.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5746g0418o.357453.wrp.453912.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5746g0428o.357464.wrp.453927.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5746g0441o.357477.wrp.453937.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5746g0450o.357486.wrp.453947.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5748g0357o.358780.wrp.453989.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5748g0368o.358778.wrp.454010.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5748g0379o.358756.wrp.453971.skycell.1495.014.fun.fits
  /home/heather/sas_v8_stacks/o5748g0389o.358765.wrp.453992.skycell.1495.014.fun.fits

  Output image = sastestkenmed2, ncombine = 12

Here is a comparison of IRAF to IPP for gpc1

(left IRAF right IPP)

Now for the big question: how to make ppStack look more like IRAF?