IPP Progress Report for the week 2010.01.18 - 2010.01.22

(Up to IPP Progress Reports)

Eugene Magnier

We have primarily been working on preparations for the demo month. We built new darks and flats - this took a couple of iterations to address some minor issues, including treating missing GAIN values in the camera as a default value of 1.0, rather than evidence of a dead cell. We built a new darkmask, and over the weekend, the whole team here worked on manually masking features missed in the automatic mask generation. In this round, we have added an automatic masking of the roll-off on the cell edges by checking for the bad flatting in both a science and sample flat-field image. We also found that we needed to mask some corner glow more than in the previous round. Also, we found that we needed to mask the 0-3 columns of ota14. We did not particularly find more hot column / glow structures than previously, and in a few cases we reduced the masking around some glow structures. Overall, the total masked fraction in the vignetted area is now 9.7%, rather than the 5.6% we had in the previous round.

https://svn.pan-starrs.ifa.hawaii.edu/trac/ipp/wiki/GPC1_Mask_20100122.

Heather Flewelling

  • set up ipp account, ipp software, ipp distribution client on psifa machine.
  • processed a few exposures (old, before 12-09) with current IPP build for Joel Welling.
  • processed a few exposures (on the 14th, 18th, 19th and 20th, to have good coverage in various bands and good tests of new and old burntool) with diff and magic.
    • looked at these to see what the major causes of streaks were. Some chips just went nuts (horizontal streakies, for example chip 32, other chips as well).
  • processed various darks, and babysat a few for Gene (made sure they kept processing, kicked them when they needed reverting).
  • pixel masking

Bill Giebink

Roy Henderson

  • now outputting stack-shaped FITS files, but as yet have no test data to populate
  • more changes to the FITS output for Conrad
  • figured out how to run addstar so that I can generate more data
  • wrote perl script to automatically generate data from Bill's large throughput DVO database
  • potential bug in DVO access functions: bogus rows in detection FITS files. Presents problems when mapping to smf detections
  • pixel masking...

Paul Price

  • Image subtraction tuning
    • Problems with adding larger radii basis functions were due to kernel and stamp size set in simtest recipe.
    • Adding a large radius basis function (e.g., 18 pixels FWHM) is important for matching images with large seeing differences (as we sometimes get), but it requires a large (e.g., 40 pixels half-width) kernel and stamp, which may make finding suitable stamps difficult in the masky GPC1 images.
    • Implemented scaling of kernel basis set, kernel size and stamp size.
    • Moved detection efficiency call earlier in psphotReadoutMinimal so that it gets done even if there are no sources detected.
    • Testing on real PS1 images.
    • Setting SYS.ERR=0.1. Since USE_WEIGHT is off in pmSubtractionCalculateEquation.c, this only affects the rejection stage: brighter stars aren't rejected as easily.
    • Now getting reasonable dual convolution subtractions. Some more detected sources in dual convolution version with new code than with old code, due to low brightness background features, presumably bad dark subtraction or streakies (additional smoothing in dual convolution makes them more visible) which should be dealt with in other places.
    • Test on 10 random 3pi diffs yielded 4 good and 4 awful subtractions, with the other two intermediate.
    • Revised recipe kernel parameters to fix the awful subtractions.
    • Got ppStack working with revised PSF-matching code, added kernel parameter scaling.
  • Fixed ppVizPSF to set CELL.[XY]BIN, CELL.[XY]PARITY in pmReadoutFake.
  • Going pixel-crazy on manual masking

Bill Sweeney

At MPG.

Chris Waters

  • Finished up major work on nightly stack automation. Needs a few tweaks to better detect faulted exposures and move on without them.
  • Tracked down crosstalk issue with nightly stacks to the GHOST mask not being respected properly. Should be resolved now.
  • Checked the newest burntool code to see if it catches the cases that seemed to be missing before. Those cases seem to have been fixed, although the impact on magic detections is still unclear.
  • Update static mask.