IPP Progress Report for the week 2012.02.26 - 2012.03.04

(Up to IPP Progress Reports)

Eugene Magnier

Lots of work on relphot to ensure we understand the object averages and the non-ubercal images

  • finalized the outlier rejection and variability flagging in relphot
  • defined per-photcode poor and bad detection masks
  • added per-secfilt flags describing the photometry averaging
  • updated the DVO on-disk schema to include the above, and added a feature to relphot to migrate the schema
  • ran a large-scale test of the analysis to understand the quality. Comparison of the mean magnitudes with the synthetic magnitudes shows that non-ubercal'ed images are not straying from their mean

Serge Chastel

Heather Flewelling

  • sick 1.5 days
  • czar 1 day
  • jtrp stacks
  • ssp
    • transferred to ipp004 from atrc
    • ingested all but last few days of feb into ssp db
    • registered all but a few thousand (they have errors, need to investigate)
  • liason
    • grabbed files for tom dixon - need to transfer to his disk
    • tossed files for larry on ifaps1
    • investigated (lack of) space usage on ifaps1
    • finding "cool" images for Conrad - discovered dsa was down (steph fixed that)
  • ipp005
    • continuing to rsync/clean up old dvodbs on ipp005 to clean up space
  • dvodb
    • threepi paused because of gene (waiting for his dvodb)
    • lap paused because of lack of space on ipp005
    • sas being built
      • cam stage ingested
      • running staticsky stage (when that finishes, will add staticsky to dvodb)
  • isp
    • gave niall an isp dvodb, he has been looking into photometry/astrometry of it
    • still in process of cleaning up raw isp nebulous mess

Roy Henderson

  • 2-day week (3 days of vacation)
  • some big code changes to ippToPsps so that different clients load by stripe of RA:
    • new database tables where each client 'registers' the stripe it is working on
    • code changes to support database changes
    • introduced more cascading delete constraints so that dying clients remove all foreign-key related rows in other tables
  • some work on cleaning up PSPS public wiki
  • started new 'configmaker' program to make ippToPsps config creation and alteration easier
  • continued to load and monitor LAP/3PI data

Mark Huber

  • ppStack testing
    • using SYS.ERR=0 for stacks to avoid injecting a largely scaled model flux profile into the variance/weights results in ~no PSF growth for simulated images, <10% with 20 MD04 inputs (5.3->5.8), ~10% for 100 inputs (6.9->7.8) and ~40% for 400 inputs (7.1->9.8). Started looking into the driving factors in the very large N input case.
    • started looking into the setting of the target PSF for the convolved stacks and for not using roughly the largest PSF from the input images.

Bill Sweeney

  • Spent the week debugging psphotStack problems and building tools to help that effort.
    • Made some refinements to dlapstacks (display LAP stacks) a tool for finding and displaying LAP stack images that contain a point on the sky. Did some virtual sky gazing looking at various NGC galaxies. Some of the larger ones really show the effect of bad background correction. The newer stacks are improved in this regard.
    • The primary problem with psphotStack is that it often tries to consume infinite memory. Since psphot is quite complex I spent two full days reading code and stepping through the program. For the example case (sky_id 138369) the memory is being used by images created for each of the sources. With some effort discovered that this happens because a bit of code is calculating an unreasonably large size for the source. This caused the kron code to allocate very large images - many larger than the input image! Correcting the calculation stopped the explosion in this case. It is quite likely this is the general cause of the memory over consumption.
  • Debugged a distribution problem that NCU ran into.

Chris Waters

  • SAS: finished rerunning SAS through the current tag code, and have talked to Heather and Roy about getting this into DVO and PSPS.
  • MOPS: Tracked down various issues that were causing MOPS trouble, the most important of which seems to be that MOPS was not respecting the crosstalk ghost detection flag.
  • Background: Looked into remaining gradients identified by Nigel. These are small in the input image (1-2DN), but after stacking in the footprint, they become significant. The row-to-row variations appear to be variations on top of this gradient, and because of that, the bright variations trigger false detections. This has led to looking into the dark model, as it appears that our current dark is introducing some of these gradients.