IPP Progress Report for the week 2012.01.09 - 2012.01.15

(Up to IPP Progress Reports)

Eugene Magnier

  • working on integrating ubercal + DVO. I added a flat-correction table to DVO (coded I/O APIs) and added code to convert the ubercal format flat-correction information into the format needed by DVO. I added this information to the 'setphot' program, which also is able to load the ubercal zero points and apply them to the specific measurements. I have updated relphot to refer to the flat-correction, but need to verify that it is working correctly.
  • Doug F. and John T. convinced me that there could be a (possibly significant) faint-end bias due to the way I determine the Poisson weighting. I am using the value of the variance as determined from the image (ie, counts + read noise2). They make the case that in the low count regime, it is better to use the model as a proxy for the counts, as that is closer to the exact value claimed by a maximum likelihood analysis. We agreed (and, I think, demonstrated in the real data) that the 'constant' weighting scheme is fundamentally equivalent to the model-based variance in the faint limit, but has worse noise properties in the bright limit. I spent some time coding up this 'model-based variance' in psphot. The code is in place and runs more or less correctly, but I need to run a more complete demonstration set to show that it has the desired (or any) effect...
  • finalized WISE DVO ingest code and wrote up documentation on the ingest transformations.
  • discussions with Chris about the background issues.

Serge Chastel

  • Czar on Wed, Thu, and Fri
  • Out on Thu morning
  • Nebulous shuffling

Heather Flewelling

  • attended aas and manned booth
  • grabbed smf files for monet

Roy Henderson

  • Sick 1 day
  • Support:
    • lots of conversations and emails with Kent Wood, including help with code
    • discussions with Robert Senger about photo-Zs
    • responded to questions from Eric Morgenson regarding stack magnitudes

  • PSVO:
    • now fully platform independent, so works on Windows
    • fixed bug when logging-in as new user, using last user's Derby Db
    • fixed bug where changing user would stop Derby network server
  • Documentation:
    • updated PSVO documentation
    • updated PSPS schema changes page
  • finally removed reverts on/off capability from czartool

Mark Huber

  • Czar for Monday and Tuesday, occasionally kicking LAP throughout week
  • Worked with Ken Smith to get the TSS-event/name server test setup here
  • MD.GR continued: released full SMF camera files for available MD fields (MD04,06,07,10), full nightly stack reprocessing for MD06 running
  • Worked with Kent Wood looking into some images of Fermi sources
  • Continued looking into making a sample of MD04 JPEGs with WCS for use with PSVO+Aladin but have run across multiple troubles with image sizes
  • A little time spent on collecting data samples for testing for ppsub and the choice of convolution direction

Bill Sweeney

  • Reworked some of the psLib thread functions to work correctly in certain error conditions. This repairs a long standing bug where jobs would fault with assertion failures
  • started working on a database table to describe the IPP skycell tessellation. Surprisingly it is currently non-trivial to answer the question "Where is skycell.NNN.MMMM in the sky?"
  • Investigated the high number of multi filter staticsky failures. It turns out we need to tweak a recipe value a bit
  • Took two days off in compensation for New Years and Christmas holidays on which he worked.

Chris Waters

  • Removed magic requirement from LAP processing. This has allowed the processing rate to increase somewhat, as runs can now complete without waiting for destreaking to finish.
  • Fixed nightly science cleanup bug that was introduced with the end of magic. We previously only cleaned destreaked data, but as no data is destreaked now, everything was left on disk. This was fixed, and the backlog of old data was manually sent to cleanup.
  • Offnight diffs: checked the database for potential offnight diff pairs, and discovered that we should have created far more of these than we currently are. This appears to be an issue with difftool not allowing multiple diffs to be created with the same input warps. This will require some development work in the nightly science script to fix.
  • Background: Tracked down the bulk of the background issues observed in the stacks to cell-to-cell discontinuities that distort the background model. The improved gradient restoration implemented last year in the row-to-row PATTERN.ROW correction seems to have improved this, but not to the point of completely resolving the issue. The proposed solution is to artificially offset each cell to ensure that adjacent cells have similar boundary values, and then fit the smooth background that remains. Working on coding this correction.