Verification Datasets meeting
Attendees: Colin, Simon, Hsin-Fang, David, Zeljko
Regrets: Angelo
Colin:
- manually classifying diff im sources, illustrative
- good idea of how many false positives, how many are real how many are junk
- a lot of them look real, lots of rectangles, 1-2 columns of pixels have a short little bar
- maybe 50% are real, still need to add all of them up, a few hundred false positives per square degree
- just on moving object, excludes cases where there is an object detected in both visits
Simon:
- no update on twinkles
- working on ISR this week, xtalk, linearity and fringing need to brought in from HSC
- for example, HSC talk has hard-coded values, need to make it general
Zeljko:
- working on DPDD
Hsin-Fang:
- no updates
- spent some time looking through Jim’s Level 2 flowchart
David:
https://confluence.lsstcorp.org/display/SQRE/COSMOS+DECam+data+reduction
- trying to dig down into why the cosmos photometric scatter of individual objects over multiple visits
is larger than expected - looked at relative fluxmag0 variations versus ccdnum, there are persistent chip-to-chip variations but they are only at the ~1-2% so not enough to explain the scatter of individual object photometry over multiple visits
- Simon suggested plotting fluxmag with radius of focal plane, maybe vignetting
- Simon suggested cutting on flags, cut on saturation flags, cut on extendedness parameter, something close to 1
- Colin also suggested, psf-aperture also does a pretty good job
- David, maybe throughput changes spatially with the filter
- will dig into this some more
No verification datasets meeting next week, David will be in Pasadena. We will pick up again on April 27th.