Verification Datasets Meeting 2016-06-01 Minutes

Verification Datasets Telecon

Attendees: Colin, Simon, David


  • LDM-151 has taken over, has been distracting from verification datasets

  • hasn’t touched data for a while

  • HSC data used to search for Kuiper belt object near pluto, 1/4 terabyte of HSC data willing
    to give us, run the stack on all of that

    • tricky part is that it’s all in the bulge, worst-case scenario for image differencing
    • will have to buld templates from the data itself
  • Simon, just need to know the psf in difference image, figure out kernel needed to match one “scene” to another “scene” using a basis set for the transfer kernel, ignores the individual visit PSFs (only used for detection) directly make a matching kernel based on two sets of data instead of two (psf) models, less steps

  • verification cluster and petabyte at disk at NCSA, Colin could run HSC

  • David: was planning on originally running DES data on the new hardware, but there’s no photometric reference catalog

  • Some discussion of jointcal vs. ubercal

  • jointcal currently only can be run on a patch

  • ubercal not in calibration WBS, it’s in it’s own 02C.07

    • astrometric calibration/solution (UW)
    • photometric calibration/solution (UW)
    • calibration (Princeton), more absolute calibration
  • solving small-scale problem now with jointcal, not sure about global one, will use Gaia to inform

  • Peter Yoachim worked on photometric problem and found that we can do it, need edges to meet,
    can get it to the level that we need

  • Zeljko and Robert discussing the calibration plan, should eventually go into LDM-151, but no section for
    this right now

  • maybe run northern part of DES through stack using PanSTARRS as reference catalog, maybe ask Paul about getting a reference catalog


  • twinkles, did some difference images

  • template is a coadd, found it difficult, don’t have database of quality parameters
    no way to select good images based on any heuristics

  • PSF box size varies depending on seeing

  • psf-matched coadd, problems because it only uses visits with the same PSF-size, bug

  • some residuals in difference image in one region, maybe a PSF problem, still using PCA, will try PSFEx

  • coadd is from ~5 visits from the same night

  • still not sure what data should go into the coadd

  • David: what’s the plan for artificial stars in the stack, it’s in processCcd, right?, but also need it in coadd/forcedPhot

  • Simon: only on HSC in processCcd, not ported over yet, should add to coadd/forcedPhot but not sure when that will happen

  • Colin: “sky sources” are currently implemented


  • looked at photometric residuals in the COSMOS data, individual phot measurements per object (per band) minus the median measurement, plotted all of those on the focal plane where they were detected (so multiple values per object). no obvious spatial trends (i.e. near the edges), but some chips look worse than others
  • would like to make figure similar to what Bob Armstrong made with residual sized (object measured size - PSF size at that position), showed rings in HSC data
  • Simon suggested looking at PSF - aperture mags
  • use alpha!=1 so they are transparent a bit you can see through them, poor-man’s density diagram
  • will start rerunning processCcd on COSMOS data with PSFEx

David will be traveling in June, verification datasets telecons are canceled in June. Will start up again in early July.