wiki:Summer2013/ImSimCoaddAnalysisNotes
Last modified 6 years ago Last modified on 06/26/2013 09:43:16 AM

Notes:

Plans:

The main goal of this Analysis is to assess the quality of the Coadd produced by pipe_tasks and CoaddPsf? in actual detection and measurement tasks. At the same time, we will attempt to tweak the configs for the following pipe_tasks to produce the best measurements possible:

  • mosaicTask
  • makeCoaddTempExp
  • assembleCoadd
  • processCoadd

Deliverables

  • Config files AND defaults in obs_lsstImsim
  • Required changes to pipe_tasks and meas_mosaic to produce analysis plots
  • Plots as described in Summer2013/Coadd Testing review

Coadd Measurements with CoaddPsf?

Third Trial: (tract 3, all patches [83:86, 55:58])

Photometric plots: Aperture mag (aperture 20) and Psf mags for Coadds

http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/apermagall_mdm.png

http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/psfmagall_mdm.png

Astrometric Plots: Comparison of coadd src positions with "real" positions from Imsim inputs

Overall scatter and scatter vs. position (calexp sources):

http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/src_abs_respos.png http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/src_abs_posdpos.png

Overall scatter and scatter vs. position (coadd sources):

http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/coaddsrc_abs_respos.png http://dls.physics.ucdavis.edu/~pgee/coadd/imsim3/coaddsrc_abs_posdpos.png

Meas Mosaic Diagnostics

These are the primary diagnostics which came out of a complete run on all 111 measMosaicData visits. These plots eliminate about 2% of the objects deemed "bad", and use a 5 sigma, 2 pass clip.

  • Overall Astrometric Scatter
  • Astrometric Scatter vs. Position
  • Photometric Delta from Reference Image vs. Magnitude

http://dls.physics.ucdavis.edu/~pgee/coadd/mmsig5/ResPosScatter.png http://dls.physics.ucdavis.edu/~pgee/coadd/mmsig5/PosDPos.png http://dls.physics.ucdavis.edu/~pgee/coadd/mmsig5/MdM.png

Deblender

After running trials with and without the deblender, it was obvious that the deblender vastly improved the detection catalog. These are the deblender setting being used:

config.deblend.psf_chisq_2=1.5
config.deblend.psf_chisq_1=1.5
config.deblend.psf_chisq_2b=1.5
config.deblend.maxNumberOfPeaks=20