DMTN-036: The Jointcal algorithm

  • John Parejko

Latest Revision: 2017-02-16


This technote is not yet published.

This note describes jointcal’s algorithm, fitting process, and currently available models.

1   Other simlar work

1.1   ubercal (from SDSS)

1.2   ubercal (from Pan-Starrs)

1.3   meas_mosaic (from HSC)

meas_mosaic fits a 7th order Chebyshev polynomial on the focal plane, plus a zero point offset per CCD. The polynomial coefficients are written to the header of fcr-[visit]-[ccd].fits files as C_N_M values, while the zero point and its error is written as FLUXMAG0 and FLUXMAG0ERR. That calibration is applied to all of the fluxes in the catalog (converting them to magnitudes in the process).

2   Jointcal Astrometry

3   Jointcal Photometry

The only photometric model currently implemented in jointcal is a single zero point per (visit,ccd), which is the same as the current singleFrame processing fit, and thus does not improve the photometry. The new PhotoCalib object will be used to persist a spatially-variable zero point, and is built on a ~lsst.afw.math.BoundedField. The stack already has implemented ~lsst.afw.math.ChebyshevBoundedField, so we will be able to directly persist an nth-order Chebyshev polynomial and directly compare such a fit to that of meas_mosaic. The key, as with any new model, will be implementing that polynomial in a way that jointcal can fit it. The next step of a new model could be a composite model of a radial polynomial plus a zero point shift per ccd, to account for both ccd-to-ccd variations, and the overall optical system, where radial distortions likely dominate.