2MASS V3 Prophot PSF Photometry

2MAPPS V3 Prophot PSF Problems - Variance Floor Tests

  • Nonlinearity as a function of magnitude has been observed between the PROPHOT psf magnitude minus the aperature photometry magnitude for the PROPHOT V3 psfs. In addition, the PROPHOT magnitude uncertainties have been shown to be systematically high for high snr sources. It has been discovered experimentally that adding a floor to the PSF variance map drops the chi^2 significantly, and scaling the variance map down to renormalize the high snr chi^2 reduces the psf sigmas by approximately the square root of the scale factor.

    Tests have been run with a range of variance map floor values, rescaling the variance map to renormalize the chi^2 on scan 980811s_065, a scan with shape values near 0.95. For the psfs used for this scan, the variance map scale factors required for renormalization are shown vs the log of the vmfloor used in this plot:

    These scale factors are imperfectly determined, being systematically high in J & H at the higher values of vmfloor.

    The resulting psf sigmas after renormalization vs the log of vmfloor are:

    In the above plot, the anomalously high points in H band are due to psfs 9325 and 9505. These psfs also show anomalously high variance maps.

    The high snr dispersion of the psf photometry minus aperture photometry:

    Runs were made with a special test version of PROPHOT compiled by Bill Wheaton which can apply a minimum floor value to the input variance maps, and then apply a scaling factor to the variance map. The constants are read from NAMELIST input.

    The values above were calculated from the PROPHOT output files (????.ps) with the following cuts:

  • 9.0< mag < 12.0
  • |psf-ap| < 0.2
  • chi^2 < 2.0

    The following conclusions are drawn:

  • A suitable combination of variance map floor and scaling can adjust the PSF sigmas to values reflecting the measured dispersion.
  • The rescaling required to normalize the high snr chi^2 looks to be (to first order) a function of the ABSOLUTE value of the vmfloor.
  • The high snr photometric dispersion is relatively insensitive to the shape of the variance map, up to floor values of 10^(-4.5), or about 10% of the 99 percentile of the variance map.
  • Chi^2 at high snr is sensitive to the variance map floor, dropping rapidly as the floor is increased.
  • The psf sigmas (to a first order) scale with the square root of the variance map, but are insensitive to the floor.
  • Anomolous psfsigmas are related to particular psfs and probably the psf variance maps.

    Tests are underway for a scan with shape values near 1.1, as the dependance vs shape is unknown.

    For detailed plots and additional information, see Bill Wheaton's V3 PSF Test Summary

    No rational basis for these results has been put forth at this time.

    This page last updated on Jul 11, 2001.

    Gene Kopan - gene at ipac.caltech.edu