5 Simulation test of different fwhmr50whisker measures:

I generated simulated PSF from both bivariate Gaussian profile and Moffat profile and then run the various fwhm/whisker/r50 measures. In the simulation, I generated 500 PSF stars for each profile and assume the stars art 16.5 mag for an 100 second exposure. My tentative summary based on these analysis are:

  • For whisker, we may want to use the weighted moments, as it gives more robust results. There could be a bit off from the true value for moffat like profiles, but this offset is liear.
  • We should use R50 instead of FWHM and the results from sextractor on R50 is pretty consistent. Or alternatively, we can use the weighed moments with a correction:
    • fwhm_input = 0.955* fwhm_weighted -0.0898
  • In general, the weighted moments (use 2 pix kernel weight) is more robust than the other methods.

The following are more details about the comparison results:

  • For Gaussian PSFs:
    • weighted moments gives very consistent whisker, adaptive moments and sextractor underestimate the whisker.
    • sextractor sometimes gives very wrong whisker estimates
    • most methods give consistent fwhm, except sech2 profile fitting.

  • For Moffat PSFs:
    • all methods have some systematic non-zero whisker when measured on the zero whisker input.
    • all methods have slightly off the true fwhm, fwhm_weighted = 1.033* fwhm_input + 0.107.
    • for the r50, weighted moments gives r50 dependent results while the others works pretty consistently.