Flux-dependent PSF

Work plan ideas

(GaryB 25 April 2013)

The goal is to
  1. have a model that can explain the flux dependence of PSFs, in 2 dimensions, to the measurable accuracy.
  2. Develop a means to correct for this when creating a PSF model from bright stars.
  3. Use the model to check that this is not influencing our observations of faint galaxies at significant level.

Physical model

Two effects have been put forth as possible origin of this. In the first, the collected charge reduces the voltage from backside to the well, increasing the diffusion. So the model would have charge diffusing isotropically from bright pixels. How would this die off with distance from a pixel with a lot of charge? How should effect scale with amount of collected charge?

Second possibility is transverse electric fields due to collected charge which causes drift away from bright pixels. This would predict (I think) 1/r^2 dropoff of charge shift with distance from the bright pixel, and also linear dependence of charge shift on the ADU in a pixel.

On-sky analyses

These are the things we can do:

  1. Start with 1-parameter description of PSF, namely a half-light radius.
    1. See if all DECam CCDs have identical amplitude of d(radius)/d(peak ADU) at same seeing. If so, can combine all CCDs in tests that follow to improve S/N, and can also trust that devices in SiDet will be similar to those in DECam.
    2. At fixed seeing, more careful map of radius vs peak ADU. What is function form of the spread?
    3. Look at d(radius)/d(peak) for variety of seeing conditions. Compare with models.
  2. 1d PSF study: look at azimuthally averaged PSFs
    1. Compare bright to faint radial profiles for variety of seeing conditions
    2. See if one or both models predict observed behavior in 1d.
  3. 2d PSF study: look at PSFex-derived 2d PSFs for bright vs faint stars.
    1. See if model(s) predict amplitude dependence.

Lab analyses

  1. Using SiDet projector spot:
    1. Size vs flux for fixed spot geometry. Is effect on half-light radius linear with flux?
    2. Do models predict dependence of full 2d spot on flux?
    3. Do models still work when the spot is defocused?
    4. Does level of background illumination change the brighter/fatter relation?
  1. Using Spots-o-matic
    1. Can we produce a nearly-delta-function illumination to provide cleaner test of the effect?