Relative Astrometry in SVA1 Finalcut

I looked at the relative astrometry within SVA1 Finalcut images, and am investigating the impact on photometry.


The Issue

The Match Structures

(Will create links)

For each tile, match all bright (i<21) stars to check astrometry and photometry. The match structure has a light curve for each star, using the GCM zeropoints.

Then, I can calculate statistics for each CCD image in the tile: for example, what is the average offset in mag_psf for each star in an image relative to its average mag_psf? And what is the average offset in RA and Dec for each star in an image relative to its average position?

Fortunately, the mag_psf offsets look fine: locally, GCM is working as expected. But for astrometry...

Sample Tiles: DES0453-4748 and DES0434-5914

Looking at individual CCDs, I've plotted the total offset:

We have offsets (total in RA and Dec) of over 150 milliarcseconds...this is 0.6 pixels! Yowch! Thus, the coadds are going to be significantly smeared out, and this is going to exacerbate the problems of psf discontinuities.

Next, are these offsets correlated systematically with ccd position in the different bands?

The answer is yes!

Impact on Photometry


Source of the problem and Y1A1 mitigation

I think there are 3 sources to the problem...

  1. The .ahead focal plane distortion pattern used depends on band (see Gary's plots...need to find link.) And we used the i-band as a basis in SVA1, and this just isn't good enough for the other bands given the degrees of freedom for the fit.
  2. The .ahead in i band shifted when we had a warmup/cooldown due to ccds shifting.
  3. Finally, we are only doing one astrometry run relative to UCAC4 without enough degrees of freedom. We need the relative astrometry at the coadd step (which is planned for Y1A1). Note that we need to get these offsets into the multiepoch fits!