The WCS being placed into image headers by SISPI is accurate to <1” across entire science array (plus any errors in telescope axis pointing received from TCS). [will have to include corrections for lateral color and differential refraction].


  1. Imager is functional and focussed.
  2. Overscan/bias/dome flattening can be done and catalog with stellar centroid pixel coords measured.
  3. Need to write SISPI code that takes nominal astrometric solution, tweaks it for lateral color depending on which filter is in, and for differential atmospheric refraction given airmass and HA/Dec of observation.


During Commissioning: set up the nominal solutions

  1. Take a single exposure in each filter of a field with many astrometric reference stars (e.g. Stripe 82) & produce stellar catalogs.
  2. Use SCAMP or some other code to produce initial astrometric solution in TNX format for single exposures to match reference catalog, and the effect of lateral color in each DES filter. This code needs to be able to work from "cold start," i.e. there is nearly no information about the distortions before using this code. May need to force solutions for all CCDs to have common CRVAL values.
  3. Give the nominal solution files to SISPI and make them the ones used to write new headers.

During Commissioning and/or SV: validate the solutions.

  1. For a subsample of the images at a range of airmasses and filters, match First-Cut stellar position to USNO or SDSS or 2MASS stellar catalogs. May require special software since need to create matches with >1" errors in WCS, and because TCS may not be reporting correct pointing if lookup tables are not good yet.
  2. After adjusting CRVAL for possible mis-reported pointing by TCS, calculate each CCD's sigma-clipped mean and RMS displacement of positions using SISPI header WCS relative to the reference catalog. Need to code that does this from the matches to reference catalogs.

Note that the code that does the differentials can also report the overall shift of the camera axis relative to the value given by TCS, which can be used to refine the pointing lookup tables.


Each CCD of each exposure should have mean positional error <1" in each coordinate, (and RMS about this mean should be <<1" RMS too).


If the test is failing, we will need software to plot the residual errors vs position on the array. Then we will need to diagnose whether errors are some affine function (e.g. we are getting refraction incorrect), some radial function (distortion / lateral color to blame) or something more complex (nominal solution is changing). Would need to fix either the input nominal solutions or amend the SISPI code until it works out correctly.