Notes

9/16/14 Notes:

• in tractor/sdss.py, Brenton implemented a photon count computation that isn't included in the github tractor repo. This is in the get tractor image dr8 function. We need to:
  • Verify that this correctly computes photon counts that make up our observations
  • Decide to put these computations within a single, DESI-MCMC based tractor repo, or to make the calculations outside given the existing tractor.

10/3/14 Notes:

• Sources:
  • The catalog sources in the cat-*-.fits files that dustin is providing give U = (ra, dec) and flux in all five bands (in nanomaggies). flux in counts = (flux in nanomggy / header["CALIB"] + header["SKY"]) * header["GAIN"]
  • This flux value appears to incorporate information from the Temperature, distance, luminosity, etc. of the star, and just models how many photons you expect to see given this source and some background noise.
• Celeste model updates:
  • Not really modeling or inferring a posterior over b_s.
  • Photons_expected compute the number of photons you expect to see during a 54 second exposure with a 1.25^2 meters squared lens given temperature, luminosity, distance, and band. This number is then spread out by the point spread function over the entire image (otherwise, it would hit like one pixel and be tough to infer properties of it).
• Epsilon_n is being inferred by jeff for the individual images. It looks a lot like the 'SKY'*'GAIN' for me.