Holmes and I worked out the full computational formalism for ubercalibration, working with observed stellar fluxes and an unknown (but smooth) flat-field map. We are not taking the log and linearizing around that; we are doing the bilinear problem, where we estimate true fluxes at fixed flat-field map, and then the flat-field map at fixed true fluxes, iterate to convergence. Holmes is writing documentation (the paper) in parallel with the code so that we have a clear correspondence between the one and the other.
In many ways, this makes me wonder if we should redefine 'standard fields' for flux calibration, since in essence you could replace 'flat field' with 'flux calibrate' without any real changes (I'm only referring to imaging here, once you add a slit, shit get's weird). One has to wonder if you take your idea to the extreme if we shouldn't simply be observing globular clusters or Andromeda to get fluxing right, not some sparse Landolt field......
ReplyDelete