Project

General

Profile

Cal-R1

A photometric model (i.e. flat fields and sky-­‐subtraction methodology) is in place which produces relative magnitudes between bright stars that are reproducible to <0.02 mag RMS on different exposures taken (a) at different times during a cloudless night (b) on different cloudless nights up to 3 weeks apart. [This is essentially a test of whether we can construct a flat field that describes system response to light and remains stable. DECAM TG-­‐12. DECam TO-­‐8 covers pupil ghosts, TD-­‐10 is QE gradients]

Cal-G4 is same but cuts the tolerance in half to 0.01 mag RMS.

Prerequisites

  • Observation of rich stellar density field during photometric conditions for model construction. Dither pattern?
  • Use overlapping DES survey data to test model. Need repeated observations separated by 3 weeks.
  • Need single epoch catalogs with astrometric solution <1" (and pixel-area photometric correction image)
  • Bad-pixel maps

Procedure

  • During Commissioning or early SV
    • Use observation of dithered, dense stellar field to derive large scale variations of the stellar photometric response
  • During Science Verification
    • Test model on DES survey data with repeated observations on same night or across multiple nights.
    • Repeat derivation of star flat with dithered star field, check for variation.

In detail: the photometric model will initially be that small-scale response is accurately given by the flats and is independent of wavelength across a given band (except for very narrow-band fringing, which will wash out for any modest bandwidth of illumination). Large-scale photometric flat field will potentially differ from the flats (due to scattered light, ghosts, dome illumination, etc.) and will need to be derived from stellar-field observations. This photometric flat field is the thing that should be stable, whereas large-scale behavior of flat fields may vary with time.

*Analysis steps:
  • Create high-S/N robust combined dome flats, twilight flats, and DECal flats at blue/red ends of each filter.
  • For each filter: Take ratio of twilights and DECal flats to the dome flats. Apply high-pass filter to these images.
  • Bin flat-ratio images on scales from 1 to ~20 pixels. Note any locations where small-scale flat structure deviates by >1% (TBD). Deviations at bad pixels are not a problem, but >1% deviations at valid pixels may cause problems for our flat-fielding paradigm.
  • Next step is to create star flat by deriving large-scale multiplicative correction to the dome flat. Take the dithered star-field images and flatten them with domes.
  • Get stellar instrumental mags for all dithered star-field images. Apply pixel-area corrections.
  • Use ReMatchPhoto or other solver to derive the star-flat correction vs array position that forces agreement among dithered stellar magnitudes. Will need to include a few reference ("standard") star magnitudes to break the linear-gradient degeneracy in solution.
  • Report RMS photometric disagreement in the dithered star-field images. Should be <0.02 mag. Plot residuals vs array position to check for anomalies.
  • Apply dome flat, star-flat correction, and pixel-area corrections to generate stellar photometry for "standard mode" DES observations of other fields, e.g. Minisurvey
  • Use ReMatchPhoto to generate zeropoints for all the Minisurvey exposures.
  • Report RMS photometric deviations for stars in Minisurvey using the starflat. Should be <0.02 mag.
*Temporal stability check:
  • Take 2nd set of flats, dithered star-field images >1 week after first.
  • Measure ratio of new flats to old: after high-pass filtering, changes should be <<1% (TBD). Maybe features from dust motes on filter, etc. How to handle these?
  • Create new star flat. Ratio of new (star flat) x (dome flat) to old one should be constant to <1% (TBD). How would we handle dust motes etc given that we do not expect large-scale illumination of flats to be stable over time?

Verdict

  • Relative photometry of bright stars < 0.02 mag RMS

Consequences

If the requirement is not met we may need to refine model or reexamine flat-fielding and sky subtraction methods.