"Ratio of dome flat to twilight flight is constant to <1% RMS after high-pass filtering with SExtractor sky-subtraction algorithm. Similarly for night-sky flats in g,r,i bands (where fringing should be negligible), and for DECal flats spanning each filter band (to the accuracy allowed by DECal S/N). [This is a test of whether the small-scale responsivity variations are properly traced by dome flats, and whether pupil ghosts and other additive sources of contamination are smooth enough to be removed by sky subtraction algorithms. Failure indicates that better pupil-ghost removal, etc., must be developed, and/or sources of scattered light mitigated. Also might change the requirements for daily calibration images during main Survey.]"
Note that this is a test of small-scale trends. The twilight sky typically shows gradients on scales of 1 deg and are generally not ideal for use in de-trending images from wide-field imagers (see, e.g., Chromey and Hasselbacher 1996).
Cal-R1 writeup covers much of this procedure as well.
- Afternoon (for dome flats) and early twilight test (for twilight flats).
- Clear (but not necessarily photometric) conditions for the twilight flat observations
- Also, preferably with little or no moon for the twilight flat observations.
- Nighttime and clear (but not necessarily photometric) conditions for deep sky flats.
- Obtain at least 10 high-S/N (but not saturated) dome flats in each filter during an afternoon. (This is part of standard operations, so should be easily accomplished -- many times over).
- Obtain at least 10 high-S/N (but not saturated) twilight flats in each filter (This is not part of standard operations, and obtaining good twilight flats during the short window of time after the sun has set but before the sky gets too dark is a non-trivial task. It will probably take several nights before we obtain a full set of 10 or more high-S/N (but not saturated) twilight flats in each filter.)
- If possible, obtain at least 10 deep sky flats in each filter in an "empty" part of the sky (maybe in the CDFS field? Need to dither to median-filter out stars).
- Perform overscan-removal and bias-subtraction of the individual flats.
- Median-combine each set of dome flats to create master dome flats in each filter.
- Median-combine each set of twilight flats to create master twilight flats in each filter.
- High-pass filter each master dome flat and each master twilight flat with the SExtractor sky-subtraction algorithm.
- Divide the high-pass-filtered master dome flat for each filter by the high-pass-filtered master twilight flat in that filter.
- Measure RMS of small regions (say 100-pixel by 100-pixel boxes) of the ratioed image in each filter.
- Do the same analysis replacing deep sky flats for the twilight flats.
Analysis relevant to this requirement is available on this wiki page: Flat Fields
The RMS of 100-pixel by 100-pixel boxes of the ratioed image in each filter must be less than 1%.
If the requirement is not me, better pupil-ghost removal, etc., must be developed, and/or sources of scattered light mitigated. It might also be necessary to change the requirements for daily calibration images during main Survey.