Year 1 star flats

(6 Sep 2013 Gary Bernstein)

Star flat observations (and dome flats) were taken during the engineering nights 15 Aug and 29 Aug 2013, for purposes of generating a star flat / illumination frame to be installed in the DESDM pipeline for the start of Year 1 processing. This page describes the comparison of the dome and stellar response of these two dates to each other and to the SV-season response.

Terminology to keep clear:
  • Dome flat is simply the (relative) counts across the array in the dome flats for a given filter.
  • Star flat is the correction (in mags) that must be applied, as a function of array position, to instrumental magnitudes from dome-flattened images, in order to make the mag of a star be independent of where it is imaged on the array. It is the ratio of response to focussed starlight vs diffuse dome light, and is basically a map of stray (unfocussed) light falling on the array.
  • Sky response is the product of the dome and star flats, which gives the response of a given pixel to focussed light from the sky. It is ultimately the thing we care about.

Conclusion: The Aug 15 dome and sky response patterns are both different at >2% level (p-p) from the SV season flats. But this change appears largely reversed on the Aug 29 flats, suggesting that something was different on 15 August and we should not use those flats. The dome response on Aug 29 shows up to +-1% changes from January SV data, however the sky = dome*star flat response is very similar to the Jan/Feb data, with only ~2 mmag RMS difference. This suggests that the dome illumination is a little different than last season, but the response to the sky is (as expected) pretty close. Domes from 5 Sep are the same as those from 29 Aug to within 0.5 mmag RMS, suggesting that the 29 Aug configuration is indeed the current stable response of the camera.

My recommendation would therefore be that we adopt the Aug 29 star flats for the initial processing of Year 1 data. The seasonal change in dome illumination is not a problem. We should try to figure out why the dome and sky response would have differed on 15 Aug; the PF cage was flipped to f/8 and back both before and after this run, perhaps that is an issue.

Details: I retrieved the combined dome flats from these DESDM runs per instructions from Margaret Gelman:

  • 15 Aug: 20130819150900_20130815
  • 29 Aug: 20130903172904_20130829

The star flat observations' exposure numbers, and the DESDM runs in which they were reduced, are:

  • 15 Aug: Exposures 226510-226621, DESDM runs 20130820071917_20130815, 20130823084219_20130815, 20130821081604_20130815, 20130823111122_20130815, 20130823084101_20130815
  • 29 Aug: 228544-228685, DESDM run 20130903224504_20130829. This includes u-band data which I will not address here.

First we can examine the dome flats. These three plots show the ratio (expressed as mags) of the dome flats for 15 Aug and 29 Aug to each other and to superflats made from late-Jan 2013 data (or Nov 2013, which is very similar)

You can see a substantial change (2% shift along the left and top edges of the flats) from last season to the 15 Aug flats. Then it changes back for 29 Aug. The last plot is the ratio of 29 Aug to last season, and milder changes are apparent. These milder changes are filter-dependent whereas the deviations on 15 Aug show a common pattern in all filter. The variation of individual CCDs seen in Y band was observed during SV to occur whenever the focal-plane temperatures change, and so this type of change is expected to have occurred between seasons.

I have examined the dome flats from 5 Sep: they are the same as those from 29 Aug to within 0.5 mmag RMS, suggesting that the 29 Apr configuration is indeed the current stable response of the camera.

Next we examine the sky response = product of dome and star flats. Here again are intercomparisons between 15 Aug, 29 Aug, and Feb 2013 star flat run results.

We see here that there is again a significant change from SV season to 15 Aug, but this has changed back for 29 Aug. And the most stable comparison of all is that the sky response on 29 Aug is very close to that in SV season, with just 1.5-3 mmag RMS difference (depending on filter). This is similar to the level of variation seen during SV.

It therefore appears that the 15 Aug domes and star flats have some kind of aberration from the other epochs. Since this aberration is present in both dome and sky response, it is not just a change in dome illumination or positioning of the telescope during the dome flats. Since it has a similar pattern in all filters, it is not related to positioning of the particular filters.

My recommendation would therefore be that we use the 29 Aug star flats in initial Year 1 reductions. When combined with current-epoch dome flats, they show a sky response very close to last season's, giving some confidence that this is the "correct" stable response of the instrument.