Here is a comparison of the relative response of the array to stars across the FOV during the three star flat trials: Nov 20 & 21 2012, Dec 12 2012, Feb 22 2013.
- All plots are showing the ratio of response in one epoch to response in another. Units are magnitudes. All of the plots below have color scale spanning +-3 mmag. All plots have had the best-fit constants and linear slopes across the field of view removed, as these can be produced by changes in atmospheric extinction.
- The dome flat is the response to the dome lamps. These come from the DESDM supercal runs done for SVA1. All CCDs are given a common normalization factor, i.e. I undo the individual CCD-to-CCD normalizations done in DESDM using the SCALMEAN values in the superflat headers.
- The star flat is adjustment that must be made to stars to get correct magnitudes after applying dome flats to the images. It is essentially the fraction of the dome illumination that is caused by scattered light instead of focused light.
- The stellar response is the product of star flat and dome flat: it measures the response of the array to focused light from celestial objects.
- The plots below use MAG_AUTO to derive the star flat. Results from using fixed 8" diameter aperture are insignificantly different.
- The stellar response is more stable than the dome flats! In other words, using a different dome flat each night or week is actually degrading the photometric stability because the domes are changing for reasons (maybe light in the dome, etc.) that do not actually change the camera's sensitivity to stars.
- The stellar response changes by <1 mmag RMS in riz and ~1.5 mmag RMS in gY between epochs. The biggest changes are about +-3 mmag peak-to-peak. [There are larger changes on the flaky amplifier of CCD #31].
- The dome flats had a change in mid-Feb of about +-10 mmag. and have about +-3 mmag changes on other epochs. I suspect that close examination would find that there are single nights that have discrepant dome flats at ~1% level due to some change in procedure. Worth further investigation!
Dome flat variations¶
Here are the changes in dome flats between 5 different epochs of superflats produced by DESDM for SVA1. These plots show the differences between adjacent pairs of these flats (note I have not looked at all the epochs produced by DESDM):
What you can see below is that
- in griz, the 1st, 2nd, and 4th epochs are very similar - ~0.5 mmag RMS change between them. But there is a common pattern that appears in the 3rd of these epochs (and then disappears) at +
3 mmag peak to peak. Then in the last epoch there are +10 mmag patterns that show up.
- in Y band there are stronger changes from epoch to epoch after the first two. Not sure what these are, but it's a different pattern and mechanism from the changes seen in griz. Maybe temperature changes in the devices have an effect here?
Star/sky response variations¶
This first plot shows the change between the Nov and Dec star flats. The dome flats had negligible change between these two (first plot above) so this is also the change in response to the sky. The RMS variation is <1.5 mmag, except in g band where I believe the bad amplifier on CCD #31 is inflating the statistics. There are some coherent patterns to these changes but they are small.
Next are two plots: the first is the change in the star flat between Dec and Jan, and we see larger changes than the above case. But in fact the dome flats used in Feb are different from those in Dec. The second plot shows that when we multiply star flats times dome flat to get real stellar response, the camera is in fact quite stable between the two epochs, again with changes of <1 mmag in riz and somewhat larger in gY.
A movie of Y band dome changes¶
The movie file attached below shows the Y band dome flats of many mornings/evenings relative to a template for each "epoch". There are 4 epochs, between each is a temperature excursion of the focal plane that changes the chip-to-chip normalizations of the domes by up to 1%. This movie shows that changes aside from these temperature-induced shifts is small - generally < 1mmag RMS - but there are significant patterns and changes.
Addendum 8 Sep 2013 (Gary B.):
There was a 5K focal plane temperature excursion on the evening of Sep 6 (after that evening's flats). Below is a graph of shifts in the means of each CCD in the Y band flat fields (means provided by Ricardo Covarrubias). Vs CCD number, I plot the ratio of Sep 7 to Sep 5 flats as the filled symbols. The ratio of mean flats is normalized to the ratio for CCD #29, to take out any illumination-level shifts. Changes up to +-0.002 are seen, and are common to A and B amps of each CCD.
The x symbols show similar ratio for Sep 3 / Sep 5, as a control, since temperatures were constant in this interval. Changes are 4x lower, +-0.0005 or less.
Conclude that 5 K temperature excursions might cause 2 mmag shifts in some CCDs dome levels. If past experience is a guide, sky response will have changed by less.
A 0.5 K glitch on 18 Jan 2013 did not cause detectable changes in Y domes. Also there is no evidence of changes in Y domes when the monsoon elex are power-cycled, as long as there is no temperature glitch.
!Y flat changes 7 Sep 2013.png!