This page is the work area for flat field (detrending) analysis for requirement Cal-R4. The subsections below contain analysis of
- Comparison of Twilight Flats and Dome Flats
- Comparison of Dome Flats obtained on a single day
- Comparison of Dome Flats obtained with different filters
- Comparison of Dome Flats over the course of SV
- Comparison of Dome Flats and Deep Sky Flats
Comparison of Twilight Flats and Dome Flats¶
The goal of this comparison is to determine if there are differences between the twilight and dome flats. The presumption is that if there are no differences, there is no information in the twilight flats that is not also in the dome flats. This does not mean the flats contain all the information necessary to 'detrend' the data, just that the twilight flats do not add any information not already in the dome flats.
To compare the flats, I compared the noise properties of the ratio of the dome and twilight flats on three scales: single pixels, 10" square boxes (characteristic of a single object) within a 60" square region (characteristic of the smallest scales on which the photometric calibration may have to rely on the flats), and 60" square boxes over an entire amplifier on a CCD.
To measure the rms for single pixels, I measured the rms per pixel in several hundred 10"x10" boxes randomly distributed across the amplifier. In all cases the rms per pixel is consistent with the expectations of Poisson statistics (modulo the uncertainty in the gain).
I then measured the median value in 20 10"x10" boxes that were randomly distributed within 60"x60" regions, calculated the rms of these 20 median values, and repeated this for a total of 20 60"x60" regions distributed across the amplifier. The median rms value within these 60"x60" boxes is within a factor of two of the expectations from Poisson statistics (37x smaller than the per pixel rms -- I adopted 0.27"/pixel). These values are on order 2e-4 and always less than 1e-3.
Finally, I measured the median value in each of the 20 randomly distributed 60"x60" regions on the amplifier and calculated the rms of these 20 measurements. These values exceed the expectations of Poisson statistics (220x smaller than the per pixel rms), which indicates that large-scale gradients are present. However, the vast majority of the variations are less than 1e-3 and they are always less than 5e-3.
I ran this analysis on both amplifiers of all CCDs and on flats obtained in the ugrizY filters. Based on this analysis, I conclude that there is no additional information in the twilight flats that would help achieve 1% (and probably even 0.1%) photometry. This meets the dome-twilight requirement Cal-R4.
This analysis was performed with twilight flats obtained on 20121104 and 20121105 and dome flats obtained on 20121104.
-- Paul Martini, 19 Nov 2012
Dome Flat Comparison (Single Day)¶
I ran the same analysis described above on pairs of dome flats obtained in the same filter on 20121104. The ratio of two dome flats is consistent with the expectations of Poisson statistics on scales of single pixels, 10"x10" boxes, 60"x60" boxes, and over the entire amplifier. This is true for all six filters: ugrizY.
-- Paul Martini, 19 Nov 2012
Dome Flat Filter Ratios¶
I compared the ratio of different filters (same chips) to one another to test the uniformity of the flat fields as a function of wavelength. I started with the 'flatcor' output from DESDM and took the ratio of the u/g, g/r, r/i, i/z, and z/Y flats. These show some large-scale structure. In the case of chip 37 (N4), there are gradients along the long axis of the CCD of 4% (u/g), 2% (g/r), 1% (r/i), 1% (i/z) and 1% (z/Y). The gradients are somewhat larger, particularly in u/g, toward the edge of the field.
This image shows the ratio of the u/g, g/r, r/i, i/z and z/Y flats for chip 37:
The grayscale map is from 0.9 to 1.1.
All of this analysis was performed with the 20121119091456_20121118 processing run from DESDM (dome flats from 20121118).
-- Paul Martini, 20 Nov 2012
Dome Flat Comparison (Multi Day)¶
Comparison of Dome Flats and Deep Sky Flats¶