Flat Exposure Properties of VIRCAM Detectors

Document number VDF-TRE-IOA-00008-0013 (Draft 20061122)

Jim Lewis

1. Introduction

In this document we are looking at the flat properties of some of the AIT data that was taken at RAL with the 16 VIRCAM detectors. In what follows we are looking at the frames VIRCAM_IMG_FLAT255_0043 to 56 and their associated dark frames, VIRCAM_IMG_DARK255_0124 to 137. All were taken with single integrations using torch in a darkened room. The exposure times for each frame are:

Frame	Exposure time (s)
VIRCAM_IMG_FLAT255_0043	1
VIRCAM_IMG_FLAT255_0044*	10
VIRCAM_IMG_FLAT255_0045	5
VIRCAM_IMG_FLAT255_0046*	10
VIRCAM_IMG_FLAT255_0047	10
VIRCAM_IMG_FLAT255_0048*	10
VIRCAM_IMG_FLAT255_0049	15
VIRCAM_IMG_FLAT255_0050*	10
VIRCAM_IMG_FLAT255_0051	20
VIRCAM_IMG_FLAT255_0052*	10
VIRCAM_IMG_FLAT255_0053	25
VIRCAM_IMG_FLAT255_0054*	10
VIRCAM_IMG_FLAT255_0055	30
VIRCAM_IMG_FLAT255_0056*	10

The starred frames were taken in order to monitor the flux of the light source over time. The dark frames have matching exposure times. Below are the images resulting from dark subtracting and combining all of the 10s exposures in the list above.

Figure 1.1 Detector 1	Figure 1.2 Detector 2	Figure 1.3 Detector 3	Figure 1.4 Detector 4
Figure 1.5 Detector 5	Figure 1.6 Detector 6	Figure 1.7 Detector 7	Figure 1.8 Detector 8
Figure 1.9 Detector 9	Figure 1.10 Detector 10	Figure 1.11 Detector 11	Figure 1.12 Detector 12
Figure 1.13 Detector 13	Figure 1.14 Detector 14	Figure 1.15 Detector 15	Figure 1.16 Detector 16

2. Self Flat Correction

As a first test for the repeatability of the flats, we flat corrected each of the 10s flats that went into the master 10s flat.
The figure below shows the result for one of the input frame (number 44) -- this is very typical.

Figure 2.1 Detector 1	Figure 2.2 Detector 2	Figure 2.3 Detector 3	Figure 2.4 Detector 4
Figure 2.5 Detector 5	Figure 2.6 Detector 6	Figure 2.7 Detector 7	Figure 2.8 Detector 8
Figure 2.9 Detector 9	Figure 2.10 Detector 10	Figure 2.11 Detector 11	Figure 2.12 Detector 12
Figure 2.13 Detector 13	Figure 2.14 Detector 14	Figure 2.15 Detector 15	Figure 2.16 Detector 16

The table below shows the normalised RMS of the flat corrected data (all bad pixels removed from stats). This shows that the small and large scale variation in the flat field is correctable to within less that half a percent in most cases. There is some striping visible on a few
of the images in figure 2 (detectors 2 and 3 for example) and this is more easily seen by bumping up the contrast on a display server. Removing the stripes using the VIRCAM destripe algorithm does improve the RMS estimates slightly, but not to the point where it's worth bothering in this case.

Detector	RMS
1	0.0048
2	0.0055
3	0.0059
4	0.0049
5	0.0042
6	0.0045
7	0.0043
8	0.0047
9	0.0038
10	0.0040
11	0.0042
12	0.0041
13	0.0044
14	0.0036
15	0.0037
16	0.0051

3. Linearity

The images mentioned above can be used to measure the linearity of each of the detectors. The 10s exposures that were interspersed with the others (starred in the first table) were used to measure the drift of the light source during the series. The illumination of the dome for the real VISTA telescope has been specified to be constant to within a level of well below 1%, so monitoring source drift will probably not be necessary during normal operations. In the case of this AIT data, however the source did drift by more than 3% during the course of the observations. The VIRCAM recipe for linearity measurement works on images that have not be dark corrected, so to simulate data taken with a steady light source each exposure in the series was corrected by:

Removing the dark for that exposure
Multiplying the dark corrected image by a factor to account for the source drift
Adding the dark back on

The correction factor for a given exposure was calculated by the ratio of the flux of the first monitoring exposure divided by the flux of the monitoring exposure taken just after the exposure of interest.

Although we did our best to correct the flickering of the light source, it is obvious when looking closely at the data, that our simple linear correction isn't quite enough and that we don't have enough information to correct things any better. Looking at the curves in the figure below also shows that the linearity curve for each detector is very different. Modelling the curvature properly in many cases will require more than the 7 observations we have, especially if the flux of the light source is a little unreliable (again hopefully this won't be the case at the telescope). The x and y axes respectively represent exposure time and median flux in a small box on each detector. The line is a linear fit to the first 3-4 points on the graph, as it shows the curvature in each linearity relationship. Restricting the linearity fit to the first 4 exposure times can at least get us a reasonable guess at the level of the non-linearity in each chip at the nominal 10000 count mark that we are using for a QC parameter.

figure 3.1 Detector 1	figure 3.2 Detector 2	figure 3.3 Detector 3	figure 3.4 Detector 4
figure 3.5 Detector 5	figure 3.6 Detector 6	figure 3.7 Detector 7	figure 3.8 Detector 8
figure 3.9 Detector 9	figure 3.10 Detector 10	figure 3.11 Detector 11	figure 3.12 Detector 12
figure 3.13 Detector 13	figure 3.14 Detector 14	figure 3.15 Detector 15	figure 3.16 Detector 16

For each detector a linearity fit is done for each channel. By and large the results suggest that the linearity is independent of channel number, although for some there does appear to be some significant variation. Below is a table showing the average and standard deviation non-linearity at 10000 counts over all channels in each detector. The linearity recipe also sets a quality of fit parameter for each channel. This is calculated by first using the fit coefficients and the input flux to calculate a 'linear' corrected flux for each exposure. Then an RMS is worked out which indicates the deviation of the exposure time vs corrected flux from a linear relationship. This value is also included in the table below.

Detector	Linearity at 10000 ADU (%)	Fit Quality (%)
1	1.30 +/- 0.09	0.52 +/- 0.04
2	2.09 +/- 0.11	0.53 +/- 0.03
3	2.60 +/- 0.42	0.51 +/- 0.03
4	2.14 +/- 0.18	0.55 +/- 0.06
5	1.87 +/- 0.11	0.52 +/- 0.03
6	1.69 +/- 0.07	0.58 +/- 0.02
7	1.32 +/- 0.05	0.53 +/- 0.01
8	2.23 +/- 0.24	0.61 +/- 0.03
9	2.12 +/- 0.11	0.64 +/- 0.01
10	1.75 +/- 0.06	0.57 +/- 0.02
11	3.26 +/- 0.52	0.71 +/- 0.03
12	1.61 +/- 0.05	0.57 +/- 0.02
13	5.98 +/- 0.34	0.62 +/- 0.08
14	2.04 +/- 0.03	0.60 +/- 0.01
15	1.60 +/- 0.02	0.49 +/- 0.01
16	2.45 +/- 0.11	0.65 +/- 0.01

In order to test whether the linearity correction is any good, we did a linearity correction on all the 10s flats and combined them into a master corrected 10s dome flat. The we linearity corrected a 5s flat and divided it by our corrected master. The result is below in Figure 3. The table below shows the RMS for the resulting linearity corrected and flat fielded image. In order to compare this result with the table above in section 2 we divided the RMS by sqrt(2) to take into account the shot noise difference between this 5s exposure and the 10s exposure presented above. This corrected RMS is comparable to that from the self flat operation, indicating that by and large the flat field correction is not degraded by the linearisation operation. There are some small areas in, for example, detectors 4 and 13 which raise some questions and these will have to be investigated at a later time. A priority for daytime commissioning should be to redo this exercise with the proper dome lines and with many more exposures in the sequence.

figure 4.1 Detector 1	figure 4.2 Detector 2	figure 4.3 Detector 3	figure 4.4 Detector 4
figure 4.5 Detector 5	figure 4.6 Detector 6	figure 4.7 Detector 7	figure 4.8 Detector 8
figure 4.9 Detector 9	figure 4.10 Detector 10	figure 4.11 Detector 11	figure 4.12 Detector 12
figure 4.13 Detector 13	figure 4.14 Detector 14	figure 4.15 Detector 15	figure 4.16 Detector 16

Detector	RMS	RMS(10)
1	0.0075	0.0053
2	0.0087	0.0062
3	0.0091	0.0064
4	0.0107	0.0076
5	0.0070	0.0049
6	0.0068	0.0048
7	0.0069	0.0049
8	0.0079	0.0056
9	0.0064	0.0045
10	0.0070	0.0049
11	0.0073	0.0052
12	0.0067	0.0047
13	0.0090	0.0063
14	0.0060	0.0042
15	0.0062	0.0044
16	0.0085	0.0060

4. Bad Pixel Masks

The linearity_analyse recipe also writes out bad pixel masks. Figure 4 below shows a full grid of the masks for each detector. The actual percentage of bad pixels is listed in the caption.

figure 5.1 Detector 1 (1.93%)	figure 5.2 Detector 2 (1.30%)	figure 5.3 Detector 3 (0.91%)	figure 5.4 Detector 4 (0.63%)
figure 5.5 Detector 5 (0.14%)	figure 5.6 Detector 6 (0.23%)	figure 5.7 Detector 7 (0.22%)	figure 5.8 Detector 8 (0.32%)
figure 5.9 Detector 9 (0.27%)	figure 5.10 Detector 10 (0.10%)	figure 5.11 Detector 11 (0.24%)	figure 5.12 Detector 12 (0.22%)
figure 5.13 Detector 13 (0.90%)	figure 5.14 Detector 14 (0.97%)	figure 5.15 Detector 15 (0.53%)	figure 5.16 Detector 16 (1.43%)