I show below an analysis of the first two years of data. I split the flat data into year 1 and year 2, and show that the flat has not changed in any detectable manner. I then make new superflats for the mission to date.

Comparison of Year 1 and Year 2

I began by taking all of the flatfield data from the IST archive on /iracdata/flight/calprog. This includes all processing versions of all the flatfields throughout the mission. I examined all of these using a small idl script that steps through all of the flats, allowing me to toggle them on or off. The flats were examined for obvious latents or other defects. At most one flat per AOR was chosen (no multiply processed data), so there are truly independent inputs to the stacking. The stack of "good" data was then split at campaign IRAC005100. The input data were then combined using IRAF IMCOMBINE.

Several things about this are notable. First, because there was more data I rejected more flats as "bad" than previously. Technically, this lowers the photometric S/N, but at the same time lowers the amount of residual star contamination in the resulting superflat. Also, in the second year of the mission we took flatfield data much less frequently, hence there were fewer images in the input stacks.

Figure 1 shows the superflats made from the data in the second year. Figure 2 shows the difference between the first and second years. There are no obvious large systematics.

• Channel 1 - there are no systematics. The marbled appearance is probably correlated noise resulting from stellar residuals. The amplitude of these difference are about 0.5%.
• Channel 2 - Same as channel 1, but the differences are about 0.2%. There is one coherent structure related to the column that is periodically pulled down.
• Channel 3 - Differences are almost all jailbar stripes, and some incoherence in handling the scattered light near the top of the frame. Max differences are 0.3%.
• Channel 4 - There is a fairly clear gradient. However, it's amplitude is tiny, less than 0.05%.

Fig 1. - superflats for the second year of operations. Channels 1 to 4 are shown left to right.

Fig 2. - year1 superflat minus year2 superflat.

2-year Superflat

I took all the data, and contructed a superflat appropriate for the last two years. This superflat is shown in figure 3. Note that it looks identical to figure 1 - which it should, since the flat does not change! Figure 4 shows the sigma for this flat derived from the input pixel stacks. Again, because the larger amount of data allowed for more careful discrimination against stellar residuals in the input data, the resulting sigma images are much more well-behaved. Most of the regions of high scatter appear to be related to the handling of the stray light, which is expected. There is also some increased uncertainly around the amplifier glow in channel 3.

The new flats may be dowloaded here.

An additional minor change is that pixels that have no valid values (were all NaNs when processed, usually due to bad pixels) are now set equal to 1 instead of NaN, as was the previous convention.

Fig 3. - superflats made from the first two years of data.

Fig 4. - sigma images for the above superflats.