Bad Pixel Maps for IRAC Flight Arrays
Note: the following information is intended for the IRAC instrument team
only. If you are not such a person, you should not be reading this. No
guarantees are made as to the applicability of this information to the
real IRAC.
I have done a quick analysis of which pixels are "bad" in the IRAC flight
arrays. The techiques used to derive them are driven mostly by our lack of
complete datasets to use in the manner that we intend to during flight.
Namely, during flight LINCAL (or something similar) will tag bad pixels
based on the linearity test datasets - pixels with abnormal responses will
be identified based on how they respond (or not) to light.
To get this initial identification of the bad pixels, I have used a technique similar to how observers pick out bad
pixels "by eye". The flood calibrator data from the January A-side LFF was
used, since the transmission calibrator data does not illuminate the entire
array. I only looked at Channels 1,2, and 4 since Channel 3 has been
replaced. 5 images with the same exposure time in each channel were chosen.
They were processed using IRAF and the SSC in-house pipeline modules in
order to convert them to 32-bit format, recitify the InSb data such that positive
flux yields increasing DN, and then corrected wraparound issues using a
threshold of 40,000 ADU, which appears safe for this data.
The 5 images were median stacked in order to reject cosmic rays or other
transient glitches.
The bad pixels were identified on the basis of their being too high or low
relative to their immediate neighbors. The comparison must be done locally due
to the variable illumination. The median-stacked image was smoothed using a
running boxcar 5x5 median filter, and hence each pixel in the smoothed
image is an estimate of the local pixel value. The ratio of the median-stacked image
and the smoothed image was taken. Pixels with values less than half of
their immediate neighbors were tagged as "dead", and more than twice the
neighboring values as "hot". The identified pixels were checked by eye vs.
both the median-stacked image and the individual frames for aparrent
validity.
|
Examples of Bad Pixels
|
 |
 |
| Channel 2 Hot Pixels |
Channel 2 Dead Pixels |
The arrays are, in fact, startlingly good. Less than 1 in 1000 pixels are
bad. The InSb arrays seem to be worse than the SiAs. The most significant
problem is the "fiber" on channel 3, which effectively eliminates a bit
more than 100 pixels. The pmask for this file was made by my blanking
areas around the fiber, since technically they pass the above bad pixel criteria
even though no useful information is carried in them. They are illuminated
by light diffracting around the fiber. Otherwise, this array has no defects. Also, note
that the bad pixels may change when the array voltages are changed.
Channel Hot Dead
---------------------------------------
1 10 6
2 20 8
3 0 122
4 1 1
By clicking on the images you can download a FITS format mask file. The
masks are zero everywhere except where the bad pixels are located, which
are set to 1.
You can also download SSC "pmask" files, which are the above encoded into a
16-bit mask file according to the (current as of this writing) key:
Bit # Condition
----- -------------------------------------------------
0
1
2
3
4
5
6
7 Linearity model could not be computed
8 Dark current could not be computed
9 Dark current too variable
10 Dark current too high
11 Responsivity could not be computed
12 Responsivity too variable
13 Responsivity too high
14 Responsivity too low (pixel essentially dead)
15
