Monitoring bad pixels with UFTI

Monitoring bad pixels with UFTI

The data tabulated below illustrate the degree to which bad pixels “develop” over time. At present, every night of UFTI observing begins with a sequence of four dark exposures; these observations are known as the “Array Tests” and are used to calculate the NDSTARE read noise. The fourth exposure, a 50-second dark, will be used periodically to check the accuracy of the bad-pixel mask being used on all data. The DR recipe REDUCE_DARK, when applied to this frame, yields a dark exposure corrected for bad pixels. Any remaining bad pixels will thus be either new or “transient pixels”. Results are tabulated below:


Bad-pixels remaining in a dark that has been “corrected for bad pixels”.

Typically about 0.3% of the 1,048,576 pixels in the UFTI array are flagged as “bad” by the bad-pixel mask. Initial results indicate that, over a short period of time (a few days) some pixels “migrate” across the good/bad cut-off; the bad-pixel mask nevertheless remains accurate to within a few percent for a period of a few months.


Making a Bad Pixel Mask (Starlink Software)

From 3 or 4 nights of UFTI data. Produce “QUICK LOOK” versions of the 50sec dark taken as part of the array tests (the 4th frame in the sequence) with ORAC-DR; these aren’t corrected for bad pixles (i.e. they are raw NDF frames). Take the average of these raw dark frames, and flag all pixels above 200 and below -20 as bad (should be a few thousand pixels). Finally, multiply the frame by zero so that the resulting bad-pixel mask has data values of “0” and “bad” only.

For example:

  > oracdr_ufti 20010101
  > setenv ORAC_DATA_OUT `pwd`
  > oracdr -list 4:4 QUICK_LOOK -nodisplay
  > oracdr_ufti 20010102
  > setenv ORAC_DATA_OUT `pwd`
  > oracdr -list 4:4 QUICK_LOOK -nodisplay
  
  > kappa
  > add f20010101_00004_raw f20010102_00004_raw add_darks
  > cmult add_darks 0.5 av_darks
  > thresh av_darks av_darks_thresh -20 200 bad bad
  > cmult av_darks_thresh 0 bpm title=\"UFTI bpm, January 2001\"

The above sequence produces a bad-pixel mask NDF (bpm.sdf) that can be viewed with GAIA; note the pixel values (zeros and bad). To re-reduce your data in ORAC-DR with the new bad-pixel mask, specify the bpm on the command line, e.g. oracdr -list 5:500 -cal mask=/your_directory/bpm.