ICal-related
------------

Last modified: F. Masci, October 18, 2010

 * mitigate/reduce w1,w2 channel, amplifier bias fluctuations and 
   checker-boarding
[1 week]

 * pixel de-saturation algorithm using saturation encoding - for *soft*
   saturated cases only.
    - Note: during W3 warm-up, saturation encoding stopped working
      for W2, and maybe became unreliable for W1. Note sure about W3,W4 close
      to their warm-up phases.
[1 week]

 * flat-field (and possibly w3,w4 dark) picking infrastructure:
    - vs. period of application? (e.g., saw that w3,w4 flats changed ~ 2/1/10)
    - vs. zodi color? (analysis pending)
    - vs. time since anneal or treat post anneal periods separately?
    - darks (and linearity - see below) are coupled to all the above
    - need to handle warmup phase(s)
[3 weeks]

 * update tempcal module (for dynacal) to create products using a single-frame
   step moving median. Also update/optimize dynacal pipeline infrastruct.
[2 weeks]

 * refine droop corrections and algorithm:
    - adapatively inflate/deflate threshold for split corrections according 
      to background level. Now: for high backgrounds => unnecessary
      split correction is applied => mess's up global Q-to-Q droop correction
      using ref pixels.
    - consider not applying split-droop correction if actual correction is
      "high" due to confusing structure in frame?
    - quadrant droop gradients (e.g., Cat's Eye in touchstone field!). See
      analysis charts and characterization code.
    - temperature dependencies during warmup [calibration only]?
    - ***NOTE: for non-nominal exptimes in warm-up, can't key off L0 
      sat values >=32,753.
[1 week]

 * consider applying smoothed running scan-flat "delta" corrections on top of
   static flats, before final skyoffset calibration. Can be built into
   dynacal infrastructure.
[2 weeks - analysis pending]
      
 * single-frame glitch detector needs tuning for w1,w2. Needed to tag
   CRs for low-depth regions in coadds where temporal method is unreliable
[0.5 week - calibration/tuning only]

 * linearity: 
    - reanalyse on-orbit experiment using single frame extractions:
      improved registration/photom expected to make a difference.
    - continue to analyse WISE vs. IRAC+MIPS24 photom. (NEP, SEP etc.)
    - continue to monitor CMDs (from science applications).
    - continue to monitor MAGZP's vs flux for all bands
    - idea if time permits: derive NL, dark + relative flat from one super
      self-cal fit offline?
    - assess warmup phase(s)
[2 weeks - calibration only]

 * check gain and read-noise estimates throughout mission, especially during
   warm-up phase(s)
[1 week - mostly analysis/calibration]

 * static bad-pixel mask checks throughout mission by searching for
   x,y correlations in source space. Also use N/M metric. Thankfully,
   dynacal masking does a pretty good job.
   - asess changes due to warmup. Note FEB offset was changed for W1,W2 
     to reduce bad-pixel count. Pedro has provided binary bad-pixel maps for
     W3 where there was _no_ FEB offset adjustment
[1 week - calibration/sanity check only]

 * level 1b frame "thin" streak detector. Co-adder outlier rejection not
   sensitive to thin, faint streaks.
[1 week]


Multiframe
----------

 * optimize frame filtering:
   - further moon-mask optimization
   - bad droop corrections, including the worst of the Cat's eye.
   - dynacal failures? - if not run or bad correction applied - how can we tell
     in an automated manner?
[0.5 week]
   
 * ensure proper throughput matching of frames with different exptimes:
   assumes frame ZPs properly calibrated - especially during warmup.
[0.5 week: ensure follows through to wphot]

 * automatic nominal-saturated pixel replacement with shorter ('unsat')
   exptime data?
[0.5 week: ensure follows through to wphot]

 * use multi-epoch (scan-direction dependent) data to mitigate latents in
   coadds - especially short term w3,w4 latents?
[0.5 week]

 * any way to suppress short-term latents at some |elat| < max using
   temporal matched filtering in stacked pixel space?
[no idea]