Calibrated (L1b) frame metrics from INCREMENTAL v7.5 processing 

The plots below show scan-collapsed L1b-frame metrics (either scan-medianed 
or scan-averaged, as indicated). These serve to monitor pixel calibration
performance in a relative sense. For the first three plots, only framesets 
satisfying the following critiera were used: |glat| > 20deg, saadist > 5deg, 
and moondist > 30deg. An orthogonal trending analysis using L0-frames 
averaged over 3 to 4 day windows can be found here. The latter are primarily
for detector trending and creating calibrations.

* median of robust spatial frame-pixel RMS per scan as a function of MJD:
  medrms_vs_mjdL1b_v7.5.pdf
  red lines are "optimal" local nonparametric regression fits. 

* median number of low+high sky-offset pixels with |value|>20 DN over scan:
  ** NOTE: these are eventually subtracted when making final L1b products. 
  These trends primarily serve as an indicator of how far/close the input
  static darks were to the true state of the detector. I.e., the dynamic
  sky-offset calibrations will need to catch/correct "fewer" low/high dark 
  pixels if the input darks were perfect. This can be seen soon after a 
  calibration update (the discontinuous jumps).
  medskyLoHi_vs_mjd_v7.5.pdf

* median of ratio: "robust frame-pixel RMS/mode{error-model pixel uncertainty}"
  over all frames per scan:
  medchi_vs_mjdL1b_v7.5.pdf

* median number of hi-spike (glitched) pixels per frame over scan:
  medspikeHi_vs_mjdL1b_v7.5.pdf

* maximum number of hi-spike (glitched) pixels over all frames in scan:
  maxspikeHi_vs_mjdL1b_v7.5.pdf

* Residual low spatial-frequency responsivity maps created by
  stack-averaging ~200 scan-collapsed X,Y binned "NEOWISER - ALLWISE" delta
  magnitude tables generated in operations per scan. The delta-mag
  estimates in these tables are first converted to flux-space, renormalized
  to have a spatial-average of unity, then stacked using trimmed averaging
  and smoothed using a constant Gaussian kernel to remove the binning pattern.
  These products illustrate the performance of the pixel calibrations
  (flats, darks, sky-offsets etc.) and hence the quality of the L1b frame
  products from which sources are extracted. The min-max stretch on all 
  jpegs is 0.995 --> 1.005, i.e., 1 +/- 0.5%
  Relative "NEOWISE-ALLWISE" residual maps


Following discontinued for regular trending on 01/13/15 (not much use)

* number of residual hot (> 20 DN) pixels from median-stack of
  filtered regularized L1b frames per scan:
  stackhot_vs_mjd_v7.5.pdf
  frame filtering: moondist > 23deg; |b| > 10deg; saadist > 2deg.
  
* number of residual noisy (rms > 25 DN) pixels from trimmed 
  stack-standard-deviation of filtered regularized L1b frames per scan:
  stackrms_vs_mjd_v7.5.pdf
  frame filtering: moondist > 23deg; |b| > 10deg; saadist > 2deg.


Calibrated (L1b) frame metrics from YEAR 1 processing: 12/13/13 - 12/13/14

** NOTE: these are not updated as regularly as the incremental trends above.

* median of robust spatial frame-pixel RMS per scan as a function of MJD:
  medrms_vs_mjdL1b_v7.5_yr1.jpg
  verticle red lines demarcate the calibration delivery dates.

* median number of low and high sky-offset pixels with |value|>20 DN over scan:
  These lo/hi pixels were subtracted when making final L1b products.
  These trends primarily serve as an indicator of how far/close the input
  static darks were to the true state of the detector. I.e., the dynamic
  sky-offset calibrations will need to catch/correct "fewer" low/high dark
  pixels if the input darks were perfect. This can be seen soon after a
  calibration update (the discontinuous jumps).
  medskyLo_vs_mjd_v7.5_yr1.jpg
  medskyHi_vs_mjd_v7.5_yr1.jpg
  verticle red lines demarcate the calibration delivery dates.

* median of ratio: "robust frame-pixel RMS/mode{error-model pixel uncertainty}"
  over all frames per scan:
  medchi_vs_mjdL1b_v7.5_yr1.jpg
  verticle red lines demarcate the calibration delivery dates.


F. Masci, 01/27/2016