On May 24, 2011, at 5:10 PM, Roc Cutri wrote:
WISE-Ones,

We've started a web page to track 3-band cryo processing with
WISE v4.5.  Please see:

http://wise2.ipac.caltech.edu/proj/roc/processing/3band_processing.html

ICal strategies, first thoughts:
---------------------------------------

* isolate scans that contain useful (calibratible) data. Metrics to determine what's usable are TBD.

* perform a test with our current v4 dark/flat/linearity/mask/droop/gain/read-noise calibrations + nominal dynacal + other params to see how far they take us according to some TBD metrics. First epoch data could be used as a benchmark and to set tolerances on what's acceptable. It may involve making coadds. The plan would be to reprocess sets of scans spread over this time period.

* second test: open up windows in dynacal to make higher S/N dynamic flats, skyoffsets, and tweak params to create *high spatial-frequency* responsivity maps for W3, and perhaps W1,W2 too if necessary. Recall that in v4 we created dynamic low-frequency responsivity maps. This scheme may take care of relative gain (flat) and dark variations - a huge step forward. This will require much testing and analysis.

* changes in exposure-time and SUR-weighting: need to adapt linearity calibration (see below) and noise model in an exposure-time-independent manner.

* for non-linearity: it's complicated and we can't assess the situation until we do some source-space analysis. We have tools to rescale linearity model coefficients to give us what we expect to see empirically in source color-magnitude diagrams. The hope is the linearity does not evolve too quickly, although we do expect changes in ramp shape due to changes in the effective exposure time, i.e., the usable dynamic range.

* for bad-pixel masks: new static array masks may need to be made (all bands). Nonetheless, if things evolve quickly, the transient bad-pixel detector in dynacal will tag new bad pixels (the variety that stay hi/lo for some period, not the new noisy ones). It's worth checking the noisy pixel status throughout this period.

* electronic gain and RN maps to support ical noise model. These uncertainties feed into sdex, mdex, and co-addition. New static calibrations may need to be made here. Nonetheless, ical also has a dynamic and robust local RMS scaling method we can use.

* Photometric ZPs in L1b headers should be updated to track changes in throughput over this period, including intentional exposure time changes. The coadder uses these ZPs in a self-consistent manner. The new ZP's will have to be derived using data processed with the latest scan/frame pipeline (i.e., v4+).

* saturated pixel masking. This information is used in ical for droop corrections, and tagged in masks for propagation to WPHot, scansync, coadder. Other downstream modules read the WPHot saturation indicators (e.g., ARTID and PCAL). How do we account for the loss in the saturation encoding? Can we do this at the pixel level, coarsely but conservatively? Use prior knowledge from pass1 processing? Analysis is in progress.