Index: /trunk/doc/design/readnoise.map.txt
===================================================================
--- /trunk/doc/design/readnoise.map.txt	(revision 24697)
+++ /trunk/doc/design/readnoise.map.txt	(revision 24697)
@@ -0,0 +1,44 @@
+
+need to update the following functions:
+
+ppImage/src/ppImageDetrendReadout.c:        pmReadoutGenerateVariance(input, true);
+ppSim/src/ppSimLoop.c:                    if (!pmReadoutGenerateVariance(readout, true)) {
+ppSim/src/ppSimMergeReadouts.c:      if (!pmReadoutGenerateVariance(inReadout, true)) {
+psphot/src/psphotMaskReadout.c:        if (!pmReadoutGenerateVariance(readout, true)) {
+
+plus: 
+  psModules/src/camera/pmFPAMaskWeight.c:bool pmReadoutSetVariance(pmReadout *readout, bool poisson)
+  psModules/src/camera/pmFPAMaskWeight.c:bool pmReadoutGenerateVariance(pmReadout *readout, bool poisson)
+  psModules/src/camera/pmFPAMaskWeight.c:bool pmReadoutGenerateMaskVariance(pmReadout *readout, psImageMaskType satMask, psImageMaskType badMask, bool poisson)
+
+
+i'm looking into defining the readnoise for ppImage at a finer level
+of resolution than by cell: this will let us set thresholds which are
+prevent excessive magic false positives in the regions with bad
+correlated noise, but still allow detection of streaks and faint
+sources in the regions on the same cells with low correlated noise.
+
+my proposal is as follows:
+
+* define the readnoise map using a superpixel map image for each cell,
+equivalent to the background model, with data saved in a FITS file
+collection parallel to the other camera images -- for GPC1, that would
+mean an extension per cell and a file per chip.
+
+* save the readnoise map as a detrend type, and load it with the
+detrend loading mechanisms.
+
+* pass the information to pmReadoutSetVariance with a pointer to the
+psImageMap structure, which defines the super-pixel image and the
+information about overlaps needed to line up the fine and rough
+resolution images.
+
+* trigger the selection based on a concept CELL.READNOISE.MAP (a bool).
+
+I think the mechanism will be useful even if it turns out that JT can
+correct the images -- there may still be sub-cell variations in the
+effective readnoise level.
+
+* how to read/write the map:
+  * ppImage output file?
+