I have looked at a set of 4 sub-images which Tom Droege acquired and sent out last night. Each is a small subsection of a Mark IV dark image, taken at a different temperature. You may download the images themselves if you wish:
I created a histogram of the pixel values in each image. It turns out that there are a set of regular peaks in each histogram. The spacing of the peaks decreases with temperature. Here's a graph showing the effect:
And here's a table with my rough measurements of the peak positions:
Positions of the Peaks in Noise Distribution
----------------------------------------------------------------
peak 1 peak 2 peak 3
+15 C -18,300 -16,123 -13,989
+1 C -23,465 -23,075 -22,679
-10 C -23,758 -23,567 -23,377
-20 C -23,847 -23,793 -23,738
----------------------------------------------------------------
As you can see from the graph, or from the table, the peaks at each temperature are spaced at equal intervals:
Spacing of Peaks vs. Temperature
---------------------------------------------------------
Temp (deg C) Spacing of peaks (DN)
+15 2150
+1 390
-10 190
-20 54
---------------------------------------------------------
Here's the same information in a graph, with the spacing plotted both linearly and logarithmically against temperature:
