As the clouds and the rain are preventing me from taking the camera outside, I have been playing with various measurements. Data for the dark noise vs temperature is below.
Few interesting observations:
I did two noise measurements, one on an image region without hot/warm pixels (noise1) and one where almost the whole image was selected (noise2), i.e. including hot/warm pixels.
The noise1 (without hot pixels) slightly goes down (2% with measurement errors ignored) if the camera is cooled below 5C. As it was suggested before, there is absolutely no point going below zero.
However, noise2 goes down when cooled below zero. Actually there is still a significant (25%) difference between the -4C and the bias file. I would say that the hot/warm pixels get less hot/warm as the camera is cooled below zero. Not really of significance as the dithering removes the hot pixels.
Also I noticed that the white pixels/signal cut off at 15846. I have expected a well depth of 16383 (2^14) as it is a 14-bit camera. It is always 15846, no matter what exposure or ISO is used.
Also the dark signal starts at about 1000 - below it is much less or zero which cannot be real.
The camera is clearly doing processing of the RAW files
I suppose you get what you pay for
