well the dark scaling is done linearly, by multiplying the dark frame by some number. what i am saying is that if canon has played tricks with the dark signal in the light frame (and also the dark frame) then the dark is not truly a dark. we don't know what they have done to it. i guess you could probably give ImageCalibration some bogus "dark" frame and it would then scale it until the noise in the result was minimized. but that does not make the result correct...
i have scaled DSLR darks in the past and mainly the problem is that hot pixels are either under- or over- corrected. generally these are handled fine by dithering and rejection during integration, though. i have not done any empirical experiments to see if the results are superior with scaled DSLR darks; i have just been put off of it due to the lack of knowledge about what canon is really doing behind the scenes to the dark current.
Yes, scaling is multiplication by a constant. What I meant by linearity is the duration and temperature dependence of dark current. In this sense, PI could take the dark current form the master dark and use its duration and temp as a basis to extrapolate/interpolate linearly and come up with a scale factor. Instead, it just empirically minimizes the resulting noise.
If you throw some random dark, there will likely be no correlation with the light noise, so k will end up being zero. So I guess, even with DSLR that somehow messes up the raw files, there is some value in dark scaling. And, of course, the best thing is to just match time and temp, and avoid scaling altogether. In my experience, temp scaling within +/- 2°C works very well. Never tried time scaling.
Regarding hot pixels, them being highly non-linear, they cannot be calibrated whenever there is a linear scaling factor involved, independent of the camera being a dslr or an astro ccd. I understand that PI is working on a solution to this.
Ignacio
