DSLR lights calibration, optimize or not to?

[AstroScience]

This is starting to be confusing, as different users report different results with their cameras and some do suggest to scale master dark and some not. Being curious I decided to experiment my self too. My DSLR is cooled and for example I took single uncalibrated frame of 720s exposure with camera reporting -3C, you can see it on the upper left with it's dark current hot pixels.

To calibrate this frame I have created master dark from 21 precalibrated with master Bias frames, which were in range of -4C to -2C. So the master Dark matches as close as it can with the exposure time and temperature.
   
Since the exposures are matching up I should not use scaling when calibrate, right? So the Optimization wasn't ticked and the result is on the upper right. Noise evaluation reported a bit lower noise levels, but the hot pixels wasn't cleared.

Next time I did tick the Optimization and Dark scaling factor reported was high as 1.7 and the result is shown on the bottom. It's clear that the hot pixels now removed and noise estimates got even lower. So, are we, DSLR users, should always use Optimization? Even when master dark matches the lights in exposure times?

Why PI didn't clear those hot pixels without Optimization ticked?

[georg.viehoever]

I have an uncooled DSLR, so my experience may or may not apply to you:
- hot pixels are usually eliminated by PixelRejection (during ImageIntegration). Optionally, you can also add CosmeticCorrection to your calibration procedure.
- my results are usually better without Dark Scaling. Even for Darks created with the same exposure time as the lights, I see scaling factors from 0.3 to 1.8, when the really should be close to 1.0
- There have been several posts that say that Canon DSLR images are not fully linear. I can confirm that from my own investigation: For RAW images there happens some magic in the background, and Canon is not willing to share details. However, Juan says that dark scaling works independently of linearity assumptions. This is true, the algorithm does not need it. Still, in my experience, scaled Darks cause worse results.

My recommendation: Try both options for yourself and use what is better. Remove hot pixels with pixel rejection, or by adding CosmeticCorrection to your callibration run.

Georg

[AstroScience]

Hi Georg,
 seems like I have more testing to do, thank you.

[Ignacio]

I also work with DSLRs cameras, and my experience has been very good with dark scaling, meaning that if the master dark is built to match lights (temp and duration), then the scaling is very near 1 (to 2 or 3 digits). In fact, I always optimize the master dark, even when there is perfect match, as a verification that the master dark was correctly built.

There are a couple of thing I do to get to these results. First, I build a very accurate master bias, using 200 frames, and sometimes Carlos's superbias script. Then, I build a master dark without calibrating each dark frame (ie, without subtracting the master bias). Finally, during light calibration, I check "optimize" and "calibrate" master dark, use the whole frame to estimate the scale factor (by entering zero in the corresponding dialog), and choose "Force CFA" (I use monochrome CFA format in all my frames). With this workflow I get very consistent results.

Hope it helps.
Ignacio