Dark scaling always conservative?

[papaf]

Hi all,
in the process of doing some video tutorial in italian language, which I will link somewhere (where??) when finished, I saw that, using data coming from a friend's Canon camera, the dark scaling function was always reducing their weight way too much. It was something in the neighborhood of 0.08. I thought this was because the darks were taken at a different ISO setting than the lights, but the final resulting image was still showing the typical glow. All of this using BatchPreProcessing script. Anyway, the result was still ok, so I didn't bother too much.

Now though the same happened with data coming from a cooled CCD, namely a QHY8L. The dark is still scaled, albeit not so much: 0.2. Being a cooled camera, I expected the difference to be minimal. Needless to say, the final image still shows the amp glow in a corner. I tried disabling the dark scale altogether, and the result was much better, with the amp glow no longer visible.

What could be wrong here?

Thanks!

[georg.viehoever]

If your lights have an exposure time of n*dark_exposure_time, the darks should be scaled by the factor n - if the theory that dark current is proportional to exposure time is right. Some experiments I did with my CanonEOS40D showed that this is not strictly true. Also, there seem to be some non-linear effects going on. At least for DSLRs, I would recommend to use darks with similar exposure times as the lights.

The algorithm used by ImageCalibration minimizes the patterned noise caused by dark current using a mathematical minimization algorithm. The general procedure is very robust, however I have the impression that sometimes the noise estimation is a bit fragile. However, using darks with similar exposure times as the lights I always got very reasonable results.

Georg

[papaf]

Thanks for your reply.
When you use such similar dark frames, you do disable the scaling feature, right? Because in the qhy8l case, I did in fact use the same exposure for both darks and lights, but still got it reduced by 3/4.

[georg.viehoever]

For my DLSR images and darks_exposure_time==light_exposure_time, using scaling often gave better results than without. But I always try both on and off settings to see which results looks better, and I have a look at the scaling factors reported in the console window. If scaling factors are 0.7..1.3, scaling usually improves results. If scaling factors are far away from 1.0, usually the unscaled darks worked better.

Georg
 

[papaf]

I understand.
Thanks Georg!