Every once in a while, I will see someone assert that they don't bother taking darks any more. They just dither and let modern stacking methods do their work. I have always greeted this claims with more than a certain amount of skepticism. And when they are mentioned by the DSLR crowd, I just figured they had poorly matched darks and those can indeed be worse than nothing.
Still, I'm a big believer in experimentation with your own equipment. I figured I would give this a try with the ACO 2151 data I collected recently. It would be pretty easy (or so I thought) to simply not use my master dark in calibration and then compare results.
I used the PixInsight batch preprocessing script and created a Master L using my 57 light subs but leaving out the master bias and master dark. I did use my master flat. No cosmetic correction was done. As anticipated the result was pretty horrid, but not for the reason I expected.
Still, I'm a big believer in experimentation with your own equipment. I figured I would give this a try with the ACO 2151 data I collected recently. It would be pretty easy (or so I thought) to simply not use my master dark in calibration and then compare results.
I used the PixInsight batch preprocessing script and created a Master L using my 57 light subs but leaving out the master bias and master dark. I did use my master flat. No cosmetic correction was done. As anticipated the result was pretty horrid, but not for the reason I expected.
As you can see, the flat way over corrected the vignetting. As I thought about it, my conclusion was that the way PixInsight created the original master flat, it was probably depending on the master bias. Thus I decided to reprocess, this time using both the master flat and master bias.
Now that looks much more promising. But how does using just the master bias and master flat compare to also using the master dark?
Here is a 100% crop using the master dark, master bias, and master flat.
Here is a 100% crop using the master dark, master bias, and master flat.
Here is the same area using just the master flat and master bias. Identical histogram stretches were applied to both.
Any difference between them is very very minor.
Here are some stats from PixInsight before the images are stretched.
Here are some stats from PixInsight before the images are stretched.
Now for some caveats. First, there still may have been significant differences if I had created the flat in a way that did not depend on the bias to work properly. Second, notice that I had 57 subs. That gave the Winsorized sigma clipping algorithm LOTS of data to play with. Things might have looked quite different using only 8 or 9 subs. Third, these subs were 5 minutes long. If I had been doing 10 or 20 minute subs, the answer might have been different. Fourth, I was shooting at -15c so there isn't a lot of dark current, which is what darks correct.
Still, the results are interesting just as is. It is certainly easier to shoot 100 bias subs (what was used for my master bias), than it is to shoot 20 dark subs (what was used for the master dark).
I will need to be evaluating this further. But I certainly can't dismiss the claims based on what I'm seeing here.
Still, the results are interesting just as is. It is certainly easier to shoot 100 bias subs (what was used for my master bias), than it is to shoot 20 dark subs (what was used for the master dark).
I will need to be evaluating this further. But I certainly can't dismiss the claims based on what I'm seeing here.