What is the best way to do Darks?

[Spookyer (Brett)]

Hi all, I thought it was time to update my CCD dark library. Done a bit of research and can't determine whether I just do darks for 20mins (the longest subs I do) and then scale them using some PI wizardry I have read about or whether I do what I did in the past and do darks for each sub length I am likely to use 2,5,10,20 mins etc ?

Second question is how many darks should I be doing for each sub length. I have read advice from 3 to 50.

Note the camera in question is a SBIG STT-8300 mono (KAF8300 chip)

Cheers
Brett

[billdan (Bill)]

Hi Brett,
I do darks for each sub-length I am liable to use, in my case 1,2,5,10 and 15 mins.

Signal to Noise improves by the square root of the number of subs taken, the number of dark subs I use is 37. As darks are a median stack I always use an odd number of subs, so the middle number is easier to calculate.

Cheers
Bill

[RickS (Rick)]

Brett,

Dark scaling in PI works pretty well. I have done 30 minute darks and used a 30 minute master for all my dark calibration needs with a KAF-8300 and a KAF-16803 sensor with good results. Doing darks that match all your sub lengths is ideal if you have the time (and is unavoidable if you're doing RBI preflash.)

There's a rule of thumb for darks that comes from the Handbook of Astronomical Image Processing (by Berry and Burnell) that says you should do darks for five times your total imaging time. The book includes the maths behind this showing that you get an "acceptable" additional noise contribution of 10% by following this rule.

I typically just collect darks for a few days. Any longer becomes impractical.

Originally Posted by billdan

As darks are a median stack I always use an odd number of subs, so the middle number is easier to calculate.

Bill: a median combine is good for rejecting outliers but doesn't give you optimal SNR. You would get a better result using an average combine and sigma clipping. The difference is significant. You lose about 20% SNR with a median combine.

Cheers,
Rick.

[rally]

Brett,

Brad Moore wrote "Dark Frame Manager" - it sold for $20 and is by far the easiest method to create (and manage) a complete Darks library.

You give it the list of all the exposure times and all the temperatures you want on your SBIG camera (I think it does others but never needed that) and it just goes away and takes all the 1000's of darks you need
Specify 20 or 25 darks per Master Dark you want in each temperature & exposure time you might ever need and then generate master darks.

A good library can take days to complete - in your absence and completely automatically.
If you are doing 30 or 60 minute narrowband subs then 25 x 60 minute subs for say 4 or 5 different temperatures is a lot of tiime if you have to babysit.

Do it again in 6-12 months without being afraid of all the time and effort !

He used to have a website - not sure how you get it now, but Im sure you can contact him directly.

Cheers
Rally

[Spookyer (Brett)]

Thanks guys for the advice.

Brett

[gregbradley]

Bill: a median combine is good for rejecting outliers but doesn't give you optimal SNR. You would get a better result using an average combine and sigma clipping. The difference is significant. You lose about 20% SNR with a median combine.

Cheers,
Rick.[/QUOTE]

I wonder if that is true only for PI. In CCDstack I have done stacks using both and saw no real difference in the reported stats.

Greg.

[RickS (Rick)]

Originally Posted by gregbradley

I wonder if that is true only for PI. In CCDstack I have done stacks using both and saw no real difference in the reported stats.

Hi Greg,

The difference in SNR is a consequence of the mathematics and not due to any specific implementation of stacking. It's provable for those who are interested in the theory (see the PI doc for ImageIntegration.) It's also borne out in empirical testing: http://www.clarkvision.com/articles/...cking-methods/

Cheers,
Rick.