Page 1 of 1

Optimum ratio of dark frames to light frames?

Posted: Thu Aug 03, 2017 11:31 pm
by oopfan

When I launch the dialog to begin dark frame acquisition it defaults to 10 frames. I have always wondered if there is a magic ratio of darks to lights that achieve the best result. My gut tells me that if I aim to only collect 100 lights then 10 darks are acceptable, but if my goal is to collect 1000 lights then I might want to increase the number darks. Am I onto something or is my gut wrong?


Re: Optimum ratio of dark frames to light frames?

Posted: Fri Aug 04, 2017 8:55 am
by admin
Hi Brian,

it all comes down to a moderately complicated noise calculation. Noise in an averaged frame is SQRT(n) times smaller than the noise in an individual frame, so if you take 16 frames to make your master dark the noise is 4 times smaller, if you take 100 it is 10 times smaller.

Ideally you want the noise from the master dark to be smaller than the noise in the stacked light frames.

For an individual frame

Dark Noise = Read Noise + Thermal Noise

Light Noise = Read Noise + Thermal Noise + Shot Noise

If you have a dig around on the web you will find people have considered this in great detail already and you'll find calculations you can apply to your situation.



Re: Optimum ratio of dark frames to light frames?

Posted: Fri Aug 04, 2017 12:59 pm
by oopfan
Hi Robin,

Thanks for all your help.

I dug around on the web and found some useful information at the DSS website. I could be wrong but my takeaway was that you can never have too many darks!

Well I allowed that to bounce around in my empty skull for a while, and then I came to the frightening realization that when using SC's dark frame capture that I only get one camera-settings metadata file for the entire run. Is the reported sensor temperature for the beginning of the run or the end or is it the average temperature? After being bitten by the temperature-mismatch crocodile the other night I am a bit skittish about this.

Maybe the answer is that temperature drift isn't that important when you have captured a lot of darks and a lot of lights. Noise is one thing but hot pixels are another and it seems like I get a lot more hot pixels as the temperature rises. I have never seen a cooled camera so maybe these issues that I am working through don't apply that much when you have one.

At this point I am thinking that in my situation with my uncooled camera that I am best off not to use SC's dark frame subtraction. Perhaps I should capture darks like any standard frame with its associated metadata file. And then afterwards, group the frames into separate folders according to temperature, and then do dark frame subtraction in DSS.