How many Darks?

[JohnnyD333]

Hello SGL,

A new question for the day, actually its more of a poll type question. I am curious about the amount of Dark frames everyone takes in their photos. I feel that the average I see in SGL is around 10. Is this a good amount, should I use more or less? Does it depend on the DSO I am imaging? Any advice is much appreciated.

[Merlin66]

IMHO five or seven into a master dark works for me.

[roundycat]

I use thirty to get the added noise down to around 2%. That means it is likely to be as low as in the rest of the picture so the master dark will not add any noticeable noise.

The reasoning behind all this is almost entirely mathematical. Your chosen subject has no bearing on the number of darks you use but if the target is faint or your exposures too short you may well end up wallowing in excessive noise. It will not neccessarily be a function of dark subtraction. See here Astrophotography.

You can get away with just subtracting a master bias from each light provided your camera does not produce lots of thermal pixels. If you do this you should certainly dither between frames.

Dennis

[JohnnyD333]

I learn something new everyday Roundy. I could never figure out if the object had any bearing on the number of Darks I needed to add. But I think I have to experiment with this over time. You use around 30 Darks, Merlin said 5-7, I just need to find that sweet spot between too little and too much. Thanks for the info guys.

[ollypenrice]

Listen to Dennis for he is wise and well informed on these matters! I take about thirty as well.

However, if you are having to do them each night with a DSLR to match temperature then I would say that you could adjust the number in the light of the target. You shouldn't but you could. For example, the Double Cluster. Here the pixels are either bright or faint. Maybe the edges of stars will be in between but the area these cover is small. So I think noise will be less of an issue than in an image in which you know you are going to be stretching the socks off the extended faint stuff and stretchng until you reach the noise limit. Here you need to be more meticulous. And, of course, you need a lot more subs.

If you can contrive a temp contollable environment, like a fridge, you can create a darks library at a variety of likely real world imaging temperatures and this would make taking thirty less onerous.

Olly

[JohnnyD333]

Thanks for the great info Olly. I have learned over the past year, going from winter to summer imaging, how important temp control can be. And often I would read that people would take their darks after their imaging session. I now know why software is used to take looping exposures with their cameras. Taking 30 darks at just 300 sec would take an hour and a half, let alone going 600 sec:eek: My new theory is to take as many darks until my brain starts shutting down:) Anyway, I really appreciate you and Dennis helping me on this.

[Psychobilly]

Trouble is if there is a significant temp change across an imaging session and the cameras noise levels change radically with changing temp then how do you cope?

Do you apply a temp matched "sub-master" dark to each frame?

Peter...

[JohnnyD333]

Psych, I have to sit outside and take the darks because my house is air conditioned and the temp outside is too hot. I just imagine making specific temp matched subs could be quite a task. You are much more knowledgeable in this field than myself, if you ever test it out make sure you write a book and send me a copy:)

[Psychobilly]

I Haven't used darks for ages... but that flies totally against conventional wisdom

excuse the link to a previous thread... the Heretical No darks bit starts a little way into it...

http://stargazerslounge.com/imaging-image-processing-help-techniques/136090-dslr-horizontal-banding-4.html#post1761948

[RobH]

I'm with Dennis and Olly......30 minimum, same for flats, and for bias, as you only need to shoot them once every few months, I go for 100.

Use a fridge, as Olly suggests. That way, you can shoot your darks all day long and build up a good library.

Rob

[Psychobilly]

Mis-matched Darks can be worse than useless...

I suppose that the ideal situation for a DSL ris to leave the in camera long exposure noise reduction on taht way each "light" will ahev a v ery closely matched dark applied.. however this waste valuable "light Sub" exposure time... and give me more signal to play with anyday...

I would get to know the noise performance of your particular camera.. not the Generic Chatter you read online about a particualr model .. but do some tests on the actual camera you are goign to be using... I do this with every DSLR I buy for astro use... and I have had a few... some haven't hung around for very long...

Peter...

[RobH]

Mis-matched Darks can be worse than useless...

Peter...

Quite true

I don't know what it's like with a DSLR, but with my old Meade DSI, if the darks were more than 2 degrees out with the lights the results were awful. Both CCD's I have now are much more forgiving, the 16HR, unless you're doing very long exposure binned subs, is so quiet that you can get away without darks if you have to. The H18 absolutely needs darks, but you can be a bit more flexible on the temperature range (the cooling tendss to waver a fair bit so it's a good job!)

Rob

[rocketandroll]

A few people have told me that 20 is about the right number.... you hit the law of diminishing returns after that.

I'd say 20 - 30 sounds about right if you can, but at least 10 minimum.

I've never used more than 20.

Ben

[RobH]

The law of diminishing returns really kicks in at about 50, but noise is down a long way at 30...a good bit more than 20. Shoot a bunch odf lights and stack different numbers of them and then compare....it's pretty obvious when you do this.

If you want a really low noise floor, then 100 is about the maximum....above this you really don't notice any difference.

I shot 100 x 10 minute Ha subs of the rosette a few years back....took ages, and, due to not getting the right focal reducer/spacing, I had rotten coma and the image was useless.......useless but very noise free though!

[RikM]

30x darks for me too. 30x flats, 90x bias and I try and get 60 or 90 usable lights. I have a thing about threes and multiples thereof...

[dph1nm]

Just remember - if you take n darks and n lights then by subtracting your master dark you will increase the read noise contribution to your final image by sqrt 2. So if your subs are short enough to be affected by read noise then you really want many more darks than lights.

NigelM

[obscura]

Shouldn't darks, bias, flats be take with the same scope/mount config as the lights?

Using a fridge flies against this. Or have I missed a trick here. But I like the sound of it.

[RobH]

Shouldn't darks, bias, flats be take with the same scope/mount config as the lights?

Using a fridge flies against this. Or have I missed a trick here. But I like the sound of it.

Since both darks and bias are taken with no light entering the camera, it's irrelevant whether the camera is actually on the scope or not, the only important thing being that the dark temperature matches that of the lights pretty closely.

For flats though, you are making a record of any imperfections in the optical train, and so everything should be exactly the same as when you were taking your lights....in other words, change nothing before you take your flat frames.

HTH

Rob.

[ollypenrice]

[quote

I shot 100 x 10 minute Ha subs of the rosette a few years back....took ages, and, due to not getting the right focal reducer/spacing, I had rotten coma and the image was useless.......useless but very noise free though!

Awe, that's terrible!!!

Olly

[johnb]

How many well I use 50% of the number of lights taken so 50 lights, 25 darks etc - I do however use Dark Library which is a real bonus and it works well as long as you keep it fresh

Have a look: Telescope Reviews: Dark Library V1.2

John B

[JohnnyD333]

John,

That is a really awesome piece of software. Although I dont have a Canon, Im very impressed that someone developed this.

[obscura]

Thanks Rob. That's very useful to know. That makes imaging less time consuming than thought. Assuming I can match light temps to fridge temps as near as possible.

This could be done in advance at various temps as indicated above.

That is a very big step forward.

Thank you

[RobH]

How many well I use 50% of the number of lights taken so 50 lights, 25 darks etc

John B

The problem with that approach is that if you only have, for example, 20 lights, then you just use 10 darks, which isn't enough to get the noise down in the master dark, so you end up adding noise to your final image. For 20 lights, I'd still go for 25-30 darks.

With a dark library, after some time has been spent acquiring them, this isn't a problem of course.

Rob

[blackparticle]

Trouble is if there is a significant temp change across an imaging session and the cameras noise levels change radically with changing temp then how do you cope?

I've found that with a DSLR I can't.

Combined, I ended up with sets of dark frames such as this -

With the difference in ambient temp and the change in sensor temp I've found that a stack of separate darks, even with histogram matching etc, generally aren't working for me. Most of the time they end up adding an amount of inverse colour noise which builds up the more subs I add.

I've gone back to turning the DSLR noise reduction feature back on and letting the camera do it on a frame-by-frame basis.. at the sad expense of imaging time I know but the overall "noise" appears to be much more even and controllable in post-processing.

Alan.