Darks, Flats, Bias Ratios

[johnb]

Hi

Is there a rule of thumb about the ratio of Darks, Flats and Bias Frames so if I have say

10 x 30 secs - 300 secs of images

How many darks etc do i need ?

Regards

John B

[daz]

For darks, the same number of exposures and the same length - eg 10 x 300 lights, 10 x 300 darks - and ideally at the same temp.

Bias - depends on the camera, but I used 50 for my SXV and created a master bias

Flats - lots and lots of discussions around an optimal number! Go for at least 50% of your lights and see how that works for you... DO REMEMBER that you will need flats for each filter (if using) and the image train should not be moved between the lights and flats.

[Psychobilly]

Whats the camera...?

[johnb]

Thanks for that, really helps

Im using a Sony A350 (un modded)

Regards

John B

[Psychobilly]

I always think its worth people getting to know their own cameras... especially how they behave from a "noise" point of view...

You want to check how the darks change with temperature, and the length of the gap left between "subs"...

The best but least efficient way to handle dark frames is to take interleaved ones as these should provide the best match to the lights in terms of temp etc.. but this is a real waste of time when you would be better of capturing more lights...

If the dark noise is fairly consistent over a range of temperature then you can grab the darks at the end of the session (or in the UK) during those infrequent cloudy interludes....

Peter..

[RobH]

With all calibration frames, the same rules apply as with light frames....more individual subs means less noise when stacked.

I go for 50 bias, and 25-30 darks and flats.

Cheers

Rob

[themos]

If you can, beg borrow or steal the Handbook of Astronomical Image Processing. It will explain the maths behind this question so you can work out the answer yourself.

To summarise, the reason you want to shoot many calibration frames is the same reason you shoot many data frames: reduce noise. If you know how much noise there is in your individual data frames and your individual calibration frames you can then work out how many to shoot so that the calibration does not appreciably reduce your signal/noise ratio.

The calibration frame noise depends only on your sensor (including its temperature) but the data frame noise depends on the quality of your skies as well so the answer to "how many calibration frames?" depends on your light pollution. Any "rules of thumb" will thus apply to a particular equipment setup and location.

Perhaps a better question would be : "How do I measure the signal/noise ratio of my data and calibration frames?"

The Handbook of Astronomical Image Processing with AIP4Win Software by Richard Berry and James Burnell

[Psychobilly]

I knew the mathematicians would be a long with the explanation....

[roundycat]

the problem with AIP is that they were working on the book when computers could barely manage to work out 14 squared. They suggest ten darks I think and that will reduce the added noise (uncertainty in the master dark) to about 5%. That is not low enough and could well be on a par with the read noise. All noise contributions should be kept low, I use 20 darks as the added noise is then about 2% and together with low read noise you will get a fair result. To take as many darks as lights is quite wrong, if you only have five lights your master dark made from 5 darks will be a mess.

The same basic thing applies to bias and flats.

Dennis

[themos]

I am pretty sure that they don't suggest a single value (like 10) for the darks but explain how mathematical considerations (quantifiable imaging parameters) lead to a certain choice for a desired result

if you only have five lights your master dark made from 5 darks will be a mess.

I think they suggest using a longer exposure time for the darks and then scaling them back. Isn't that what their "Image-Times-Five" rule is all about?

[dph1nm]

With 30sec exposures I wonder if you nned darks at all. Check they are doing something useful before you spend ages taking them. I certainly don't need them for my Canon 1000D in that time.

NigelM

[RobH]

With 30sec exposures I wonder if you nned darks at all. Check they are doing something useful before you spend ages taking them. I certainly don't need them for my Canon 1000D in that time.

NigelM

I agree. At 30s bias will deal with most issues, and stacking using median combine, SD mask or sigma clip algorithms should get rid of any hot pixels, unless your tracking is absolutely perfect.

Cheers

Rob

[ngc2403]

Maths Yeah .......

I agree with Rob, darks might be over kill for 30second frames,

Bias frames will sort out most of that noise.

flats ...... do i really want to step into that again?

well as a rule of thumb they are hard to understand.

if done wrong make an image worse.

for long exposures flats are the only way to remove the bulk of the noise.

take as many Bias as you do Flats, like so many things in astronomy, more is better.

i do try but usually can't be bothered taking more than 30 or 40 flats However if i think the data is really good then i would make the effort

alasdair

[RobH]

Alasdair, sorry, I had to correct a couple of statements that you made which could easily confuse matters

Maths Yeah .......

flats ...... do i really want to step into that again?

well as a rule of thumb they are hard to understand.

if done wrong make an image worse.

Correct, but they are easy to do and people shouldn't be put off from using them. Just make certain that they are bias subtracted before applying them or they won't work properly.

for long exposures flats are the only way to remove the bulk of the noise.

This is not right. Flats are used for subtracting optical effects such as dust bunnies and vignetting from the images. Darks are used to remove noise from long exposures.

alasdair

[themos]

I think calibration is not meant to remove noise if by "noise" we mean inherently unpredictable variations. Calibration removes any systematic effects that we can predict (dead/hot pixels, amplifier glow, sensor patterned response, dust doughnuts, vignetting). That's why calibration frames need to be averaged, we don't want them to have significant unpredictable patterns in them. The only way to remove "proper" (i.e. unpredictable) noise is to shoot many data/calibration frames and combine them.

[FraserClarke]

I think calibration is not meant to remove noise if by "noise" we mean inherently unpredictable variations.

Absolutely

You can't remove noise -- that is why is it noise... what you can remove is unwanted signal from sources other than your object. but that's all in the semantics

re: RobH. Flats do remove pixel-to-pixel sensitivity effects, as well as optical effects. Depending on the camera and the type of exposure, this can be more important than the dark correction.

[adamsp123]

I have taken 40 darks at various exposure lengths in my fridge, Ice box = 0 degrees C and bottom shelf +5 degrees C , which at the mo seems to cover most recent nightime temps. These are in my Dark library as well as 50 bias frames.

After imaging I use the white T shirt method and take 40 flats after each session (camera set to auto exposure) without moving anything in the imaging train.

It seems to work nicely for my canon 1000D

[RobH]

re: RobH. Flats do remove pixel-to-pixel sensitivity effects, as well as optical effects. Depending on the camera and the type of exposure, this can be more important than the dark correction.

Hello TD....You are quite correct of course. Pixel to pixel variation are removed by flats.

The main use of flats though, certainly as it needs to be understood by folks who are starting to get into calibration, is to remove optical train issues...all those horrible dust bunnies that folk who don't keep their kit as clean as it should be (like me ) get, plus vignetting etc. This is why I've steered clear of getting bogged down into technicalities with my explanation....the whole calibration issue can be daunting enough for newcomers as it is!!

Cheers

Rob

[themos]

On a slightly pedantic note, do we think that shooting calibration frames comes under "Imaging techniques" or under "Image Processing techniques"? I think the former.

[RobH]

I'd say the latter, my logic being that calibration is part of processing as opposed to imaging itself. A very grey area though, as calibration is obviously done using the scope.....a subsection or sticky devoted to calibration might not be a bad idea...there are always a lot of questions about this.

[roundycat]

I wrote this Astrophotography a while back, it might be helpful.

Dennis