How many darks required???

[sergeC14]

Hello Gents,

I read somewhere on the web that using too few darks for calibration actually ADDED noise to the image...

Is this true?

I just spent 2 nights acquiring Hickson 44 and I have 36x10 min subs. How many darks should I use for calibation?

It's just that usually I use less darks than subs and my background is pretty noisy.

Thx

Serge

[Freddie]

I aim for a minimum of 20. You can keep a library of darks if you have a set point cooling camera to save a bit of time.

[sergeC14]

thx freddie.

Ok, but would 20 be enough for 36 subs?. Also, one thing I have noticed is that the sensor changes with time, so darks library is only useful for a couple of weeks of intensive CCD use.

Serge

[Merlin66]

Yes,

20 darks is more than enough.

I use AA5 and an ATik314L and usually only use 9 flats (for a master flat) with good results with the spectroscope.

[Gina]

Yes,

20 darks is more than enough.

So how many darks (at each temperature and exposure) would you recommend?

[Merlin66]

Gina, et al,

I've been using AstroArt since I moved to CCD/DSLR a few years ago.

I find for spectroscopy a law of diminishing returns....the added noise from multiple darks v's the outcome. Yes, up to a point the summing of darks for a master adds more signal than noise (sqrt signal) but I find 8-10 max darks does the job as well as any.

(see Section 9.2, p131 of the AA Manual)

Just my 0.02 euro

[Gina]

Thank you

[sergeC14]

Thx you, that's very informative.

I get alot of baseline "background" noise on my images (black background) whenever I try to bring out the faintest details (e.i faint galaxy arms), by having to stretch the lowest part of the histogram to the max. So, I was hoping that the number of darks I am using was the problem, but apparently, I already use plenty enough, usually around 15 regardless of the number of subs I have to calibrate...

I guess I now have to look elsewhere, but, that being said, my sky is not extremelly dark, so maybe I need look no further...

Regards

Serge

[Ant]

I've always been told to use an odd number of calibration files.

I don't need darks, but I do 31 Flats and normally 101 bias.

I actually don't know why an odd number...

Cheers

Ant

[alpal]

I always thought the rule was " as many darks as lights."

Remember there is noise in the darks too.

This can be cancelled out by using more of them.

The same function applies - the noise is a square root function.

16 darks will give 1/4 the noise.

100 darks will give 1/10th the noise.

[Gina]

I've always been told to use an odd number of calibration files.

I don't need darks, but I do 31 Flats and normally 101 bias.

I actually don't know why an odd number...

Cheers

Ant

Now that is really odd/strange

[Merlin66]

Your arithmetic is correct, but the low level of dark noise in todays CCD just doesn't justify redoing 100's of darks....

Ant,

Why so many Bias frames???? You only need to use them to compensate for adding darks when the exposure times are different?

[Earl]

I use 100 darks and 100 bias.

[Ant]

Your arithmetic is correct, but the low level of dark noise in todays CCD just doesn't justify redoing 100's of darks....

Ant,

Why so many Bias frames???? You only need to use them to compensate for adding darks when the exposure times are different?

When I used to have more time I used to do 101 flats and 101 bias. When your flats (in Ha) take somewhere around the 15-20 second mark - thats 35 minutes if time that I could have in bed

So I reduced the Flats to 31 and have always done the 101 bias as they are really quick.

Would be interested to know how many is best... Maybe I can get to bed even earlier

Ant

[Merlin66]

Feel for me....

My usual exposures are 300 or 600s!!!!

But, seriously - why do you need so many bias frames? What do you do with them??

I'd be interested to see the SNR difference in a typical image reduced with say a 7 frame master dark and a 31 frame master dark.

[Gina]

Yes, that would be interesting With the variety of temperatures and exposures I seem to be using, it would be helpful to reduce the number of darks.

[derrickf]

There are two ways to answer the OP's question:

1. A combination of "suck it and see", ask Dave down the pub and copy what others have done

2. Get an understanding of the underlying maths governing signal to noise ratio in images then calculate the appropriate parameters.

Solution 1 can work well and there has been a variety of good responses to the OP.

A good intro to the mathematical approach (the maths is not too difficult despite lloking a little daunting at first sight) can be found here http://www.hiddenloft.com/notes/DarkSubExp.pdf and here eXtreme Imaging – going deep

HTH

Derrick

[Earl]

i go with 100 as once I have a library which takes a day or two I consider it job done until im not happy with the results anymore..

or get a new cam

[Merlin66]

Derrick,

Good reference there...

I'd say the example needs to be updated to refect the smaller read noise on current cameras (RON e^-1), also the dark signal per pixel (e/s^-1):

Atik314L = 4.5, 0.0002 @ -10c

Titan = 7.4, 0.014 @ 0c

QSI-532 = 11, 0.002 @ -10c

SBIG STX-16803 = 17.7, 0.16 @ -10c

The above data is taken from a recent review of cameras used for spectroscopy.

Noise and electronic gain evaluation of a sample of astronomical CCD camera

[derrickf]

Merlin,

I totally agree the correct parameters for a specific camera model or camera should be used. I characterise the read noise, linearity and dark current etc. of my own cameras using the methods in Berry and Burnell's handbook of Astronomical Image Processing.

Derrick

[sergeC14]

Dear Me!

I still don't really know how many darks to acquire... But 101 darks is way out of the question or I'll still be doing darks by next week with 600 Sec subs... Well, I'm just going to try to use the same number of dark as subs (oops, that's 36 +1!!!) and hope for the best. It's sure is starting to look like black mangic, even though we all do calibrations all the time!

I think I'll go to the Astrosurf french forum and ask the question:). Most of them are a bunch of extremelly boring, ego-inflated, troskist guys (university types), but they are quite experts in this field. Once I get an answer, being a traitor to mother and country, I'll come bak and report on here....

I'll be right back

Serge

[derrickf]

Serge,

there is no real correlation between the number of light subs and the number of darks you need; you will get a reasonable increase in SNR with 15 - 20 dark subs and, if you are using some kind of statistical stacking algorithm for the darks e.g. sigma rejection, you will also have enough subs to reject any anomalous pixels in the darks (cosmic ray hits etc.).

Of course if it makes you feel beter to take as many darks as lights, it will do no harm so it's your choice

HTH

[Merlin66]

Take 101 and be done with it!!

(I still think the answer - if you look at the quality of your images is around 10!)

[sergeC14]

Right!

I am back from the French enemy line!

so, And the answer is:

15 darks, signal to noise ratio is 3.87

20 darks, signal to noise ratio is 4.47

50 darks, signal to noise ratio is ... 7

So, the winner is... Earl!

the more darks the better....

so, a good average would be 20 darks, regardless of the numbe of subs.

So I have to look elsewhere for my noisy background.

Rgds

Serge

[alpal]

I think 100 darks is a nice round figure & about right for a purist

who wants to image right down at the noise level of their system -

especially true for faint galaxies.

It's not so important with super cooled CCD cameras but for a DSLR

in summertime it's essential.

The noise approx. doubles for each 7° rise in temp.

However I don't believe in using old dark libraries as the pixels age over time

so that a pixel that wasn't noisy can become noisy 2 years later.