A stupid astro photography question!

I do of course see the merits in stacking different subs in order to minimise noise and it makes sense to me that if you have 20 copies of a single sub the software sees the noise as part of the image and amplifies it like in John's example- whereas noise being inherently random it will be different in each individual sub and therefore easier to subtract with a stack of different subs.

But what about these new Sony sensors 694 and 814 that Atik and SX claim are so quiet that dark calibration frames are not required?? If there is no noise to remove and only clean signal does that mean the image can be processed slightly less traditionally given the new clean chip and the greatly improved signal to noise ratio?

If we took 15 relatively short subs with these chips and copied each one so we had 30 (i.e.2 out of 30 are identical)- can I then cheat the system and go to bed at a reasonable hour? I know the answer is likely "No- have some patience and man up, it takes as long as it takes" but I won't rest 'till I see a side by side comparison!!! :laugh:

The difference between stacking multiple images - and the same image multiply is this:-

Each separate image is a sample of data (each image is a measurement of how much light a particular pixel should have). This is subject to Poisson statistics - there is a variation between values in one image compared to another. If any particular pixel value is "meant" to be N - we would expect each image to deviate (Poisson noise) such that the captured value is typically within N +/- SQRT(N). Increasing the number of unique images that are combined means the square root of the summed N is a smaller fraction of a bigger N compared to a smaller N. If N = 4 the square root is 2 - Poisson noise is 50%. If N = 100 then square root is 10 - Poisson noise is 10%, etc.

By stacking the same image multiply we are not increasing our sample size so we are not getting better signal to noise ratios - we might as well take a single image and multiply all pixel values by some factor - clearly not a useful thing to do.

One can't subtract noise - we can reduce it's effects by increasing the volume of data. When we subtract dark images from light images we are not subtracting noise we are subtracting signal associated with dark current (thermal electrons) - the noise within our dark images will always be *added* to our light images - but this is better than having dark signal in our images.

Combining a small number of images multiply is a bit like taking a very small number of persons and making a judgement on what the nation will vote for at the next general election.

Dave is right - no such thing as a stupid question.

Mark C.

stang wrote: But what about these new Sony sensors 694 and 814 that Atik and SX claim are so quiet that dark calibration frames are not required?? If there is no noise to remove and only clean signal does that mean the image can be processed slightly less traditionally given the new clean chip and the greatly improved signal to noise ratio?

There is more than one source of noise. See www.mssl.ucl.ac.uk/www_detector/optheory/darkcurrent.html

Even without dark calibration frames, you will still have noise in the image. It cannot be removed entirely but can be reduced (e.g. cooling) and taking multiple images (subs) and increasing the s/n ratio.

stang wrote: but I won't rest 'till I see a side by side comparison!!!

Fill your boots - but make sure to post back here with your results...

Mark - nicely explained thanks.

Yes guys, Thanks for the explanations - makes a bit more sense to me now. Will still probably have a play with some old data for pig iron!