How many subs is too many subs?

[vanwonky (Dave)]

Howdy All,

Just trying to understand the whole multiple subs thing.

I see some people are taking a huge amount of subs with the same timing and ISO. I can understand the taking of subs of different lengths and maybe ISO so you can use them for masking etc. but the use of images with the same length/ISO does do my head in a bit even though this is standard practice.

Lets say I take a 60sec/800 iso image of M42. Why not just use that same image 20 times in your stacking program? Is it not the same data if I take it once or 20 times?

Admittedly I am yet to understand what 'magic' goes on with the software when stacking like images so I might be missing something quite basic here. Is it just noise reduction that is the benefit?

Otherwise what is an optimal amount of subs?

Cheers
Dave

[Peter.M]

Originally Posted by vanwonky

Lets say I take a 60sec/800 iso image of M42. Why not just use that same image 20 times in your stacking program? Is it not the same data if I take it once or 20 times?

Using the same sub 20 times will not do anything to the image, the noise in the image is the same and so the softwares stacking method has nothing different to work with.

Originally Posted by vanwonky

Admittedly I am yet to understand what 'magic' goes on with the software when stacking like images so I might be missing something quite basic here. Is it just noise reduction that is the benefit?


Cheers
Dave

Yes noise reduction is the benefit, the software exploits the fact that signal is not random and noise is. As a simple example think of one pixel that is bright in a picture, if this due to signal then it would be in any subsequent frames and the average of those pixels would be a bright pixel. Think now that the bright pixel was due to noise, it may not be in subsequent frames (because it is random) so a bright pixel will be averaged with other sub frames that are dim, this will result in the pixel looking dimmer. And the more subframes you take that dont contain the bright pixel the dimmer it would look.

Obviously this is very simplistic but it should give you an idea of whats going on.

[MrB (Simon)]

Damn, there is a great website that shows examples of the results of stacking with mouse-over images, but buggered if I can remember it!

[vanwonky (Dave)]

Originally Posted by Peter.M

Yes noise reduction is the benefit, the software exploits the fact that signal is not random and noise is. As a simple example think of one pixel that is bright in a picture, if this due to signal then it would be in any subsequent frames and the average of those pixels would be a bright pixel. Think now that the bright pixel was due to noise, it may not be in subsequent frames (because it is random) so a bright pixel will be averaged with other sub frames that are dim, this will result in the pixel looking dimmer. And the more subframes you take that dont contain the bright pixel the dimmer it would look.

Obviously this is very simplistic but it should give you an idea of whats going on.

Right so noise is the key. But are you saying that the stacking software will do nothing significant if anything at all (with exception of noise removal) with the image data (nebula etc) as far as improving it in these similar images?

Originally Posted by MrB

Damn, there is a great website that shows examples of the results of stacking with mouse-over images, but buggered if I can remember it!

That would be interesting to see Simon.

[MrB (Simon)]

Originally Posted by vanwonky

That would be interesting to see Simon.

Found it!
It's on the DSS (Deep Sky Stacker) website:
http://deepskystacker.free.fr/english/theory.htm

[vanwonky (Dave)]

Originally Posted by MrB

Found it!
It's on the DSS (Deep Sky Stacker) website:
http://deepskystacker.free.fr/english/theory.htm

Thanks Simon. That is a good read and the roll-over image examples are a great way to see the process.

"This is the reason why you should use as many dark/bias/flat frames as possible. On the practical side, 20 frames is a minimum if you want to not add too much noise, and 50 to 100 will give you really nice and (almost) noise free masters.."

100! That is quite some investment in time there after your subs. Might have to try it one week

Cheers
Dave

[whzzz28 (Nathan)]

I know a lot of people like to just make some masters and re-use them which will certainly the time.

Going to tack on a bit here as i was thinking about this last night:
I often take shots at 1600 ISO for 300seconds which works well.
I see a lot of people however shooting at ISO 800 and even 400.

So what am i better off doing:
a) 1600 ISO for 300seconds, 10 subs or,
b) 800 ISO for 600seconds, 10 subs or,
c) 800 ISO for 300seconds but take double the amount of subs (20)

According to the DSS link above, the longer exposure times are recommended, but with half the amplification, will ISO 800 get as much detail as 1600 (at double the time or double the amount of subs)?

[E_ri_k (Erik)]

I was wondering the same thing when i started imaging. I read the DSS website, and i was thinking, is say a 1 hour exposure made up of 6 x 10 minute subs really a 1 hour exposure? Or it realy only a 10 minute exposure but with a very high SNR?

[tlgerdes (Trevor)]

Originally Posted by vanwonky

Right so noise is the key. But are you saying that the stacking software will do nothing significant if anything at all (with exception of noise removal) with the image data (nebula etc) as far as improving it in these similar images?

Longer subs will give you more signal, as will more aperture or a more sensitive camera.

More subs, give you better signal to noise ratios. Signal stays constant while noise lowers. A better signal to noise ratio will allow you to stretch the data in post processing and bring out more detail before you start lifting the noise in the picture.

[tlgerdes (Trevor)]

Originally Posted by E_ri_k

I was wondering the same thing when i started imaging. I read the DSS website, and i was thinking, is say a 1 hour exposure made up of 6 x 10 minute subs really a 1 hour exposure? Or it realy only a 10 minute exposure but with a very high SNR?

That depends on the algorithms used to stack your subs. Remember you are just counting photons.

Are 100 photons of light and 10 photons of noise any better than 1000 photons of light and 100 photons of noise?

[Peter.M]

Originally Posted by whzzz28

So what am i better off doing:
a) 1600 ISO for 300seconds, 10 subs or,
b) 800 ISO for 600seconds, 10 subs or,
c) 800 ISO for 300seconds but take double the amount of subs (20)

As far as I am aware from trolling through all the mathmatics when I was using a DSLR. The ISO reading does not make any difference to the signal generated.

If you think about this it makes sense, in film cameras the ISO would make the film more sensitive to the light but in DSLR photography this is not possible. ISO cannot make the sensor more sensitive. It also cannot make more light hit the sensor, so what it does is amplify the signal. This also amplifys the noise and you end up with a situation like tlgerdes suggests, 10 noise and 100 signal at iso 400 would go to 100 noise and 1000 signal at iso 4000.

Using the above example it would suggest that the lowest amplification factor is best (iso). But it turns out that the iso rating of a camera changes its internal noise. For this reason there is a optimal ISO rating for each camera that has the lowest noise. The theory suggests that you should always use this optimal iso rating for the best SNR images.

[whzzz28 (Nathan)]

Interesting.
I think the next time i get out there ill try 800 ISO and see how i go.
I guess i was thinking that 1600 @ 300seconds will be linear to 800 @ 600seconds without understanding how ISO actually works.

Thanks all.