Stacking v PE

[traveller (Bo)]

I am not sure if this was discussed before, but what are the advantages and disadvantages of stacking shorter duration images from a mount with larger PE versus a more accurate mount and fewer but longer exposures?
Logic tells me that a more accurate mount with longer duration will give you more details, but is that all? Would stacking give you similar results?
Can someone assist?
Thanks, Bo

[WingnutR32 (Sam)]

Stacking will only give a better signal to noise ratio. If you have 30 images of 30 seconds stacked, you still only have 30 seconds of signal, but your noise will decrease.

If you have 1 image of 30 minutes, you will have a heap of detail, but possibly high noise (depending on hardware).

So no, stacking won't give similar results, stacking is only used to improve the signal to noise ratio.

[traveller (Bo)]

Thanks Sam,
So more accurate mount, short bursts to decrease noise and longer bursts to increase detail.
Heaps to learn still.
Bo

[RickS (Rick)]

Originally Posted by WingnutR32

Stacking will only give a better signal to noise ratio. If you have 30 images of 30 seconds stacked, you still only have 30 seconds of signal, but your noise will decrease.

If you have 1 image of 30 minutes, you will have a heap of detail, but possibly high noise (depending on hardware).

So no, stacking won't give similar results, stacking is only used to improve the signal to noise ratio.

I disagree with the assertion that stacking short exposures results in less noise than a single exposure of the same total length.

There are several different sources of noise. A major and unavoidable source of noise is called "shot noise" and is a consequence of the quantum nature of light. The magnitude of the shot noise is determined by the square root of the number of photons detected and will be the same so long as the total exposure time is the same.

Other sources of noise are internal to the sensor itself, e.g. read noise. This type of noise will be worse in the situation where you are doing multiple exposures as you will incur this noise multiple times.

You will get the best S/N ratio by doing one long exposure, however, that may not be practical for a number of reasons, e.g. exceeding the full well depth of the sensor, inability to guide accurately for long periods, the possibility of losing an exposure due to clouds, etc.

You'll get the best results by making your exposures as long as you can practically make them given the constraints such as those mentioned above...

Cheers,
Rick.

[bmitchell82 (Brendan)]

I was under the impression rick that with read noise you can partially null the effect by the use of bias which is a map of your read noise.

Flats null the effect of light drop off and optical defects

Darks map your Thermal noise

combinations of these help you sort out and make the cleanest possible data for your effort.

Shot noise as your calling it am i right of the assumption that its what we normally call colour noise? and hence the reason for taking many long exposures to average this out and smooth the end image.

In any case ide rather lots of long exposures and now i have started seeing the effects of taking 10-15min RGB's, 15-20 min Lum and 15 min right upto 30 min for Narrowband unless its to do a masking image then ill never go back to short subs and lots of them as it just doesn't show details!

[WingnutR32 (Sam)]

Thanks Rick,

In regards to, "The magnitude of the shot noise is determined by the square root of the number of photons detected and will be the same so long as the total exposure time is the same", I suppose I would have to agree as you lost me in to many big and complex words.

I suppose I was being to 'short and sweet' in my reply as this was the beginners section.. I say this very light heartedly as I am very much a beginner myself having only done a few short exposures myself so far and knowing very little.

I was more answering the question directly from the two examples the OP had given.

I would think that the best result will come from exposing the region in concern for a length of time that will fill up the wells without topping out and then repeat for as many times as you want to 'smooth' out the image (I am sure there would probably be a limit to the amount of stacks used, but I cannot answer that).

But the direct answer to the examples given would still be more detail in a longer exposure (assuming no stack) with lower signal to noise ratio (again, depending on hardware) and less detail and higher signal to noise ratio (smoothness) with shorter exposures with many stacks.

As Brendan said though, many long exposures = WIN!

[RickS (Rick)]

Originally Posted by bmitchell82

I was under the impression rick that with read noise you can partially null the effect by the use of bias which is a map of your read noise.

Flats null the effect of light drop off and optical defects

Darks map your Thermal noise

combinations of these help you sort out and make the cleanest possible data for your effort.

Shot noise as your calling it am i right of the assumption that its what we normally call colour noise? and hence the reason for taking many long exposures to average this out and smooth the end image.

Brendan, I think a lot of people are confused about the difference between unwanted signal and noise. I know I used to be...

Unwanted signal includes things like sky glow, sensor bias voltage and dark current. These are things you don't want in your image, but they are not noise in the technical sense. Bias frames are used to remove the unwanted bias signal from your sensor. Dark frames are used to remove unwanted dark current. Neither removes noise. In fact bias and dark subtraction actually adds some noise! We minimize this noise by taking many bias and dark frames.

Noise is about uncertainty and sampling. Think about doing a series of one minute exposures of a dim object and looking at one specific pixel. Because photons behave probabilistically, you might capture 7 photons in the first exposure, 11 in the second, 13 in the third, 9 in the fourth, etc. This variation is shot noise. By averaging the value over many exposures you reduce the contribution of the noise and get a better quality image.

There is also noise associated with bias voltage and dark current. Read noise is noise generated in the sensor and camera electronics when you read an image. Even if you received exactly the same signal several times, the image you read out of the camera each time would be different due to these (hopefully small) random variations.

Originally Posted by WingnutR32

I would think that the best result will come from exposing the region in concern for a length of time that will fill up the wells without topping out and then repeat for as many times as you want to 'smooth' out the image (I am sure there would probably be a limit to the amount of stacks used, but I cannot answer that).

Sam, I mostly agree with that. You probably don't want to fill the wells completely because they may behave non-linearly when they start to get full, especially if the sensor is anti-blooming.

There isn't a limit on the number of exposures you can stack, but you will find you get diminishing returns. The improvement in signal to noise is proportional to the square root of the total exposure. To make your S/N twice as good you need to increase exposure by 4 times. To make it three times better you need 9 times the exposure, etc.

Originally Posted by WingnutR32

But the direct answer to the examples given would still be more detail in a longer exposure (assuming no stack) with lower signal to noise ratio (again, depending on hardware) and less detail and higher signal to noise ratio (smoothness) with shorter exposures with many stacks.

Still don't agree with this

Originally Posted by WingnutR32

As Brendan said though, many long exposures = WIN!

Amen

[bmitchell82 (Brendan)]

Lets draw it there then

Many exposures @ long intergration times + associated flats = Call it a day boys

[irwjager (Ivo)]

That's an excellent explanation Rick. Fundamentally understanding exposure, the different types of noise and CCD/CMOS characteristics are very important areas. There are a lot of misconceptions and myths around the different types of noise (I have to say that I consider unwanted signal and noise the same thing - not all noise is per definition randomly distributed). There is no one-answer or one formula that will work for all your imaging sessions and every decision you can make will be a trade off.
A mount with a large PE will be a consideration in that trade off, but it will be just one of many factors - it may not be a major factor at all!
Purely mathematically, adding (notice I'm not saying "stacking" here) multiple short frames together is the exact same thing as recording one long exposure. But as Rick so eloquently pointed out; throw the different types of noise in the mix and things become a little more complicated.
One nice thing about having multiple sub frames though, is that, as opposed to one long exposure, you have after-the-fact control of the integration ("stacking"). This allows you to tailor the integration process to the data, gear and its noise profiles.
For example, instead of simply adding the sub frames (which is the same as averaging all frames, then multiplying the result by the amount of frames), there are a variety of other ways to determine the 'true' value of a pixel. Instead of the mean, one could take the median, or the mean of 2 samples around the median, or any other use any other type of rejection algorithm. Using different integration/stacking algorithms some types of noise (cosmic rays, 'hum', inconsiderate neighbors, atmospheric turbulence, etc.) can be better overcome than when using "nature's" integration method.

Cheers,

[traveller (Bo)]

I am sure I just heard several A380s fly overhead, but thanks for the detailed explanation

[Peter.M]

I'm not sure people here have addressed the question asked here, while your answers were what I think he wants to ask. My answer to your question would be an accurate mount with longer exposures will always beat a less accurate mount with periodic error. The reasons why are outlined above. To add to this the good mount can always do shorter exposures but the bad mount cannot do the longer ones.

[RickS (Rick)]

Good point on the different integration options, Ivo.

Bo, The concepts can be a bit tricky at first but it's worth trying to figure it out. The five Signal to Noise articles here are a good read: http://www.stark-labs.com/craig/articles/articles.html

Cheers,
Rick.

[Peter.M]

not to mention the nature of periodic error will ruin long exposures but also some of the short exposures and if you have multiple defects in the gears then short exposures may show signs of pe