Is there a limit to stacking or should you keep gaining data regardless?

[Trippelforge]

My son and I took 30 x 30 second exposures of Andromeda. Once stacked the image we got was pretty amazing, so the assumption we had was that taking even more images would keep increasing the detail and quality overall. So we went out and took another 30 x 30 and stacked those on top of the others, tet we couldn't discern much in the way of any changes. 

So back to the title question; is there a limit or massive jump that needs to be taken to really improve the end result? Such as perhaps 30 more exposures isn't going to do much, but for example 90 more exposures should?

Hopefully that makes sense, as I always assumed gathering more and more images could increase quality.  

 

[Elp]

There is a point of limited returns yes, if you think about it logically. You'll only pick up photons as they're being received, what you're effectively doing by taking lots of images and stacking is increasing the probability and occurrence of photons being captured and the more you expose the better chances of fainter ones being captured and you end up averaging out the total signal within the stack and also the noise so the stack eventually gets better defined and faint signal shows through. You can experiment with this yourself, try a short stack and compare it with a longer one. When post processing increase the contrast or levels a lot so you can see the signal and noise clearly, the short one will look blotchy and less defined.

Dr Robin Glover of Sharpcap has a good video talking about settings to use and max exposure in one of his lectures.

Astro images get their "wow" factor after post processing where you can really push the data and apply some sort of noise reduction to make them smoother.

[newbie alert]

So you have 30 mins of data...  You will get a far better image if the integration time runs into a good few hours.. the idea of stacking is to add signal above the noise from the camera.. so the point of deminished returns is by far in the 10's of hours

[Ouroboros]

The statistical mathematics of this is that noise is proportional to the square root of the number of samples.  So you have to image for four times longer to reduce the noise by a factor of two.   So one would expect the average of 60 x 30 second sub images will have a noise level  of 1 divided by the square root of 2 times better than the average of 30 x 30s subs.

i.e. 

Noise of 60 subs = (1 / 1.41) x Noise of 30 subs.

There is a diminishing return of taking more and more subs. If you’ve already got 2 hours of data you have to image for 8 hours in total to see a factor of 2 improvement in noise.  In the end it comes down to how much time are you willing to devote to this image. 

Edited November 3, 2022 by Ouroboros

[Trippelforge]

15 hours ago, newbie alert said:

So you have 30 mins of data...  You will get a far better image if the integration time runs into a good few hours.. the idea of stacking is to add signal above the noise from the camera.. so the point of deminished returns is by far in the 10's of hours

Shorter exposures = less noise, so a ton of shorter exposures will keep building up the image. If I try take longer exposures the noise will increase to the point of not getting as much "good" data out of each frame.  So if I am understanding right it's all about balancing and 2 hours vs 30 minutes will net a better image. So I probably should get more. Sorry just trying to make sure I am understanding. 

15 hours ago, Ouroboros said:

The statistical mathematics of this is that noise is proportional to the square root of the number of samples.  So you have to image for four times longer to reduce the noise by a factor of two.   So one would expect the average of 60 x 30 second sub images will have a noise level  of 1 divided by the square root of 2 times better than the average of 30 x 30s subs.

i.e. 

Noise of 60 subs = (1 / 1.41) x Noise of 30 subs.

There is a diminishing return of taking more and more subs. If you’ve already got 2 hours of data you have to image for 8 hours in total to see a factor of 2 improvement in noise.  In the end it comes down to how much time are you willing to devote to this image. 

Thanks for the detailed explanation. So "how much time" basically means I probably can keep going until I hit the diminishing return wall. Which I probably will notice eventually when the stacked image because crap. I told my son that we may want to move to Andromeda each night, and snap 15 minutes just because. And eventually after several hours may have a much more detailed result. I just didn't want to keep going into the several hour range if there wasn't much point.

 

[Ouroboros]

@Trippelforge Noise is quite a complex subject, as you might expect.  There are different contributions to noise in an image. Some of the noise is random and some is systematic.   We can point you at various sources of information if you’re really interested in getting into the subject.  But in the end it often comes down to a compromise in selecting the best exposure times that suit your equipment and sky conditions.

It would be helpful if you told us what sort of telescope and camera you’re using, and the type of mount. Is it an equatorial mount? Are you able to guide the mount?  What’s the light pollution like where you are? 

Edited November 4, 2022 by Ouroboros

[ollypenrice]

Yes, as above. You would certainly do better to take longer sub exposures if you can. This is less true of very modern cameras with low read noise but that probably isn't what you're using?

We do take a long time over our pictures in AP, though. My own Andromeda probably had about 20 hours.

Olly

[Trippelforge]

2 hours ago, Ouroboros said:

@Trippelforge Noise is quite a complex subject, as you might expect.  There are different contributions to noise in an image. Some of the noise is random and some is systematic.   We can point you at various sources of information if you’re really interested in getting into the subject.  But in the end it often comes down to a compromise in selecting the best exposure times that suit your equipment and sky conditions.

It would be helpful if you told us what sort of telescope and camera you’re using, and the type of mount. Is it an equatorial mount? Are you able to guide the mount?  What’s the light pollution like where you are? 

 

Sorry, I am not sure how to put a signature in my posts here. I noticed the about me section, but nothing else.

Barska 80mm ED @ f/7
Celestron CG-4 with tracking (non-connecting)
Canon 500D
Bortal 5

 

[Ouroboros]

26 minutes ago, Trippelforge said:

 

Sorry, I am not sure how to put a signature in my posts here. I noticed the about me section, but nothing else.

Barska 80mm ED @ f/7
Celestron CG-4 with tracking (non-connecting)
Canon 500D
Bortal 5

 

Looks not unlike my first set up which was Canon 450D on an EQ5 tracking (unguided) mount under similar skies. Mine was a faster scope f/5.

So the duration of your subs is limited by the mount and how well you can polar align. I reckon I was limited to exposures between 60 and 120 seconds when well polar  aligned.   Without a polar scope your 30 seconds might be about what’s achievable.

Then we must consider ISO. As this site puts it “There is an optimal ISO value for each DSLR, where your specific sensor provides the optimal balance between read noise and dynamic range.”  Read noise occurs each time an image is transferred off the camera. 

They reckon for the 500D the optimum ISO is 1600. For my 450D I know it is 400 or 800.

As Olly says, it’s better to take longer subs if you can (up to a point).  You don’t want light pollution to become the dominant signal.  But I think your limiting factor is the mount. If you can tweak the polar alignment to allow for 60 second or more that would be worth doing. 
 

 

[newbie alert]

4 hours ago, Trippelforge said:

Shorter exposures = less noise, so a ton of shorter exposures will keep building up the image. If I try take longer exposures the noise will increase to the point of not getting as much "good" data out of each frame.  So if I am understanding right it's all about balancing and 2 hours vs 30 minutes will net a better image. So I probably should get more. Sorry just trying to make sure I am understanding. 

Depend on the noise, light pollution and target brightness... There a few different types of noise which will all have an effect on your final image.. background sky plays a part in how much ( good) signal hits the sensor

Depending on how bright your target is, some targets are so dim that you won't get much signal from, so you can't build on signal if it's not contributing to filling the well..

 

 

 

[Trippelforge]

2 hours ago, Ouroboros said:

Looks not unlike my first set up which was Canon 450D on an EQ5 tracking (unguided) mount under similar skies. Mine was a faster scope f/5.

So the duration of your subs is limited by the mount and how well you can polar align. I reckon I was limited to exposures between 60 and 120 seconds when well polar  aligned.   Without a polar scope your 30 seconds might be about what’s achievable.

Then we must consider ISO. As this site puts it “There is an optimal ISO value for each DSLR, where your specific sensor provides the optimal balance between read noise and dynamic range.”  Read noise occurs each time an image is transferred off the camera. 

They reckon for the 500D the optimum ISO is 1600. For my 450D I know it is 400 or 800.

As Olly says, it’s better to take longer subs if you can (up to a point).  You don’t want light pollution to become the dominant signal.  But I think your limiting factor is the mount. If you can tweak the polar alignment to allow for 60 second or more that would be worth doing. 
 

 

 

I believe it is 1600, that is what I have been shooting at. I do have a polar scope and have spent a lot of time calibrating everything. Polaris stays on the circle line pretty dead on when I spin it, so I am hoping it's good enough. I had a hell of a time tweaking it though with those stupid tiny set screws (lol). 

However I do notice a tiny bit of trailing when I go over 40-45 seconds. It's very slight but obviously builds from there so I am not tracking perfectly. I don't know how to get it aligned any better though, however I can't rule out the motor drive (which I am looking to replace). 60 seconds might be doable though if I don't zoom all the way in on the stars and get critical. 

 

 

2 hours ago, newbie alert said:

Depend on the noise, light pollution and target brightness... There a few different types of noise which will all have an effect on your final image.. background sky plays a part in how much ( good) signal hits the sensor

Depending on how bright your target is, some targets are so dim that you won't get much signal from, so you can't build on signal if it's not contributing to filling the well..

 

 

 

Thank you, I will keep that in mind when we start going down the ladder in regards to magnitude. I suppose at one point we will hit a limit due to LP and need darker skies. The best thing I can get to is Bortal 3, which is about 3 hours away. Although I have already been planning a little camping trip with my son in order to do that. 

[Ouroboros]

@Trippelforge Sounds like you’ve got it  more or less worked out. I got my best polar alignment performance with a polar scope when I carefully placed the little circle at the correct hour angle.  That’s kind of obvious, but can be difficult to achieve in practice. It’s a matter of correctly setting the rotation angle of the RA axis.  This requires using the graduated RA scale on your mount. You may be well aware of this already, but if not this site explains it well, including for your type of mount using the graduated scale on the polar scope itself. You should be able to improve your PA by a factor of 2 at least. 

As you’re probably aware there are little polar alignment cameras that pop into the polar scope hole. Involves having a computer at the scope which might be a hassle.  Plus it’s extra cost. 

Another alternative is to do some wide-field imaging with only the camera + lens connected to the mount. The much shorter focal length achievable will considerably increase your exposure times before star trailing becomes apparent.  Andromeda M31 looks particularly good in wide field. 

[wimvb]

On 04/11/2022 at 13:46, Trippelforge said:

Shorter exposures = less noise, so a ton of shorter exposures will keep building up the image. If I try take longer exposures the noise will increase to the point of not getting as much "good" data out of each frame.  So if I am understanding right it's all about balancing and 2 hours vs 30 minutes will net a better image. So I probably should get more. Sorry just trying to make sure I am understanding. 

Thanks for the detailed explanation. So "how much time" basically means I probably can keep going until I hit the diminishing return wall. Which I probably will notice eventually when the stacked image because crap. I told my son that we may want to move to Andromeda each night, and snap 15 minutes just because. And eventually after several hours may have a much more detailed result. I just didn't want to keep going into the several hour range if there wasn't much point.

 

Shorter exposures does NOT equate to less noise, because cameras have read noise, independent of exposure time. Longer exposures = less noise, or higher signal to noise ratio ("more signal per noise"). In theory, every time you double the total exposure time you already have, you cut the noise to 70%, as long as you've exposed long enough to drown the read noise. The point of diminishing returns is determined by perseverance, or just simply by you saying "that does it for me".

When we progress in astrophotography, most of us (including myself) go from shooting multiple targets per night to shooting multiple nights per target.

[Trippelforge]

On 06/11/2022 at 02:42, wimvb said:

Shorter exposures does NOT equate to less noise, because cameras have read noise, independent of exposure time. Longer exposures = less noise, or higher signal to noise ratio ("more signal per noise"). In theory, every time you double the total exposure time you already have, you cut the noise to 70%, as long as you've exposed long enough to drown the read noise. The point of diminishing returns is determined by perseverance, or just simply by you saying "that does it for me".

When we progress in astrophotography, most of us (including myself) go from shooting multiple targets per night to shooting multiple nights per target.

 

Some people pointed out that when stacking short exposures it increased the signal to noise ratio. So is that simply not true, or does the read noise pile on as well and make things worse unless I extend exposure time? I keep getting a bit confused, obviously I can simply experiment and see though.

My simple situation (as I probably mentioned) is that my son and I took a bunch of 30 second images. Our idea was to hit Andromeda for more 30 second exposures each night for additional stacking. Which obviously the thought process that the stacked image would improve over time. Taking into consideration of your explanation it seems signal to noise won't ever build up enough to become detrimental to the process. 

Sorry I have been getting somewhat confused, as is it seems that I am getting some conflicting information. 

[DaveS]

Ideally your exposure times should be long enough to swamp the read noise by about 3x. How long this is depends on your light pollution, the read noise of the camera, and whether you're doing OSC, mono RGB or NB. The rather noisy CCD cameras that I use tend to need longer exposures, but CMOS can get away with shorter.

[Trippelforge]

24 minutes ago, DaveS said:

Ideally your exposure times should be long enough to swamp the read noise by about 3x. How long this is depends on your light pollution, the read noise of the camera, and whether you're doing OSC, mono RGB or NB. The rather noisy CCD cameras that I use tend to need longer exposures, but CMOS can get away with shorter.

OK so there is a point when the exposure times are so long that read noise etc gets out paced by the "good" image data. I always had assumed that it worked the opposite way. Taking that into consideration I REALLY need to upgrade my motor drive system, 30-40 second crap isn't cutting it. 

[wimvb]

5 hours ago, Trippelforge said:

 

Some people pointed out that when stacking short exposures it increased the signal to noise ratio. So is that simply not true, or does the read noise pile on as well and make things worse unless I extend exposure time? I keep getting a bit confused, obviously I can simply experiment and see though.

My simple situation (as I probably mentioned) is that my son and I took a bunch of 30 second images. Our idea was to hit Andromeda for more 30 second exposures each night for additional stacking. Which obviously the thought process that the stacked image would improve over time. Taking into consideration of your explanation it seems signal to noise won't ever build up enough to become detrimental to the process. 

Sorry I have been getting somewhat confused, as is it seems that I am getting some conflicting information. 

As Dave says, when the noise is mainly from the captured light, it will diminish as you collect more data, either through longer exposures, or through more exposures. But if the noise is dominated by read noise from the camera, then stacking more subs won't help. That's why exposures need to be long enough. In practice, if you see horizontal or vertical bands in the single exposures when you stretch them, this is most likely read noise, and you need to increase the exposure time.

The amount of read noise depends on the camera and its settings. Modern cmos cameras have such low read noise that single second exposures can be long enough. Dslrs and some cmos have higher read noise and need longer exposures. Unfortunately there is no single, universal correct exposure time, and you will need to experiment.

Edited November 8, 2022 by wimvb

[Scott Badger]

I have a QHY 268M (CMOS) and a fairly dark site, and even at 120s exposures of M13, I got the vertical banding mentioned above and very low background ADU values (0.007) in the integration.

In any case, as has mostly been said, longer exposures have a higher SNR because the noise increases to the square root of the signal, i.e. both noise and signal increase over a longer exposure time, but signal increases faster. The trade-off is a higher percentage of unusable subs due to planes, clouds, poor guiding, etc. and the longer the exposure, the more imaging time wasted.

Increasing the number of subs doesn't increase signal, but increases SNR by allowing the averaging process to reduce the random noise closer to zero. So, more detail is exposed because the noise it's buried in (in the individual subs) is stripped away. Think of removing the sediment to expose a fossil (and really, our images are photon phossils...). The more subs, the less noise, but at a diminishing return -- the amount of noise reduction is the square root of the number of subs.

EDIT: I think the general rule is, as long an exposure as equipment and conditions allow, and as many exposures as patience allows.... The one exception might be If seeing is particularly bad, then a more lucky imaging approach of taking thousands of subs, each just a few seconds exposure, can work and improve resolution, if you have the computer power for it.

Cheers,
Scott

Edited November 11, 2022 by Scott Badger
add comment