Is there a limit to short exposures?

[Trippelforge]

My tracking is lackluster right now so I have to stick to shorter exposures. But I am trying to figure out my limitations, such as at what point do short exposure images for stacking become moot. Or do they not, such as would an hour worth of 30 second exposures net a better image than only 30 minutes worth?

Hopefully that makes sense!

Thanks

 

 

 

 

 

[The Lazy Astronomer]

8 minutes ago, Trippelforge said:

My tracking is lackluster right now so I have to stick to shorter exposures. But I am trying to figure out my limitations, such as at what point do short exposure images for stacking become moot. Or do they not, such as would an hour worth of 30 second exposures net a better image than only 30 minutes worth?

Hopefully that makes sense!

Thanks

Optimum minimum sub exposure times using a modern, low read noise cameras can be as low as around 5s or so in typical suburban sky conditions. Many people (myself included) tend to stick to exposures somewhere in the range of 30 - 120s to avoid building up thousands of frames to have to store and stack.

Generally speaking, the longer the total integration time, the better the image in terms of signal to noise ratio (SNR), but it's diminishing returns - I can't remember the exact numbers, but for example, it might be something like 4x the integration time for a 2-fold increase in SNR (that probably isn't right, so someone please correct me!).

[vlaiv]

There are two things to consider when using very short exposures:

1) ability to stack

2) level of read noise

Read noise is the only thing that makes the difference between say 3600 x 1s stack and one single 3600s exposure.

All other noise sources grow with time and if you add two 10 minute subs - it will be the same as one 20 minute sub - except for read noise.

Read noise does not grow with time - but instead grows with number of subs (each time you read out a sub - it gets one "dose" of read noise).

If read noise were 0 - we could use arbitrarily short exposures - as long as we are able to align them for stacking.

In order to stack subs - we must be able to align them. If you image faint stuff - you won't have any data to align on in short sub. If all the stars in your frame are faint - at some point, short enough exposure simply won't capture enough light from those stars to be able to do alignment properly.

Planetary imagers don't have this problem as planets are very bright - and they regularly use exposures that are few milliseconds long.

48 minutes ago, The Lazy Astronomer said:

I can't remember the exact numbers, but for example, it might be something like 4x the integration time for a 2-fold increase in SNR (that probably isn't right, so someone please correct me!).

That is exactly right - quadruple total imaging time to get x2 increase in SNR.

By the way - x2 increase in SNR is not insignificant - it can be difference between unable to detect object and detecting an object (it is said that SNR of 5 is needed to positively identify object, so if you have SNR of say 3 - raising SNR by factor of x2 will make that 6 - and over the detection value).

1 hour ago, Trippelforge said:

But I am trying to figure out my limitations, such as at what point do short exposure images for stacking become moot. Or do they not, such as would an hour worth of 30 second exposures net a better image than only 30 minutes worth?

By the way - they never become moot unless you can't align frames any more. Stacking more subs - however short they are, increases SNR the same way as stacking long subs.

If you stack 100 short subs you will improve SNR x10 over single short sub.

If you stack 100 long subs you will improve SNR by factor of x10 over single long sub - works the same regardless of sub duration.

[ONIKKINEN]

What scope are you shooting with?

If you pair a somewhat fast scope, like a newtonian, and a modern low read noise astro camera you can get away with very short subs under moderately light polluted skies. The more light pollution the shorter you can make the subs after shot noise from the sky overwhelms other sources of noise - called read noise swamping.

I had moderate success with an f/4.2 newtonian and a camera with about 1e- of read noise, and at the time was only taking 30s subs because of mount issues. I also deliberately chose targets at higher declinations because the sky moves slower (relatively) the closer you you shoot to the celestial pole.

Edited October 17, 2022 by ONIKKINEN

[The Lazy Astronomer]

1 hour ago, vlaiv said:

That is exactly right - quadruple total imaging time to get x2 increase in SNR.

By the way - x2 increase in SNR is not insignificant - it can be difference between unable to detect object and detecting an object (it is said that SNR of 5 is needed to positively identify object, so if you have SNR of say 3 - raising SNR by factor of x2 will make that 6 - and over the detection value).

Yes! Remembered something correctly!! 😁

Maybe "diminshing returns" wasn't the right phrase, but I think in my head I was thinking along the lines of "if I have, say 4 hours on a target, then shooting an extra couple of hours would help a little bit, but not significantly increase SNR".

[vlaiv]

10 hours ago, The Lazy Astronomer said:

Yes! Remembered something correctly!! 😁

Maybe "diminshing returns" wasn't the right phrase, but I think in my head I was thinking along the lines of "if I have, say 4 hours on a target, then shooting an extra couple of hours would help a little bit, but not significantly increase SNR".

Yep.

Another way to think about it is - if I have 1 hour, than I need 3 more hours to double SNR. After that if I want to double it again - I'll need additional 12h (to make it 16 in total) to double it again. Next doubling would need 64h total - so each doubling gets progressively more "expensive" - compared to base time.

[Trippelforge]

17 hours ago, The Lazy Astronomer said:

Optimum minimum sub exposure times using a modern, low read noise cameras can be as low as around 5s or so in typical suburban sky conditions. Many people (myself included) tend to stick to exposures somewhere in the range of 30 - 120s to avoid building up thousands of frames to have to store and stack.

Generally speaking, the longer the total integration time, the better the image in terms of signal to noise ratio (SNR), but it's diminishing returns - I can't remember the exact numbers, but for example, it might be something like 4x the integration time for a 2-fold increase in SNR (that probably isn't right, so someone please correct me!).

 I live in suburban sky conditions as well, but I find if I run way up on exposure time (1 min) the image gets extremely washed to the extent that I can't edit it well. I need to read up more on exactly the best way to wipe out some of that.

 

17 hours ago, vlaiv said:

There are two things to consider when using very short exposures:

1) ability to stack

2) level of read noise

Read noise is the only thing that makes the difference between say 3600 x 1s stack and one single 3600s exposure.

All other noise sources grow with time and if you add two 10 minute subs - it will be the same as one 20 minute sub - except for read noise.

Read noise does not grow with time - but instead grows with number of subs (each time you read out a sub - it gets one "dose" of read noise).

If read noise were 0 - we could use arbitrarily short exposures - as long as we are able to align them for stacking.

In order to stack subs - we must be able to align them. If you image faint stuff - you won't have any data to align on in short sub. If all the stars in your frame are faint - at some point, short enough exposure simply won't capture enough light from those stars to be able to do alignment properly.

Planetary imagers don't have this problem as planets are very bright - and they regularly use exposures that are few milliseconds long.

That is exactly right - quadruple total imaging time to get x2 increase in SNR.

By the way - x2 increase in SNR is not insignificant - it can be difference between unable to detect object and detecting an object (it is said that SNR of 5 is needed to positively identify object, so if you have SNR of say 3 - raising SNR by factor of x2 will make that 6 - and over the detection value).

By the way - they never become moot unless you can't align frames any more. Stacking more subs - however short they are, increases SNR the same way as stacking long subs.

If you stack 100 short subs you will improve SNR x10 over single short sub.

If you stack 100 long subs you will improve SNR by factor of x10 over single long sub - works the same regardless of sub duration.

OK so stacking can be fairly consistent if you have well defined points to do so in each shot. I wasn't completely sure how DSS ran into problems during the process. But it makes a lot of sense with exposure time effecting things to be overly faint.

I am confused about the SNR situation, I always thought less was better. But if I am understanding it seems the opposite is true, and that I may not even be able to detect objects without a higher number.  

So it doesn't matter if it's a long or short exposure, as long as the total amount of exposure time is the same? 

 

16 hours ago, ONIKKINEN said:

What scope are you shooting with?

If you pair a somewhat fast scope, like a newtonian, and a modern low read noise astro camera you can get away with very short subs under moderately light polluted skies. The more light pollution the shorter you can make the subs after shot noise from the sky overwhelms other sources of noise - called read noise swamping.

I had moderate success with an f/4.2 newtonian and a camera with about 1e- of read noise, and at the time was only taking 30s subs because of mount issues. I also deliberately chose targets at higher declinations because the sky moves slower (relatively) the closer you you shoot to the celestial pole.

I am shooting with a 80mm triplet (f/7), so it's obviously not ideal in suburban skies.  My son and I have been targeting Andromeda and M101 the past few days (to stick closer the CP). It does seem to help us out, but with my scope, plus trying to push exposure time my images are getting very washed out. That was the main driver for my question as I was hoping I could take lots of shorter exposures to stack. 

[vlaiv]

40 minutes ago, Trippelforge said:

I am confused about the SNR situation, I always thought less was better. But if I am understanding it seems the opposite is true, and that I may not even be able to detect objects without a higher number

Maybe you confused SNR with noise.

SNR is signal to noise ratio. It is important for signal to be stronger than the noise in order to be able to show it (see things in image). High SNR means that signal is significantly stronger than noise.

41 minutes ago, Trippelforge said:

So it doesn't matter if it's a long or short exposure, as long as the total amount of exposure time is the same? 

That is only true for hypothetical camera that has read noise equal to zero.

Problem is that we don't have such cameras, and even best camera has some read noise.

If you image for an hour and do a single exposure - you will have all the things that depend on time (signal and noise) accumulated and one read noise "dose" added to it as you read the sensor only once.

Now imagine you instead take two half hour exposures - everything will be the same - except this time you'll have two read noise "doses" added.

Similarly, if you take 60 exposures - each one minute long - things will add up and only difference this time will be that you have 60 read outs - so read noise is added x60 times

(you can see how if we have read noise be equal to zero - then any exposure length will produce same results - since 0 x 60 is still zero).

You might think that adding x60 or x120 read noise doses makes a lot of difference - and sometimes it does and sometimes it does not.

It depends on other factors - because noise does not add "normally" like signal does. It adds like linearly independent vectors (square root of sum of squares).

This form of addition has interesting property - if two vectors you are adding have significantly different magnitude - result is not much different than larger vector - you can see this in following diagram:

Hypotenuse is almost as equally long as longer side in above right angled triangle, yet we calculate it by square root of sum of squares of sides.

That is why we talk about "swamping" the read noise with some other noise source - like LP noise. Once there is significantly larger noise than read noise - total noise will be just a tiny bit bigger than that larger noise source and read noise contributes very little.

That other noise source can be - Light pollution noise (most often in deep sky imaging). It can be thermal noise - for uncooled cameras or it can be target shot noise - which is case in planetary imaging (planets are way more bright than DSOs).

 

[Trippelforge]

1 hour ago, vlaiv said:

Maybe you confused SNR with noise.

SNR is signal to noise ratio. It is important for signal to be stronger than the noise in order to be able to show it (see things in image). High SNR means that signal is significantly stronger than noise.

That is only true for hypothetical camera that has read noise equal to zero.

Problem is that we don't have such cameras, and even best camera has some read noise.

If you image for an hour and do a single exposure - you will have all the things that depend on time (signal and noise) accumulated and one read noise "dose" added to it as you read the sensor only once.

Now imagine you instead take two half hour exposures - everything will be the same - except this time you'll have two read noise "doses" added.

Similarly, if you take 60 exposures - each one minute long - things will add up and only difference this time will be that you have 60 read outs - so read noise is added x60 times

(you can see how if we have read noise be equal to zero - then any exposure length will produce same results - since 0 x 60 is still zero).

You might think that adding x60 or x120 read noise doses makes a lot of difference - and sometimes it does and sometimes it does not.

It depends on other factors - because noise does not add "normally" like signal does. It adds like linearly independent vectors (square root of sum of squares).

This form of addition has interesting property - if two vectors you are adding have significantly different magnitude - result is not much different than larger vector - you can see this in following diagram:

Hypotenuse is almost as equally long as longer side in above right angled triangle, yet we calculate it by square root of sum of squares of sides.

That is why we talk about "swamping" the read noise with some other noise source - like LP noise. Once there is significantly larger noise than read noise - total noise will be just a tiny bit bigger than that larger noise source and read noise contributes very little.

That other noise source can be - Light pollution noise (most often in deep sky imaging). It can be thermal noise - for uncooled cameras or it can be target shot noise - which is case in planetary imaging (planets are way more bright than DSOs).

 

 

Ya I was associating SNR with simply noise. I think I understand for the most part, I read it three times now and will be a few more (lol). My basic take away is that the answer is, "it depends". As I am seeing there are various factors involved and that things are not so simple. I think I based some of this question off of if I should upgrade my DSLR to a modified one (and snap away). Or pay a chunk of money to buy a custom drive system to increase my exposure capabilities.

Thanks for the help, I really appreciate it! 

[The Lazy Astronomer]

3 hours ago, Trippelforge said:

That was the main driver for my question as I was hoping I could take lots of shorter exposures to stack. 

You probably can. It sounds like you've got a fairly high amount of light pollution, so you've got a bit of leeway l reckon, so long as you've got plenty of storage space!

3 hours ago, Trippelforge said:

I am shooting with a 80mm triplet (f/7), so it's obviously not ideal

I wouldn't say that. Nothing wrong with a 500 - 600mm fl triplet! Assuming a pixel size of around 5um, that gives a sampling rate of around 1.8"/px, which is pretty good I'd say. 

3 hours ago, Trippelforge said:

 I live in suburban sky conditions as well, but I find if I run way up on exposure time (1 min) the image gets extremely washed to the extent that I can't edit it well. I need to read up more on exactly the best way to wipe out some of that.

A simple background extraction might just fix it - consider looking at Siril (free), or, if you're feeling fancy, Pixinsight.

[Trippelforge]

1 hour ago, The Lazy Astronomer said:

You probably can. It sounds like you've got a fairly high amount of light pollution, so you've got a bit of leeway l reckon, so long as you've got plenty of storage space!

I wouldn't say that. Nothing wrong with a 500 - 600mm fl triplet! Assuming a pixel size of around 5um, that gives a sampling rate of around 1.8"/px, which is pretty good I'd say. 

A simple background extraction might just fix it - consider looking at Siril (free), or, if you're feeling fancy, Pixinsight.

 

Does Siril and Pixensight leave GIMP in the dust? I don't own Photoshop so I have been attempting to use it. I keep watching tutorials but I always feel like a few of my images could've turned out better.

Off topic question: Do you have any experience with NINA? I have been using Back yard EOS and wanted to switch prior to the trial ending. I was curious if NINA (or anything) had the capability to look at a section of the sky, and know where I was pointed?  It sounds pretty exciting if something could actually do that.

EDIT: It seems like NINA does plate solving. Not sure how easy it is to use, but I feel like if I could figure it out it would help immensely. 

Thanks again for all of the help!

Edited October 18, 2022 by Trippelforge
added text

[The Lazy Astronomer]

2 hours ago, Trippelforge said:

 

Does Siril and Pixensight leave GIMP in the dust? I don't own Photoshop so I have been attempting to use it. I keep watching tutorials but I always feel like a few of my images could've turned out better.

I wouldn't say they leave GIMP in the dust, excellent results can be achieved by someone experienced the use of their chosen image editing software. It is difficult when you're first starting out, because it's very likely you're not getting the best out of your image - practice makes perfect!

I would say though, that a software like Siril or Pixinsight (amongst many others) are specifically designed for astro image editing so I think give an easier experience than a more generic photo editing software like GIMP or Photoshop*. However, there are some people on this forum who get fantastic results with Photoshop, so really it goes back to what l said in the first paragraph.

*There are a lot of people who would find it hilarious to see the phrase "easier experience" and "Pixinisight" in the same sentence, but I found it nowhere near as difficult to get started with as it's reputation seemed to suggest. It's not easy, but it's not completely unintuitive (to me at least, anyway).

3 hours ago, Trippelforge said:

Off topic question: Do you have any experience with NINA? I have been using Back yard EOS and wanted to switch prior to the trial ending. I was curious if NINA (or anything) had the capability to look at a section of the sky, and know where I was pointed?  It sounds pretty exciting if something could actually do that.

EDIT: It seems like NINA does plate solving. Not sure how easy it is to use, but I feel like if I could figure it out it would help immensely.

Yes, I use NINA exclusively for complete control of my whole setup during capture. To be honest, NINA was the first complete control software l tried and it's worked so perfectly I've never felt the need to look into any of the other options.

Plate solving is very straightforward, and much easier than using the mount's hand controller and doing star alignment. A small amount of initial set up in NINA and you're good to go.

Once the mount is polar aligned, NINA needs to know where you want to go, and this can be done by using the object database in NINA, or importing from a planetarium software like Stellarium, or manually entering celestial coordinates, or uploading a previous image file (so many options!). NINA will then slew the mount, take a short exposure, plate solve the current location, and make any positional adjustments as needed. It will keep repeating these steps automatically until the pointing is within your specified level of accuracy. Magic!

For me, it usually takes 2 or 3 rounds of plate solving to get my target spot on - the first one gets it close, the follow ups make fine adjustments to get it perfect. Takes maybe 2 or 3 minutes total normally. Combined with Sharpcap's polar alignment tool, I'm usually imaging within about 10 - 15 minutes.

[900SL]

On 17/10/2022 at 22:19, Trippelforge said:

My tracking is lackluster right now so I have to stick to shorter exposures. But I am trying to figure out my limitations, such as at what point do short exposure images for stacking become moot. Or do they not, such as would an hour worth of 30 second exposures net a better image than only 30 minutes worth?

Hopefully that makes sense!

Thanks

 

 

 

 

 

Watch this, and also Google 'exposure times in astrophotography,

 

 

[Trippelforge]

14 hours ago, The Lazy Astronomer said:

I wouldn't say they leave GIMP in the dust, excellent results can be achieved by someone experienced the use of their chosen image editing software. It is difficult when you're first starting out, because it's very likely you're not getting the best out of your image - practice makes perfect!

I would say though, that a software like Siril or Pixinsight (amongst many others) are specifically designed for astro image editing so I think give an easier experience than a more generic photo editing software like GIMP or Photoshop*. However, there are some people on this forum who get fantastic results with Photoshop, so really it goes back to what l said in the first paragraph.

*There are a lot of people who would find it hilarious to see the phrase "easier experience" and "Pixinisight" in the same sentence, but I found it nowhere near as difficult to get started with as it's reputation seemed to suggest. It's not easy, but it's not completely unintuitive (to me at least, anyway).

Yes, I use NINA exclusively for complete control of my whole setup during capture. To be honest, NINA was the first complete control software l tried and it's worked so perfectly I've never felt the need to look into any of the other options.

Plate solving is very straightforward, and much easier than using the mount's hand controller and doing star alignment. A small amount of initial set up in NINA and you're good to go.

Once the mount is polar aligned, NINA needs to know where you want to go, and this can be done by using the object database in NINA, or importing from a planetarium software like Stellarium, or manually entering celestial coordinates, or uploading a previous image file (so many options!). NINA will then slew the mount, take a short exposure, plate solve the current location, and make any positional adjustments as needed. It will keep repeating these steps automatically until the pointing is within your specified level of accuracy. Magic!

For me, it usually takes 2 or 3 rounds of plate solving to get my target spot on - the first one gets it close, the follow ups make fine adjustments to get it perfect. Takes maybe 2 or 3 minutes total normally. Combined with Sharpcap's polar alignment tool, I'm usually imaging within about 10 - 15 minutes.

I downloaded Siril and messed with it last night and I think I expected something like GIMP as far as interface. What does it actually do "processing" wise? I noticed a pre-processing tab, but otherwise it made me think it was just a stacking program. Obviously I need to go watch some tutorials, but was just curious.

I cannot take advantage of 90% of NINA's features it seems since I have no way to connect my scope outside of the DSLR. I plan on picking up some legit motor drives with PC connection in the near future. But right now I am trying to do what I normally do with Backyard. Such as simply focus, and setup camera settings. My son messed with the plate stacking aspect and Stellarium integration last night (outside 35'f). He says he got that part figured out but I still didn't see how to actually use a focus feature. I got a star in live view on the screen and pretty much jumped around trying to get that meter to do something. No luck yet, but obviously still learning.

Anyways, is it possible to use the plate stacking feature without it auto slewing? Such as manually moving into position with it's guidance? 

I can already tell though NINA is a more polished and feature rich program, Backyards trial ends in a few days and I am glad I found out about it!

Thanks for the help!

[ONIKKINEN]

You can platesolve without having a mount connected, yes. It will show you the coordinates you are at and you can then manually move the scope closer. Rinse and repeat until there.

[ollypenrice]

What the dedicated astro software packages have are astro-specific routines. Of these, by far the most important is gradient removal, ideally a gradient removal which leaves you with a neutral grey background sky (equal brightness in red, green and blue.) For this I use Pixinsight, after which I work almost exclusively in Photoshop - which probably won't be that different from GIMP though I don't know for sure.

When you say that you think your images may have more to give in processing, you will be right - because this is true of any image. It doesn't matter how good you are at processing today: tomorrow you will be better. There is no reachable limit in terms of processing skill. That's its great charm.

Olly

[The Lazy Astronomer]

5 hours ago, Trippelforge said:

I downloaded Siril and messed with it last night and I think I expected something like GIMP as far as interface. What does it actually do "processing" wise? I noticed a pre-processing tab, but otherwise it made me think it was just a stacking program. Obviously I need to go watch some tutorials, but was just curious

I only use Siril for stacking, and I've never actually tried to process an image with it, so can't advise there, but have a look at the tutorials on their website to get you started. This one runs through a full processing workflow, including background subtraction/neutralisation and colour calibration: https://siril.org/tutorials/tuto-scripts/#lets-start-processing-the-result-a-nametuto-3a

[Trippelforge]

19 minutes ago, The Lazy Astronomer said:

I only use Siril for stacking, and I've never actually tried to process an image with it, so can't advise there, but have a look at the tutorials on their website to get you started. This one runs through a full processing workflow, including background subtraction/neutralisation and colour calibration: https://siril.org/tutorials/tuto-scripts/#lets-start-processing-the-result-a-nametuto-3a

Thank you! I will go check it out!