60 replies to this topic

[Tony_Gondola]

I'm trying to get my head around the real reason one would want to shot long subs verses short in CMOS. Many people are convinced that long subs are better but I haven't seen an explanation as to why, other than swamping read noise.

[klangwolke]

I shoot 600 second subs because my guiding is good and I find it easier to deal with fewer images per session. It does mean I have to throw away more images though. I’m also shooting full-frame and typically at a gain of 0 because of the long per frame integration time. It’s just more convenient for me. It would be neat to see a histogram of times use, and reasons for shorter?

I also shoot at OSC no filters at 300s but sometimes 180 or even 120 depending on sky conditions and target.

[TXLS99]

Heres some past discussions on the same topic, lots of good info

 

 

https://www.cloudyni...er-longer-subs/

 

 

https://www.cloudyni...onger-exposure/

 

 

 

https://www.cloudyni...less-long-subs/

[scanner97]

5-minute subs were a holy grail decades ago and are still the common wisdom for a lot of folks.  But, yes, read noise with modern CMOS is so low, you won't necessarily get a total noise benefit - depending on a whole bunch of factors.

 

OTOH, clouds rolling in as a real risk for many imagers, especially since automation means you probably aren't outside to notice.  (Although you can even automate cloud detection along with an alarm to notify you in the house if you want to...)  

 

And, as you'll see in some of those other posts, a 10x increase in files (going from 5 minutes guided to 30 seconds unguided, say) means you need a lot more storage space and processing time.  Or more processing hoirsepower, all of which costs something.  There are complex trade-offs.  What you'll find is some folks really invested in the trade-off they made as being the right one.  It probably was the right one for them, but it might not be for you.

 

For what it's worth, there are plenty of very experienced folks shooting 5-minute subs with awesome results.  And there are guys like Steven Miller who often shoots 8-10 sec subs, mainly because he started with a big Dob.  He gets amazing results too.

 

Good luck.

Edited by scanner97, 15 March 2025 - 12:17 PM.

[markalot63]

I think the only critical issue it subs too short, generally you  want your dimmest target to be above read noise'ish.  Pretty easy to achieve.  I do 20 seconds on Orion, 2 minutes on everything else.  With an APS-C size sensor file space adds up quick but the longer your subs the less color in the stars.  Easy to work around by combining short sub stars with longer sub deepsky if you're bothered.  

 

I think it's important NOT to worry about this because your conditions will vary and I have not seen a test that proves anything other than a sub length works best for the testers unique conditions.  They have done honest work, but just too many variables.  As mentioned above, clouds, satellite trails, etc. 2 minute subs on anything but the brightest objects seems like a good starting point, which is why I use it.  If I had unlimited disk space I would probably go for 1 minute subs.  I have done 1 minute subs on partly cloudy nights and been pleased by how many usable subs I was able to luck in to.  

[kathyastro]

Granted that low-noise sensors are more common these days.  But there will always be targets that are fainter than the read noise, no matter how low that is.  For those targets, you need long enough exposures.  

 

When, as in most cases, that is not a factor, then you want exposures that are short enough that errors in guiding and polar alignment are not noticeable, and to avoid saturating bright stars.

[rj144]

As already noted, "some" targets are dim and actually require longer subs.  However, for nearly every popular target, most parts are not that dim and short exposures work just as well as longer exposures.  I always do short exposures because it gives me more dynamic range in post, I have less data thrown out due to clouds, wind, etc.. and it also minimizes tracking errors.

 

Again, outside of rare dim targets, the ONLY benefit to longer exposures is that it saves storage space and decreases stacking time.  

 

I often make suggestions here to shoot shorter exposures because quite a few just do 3 to 5 minutes for everything and I don't get it.  3 to 5 minutes on M45, M42, or M31, for example, makes absolutely no sense to me.  It's blowing out stuff all over that's very hard to fix in post.

Edited by rj144, 15 March 2025 - 02:07 PM.

[donjulio]

Check out this YouTube video, very informative. 
 

https://youtu.be/3RH...ezncoSjFSyTw0rC

[Spaceman 56]

I'm trying to get my head around the real reason one would want to shot long subs verses short in CMOS.

 

Many people are convinced that long subs are better but I haven't seen an explanation as to why, other than swamping read noise.

beginners often think long subs are better, as they want to clearly see Something in their individual subs. thats incorrect thinking.

Longer subs have the advantage that you will have less of them to deal with in stacking. also SNR will be higher on a per sub frame basis.

 

I find the longer the subs, the more time gets wasted (discarded) when sub rejection comes into the equation.

 

if you shoot 100 x 30 second subs, and 2 of them have issues, you lose 60 seconds.

if you shoot 10 x 300 second subs, and 2 of them have issues, you lose 600 seconds.

 

thats a 10 to 1 ratio in favour of short subs.  I shoot at 120 seconds because of this, as my retention rate is higher.  

[matt_astro_tx]

With longer subs there's also the advantage of having less files to store and process at the end of the night.  Just saying.

[Spaceman 56]

I often make suggestions here to shoot shorter exposures because quite a few just do 3 to 5 minutes for everything and I don't get it. 

 

3 to 5 minutes on M45, M42, or M31, for example, makes absolutely no sense to me.  It's blowing out stuff all over that's very hard to fix in post.

I agree.

 

at F5 I blow out the stars at 120 seconds in OSC on bright targets. (without filters)

 

at F6 I blow out stars at 180 seconds in OSC on galaxies. (without filters)

 

I feel people using very narrow filters might shoot longer without issues, but I have no experience with filters so wont comment.

 

you probably know more about this 

[Chrisofweden]

There are many threads in here about one the pros and cons of mono vs OCS and for the record - I shoot both.
One unexpected benefit I found with shouting mono, where I predominantly shoot narrowband, is that I can shoot far longer exposures, even on my wide field f3.8 scope, and have far less subs do deal with. Makes life easier in processing. I guess the same would be true for an OSC with the NB filters too, but I’ve always regarded that as inefficient.
The storage advantage is not that relevant, since storage is so cheap.

Otherwise, I certainly agree that there is no need to “stress” your set up with long exposures unnecessarily.

/Chris

Edited by Chrisofweden, 17 March 2025 - 03:19 AM.

[Robert7980]

It’s just lower noise in faint regions (maybe, it depends) and a lower number of subs to process, the longer the shutter is open the better the signal to noise gets, there are caveats though and nothing is free. At some point you’ll outrun the cooling power of a cooled camera and run into thermal noise or other problems such as saturated stars possibly. 
 

There is always some compromise no matter which approach you choose, every decision has them. So you have to know what is being gained and what is being lost, it’s not at all straightforward and there’s no way to say this number is ideal or that. That’s why this is a difficult question to understand. It’s because there is no definitive absolute answer, same as anything that has complexity with lots of factors, functions, inputs and outputs. 
 

I recommend experimenting so you gain experience, at that point you can make informed decisions for your circumstances. 600 seconds is a viable option for my f/8 RC with 3nm SII filter for instance, it’s far less than ideal when I’m shooting f/2 unfiltered with an OSC sensor, so the specifics and details matter a lot… 

 

Everything comes down to simply being tools in a toolbox, there’s a more practical and proper tool for every situation and there are no tools that cover all situations and circumstances in an ideal way. 
 

Just saying long subs are better or worse isn’t the right way to think about it, without knowing all other ingredients in the soup it’s useless information. The total recipe is a combination of all of the inputs including how the final dataset is processed which matters probably more than anything else. So there are raw ingredients and secret sauces. 
 

In a bad analogy; say you have two vehicles one with 1500 horsepower and one with 300 horsepower. You ask which will win a race? Well that depends what else we know, say the 1500hp vehicle is a main battle tank and the 300hp vehicle is a Honda civic. Now we know what has a better chance of winning a road race and which will cover muddy terrain faster, a question we didn’t even ask is now more important. With terrain in question neither is the best overall choice for all races. Then you ask who’s driving, because that’s also critically important, and so on… 

 

If Sammy the baby kitten is responsible for piloting the car, you probably aren’t betting it’s even going to cross the finish line, so even the outlier inputs matter… Using this terrible analogy ask yourself the significance of what vehicle has the most horsepower or the longest sub exposure, it suddenly isn’t as important on its own right, and definitely not the critical number we need to know. 

 

So numbers are just information, they don’t tell us much on their own. 
 

Unfortunately one must have some intuition about these things before we can even ask the right questions, so gaining some experience is essential even if it’s not a good experience  … Knowing how to do it wrong has as much value as knowing how to do it right! 
 

The point is get out and play with the gear, when in doubt try it out… 

 

I will say that long subs aren’t for the uninitiated, they have a lot of potential problems that need more thought and skill, better equipment and so on. So if you are beginning the journey you’ll probably fail trying to build a final image from 10 minute subs. So there’s the it depends who’s driving that Honda civic… 

Edited by Robert7980, 17 March 2025 - 04:26 AM.

[Zambiadarkskies]

What Robert says above is an excellent synopsis. As we are sharing data points though, here is an interesting one.
533mc pro on a f5.6 80mm refractor. 300s subs no filter other than IR/UV saw less than 1% saturated pixels. Target was the horsehead and flame so actually pretty bright.


It's numbers like this that make me use gain 0 and longer duration subs sometimes as I really like cutting down on the number of subs where I can. So long as my guiding is up to it. And the overall integration time is long enough.

[smiller]

I'm trying to get my head around the real reason one would want to shot long subs verses short in CMOS. Many people are convinced that long subs are better but I haven't seen an explanation as to why, other than swamping read noise.

Lots of good feedback here, I suggest you watch the Glover video and read the older referenced threads on this.

 

But I want to address to you why “people are convinced that long subs are better….”.

 

What I’ve seen is that some people just haven’t caught on to the fact that with modern CMOS cameras with the low read noise, higher quantum efficiency, that what used to be true is just not nearly as true anymore.   They are just behind the times and repeat a mantra, which is really only true in corner cases anymore, without ever bothering to do the math.

 

Why do I suspect this?  Because in several situations I did the math for them and showed them how short the exposure could be, and they are often shocked at the answer.  And I’m thinking “Dude!, you’re repeating this stuff with such conviction, and you haven’t even bothered doing the math?”

 

A more helpful approach is to simply say: “Learn how to do the math for your situation because read noise contribution and desired exposure times can vary wildly.  There is never one answer for all… just do the math as the math isn’t complicated.”

Edited by smiller, 17 March 2025 - 10:12 AM.

[smiller]

I think the only critical issue it subs too short, generally you  want your dimmest target to be above read noise'ish.  

 

As already noted, "some" targets are dim and actually require longer subs.  However, for nearly every popular target, most parts are not that dim and short exposures work just as well as longer exposures.  

Actually I don’t think the math supports these statements about target dimness mattering in the exposure time/read noise contribution calculation, at least with most exposure time calculators, and let me explain why:

 

- From an image quality perspective, longer exposure times are only needed to overcome camera read noise relative to sky background and target shot noise.

 

- In the vast majority of cases the sky background shot noise is much higher than target shot noise.  Obviously for very bright targets, target shot noise is significant but in those cases, the target is bright so your total noise from both sky background and target is very high and read noise is insignificant.

 

- But more importantly, most read noise calculators (ex the Glover Video) don’t even include the target brightness so they assume that the target you are going after is nearly infinitely dark.   In others words it’s at the limit.  So most read noise contribution calculators already assume a very dim target and they calculate the amount of read noise over the sky background only, as if the target is infinitely dark.

 

 Therefore, I believe the correct statement is this:  

 

If your read noise calculator does not add shot noise from the target, then the exposure time it calculates is independent of the brightness of the target and it can be utilized for even targets that are extremely dim to determine your read noise contribution.  If you then go on to include target shot noise in those calculations, then all that will do is further reduce the required exposure time.   Even for a very dim target, it will do nothing more than reduce the required exposure time over the baseline calculation from the calculator.  It may be a trivial amount but it won’t increase the number, only decrease it.

Edited by smiller, 17 March 2025 - 12:26 PM.

[markalot63]

Actually I don’t think the math supports these statements and let me explain why:

 

 

 

The math doesn't support read noise'ish?  

 

I get what you are saying (actually I completely agree), in a heavily light polluted area the exposure time on any target might be restricted.  The photons you want are still there, but it's going to take a long time to get them above pollution noise and too long an exposure in a bright sky might not work very well. Regardless, it seems what many are concerned about is that a short exposure time will somehow miss a photon and for some reason sub length will optimize this.

 

The only factor IMO is camera noise, not sky or target noise (whatever that is).  Camera (read) noise is so low these days it's very hard to make it a factor.  Minus any noise created by the equipment the target photons are either there or they aren't and no amount of sub length will make them appear. 

 

One might add that if you are lacking good calibration longer sub lengths might be more forgiving, maybe?  I just don't think that with modern equipment any of this is worth worrying about.

Edited by markalot63, 17 March 2025 - 11:04 AM.

[rj144]

Actually I don’t think the math supports these statements about target dimness mattering in the exposure time/read noise contribution calculation, at least with most exposure time calculators, and let me explain why:

 

- From an image quality perspective, longer exposure times are only needed to overcome camera read noise relative to sky background and target shot noise.

 

- In the vast majority of cases the sky background shot noise is much higher than target shot noise.  Obviously for very bright targets, target shot noise is significant but in those cases, the target is bright so your total noise from both sky background and target is very high and read noise is insignificant.

 

- But more importantly, most read noise calculators (ex the Glover Video) don’t even include the target brightness so they assume that the target you are going after is nearly infinitely dark.   In others words it’s at the limit.  So most read noise contribution calculators already assume a very dim target and they calculate the amount of read noise over the sky background only, as if the target is infinitely dark.

 

 Therefore, the correct statement is this:  

 

If your read noise calculator does not add shot noise from the target, then the exposure time it calculates is independent of the brightness of the target and it can be utilized for even targets that are extremely dim to determine your read noise contribution.  If you then go on to include target shot noise in those calculations, then all that will do is further reduce the required exposure time.   Even for a very dim target, it will do nothing more than reduce the required exposure time over the baseline calculation from the calculator.  It may be a trivial amount but it won’t increase the number, only decrease it.

I agree almost entirely, but someone on a discussion on Reddit showed me details from a collaboration with 100s of hours on it.  The people in the collaboration orginally took shorter exposures (all with modern sensors) and thought they were missing something.  They then took longer exposures, and dim detail was much more visible in the longer exposures.  There was even a comparison between the two.  I always thought the math and in just about every scenario, exposure length didn't matter too until I saw that.  It was a few months ago (I think over the summer), so it's going to be hard to find, but I'll see if I can dig it up later.

[AstroFromHome]

By accident I took a 3000 sec, so 50 minutes, long sub on S2 last night on the Rosette Nebula. I can clearly say that you can compose a good SHO image with less than 10 frames.

The single sub is almost like an ok stack of 300 second exposures.

 

Go as long as your sky and your gear allows. Choose a length from which you say your mount can guide that long without issues. Sure it can happen that you dump a few frames but on many targets with short subs you cannot build up signal in the faint areas of the nebula. With too short subs the signal drowns in the noise.

 

Sure you should not expose for 50 minutes for one sub but 3 - 5 minutes is a good time frame to build up the necessary signal to lift it over the noise.

180 and 300 sec are also my go to exosure times as I like to have a good number of individual frames that noise calibrates out during stacking in a reasonable way.

[scanner97]

I agree almost entirely, but someone on a discussion on Reddit showed me details from a collaboration with 100s of hours on it.  The people in the collaboration orginally took shorter exposures (all with modern sensors) and thought they were missing something.  They then took longer exposures, and dim detail was much more visible in the longer exposures.  

With many participants, it would be really interesting to see what variables they controlled for!  Was everyone shooting same f/, same skies, same filter state, etc?  For my gear and skies, and BB moonless, I wouldn't expect that going from 1 minute to 15 minutes would give me much/any more faint detail.  With a 3nm Ha under same conditions, it probably would.  Anyway, love to see this so I can decide if I need to do my own experiment!

[smiller]

With many participants, it would be really interesting to see what variables they controlled for!  Was everyone shooting same f/, same skies, same filter state, etc?  For my gear and skies, and BB moonless, I wouldn't expect that going from 1 minute to 15 minutes would give me much/any more faint detail.  With a 3nm Ha under same conditions, it probably would.  Anyway, love to see this so I can decide if I need to do my own experiment!

That’s the problem with single anecdotes.  They are making an extraordinary claim that counters the math and science of imaging, so they need solid evidence.  There are two requirements that need to be met to come up with a disruptive conclusion counter to existing science:

 

1) A robust series of controlled experiments that controls for key variables (or at least measures them and accounts for them).

 

2) A theory to explain why they get the result they get.  In other words, if the math is wrong, why?  What’s the new math?

 

To date I haven’t seen anything close to meeting these requirements..

Edited by smiller, 17 March 2025 - 12:00 PM.

[scanner97]

They are making an extraordinary claim that counters the math and science of imaging, so they need solid evidence. 

Who needs evidence when you got a good story?  

 

And thanks for your spreadsheet, which I pilfered from another thread!  It helped me realize I probably don't need to guide my BB stuff, only the NB.

Edited by scanner97, 17 March 2025 - 12:00 PM.

[rj144]

By accident I took a 3000 sec, so 50 minutes, long sub on S2 last night on the Rosette Nebula. I can clearly say that you can compose a good SHO image with less than 10 frames.

The single sub is almost like an ok stack of 300 second exposures.

 

Go as long as your sky and your gear allows. Choose a length from which you say your mount can guide that long without issues. Sure it can happen that you dump a few frames but on many targets with short subs you cannot build up signal in the faint areas of the nebula. With too short subs the signal drowns in the noise.

 

Sure you should not expose for 50 minutes for one sub but 3 - 5 minutes is a good time frame to build up the necessary signal to lift it over the noise.

180 and 300 sec are also my go to exosure times as I like to have a good number of individual frames that noise calibrates out during stacking in a reasonable way.

For something like the Rosette, with a modern sensor, a stack of 20 second subs will be virtually indistinguishable from a stack of 50 minute subs if the integration time is identical.  If fact, I would argue in that case, the 50 minute subs are going to blow out parts of the nebula and 20 second subs will result in a better stack.

Edited by rj144, 17 March 2025 - 12:25 PM.

[Yomamma]

Take one shot of each and view the histogram. That is always a good starting point to determine exposure time. I take a couple at different exposures to see where they land on the histogram and adjust time from there.

[rj144]

That’s the problem with single anecdotes.  They are making an extraordinary claim that counters the math and science of imaging, so they need solid evidence.  There are two requirements that need to be met to come up with a disruptive conclusion counter to existing science:

 

1) A robust series of controlled experiments that controls for key variables (or at least measures them and accounts for them).

 

2) A theory to explain why they get the result they get.  In other words, if the math is wrong, why?  What’s the new math?

 

To date I haven’t seen anything close to meeting these requirements..

 

 

With many participants, it would be really interesting to see what variables they controlled for!  Was everyone shooting same f/, same skies, same filter state, etc?  For my gear and skies, and BB moonless, I wouldn't expect that going from 1 minute to 15 minutes would give me much/any more faint detail.  With a 3nm Ha under same conditions, it probably would.  Anyway, love to see this so I can decide if I need to do my own experiment!

I found the post from user "Krzyzaczek101" on Reddit (this is his quote, not mine):

 

"And you'd be right. Assuming low light pollution, longer subs will result in a decently cleaner stack.

 

Short exposures (<60s) are fine if:

You're imaging from a high LP area

You're imaging with LRGB / UVIR cut filters

 

Generally, the tighter your filter FWHM and the lower your LP is, your subs will have to be longer for an optimal result. I recommend just taking the longest subs your mount will consistently produce good subs at and that will not clip your target (clipped stars are fine). You'll get more data in the same amount of time, save storage space and decrease stacking times. And as an added bonus you'll get a higher quality stack with a cleaner background. For narrowband I wouldn't go belowe 3min, with 5min+ being my go-to sub length. I also would try to get at least 20 subs to get good rejection.

 

A lot of people don't take downtime between subs into consideration. Downloading the frame, dithering etc. can take up to 30s, which if you're taking short subs decreases overall efficiency much more than a couple bad subs(a common argument for short subs is that one wind gust trashes a lot more time).

I'm leading a collaborative project where we take very long integrations times, sometimes upwards of 600h+. In a sister group, there were examples when 80h of Ha data with 300-600s exposures looked worse than 40h of 900-1800s exposures. You can read about it in this AB post. Because of that, in both of our groups we set the minimal narrowband sub length requirement to 600s, with some people even shooting 40min subs.

 

Here's a comparison

of 300s vs 1200s subs shot for this image
. It's two 35-hour-long integrations of Ha data in M81. Both images were calibrated with the exact same calibration frames (the author didn't use dark frames as their sensor doesn't need them). The only difference between them is that one was shot with 300s subs and the other with 1200s subs. The long sub length stack is obviously cleaner and shows pretty significant improvement in detail in the M82 Ha stream."

 

The links didn't copy, so here they are too:

 

https://www.astrobin.com/7hwtz0/

 

A quote from that astrobin post:

 

"Capturing this region proved to be an immense challenge. In the first stack, we had about 80h of data with relatively short exposures ranging from 300” to 600” - we quickly realised that although the integration time was high, the cliffs were barely visible and couldn’t properly be extracted in post-processing. After some discussions, we realised that the signal is so weak that the noise floor of the sensors overpowers it. In our last attempt, we shot nearly another 40h of long exposure subs ranging from 900”-1800” to get the signal to show up, which it indeed did in the end.

So for anyone who wants to try this in the future, take exposures as long as your tracking allows you to; the longer the sub-exposure the better."

 

https://imgur.com/a/2aDEuq6

 

https://www.astrobin.com/ahxx7q/

 

Here's Krzyzaczek101's reddit page:

 

https://www.reddit.c.../Krzyzaczek101/

Edited by rj144, 17 March 2025 - 12:21 PM.