Integration time / diminishing returns - Forums

Topic starter

Dec 31, 2023 · Edited: Dec 31, 2023

I’ve just started my first project with my new setup (RedCat 51 + asi533mc pro), IC 1805 heart nebula.

So far I’ve only managed to capture about 3.5hrs of total integration (44x300sec) from a bortle 7 location using L-extreme filter, and I am wondering, how much more data should I add until I reach the point of diminishing returns?

Well Written Engaging

andrea tasselli

Frequent contributor

Dec 31, 2023

A lot more than you have now. As first approximation +16 hours. But a lot depends on how you process the data.

Peyton Lorimor

Topic starter

Dec 31, 2023

andrea tasselli:
A lot more than you have now. As first approximation +16 hours. But a lot depends on how you process the data.

Yeah I’m planning to add at least 10 more hours before moving on to a new target

Ryan Faulkner

Dec 31, 2023

Here’s a good video that covers doubling exposure time for improved results:
https://youtu.be/8DhRy1MT1Qs?si=5h5xKZla-oTeNohp

Basically, you see visible improvements with every doubling of exposure time, so if you have 4 hours total, adding another 4 will be an improvement, but then you’ll need another 8 hours for next improvement.

Helpful

Peyton Lorimor

Topic starter

Dec 31, 2023

Ryan Faulkner:
Here’s a good video that covers doubling exposure time for improved results:
https://youtu.be/8DhRy1MT1Qs?si=5h5xKZla-oTeNohp

Basically, you see visible improvements with every doubling of exposure time, so if you have 4 hours total, adding another 4 will be an improvement, but then you’ll need another 8 hours for next improvement.

I’ll check out the video, thanks!

Oskari Nikkinen

Jan 1, 2024

Signal to noise ratio improves with added imaging time by the desired improvement factor squared, so as an example you get twice the SNR when you integrate 4 times the current data, or 3 times the current SNR by imaging 9 times longer. So there is a practical limit to improvement, which is probably in the 30 hour range i would say. So with your current image you would be looking at a 'twice as good' shot at 14h (its not so simple of course, but a good estimate).

Since you are at bortle 7 you do want to spend some time with each image because your SNR will be quite poor compared to nicer skies, although your narrowband filter will help a lot with that.

Helpful Insightful Concise

Robert Žibreg

Jan 10, 2024

There is also an app for this. Take a look https://www.youtube.com/watch?v=E-QBFmjUQx4

Craig Dixon

Jan 10, 2024

I aim for at least 10 hours on any target but like to get above 20 hours. Very generally speaking, anything above 20 hours and I don’t see the benefit any more.

that said, it depends on your objectives. If you want to get as good quality images as possible, get more time. If you want to image as many targets as possible, spend less time on each.

Helpful Concise

D. Jung

Jan 10, 2024

This topic is much more complex if you factor in the use of noisexterminator. In my experience, if you are not a pixel peeper, diminishing results are reached much earlier since these tools are quite effective.

andrea tasselli

Frequent contributor

Jan 10, 2024

D. Jung:
This topic is much more complex if you factor in the use of noisexterminator. In my experience, if you are not a pixel peeper, diminishing results are reached much earlier since these tools are quite effective.

There is no sobstitute to actual, good quality data. All the rest is patching over the obvious issues but not without paying the price.

Jan Erik Vallestad

Jan 10, 2024

I agree, nxt is a good tool but it will not compensate for lack of data.

The best way to "cheat" with SNR is to resample the image by half in each dimension, you effectively take four pixels and make it one pixel and then you've doubled your SNR as well. Of course this comes at a high cost as you've now reduced your image size and resolution.

But more data really is the key. How much though is very case by case depending on the target, the quality of your skies and a whole lot of other factors. Keep adding data until you're happy with the image. That's the most important thing after all.

Helpful Insightful Respectful

Gary Seven

Jan 15, 2024

Ryan Faulkner:
Here’s a good video that covers doubling exposure time for improved results:
https://youtu.be/8DhRy1MT1Qs?si=5h5xKZla-oTeNohp

Basically, you see visible improvements with every doubling of exposure time, so if you have 4 hours total, adding another 4 will be an improvement, but then you’ll need another 8 hours for next improvement.

I really like Peter Zelinka's work (been a subscriber for a couple of years), but I think there is a potential for creating the wrong impression on this topic.

For example, I decide to press on from 4 hours to 8 hours. Lets say the subs are four minutes (240 seconds) each because diving an even number by factors of two is easy ;-)

But due to circumstances, I only get 119 subs - 7 hours 56 minutes worth.

Mathematically, the SNR is not twice as good as the four hour data set. No argument there, math is math.

The math *does* say it is twice as good as a 3 hours and 58 minute image (oops! one sub was cut to 120 seconds due to approaching clouds)

Can you tell the difference between an image with 3 hours 58 minutes vs. 4 hours? Does the 4 hour image look twice as good?

Does the 7 hour 58 minute exposure really not look any different than the 4 hour data set?

If you run out to ten hours, it might actually be a good thing - NSG might weight your data such that you only net 8 hours.

Zelinka also mixes two concepts in the video -Math and "just noticeable difference" which is a characteristic of visual perception. The eye responds to light logarithmically so a factor of say, 1.6x might be a bit better target from a perceptual viewpoint.

The moral of the story - I would not get too hung up on doubling. It's a target, nothing more.

Helpful Insightful Respectful Engaging

Ryan Faulkner

Jan 18, 2024

Gary Seven:
Ryan Faulkner:
Here’s a good video that covers doubling exposure time for improved results:
https://youtu.be/8DhRy1MT1Qs?si=5h5xKZla-oTeNohp

Basically, you see visible improvements with every doubling of exposure time, so if you have 4 hours total, adding another 4 will be an improvement, but then you’ll need another 8 hours for next improvement.

I really like Peter Zelinka's work (been a subscriber for a couple of years), but I think there is a potential for creating the wrong impression on this topic.

For example, I decide to press on from 4 hours to 8 hours. Lets say the subs are four minutes (240 seconds) each because diving an even number by factors of two is easy ;-)

But due to circumstances, I only get 119 subs - 7 hours 56 minutes worth.

Mathematically, the SNR is not twice as good as the four hour data set. No argument there, math is math.

The math *does* say it is twice as good as a 3 hours and 58 minute image (oops! one sub was cut to 120 seconds due to approaching clouds)

Can you tell the difference between an image with 3 hours 58 minutes vs. 4 hours? Does the 4 hour image look twice as good?

Does the 7 hour 58 minute exposure really not look any different than the 4 hour data set?

If you run out to ten hours, it might actually be a good thing - NSG might weight your data such that you only net 8 hours.

Zelinka also mixes two concepts in the video -Math and "just noticeable difference" which is a characteristic of visual perception. The eye responds to light logarithmically so a factor of say, 1.6x might be a bit better target from a perceptual viewpoint.

The moral of the story - I would not get too hung up on doubling. It's a target, nothing more.

Good point: there won’t be a magical change going from 3:59 to 4:00.
For me, the value of his video was demonstrating that you shouldn’t expect a major improvement if you add another hour when you’ve already grabbed 8 hours. Adding that extra hour will improve your image, but unless you’re going to dedicate much more time, maybe your time is better spent beginning a new project.

Well Written Insightful Respectful Concise

Scott Badger

Jan 18, 2024

D. Jung:
This topic is much more complex if you factor in the use of noisexterminator. In my experience, if you are not a pixel peeper, diminishing results are reached much earlier since these tools are quite effective.

Noise Exterminator does a great job on the noise, but I don't think it can distinguish between background/noise and faint signal that isn't above the background level, so there's no increase in faint detail. To bring out the faintest details simply requires more time, and no shortcuts. And yes, it's a diminishing return, mathematically, but it's often the case (not just in AP) that very small differences can be of much larger effect to more subjective assessments, like human perception.

Cheers,
Scott

Well Written Insightful Respectful Concise