Optimal exposure time

Jonny Bravo:
Optimal exposure calculators give you the minimum exposure time you'll need to exceed the "swamp" factor they use. Usually it's 10x. I'm not sure exactly what Dr Glover used in his implementation of SharpCap. He might discuss it in his video. If you haven't yet, I highly suggest giving it a watch. Ruzeen posted it on his YT channel: https://www.youtube.com/watch?v=3RH93UvP358&t=1s

*The limit is adjustable. There’s a drop-down field labeled „read noise limit“.

Hi Bjorn,
 My sky is Bortle 7 so is fairly bright unfortunately. 
have re-ran the sensor analysis but this time selected a different driver to enable me to select the ultra low noise driver on the ASCOM driver
 will how I get on

The reason it's referred to as the "optimal exposure" instead of the minimum exposure is because beyond that point the SNR is subject to the law of diminishing returns. The rate at which you see improvements in SNR flattens out quickly. As you increase the length of your exposure, you also increase the chances of it being degraded by poor seeing, weather conditions, tracking/guiding issues, and artificial light sources.

Might be best to think about it from a lucky imaging perspective. Shorter exposures allow a better chance at isolating moments where seeing conditions are best. With shorter exposures, you can discard more of your subframes while retaining the best of your data. It's another way of improving sampling, with shorter frames leading to a higher frequency sample rate.

Keep in mind, you still have to aim for the same amount of total integration time as you would with longer exposures. This leads to the need for more storage space, but not nearly as much as traditional high frame-rate lucky imaging.

Another thing to keep in mind is that the "optimal exposure" value will tend to vary from night to night, and maybe even from hour to hour. To minimize the amount of work you have to do with respect to calibration frames, it can be helpful to round up to the nearest exposure time in your darks library.

Yup @AstroNikko, I agree with everything that you’ve said.  But I still think that calling it an “optimal exposure” is a misnomer because it leads to the perception that longer than optimal exposures could be detrimental to the image, when in fact the SNR continues to improve when one moves past the optimal exposure (even though it’s on the flat/diminishing side of the curve).  

There is a point where SNR stops improving though and this happens when the non-star part of the image begins to saturate.  This point is the upper limit while the “optimal exposure” point provides the lower limit.  That’s why I’d like to see both numbers presented so that users could better understand the range and set accordingly, while also taking into account the other factors that you mention (guiding errors, clouds, etc).  Calling the lower limit “minimum acceptable exposure” also makes it clear that going below this number can lead to bad (unacceptable and easily preventable noise).

Short exposure CMOS imaging is getting a lot of attention and I see many people making the mistake that it’s pointless to expose beyond the lower limit and also mistakenly applying short exposures to NB imaging. I think the naming convention contributes to this problem.

Mark Petersen:
There is a point where SNR stops improving though and this happens when the non-star part of the image begins to saturate.  This point is the upper limit while the “optimal exposure” point provides the lower limit.  That’s why I’d like to see both numbers presented so that users could better understand the range and set accordingly, while also taking into account the other factors that you mention (guiding errors, clouds, etc).  Calling the lower limit “minimum acceptable exposure” also makes it clear that going below this number can lead to bad (unacceptable and easily preventable noise).

Been trying to reconcile your reasoning with my current understanding of the optimal exposure time.  As I see it, the problem with making saturation of the background the upper limit for an "acceptable exposure" time is that by that time you have long oversaturated the brightest parts of your image.

If my understanding is correct, what the "optimal exposure" is intended to do is establish the point at which you begin to see data in the background above the noise, and at which the brightest points of light are sufficiently saturated. Beyond that, bright points of light would continue to bloat and blow out.

It seems there's little point in exposing longer than the "optimal exposure" than is convenient, because the results after stacking are basically the same with respect to the amount of light gathered when targeting the same total integration time.

I went poking around to find out more about the "read noise limit" because I didn't quite understand its impact on the Smart Histogram results, and found this post in the SharpCap forum. In a response to that thread, Dr. Robin Glover linked to a lengthy thread where he basically breaks down how the Smart Histogram works. It's worth a read, especially after watching his talks on Deep Sky Astrophotography with CMOS Cameras and Choosing the right gain for Deep Sky imaging with CMOS cameras.

To run the Smart Histogram, usually only takes a few minutes to spit out a result.

Can take a bit of work ahead of time though, as the Smart Histogram requires a sensor analysis of your camera. SharpCap comes preloaded with sensor analysis results for some of the more popular cameras, but not for my Player One cameras or my QHY268C. For the QHY268C, I needed to run the sensor analysis for each of the read modes.

@Nikkolai Davenport Thanks for the links, I'll give them a deep dive later today.   First let me say that I'm a fan of Dr. Glover's efforts. I've watched both of the video presentations you linked and examined his sensor analysis tool in Sharpcap and it's great stuff.   It's technically sound and is a fantastic resource.  It enables users to come up with an exposure that will consistently deliver good results across a mix of equipment and sky gradients.  It also allows users to get the most out of their equipment.  So my comments are not to be taken as criticism of the analysis and results.  My sole point was that the terminology of optimum exposure could be improved.  

I did skim through the summary at the end in the lengthy thread and noticed that the takeaway is the same as in the video - That a 5% noise target gives a very good exposure duration that gets the user to the stable (less noisy) part of the SNR curve.  He does go on to show that there are improvements beyond that point.   My point is that since going beyond that point yields improvements then calling the noisier, shorter, exposure "optimum" is confusing.  Since the user is in effect choosing their acceptable noise level (5%) then calling it minimum acceptable is arguably a better term.  Optimum to me means that it works like say the electrical timing in a car engine - going below this number has disadvantages as does going above this number.  But here optimum gives good results but going longer gives better results (even if the better results are small and diminishing).  

It's also important to recognize that as Dr. Glover also mentions that this analysis is targeted towards the dark part of an image that is noise limited.  Shot noise and read noise are much less important in the bright part of an image such as we find with stars, here the noise is in the noise lol.  Since stars can be processed separately courtesy of new tools like Starnet and StarXterminator they can be acquired separately and the analysis is different.  Here maximizing dynamic range is key.  Similarly, as Dr. Glover mentioned, with NB imaging long exposures are still needed and with say 3nm filters, the upper exposure limit will typically be lower than the "optimal exposure limit".  

Having said all of that -

re: "As I see it, the problem with making saturation of the background the upper limit for an "acceptable exposure" time is that by that time you have long oversaturated the brightest parts of your image". 

Well not exactly because I suggested ignoring stars since these can be processed separately and then stopping before saturation of any of the diffuse stuff.  

re: "It seems there's little point in exposing longer than the "optimal exposure" than is convenient".   

I don't disagree.  But in his hobby we often see examples where folks go to great lengths to get the best possible image of an object.  Including multi-year 40+ hr long exposures.  I wouldn't say that those who go overkill are necessarily wrong in doing so.

Mark Petersen:
I don't disagree. But in his hobby we often see examples where folks go to great lengths to get the best possible image of an object. Including multi-year 40+ hr long exposures. I wouldn't say that those who go overkill are necessarily wrong in doing so.

In his presentations, Robin Glover emphasizes that his goal is to maximise dynamic range (w.r.t. to the captured data - the dynamic range of a sensor is exposure-time independent and given by full-well depth over read noise). Just a general remark to avoid possible confusions.

Why do those much longer exposure work as well?
Simply because most objects wouldn't need that dynamic range. For most deep sky imaging, only stars saturate. For example, doubling the "optimal" exspoure time will add several more saturated stars but for most objects it wouldn't clip anything on the object.
However, for bright objects (e.g., Andromeda Galaxy, Orion Nebula), the optimal exposure time makes much more sense: you image the data in such a way that the read noise in the sky background is "controlled" but without overexposing the brightest parts of the objects (HOPEFULLY!). As we all know, M42 is a good candidate where this is likely not working. As soon as the dynamic range of the object exceeds the dynamic range of the sensor, the method will overexpose the brighter parts.

Björn