I still don't get what you equate to "diminishing returns". In the case of an image exposed 20*RN^2 you have exposed 20 times the read noise of the camera squared. For broadband, this is fairly easy to do. For Narrowband that would be a challenge and not one someone is likely to attempt. In the case of my most recent usage of the ASI6200 camera, for narrowband I was at 3*RN^2 with 15 minute subs at gain 100 offset 50 on the Wizard Nebula using a F6.7 10" iDK. I considered that to be all I was willing to do in terms of exposure time in the environment I image in. Would I consider going longer to have diminishing returns? No not really.
Now if I were imaging in broadband, I would not pay attention to the swamping of read noise much. I would determine the right exposure time based on star clipping and back it off to the point where clipping wasnt an issue. I am still not in a place where any returns are diminished per se, and given that I know that my background median ADU needs to be 704 for that 20*RN^2 swamping, it is more likely I would far exceed that then not. If I am not clipping stars and I am swamping read noise by a good margin, therein lies my optimal exposure length. My 16200 is not likely to differ much from the 6200 in terms of where those exposures will be. They wont be identical but the delta is not going to be some massive amount.
Where read noise becomes a real factor, is with narrowband. Outside of narrowband, modern cameras (which include CCD's, like the 16200 chip which was released in 2016) have such low noise that for broadband imaging even the KAI-11002 is incredibly viable.
Some people could target ADU specifically for their imaging. Once they hit 3*RN^2 or in the case of the ASI6200 at Gain 100 - 528 median ADU (28 ADU over bias), they could just stop there and stack a bunch of exposures up. In doing so it is possible some very faint detail is lost. It is also possible there is not any faint detail lost. It really depends on the target.
My brain deals in integration time.
Perhaps I totally misunderstand SNR - but In integration time, diminishing returns can penalize you when you image above those diminishing returns as the SNR of your stack for fewer subs paying a higher penalty is lower than a stack of optimized subs at lower penalty is it not?
Lets say you normalize your SNR
Sub Length | SNR
1 Second | 1.00
3 Seconds | 1.66
10 Seconds | 2.69
30 Seconds | 3.66
100 Seconds | 4.38
300 Seconds | 4.68
1000 Seconds | 4.79
3000 Seconds | 4.83
Why would I want to shoot over 300 seconds? Why wouldn't I base my subs on diminishing returns especially in the context of integration time?
To yield the most improvement, i'd want more stacks of 300 second subs than fewer stacks of 3000 second subs.
If you know the signal rate of your skyglow, the signal rate of your target, the QE of your sensor, your f-ratio and you know how much each filter impacts the signal rate (e/px/s) can't you figure out the diminishing returns of your S/H/O L/R/G/B or naked OSC or LPS filter and calculate what your optimal exposure time is? It varies on emission/filter/scope/camera/speed
when I think through these numbers and do my dumb google sheet math, a 6200 shines in more shorter subs... (or any of these clean modern BSE cmos)
I got most of my e/px/s calculations from sharpcap and its sky calculator where you can insert filters and see how it impacts the e/px/s and then figuring out a few targets to test it on... it's still best case scenario and napkin math, but i honestly feel like in integration time, diminishing returns on CMOS plays a critical roll in squeezing out value.
I do trade processing time/compute... but heck, my computer + camera costs less than most comparable cameras...
On CMOS swamping 20x the read noise SNR penalized by diminishing returns
on a CCD, you're still trending in higher SNR overcoming read noise.
Edited by sn2006gy, 19 August 2020 - 03:47 PM.