Dilema - cmos gain vs noise vs well depth

[iapa]

I have just run SharpCaps sensor analysis on my CMOS camera.

Unity gain (1e/ADU) seems to be around 119

However, for this sensor, to get to the lower noise eleven, I should probably for for a gain of 121 to bring the noise nown

That means c0.86 e/ADU

However, full well depth comes way down to 14000

which means more and shorter exposures, or relatively flat start brightness as all the brighter stars are "filling" the pixels? 

Would I not be better bringing gain to around 60, 2e/ADU, 5.5e of noise wit increased FWD allowing longer exposures to capture more "useful" electrons?

 

[vlaiv]

29 minutes ago, iapa said:

which means more and shorter exposures, or relatively flat start brightness as all the brighter stars are "filling" the pixels? 

Would I not be better bringing gain to around 60, 2e/ADU, 5.5e of noise wit increased FWD allowing longer exposures to capture more "useful" electrons?

No and no.

Full well and dynamic range are not really that important in long exposure imaging.

If you have saturated regions in your regular exposure (which you should calculate based on your light pollution and read noise and how well your mount guides and so on ...) - just take some time at the end of the session (or rather make some time) to get a few short exposures.

You won't need much - something like 5-10 of 10s exposures will be enough. You won't use these subs for anything else except substituting saturated pixels in original stack. Since pixels in original stack are saturated - that means they already have high SNR and won't be hurt by short exposures or small number of them (in fact - you can calculate number of needed short exposures based on how bright signal you want to capture and what SNR you want to achieve - but you really don't need to do that).

Only trick is to know how to combine the two.

I advocate following approach - combine while linear. Stack both to separate stacks. Multiply short stack with ratio of stack exposures and then replace pixels in long stack that have value over certain threshold - like >95% with pixels from aligned short stack - there you go - you have non clipping linear data - star shapes and color preserved (and even nebulosity if it clips - like on really bright targets).

[iapa]

Thanks

I'll take note and set up accordingly for my first pass 

Edited May 11, 2020 by iapa