For any system the really dim stuff is in the bottom few digits of your sixteen bit camera.
This is after calibrating for darks and bias. This also should remove the floor.
The approach I take is to process all corrected images in 32 bit format so the dim stuff is spread over far many more binary digits. This has the advantage when stacking many frames so that the dim stuff has more inherent dynamic range and it lowers any chance of posterisation when stretching.
I have come to the conclusion that I can do better by binning x2 to have effectively 18 micron pixels with my FPL16803 and F3 600mm RH200 Optic. The read noise is also lowered by a factor of four when binning x2.
I get resolution back by doing drizzle integration. The more subs you have the better the resolution as well as the signal to noise.
I download images at the lowest speed of 1MHz and use 100 Bias and Dark frames for correction. It is pointless to introduce systemic noise at the correction phase. The camera is also flushed after each exposure. You would be surprised what is left in the wells from bright stars.
When correcting for darks I always use darks collected at the same temperature as the lights. Even PixInsight's algorithm fails dismally for temperature differences for darks and lights for really dim stuff.
Dithering is the other thing to do routinely.
Remember folks that collecting star images are limited by diffraction for a given aperture and collecting very dim extended objects by focal ratio.
Of course aperture is king and always wins.
When they make a 600mm F1.5 optic I will get one. By binning x2 my system passes for this given I collect enough subs.
I welcome any critique of my thinking. I hope this helps others.
Bert
See this for an example.
http://www.iceinspace.com.au/forum/s...d.php?t=171515