I have seen many topics on ideal sub exposure to swamp read noise but not so many considering the final printed image.
For example: Monitor contrast ratio 3000:1 (high end); typical 1000:1. High end printer maybe 200:1 level max, 100:1 is typical.
Typical stacked and stretched object histogram (Astrobin IOTD/top picks) occupies 5%-60% with sky fog peak at the 5% level.
If my non linear stretch has large stars at the upper end, sky fog at 10% and the object highlights at 90%. I am really looking to set maximum 3000 levels in the middle 80%.
Therefore prior to stretching...
In the case of 12 bit cameras. (All number converted to 16bit range of 0 to 65536)
12bit —> 16bit Bias offset 50x16 = 640
Read noise swamping (ASI1600 at unity) = 10*1.7^2 = 29X16 = 464
Target left edge histogram 640+464 (I use the edge not average) = 1104 (bortle 5 can do this in 8s L, 32s RGB or 480s narrowband).
For a single image if I want the best visible delineation on a 12 bit camera.
3000:1 ratio implies right edge of histogram =1104 + 3000*16 = 49000/65536 (way over exposed for single sub).
If 16 subs (average) then 1104+3000*4 = 13404 is target RH histogram. Most stars blown out, galaxy cores on 1 min L subs, 4 min RGB, can’t do narrowband without 30min subs and a much bigger FWC at unity.
If 64 subs (average) then 1104+3000*2 = 7404 is target RH histogram exposure. (Can be done easily for RGB, need 15 min narrowband subs (but not enough FWC >15k-20k needed at unity.))
If 256 subs (average stack) then 1104 + 3000 = 4004 for right edge (this is high end for narrowband). The challenge is 256 exposures x 8 minutes implies 30 or more hours of total integration is needed to complete the project.
The above is for monitor display... however a typical printer is much more forgiving with only 100 levels possible but the noise gradient will be much more prominent as a percentage of the image. If my typical total noise per sub is in an sqrt(std Dev) 8-9e range I would still want around 64 subs to reduce the stacked noise to about 1e (16). The exposure should then be 16x100 + 1104 = 2744 for the right edge of the histogram to allow for reasonable printing if stretched to the print histogram of 10-90%. This would be accomplished by 64 ~4-5min narrowband exposures or 5-6 hrs.
All of this is just a thought but wondering what others have found on their finished products. Guidelines for print or nice internet display at the cost of bloated stars.
Edited by bigjy989, 14 January 2020 - 10:12 PM.