Doe anyone care to explain again how 16-bit A/D conversion (65,636 levels) could be expected to improve IQ over 14-bit (16,384 levels)? I can see a use for some margin of error beyond what 12-bit conversion has, but I see no value to going beyond the safety factor of over four in 14-bit.
I agree, and think even 14-bit is perhaps overkill. I think the main issue is how much color is needed to generate a true continuous tone with no visible transition. The human eye is capable of distinguishing about 1 billion distinct shades. Granted, many of these probably have to be right next to each other with a hard transition to detect, but if we accept that number as a constant, then:
1) 1BB colors = 1,000 per channel in RGB space.
2) 10-bit per channel color = 1024 distinct shades per channel or just over 1BB distinct shades.
So, in theory, we shouldn't need any more color than 10-bit, right? Except if the adjacent hues colors are physically next to each other we can still detect the step in hue between them, so we need a "feather factor" to allow them to blend seamlessly if we want a truly continuous tone.
So the real issue is what is an acceptable "feather factor" -
I submit that one tone separating the two adjacent colors we can visually separate should be enough to "feather" the transitions so we cannot see distinct transitions. (But since noise can act as a "feather-factor" too, perhaps having a little is not a bad thing... I feel the heat already! ) However in an ideal situation, if you want to build in a fudge factor then 2-bits/channel should be more than enough, and hence I will go out on a limb and suggest that 12-bit per channel color is about all we'll ever need to display a true continuous tone.
Now enters DR. Personally, I have yet to see any single image that contained a true 11 stops of DR that actually looked visually appealing to me, so I submit more than about 8 stops is not needed for output. (Again I feel the heat... ) However, to be able to capture it ALL, then manipulate it into a smaller output range of say 8 or so stops could definitely be an advantage, and in that respect I feel you will never have such a thing as "too much DR" in a capture medium. This of course assumes it will be manipulated well into the smaller output space, and with most of the currently used tools I don't think that happens very often...
Cheers,