ksmattfish said:
If you go to, or past, the level of 100% white you are correct; the info is gone.
That's been my usage of the word "blown". It's an area that's gone to pure white, not just a bright area.
But a 12 bit (they call it 16 bit, but it's really 12) raw file has 2048 potential steps in the stop at the highlight end. That means 2047 are not true white.
Agreed. The whole range a 12-bit file can cover is 4096 steps per color. I wouldn't consider the top half the highlights, though. It all depends on how the image is exposed. (--edit-- Since the top stop is 2048 levels, it does make sense that it is considered the "highlights". I'm not sure why I wrote that. You were talking about a linear RAW file, not a TIFF.)
Currently available software is not written to be able to make 2048 levels of tone distinction in a single stop, and current monitors probably couldn't show it anyway; you may get solid white even if you really have managed to expose it a few dozen levels short of the right side of the histogram, but not actually gone into that last step of 100% white (or over). Future technology will allow us more precise control over these borderline tones. The information exists in the raw file, but our current software can't make such fine distinctions.
At first I was going to say that I didn't follow what you meant, but then I read this:
http://luminous-landscape.com/tutorials/expose-right.shtml
Photoshop can do this, we just need to learn how to master it. It's the conversion software that I think is a bit limited.
The software doesn't care what it looks like to us. The file is just a bunch of numbers. The camera will have a certain number of stops that are covered by those 4096 steps, but Photoshop isn't limited by that. As long as there is a distinction between two pixels (like 4096,4096,4096 vs. 4096,4096,4095), I can can change their relationship between them. And the difference between 0,0,0 and 10,10,10 could be 1/1000 of a stop or 1000 stops. The image won't look much like what we see with our eye, though. I guess most digital cameras capture about 5-6 stops effective levels.
From what I've read, RAW conversion software is basically just applying a logarithmic curve adjustment to the linear 1.0 gamma data so that the image is now at gamma 2.2, plus or minus any adjustments. You could do this manually if you wanted.
I completely agree with you that there can be more data in the RAW file than we see when we run the conversion, but that's only if the conversion blows out an area.
One of the misunderstandings with shooting RAW is the idea of underexposing. Since RAW is linear, each stop has half the info that the one above it has.
Stop 0: 2048 levels
Stop -1: 1024 levels
Stop -2: 512 levels
Stop -3: 256 levels
Stop -4: 128 levels
Stop -5: 64 levels
Stop -6: 32 levels
Stop -7: 16 levels
Stop -8: 8 levels
Stop -9: 4 levels
Stop -10: 2 levels
So if you underexpose so that you don't use that top stop, you instantly lose half of the info that your file can hold. The shadows don't hold much distinction. Stops -4 through -10 hold less info (254 levels total) than stop -3 does. By overexposing as much as possible without blowing, you get as much gradient detail as possible. This can really open up the shadow detail. Instead of having it all blocked up in, say, 128 levels, you could have it spread across 510 levels.