Zeke said:
I find it somewhat confusing to talk about dynamic range in bit depth though since HDR is a floating point 32bit whereas images we traditionally deal with are integer. It's a lot easier to deal with f-stops. I think raw is about 10 stops from black to white.
It's easy: one bit equals one f-stop. For example assume a 4-bit monochrome image. The shades of gray would be represented with numbers in the range 0...15 but lets call them 1...16. 1 is four stops less then 16. (each stop is a doubling of the amount of light, double 1 four times and you have 16)
But notice the other problem with the 4-bit image. while it can hold four stops of dynamic range the grey scale is in one stop steps. You can trade dynamic range for smaler steps. For example with an 8-bit image you can store 8 stops of dymanic range but with one stop "steps" or you can have 6 stops of range with 1/4th stop steps.
Let's say you want 12 f-stops of dynamic range and 1/8th stop steps. thats 12+3 bits, you would need 15 bits to do this and a 16-bit tiff file would work fine.
I think you are right about RAW being about 10 f-stops. Typically RAW is 12 bits. But RAW files do not contain pixels and must be interpolated into images so the end result will have some bits to the right of the binary point if you are using fixed point math