Mattias Murhagen wrote:Briefly: I don't think it's correct to imply that there are "the first 8 bits" that somehow represent a signal more poorly than the "top" 8 bits.

You're right; a number is a number, and there's no division. They're ALL 8, 16, 24, or 32 bits, not mixed. Whether the number is 1 or 1,000,000,000 it's represented by the same number of digits.

Bit depth is about SNR.

No, it's not. The SNR is electronic. It comes from the preamps, and has nothing to do with the bit depth of the recording; the bit depth determines how many steps you have on your ladder, the noise floor is how much of the ladder you plunked in the water. The height of the ladder depends on its height, not on how many steps it has.

And I read the previous post too and though I understand all of the words and phrases I just don't see how it all comes together to something that has meaning. What do you mean by "linear" versus "logarithmic" in these contexts?

The log vs linear thing is similar. In a "linear" ladder every step represents an incrementally higher value. In a logarithmic scale the steps of the ladder are the log of the value they're encoding. So to represent a step whose value is 1,000,000,000 you'd use (in a log base 10) 9 -- 9 zeroes, i.e. it's 1 * 10 raised to the 9th power. The log of that is 9. (It's very simple with evenly divisible numbers, chosen because they're easier to understand.)

In sound 6 dB represents twice the loudness. So the first six steps represent, say, numbers from 1-2. The next six represent 2-4. Then 4-8. Then 8-16.

Except for the specific numbers, it's exactly the same as the way that video is log encoded, in the most basic curves. The more sophisticated curves are more complex because their encoding isn't a simple and straightforward log, but reading up on log encoding for video is conceptually the same as for audio -- and for the same reason; our ears much like our eyes function logarithmically.