This thread has been most illuminating. I've learned quite a bit here myself thanks to those knowledgeable on the subject. While I don't use Audacity for recording I have an interest in determining the quality of the recordings with the recorder I use, Wave Corrector. It claims 24 bit capability but uses the MME API similar to Audacity, which I understand (perhaps incorrectly) is limited to 16 bits and 44.1 kHz sampling.
Looking at the 24 bit recordings made with the program using a hex editor as described above, the third column is randomly distributed ones and zeros. In order to produce the blank third column as described above I had to truncate the file to 16 bits then build it back out to 24 bits using dbPoweramp. Now I see a column full of zeros, just as described.
My question is: Is it possible to determine the value of information in the third column (assuming it isn't blank data)? Is it real data or is it merely noise, perhaps the result of dithering?
In attempting to answer this I made a new recording with the soundcard 'disconnected' from the program (this was easy to do because I haven't yet figured out how to connect them...it's a new card....but that's another matter...). What I have is a silent file (-90db) with neatly ordered bits - either all ones or all zeros consecutively - in the first 2 columns of each row, and randomly distributed bits in the 3rd column. It looks something like this:
00000000 00000000 00101001
11111111 11111111 01000100
11111111 11111111 10010100
00000000 00000000 00010110
The file illustrated is supposedly 24 bits. Is it in fact so? In this case, what is the content of the third column?
My apologies to the OP if this is viewed as derailing the thread. This seems to be the best place to learn about reading these files in a hex editor, and may be of some use regarding Audacity.
How boring it would be, this endeavor
If all we heard was "perfect sound forever"