Quote:
Originally Posted by jbjb
Exactly. 
With 16-bit sampling of an analogue signal, the input maximum amplitude range is divided into 65536 (2 to the 16th power) levels, and the value of the input signal at each sample point is approximated by the nearest level. With 24-bit sampling, you're dividing the input range into 16777216 (2 to the 24th power) steps, which clearly gives a much more accurate representation of the sampled signal.
This process is known as quantisation, and the differences between the 'real' analogue input values and their quantised approximations ('quantisation error') effectively adds some noise, which is known as 'quantisation noise'. If we have smaller steps, we have, on average, less quantisation error per sample, and hence less quantisation noise.
/JB |
OK, I understand now; I know about that topic (the basics, at least). While you're right, the same dude as I quoted before essentially says: "It doesn't matter. Quantisation noise is so low with 16 bit already that stepping up to 24 bit is useless." (For listening, not editing.)
He says so in the video on this page:
https://xiph.org/video/vid2.shtml
"It *might* have been more important if the CD had been 14 bit, as originally planned.... maybe."