[Mike Rivers]

On Feb 9, 10:03 am, "Take Vos" wrote:

As most audio devices do not have AD/DA converters more accurate than
24 bit integer there have not been much use for 64 bit floating point.

There is some method to this madness. Whenever you perform some
operation on a sample like change its level or add two sample values
together (really, all audio processing can be reduced to these
operation) the word length increases. Originally all arithmetic was
done so that every operation was truncated to 16 bits and this got to
sounding pretty rough after a few operations. This got translated to
"digital mixing doesn't sound good" and people were getting better
sounding mixes of digital multitrack recordings by using an analog
mixer. The added noise and analog distortion was preferable to the
distortion caused by truncation.

But old ideas die hard (the Internet hasn't helped that) so even in
the 24- or 32-bit floating point world the perceived problems with
digital mixing still remain. And with some systems, they probably do
remain because not everyone does things right. But now at least it's
possible to construct a good digital mixer if you have enough
"headroom" (which we now have).