In article ,
Arny Krueger wrote:
A Layla20 is more like 17 bits. Cool Edit pads it out to 32 bits (24 bit
mantissa+8 bit exponent floating point). The alternative would be 16 bits
which would lead to recording with less precision than the hardware is
capable of. To me the real benefit of recording at 32 bits is the ability
to process and mixdown without clipping or other loss of whatever data is
available.

Wouldn't it make more sense to print 24 bit fixed point tracks? 32
bit float is not that common of a format, and if you decide to go back
to the raw audio tracks later on, perhaps in some deep distant future
world where 32 bit float isn't used or after Cool Edit has stopped
working, you'd be S.O.L.

Also, and this is a very tiny nitpick, 32 bit float is normalized
(i.e. in the sense that the mantissa is always between -1 and 1) and
there is multiplication involved, whereas a straight 20 to 24 bit zero
padding is _completely_ clean. Yeah, the error from this is extremely
tiny, but what the heck - why not capture the bits as delivered and
then let CE pad it to float or whatever the heck it wants for internal
processing.

Just a thought...

Monte McGuire