May I suggest that you try using plugs that work with higher
resolution. 32-bit floats (which carry 24-bits of resolution) gives
some headache in terms of rounding errors. Try to find plugs that
operate at double-precision internally and (important) make sure they
use dither back to 32-bit float/24-bit fixed. Most floating point plugs
dont dither and instead rely on the rounding errors to mask the
resulting quantization distortion. The rounding errors are
signal-dependent and is distortion.
It took me a while to clean up the editing workstation so it only use
software that do the math correctly in every stage/plug. But doing this
is a one-time job and the result sure was worth it.
2004-07-26 kl. 21.02 skrev Rob Danielson:
> I've had a
> half dozen experiences with bringing home both 16 and 24 bit files
> from the same location through the same front end. It's easier to
> work with the 24 bit files compared to the 16 bit files. There's more
> "body" providing more subtle tonal/amplitude differentiation making
> equalization changes easier and more effective. The lower the file
> saturations, the more pronounced the difference. I do up-sample 16
> bit recordings to 24 bits in post when I know they will go through
> further changes downstream, but this does not create a file that
> handles as well as a 24 bit orig. All plugs are converting/working
> with 32 bit floating regardless of orig sample rate of course.
________________________________________________________________________
________________________________________________________________________
|