As stated in this discussion, 32bit floating point has less linearity
than 32bit integer (although more dynamic range). 64bit floating point
is natural to native processing, so it is not a bad decision to keep
this precision throughout the audio paths. Native hosts are competing
to DSP based systems, tipically having 32 to 48bit integer processing.
Talking about AU plug-ins, although their internal processing precision
is a different thing than their external interface precision, it is
best not to do unnecessary conversions (with or without dithering) to
keep the maximum audio quality. So apart of having the "component
opening with doubles ASDB" check, there would be nice to know what
internal representation does the plug-in has (even for adding dithering
if needed). I know that the main problem here is to require the
manufacturers to fill this information for the host to recognize it,
but this would allow them to add a marketing added value if they add
64bit processing too!
Alejandro
|
