Re: 64bit processing - Why?
Re: 64bit processing - Why?
- Subject: Re: 64bit processing - Why?
- From: Richard Dobson <email@hidden>
- Date: Fri, 31 Jul 2009 01:28:31 +0100
Ethan Funk wrote:
..
Second, I'd like to know who (other than people from marketing
departments) need audio dynamic ranges in excess of 144 dB.
The issue is not dynamic range, but precision. You are discussing the
final delivery format, for which indeed 24bits is plenty. But for dsp,
especially filters (where small numbers may be multiplied together,
sample by sample), it is at the low end of barely adequate. Some filter
coefficients (exacerbated by the fashion for high sample rates) are
required to be ~extremely~ small, such that with 24bit precision, most
of the bits are zero. In the case of Csound we have already settled on
the double precision version as standard, as for a number of critical
processes the differences are audible.
In short: for a final delivery format, 24bit is great. For intermediate
processing (which by definition includes chains of dsp plugins), 32bit
floats are for some tasks (and the more so at say 96KHz) not enough. Of
course, quality plugins will always use double precision internally
(with modern hardware it may be the faster option anyway); but the
argument for maintaining the double precision path between plugins does
have some mileage. Internal gain changes can be extreme! And conversions
between double and single precision are not particuarly trivial.
My understanding of the VST double-precision mode is that it was first
mooted by Cakewalk for use in Sonar, and that they had in effect
extended the VST spec themselves, and then encouraged Steinberg to
endorse it as an official enhancement. Maybe the truth is more complex
than that simple version, but inasmuch as Cakewalk made it a major
promotional feature, one can say that the impetus for double-precision
plugins was at the very least 50% driven by a host developer. I suspect
the same would need to apply here. Apple may have little incentive to
introduce 64bit plugins merely as a matter of ideal or principle; but
~if~ a major host developer (which presumably has to include the Logic
team) discovers an advantage to it, then one would hope the CoreAudio
team would respond supportively.
Richard Dobson
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden