Subject: Re: [linux-audio-dev] Setting up a vocoder demo?
From: Richard Dobson (RWD_AT_cableinet.co.uk)
Date: Sat Jun 16 2001 - 00:03:19 EEST
If you would consider a full pvoc rather than an analogue-style channel
vocoder:
unless you actually want the inputs to be stereo, you can feed two mono
sources into the left and right of one card, then run the whole thing
through a bespoke LADSPA (or VST?) plugin that runs a suitably adapted
phase vocoder. You can just multiply the frames together (streamed in
complex form), and get something interesting; whether it equates to a
channel-vocoder effect is another question. To get stereo output just
interfere with the frames on one channel a bit.
e.g:
http://www.bath.ac.uk/~masrwd/pvplugs.html
= sources for three pvoc plugins (using the FFTW libraries) in VST
format. Shouldn't be too hard to adapt for LADSPA, or whatever. The only
reason I haven't already developed a real-time phase vocoder app for
Linux is lack of a full-duplex card. I think Paul Davis may already have
downloaded these - any matters arising, Paul?
On my humble Pentium II 333MHz, I can just run 2 mono pvoc plugins in
real-time (4-fold frame overlap). So on the machines you folk will be
running, stereo pvocing in real-time should be a breeze.
The only other technical issue is latency - this is determined by the
overlap; so for sr=44100 and overlap = 256 samples, latency is around
6msecs. Embarrasing for low-latency diehards, but quite useful,
nevertheless.
My pvoc work is based on CARL pvoc (Mark Dolson), but also influenced by
PVC, which is an excellent package (though with a minimum of error
checking!); one issue for Intel users is the need to deal with
byte-reversal for the .snd format. I did some work on this a while back,
available from my 'downloads' page. PVC has one advantage in supporting
multiple channels (where CARL pvoc doesn't); but it is quite a bit
slower than the FFTW-powered pvoc.
However, note that all those pvocs assume offline processing (short
sounds), and thus take liberties with the output phase accumulator,
which, er, just accumulates! This is no good for a device that may run
for 24-hours or more; my code incorporates the necessary revision to
maintain bounded output phase ad infinitum.
I also have basic working phase vocoder code in SAOL, if anyone is
interested. On a fast enough machine, sfront-generated code might well
run in real-time.
Richard Dobson
Frank Neumann wrote:
>
> Hi *,
> for LinuxTag, I thought a nice demo would be to show some kind of
> vocoder under Linux. My idea goes like this:
> - Put two soundcards in a machine, one gets input from a microphone, the
> other gets string-type chords from a synth.
> - The vocoder software uses these two as carrier and modulator to
> produce some nice alien/space/scifi type of sounds/voices
> - This is put out again through one of the two soundcards. Both are
> Emu10k1 based, so I think both are able to do duplex pcm
> recording/playing at 16 bit.
>
> - Of course, all of this should be done in realtime.
>
> The only question I have: What software to use? I have found a few links
> at Dave Phillips' page, but I have no experience with these programs; do
> any of you have actual "stage experience" with one of them?
>
> Names I found:
> - Packages from the NoTAM group
> - PVC
> - PVNation
> - XVox
> - Zerius Vocoder
>
> Opinions, anyone?
>
> Thanks,
> Frank
-- Test your DAW with my Soundcard Attrition Page! http://www.bath.ac.uk/~masrwd (LU: 3rd July 2000) CDP: http://www.bath.ac.uk/~masjpf/CDP/CDP.htm (LU: 23rd February 2000)
This archive was generated by hypermail 2b28 : Fri Jun 15 2001 - 23:56:32 EEST