Path: utzoo!attcan!uunet!mcsun!ukc!harrier.ukc.ac.uk!gos.ukc.ac.uk!mg
From: mg@ukc.ac.uk (M.W.Guy)
Newsgroups: comp.dsp
Subject: Re: Digital pitch shifting
Message-ID: <2387@gos.ukc.ac.uk>
Date: 23 Sep 89 13:16:30 GMT
References: <1989Sep21.163112.5656@ivucsb.sba.ca.us>
Reply-To: mg@ukc.ac.uk (M.W.Guy)
Organization: Computing Lab, University of Kent at Canterbury, UK.
Lines: 21

In article <1989Sep21.163112.5656@ivucsb.sba.ca.us> todd@ivucsb.sba.ca.us (Todd Day) writes:
>Is there an easy way to do pitch shifting digitally in real time?
>The only method I've been able to come up with involves repeating
>or throwing out sections, which results in clicking.
>
>I would like the method to be short and sweet, as I only have
>time for 125 instructions/sample (250 if I only do mono).

The real way to do this is the ubiquitous phase vocoder.
Process the sound into amplitude/phase components for many frequency
bands, diddle with the results appropriately (in this case, shift the frequency
components up) and then transform back.  A 1024-point FFT will do this for you.

However, in 125 instructions of a conventional processor, your best way is
probably to have a cyclic buffer, with independent read- and write-pointers,
running at different rates.  You get a click every time the pointers pass each
other, but you can arrange the size of the buffer so that this happens every
50th of a second and people will ignore it as mains noise...

Yuk
	Martin