Subject: Re: [linux-audio-dev] stream buffering on EVO & pitchbend .... opinions ?
From: eli+@gs211.sp.cs.cmu.edu
Date: Thu Sep 21 2000 - 20:20:33 EEST
Paul Barton-Davis wrote:
> >No. It is more like a very long pitchbend. The tone slides to the new pitch.
>
> right, but the slide only needs to go as far as the next pre-existing
> note. then you stop computing a bend on the previous note, and start
> using the new one. and so on.
The K2000 does something like this, but it doesn't sound like it
crossfades nicely between successive samples, so there are clicks.
Has anyone heard this done with crossfading? I bet it would be pretty
good with a well-matched sample set -- maybe a little modulated-chorus
effect, but okay. I do wonder what it will sound like when the
sample set changes in timbre or envelope shape along the keyboard,
but what can you do with only a sample set?
But how would you implement this? It's fine when everything's in RAM,
but here we are 3.14 seconds into D3 and all of a sudden we start a
portamento up, so we need 3.14 seconds into D#3. Seems you'd have to
stream samples above and below the one you're playing, three times the
bandwidth. (Or you could limit portamento rate so the time to go a
semitone is always long enough to fetch what you need off disk, but...)
About the other portamento approach, the very-long-pitchbend approach,
here's a possible way to make it easy: for portamento from C2 to C4
immediately switch over to the C4 sample. Then the pitch shift is
downwards (sliding from -2 to 0 octaves), so you don't need extra disk
bandwidth. For portamento from C4 to C2, you retain the C4 sample, so
the shift is downwards there too.
The risk is that if the sample crossover is audible, it'll happen at
different times between up and down, sounding sloppy. But if it's
audible, if there's any structure to the spectral envelope, the
very-long-pitchbend approach is going to sound odd no matter what the
crossover timing is. Imagine an "ahh" sample shifting rapidly two
octaves into chipmunk land.
--
Eli Brandt | eli+@cs.cmu.edu | http://www.cs.cmu.edu/~eli/
This archive was generated by hypermail 2b28 : Thu Sep 21 2000 - 21:39:31 EEST