3 Feb
2004
3 Feb
'04
10:38 a.m.
On Tue, Feb 03, 2004 at 03:13:54 +0100, Alfons Adriaensen wrote:
On Tue, Feb 03, 2004 at 01:52:13PM +0000, Steve Harris
wrote:
OK, thanks. Yes, that rings a bell now - I think I made it a bit better by
oversampling the input (2 or 4x). I'l have a look at your plugin to see
what you've done.
- Steve
On Tue, Feb 03, 2004 at 02:21:53 +0100, Tim
Goetze wrote:
Moving it to doubles will not help. The problem is that at high F, the
simple mapping from the continuous time domain to the sampled domain breaks
down. You need to modify the params as non-linear functions of F and Q
in order to correct for this. ps: the svf sounds better to me with the
'very slight waveshape for
extra stability' removed.
Yeah, but without it the filter tends to explode with highish cutoff
values (above fs/4 IIRC). I should probably move it to doubles and remove
the waveshape.