[music-dsp] On time-varying filters
hplus at mindcontrol.org
Thu Aug 24 13:52:44 EDT 2000
I've done this with reasonable success by using a single set of interpolated
coefficients per 4 frames, and calculating new targets every 128 frames.
This was for a 4-th order IIR. Really fast movement caused some temporary
resonance spike like glitches, but that sounded kind-of cool so I left it in
If you're doing it on a bi-quad, it works best on the un-transposed form
because its coefficients are the least sensitive, I'm told.
>On Wed, 23 Aug 2000, Alexandre F. Bouchet wrote:
>> How do you implement time-varying filters in real time?
>How about linear ramps between old and new coefficient values and precise
>calculation of coefficients from the parameters (cutoff etc) at the
>startpoint of each ramp. The frequency of the ramps would need to be
>high at times when there are large changes in parameters. Linear ramping
>is cheap, just one addition per coefficient per sample.
>The hard part is that you need to know "what's coming next" and "when".
>dupswapdrop -- the music-dsp mailing list and website: subscription info,
>FAQ, source code archive, list archive, book reviews, dsp links
dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
More information about the music-dsp