[LAU] Looking for a (fixed) filterbank plugin

Fons Adriaensen fons at linuxaudio.org
Thu Jul 1 17:45:24 CEST 2021

On Thu, Jul 01, 2021 at 02:08:45PM +0200, David Kastrup wrote:
> Isn't any LTI filter with conjugated pole pairs (Q > 1 or some other
> simple constant I'd need to look up) resonant in that it produces a
> decaying sine with fixed frequency (for each such pole pair)

A conjugate complex pole pair is indeed almost a definition of
'resonance'. Which also corresponds to a second order linear
differential equation with a negative discriminant.

Such resonances occur easily in physical systems, all you need
is a restoring force proportional to displacement, Then Newton's
law F = m * a results in an acceleration proportional to minus
displacement. Acceleration is the second derivative of displacement,
so the result must be a function which has a second derivative 
proportional to minus the function itsef. The second derivative
of sin (w * t) is - w^2 * sin (w * t), so that fits the bill.

>  when its input is switched off?

Not just then, for any transient input. The impulse response is
an exponentially decaying sine wave.

For a formant filter such resonances are the obvious choice,
but as a filter bank meant to separate a signal into almost
non overlapping bands they are not very useful. To get e.g.
the IEC class I octave or 1/3 octave bands you need at least
6th order.



More information about the Linux-audio-user mailing list