Multi Stage Biquads

matti · 1 May 2007

unkargherth wrote: Ok boys and girsl this is the scenario

I'm doing some teston building very deep filters. I came with the idea of using multistage (also called multisection) biquad filters based on RJB Cook Book Implementation

the obvious first test it's to calculate coeffs once and feed them infto a collection of serial linked Biquad engines. And YES, it gives DEEP filters as results, but...

What do you mean by DEEP filters? I couldn't get a multistage filter to cut more dB's than one single stage filter. Or do you mean you want a steeper q for the filters? That can be achieved by multistages.

unkargherth · 2 May 2007

Hmmmm.

At least in theory, Cascading stages should give you "deeper filters"If a "standard" ( read "per book RBJ") biquad is 12dB/oct. cacading two of them should give you aproximately 24 dB/oct, for three you get 36db/oct, and so on. The variation depends usually on cut frequency, and , in peak filter on gain

At least in theory

The question is. This works reasonably good on LPF and HPF. But on Bandpass i get small instabilities (basically on high Q's) and the instabilities goes crazy For Peaking and Shelving filters when cascading 2,3 or 4 Biquads ins series. Thats whay i'm asking if some math conversion should be done on Gain dependindg on number of stages

matti · 2 May 2007

unkargherth wrote: Hmmmm.

At least in theory, Cascading stages should give you "deeper filters"If a "standard" ( read "per book RBJ") biquad is 12dB/oct. cacading two of them should give you aproximately 24 dB/oct, for three you get 36db/oct, and so on. The variation depends usually on cut frequency, and , in peak filter on gain

At least in theory

The question is. This works reasonably good on LPF and HPF. But on Bandpass i get small instabilities (basically on high Q's) and the instabilities goes crazy For Peaking and Shelving filters when cascading 2,3 or 4 Biquads ins series. Thats whay i'm asking if some math conversion should be done on Gain dependindg on number of stages

I only have experience on peaking filters and if you cascade those you loose gain range. For example 4 filter cascaded together would result in maximum of only -20db cut etc(well i cant remember the correct figure for rbj filters right now).. and it goes crazy if you cut more 🙂

Use the analyser to see whats happening when you cascade the filters. That should help you solve all your problems yourself 😉

AngularMomentum · 2 May 2007

I do use cascaded filters, and do use the cutoff input of all cascaded filters, but only the resonation input of the first or last one...this way I do not have my speakers bleeding while having deeper bass and such

Acrobat · 15 Jun 2007

unkargherth wrote: Ok boys and girsl this is the scenario

I'm doing some teston building very deep filters. I came with the idea of using multistage (also called multisection) biquad filters based on RJB Cook Book Implementation

the obvious first test it's to calculate coeffs once and feed them infto a collection of serial linked Biquad engines. And YES, it gives DEEP filters as results, but...

I get alot of unstabilities, especially when using BandPass and PEAK modes. Then the question is:

Should I tweak the input freq and Q ( and maybe gain) in some special way becuase I'm cascading biquad sections?. Somebody has the correct maths for doing it , assuming that the number of biquad section to use is user-selectable inside some predefined range ( say between 1 and 8, for example) ?

Mmh.... dunno if it has some interest for you but lets focus on filter clicks, for a moment.... I experience clicks in audio using Hi Resonance values, with almost every filter I implement in my synth: the SM ones and the others I collected from the forum. Why? I mean it's aliasing or envelope or what other issue? And what do u mean exctacly with "instability"? A.

infuzion · 16 Jun 2007

Acrobat wrote: I experience clicks in audio using Hi Resonance values, with almost every filter I implement in my synth: the SM ones and the others I collected from the forum. Why? I mean it's aliasing or envelope or what other issue?

Try this please: [old link]

Acrobat · 16 Jun 2007

infuzion wrote: Acrobat wrote: I experience clicks in audio using Hi Resonance values, with almost every filter I implement in my synth: the SM ones and the others I collected from the forum. Why? I mean it's aliasing or envelope or what other issue?

Try this please: [old link]

Yes, it's useful but I already did all the things they say. And that's when caused by the envelopes, not by the filter resonance... that still makes clicks especially with bandpass and highpass (the SVF from the toolbox).

A.

infuzion · 16 Jun 2007

If the filters make sudden value changes (like +0.9 to -0.4) all the time then you'll get clicks. W/o a schematic, there's not much else I can help you with.