State-variable filters: what are they?

[pflosi]

Thanks mate, highly appreciated, especially from a genius like you!

That's how we roll in modular lands, a million ways to skin that cat

[meatballfulton]

That inversion mixing trick is probably how the Voyager gets it's HPF. I used to use it with my dotcom modular's ladder filter.

[pflosi]

There are lots of cool tricks with polarizing mixers to get other filter responses. Using a BP plus inverted original signal will result in a sweep between notch and BP, for example. With further mixers you can crossfade between all kinds of shapes. Use VCAs or, better yet, VC mixers and modulate all the levels.

Hours of fun!

[CfNorENa]

All self-oscillating filters are resonant, not all resonant filters self-oscillate.

OK, the first half of that confirms what I suspected. But the second half leaves me doubting what, specifically, a "resonant" filter is. Is it simply *any* filter that has resonance? If so, is there such a thing (in traditional subtractive synthesis) as a non-resonant filter?

(apologies, again, for this continued hijacking -- I did look around on the web for info, but can only find complicated engineering theses with, you know, like numbers and graphs and stuff )

[pflosi]

Sure, the HPF on a Juno is not resonant for example.

[Stab Frenzy]

All self-oscillating filters are resonant, not all resonant filters self-oscillate.

OK, the first half of that confirms what I suspected. But the second half leaves me doubting what, specifically, a "resonant" filter is. Is it simply *any* filter that has resonance? If so, is there such a thing (in traditional subtractive synthesis) as a non-resonant filter?

(apologies, again, for this continued hijacking -- I did look around on the web for info, but can only find complicated engineering theses with, you know, like numbers and graphs and stuff )

Yes and yes.

There are heaps of non-resonant filter designs out there, but they're generally a bit less fun than resonant filters for synth usage, so you don't see them as much on synths.

[piRoN]

To go into the technical side a little, resonance is really just adding a feedback loop somewhere in the filter circuit, that feeds the output back into the input in some way. The reason you end up with a peak at the cutoff frequency is to do with both the phase response of filter networks, and the specifics of how you implement the feedback loop (you can play with stuff like inverting the feedback signal and so on).

So on a non-resonant filter you just leave the feedback loop out. The filter works just fine without it because it's not an integral part of the filter network itself.

[Jabberwalky]

If so, is there such a thing (in traditional subtractive synthesis) as a non-resonant filter?

Regarding "Self-Oscillation"
The entire Yamaha CS range could not self-oscillate. Also the early Korg's (Minikorg, 800dv). It was generally done to keep the range of sounds more musically usable. Especially since people were just obsessed with recreating tubas in those days.

Some early digital filters were not resonant. The main one in my experience being the EPS16+ filter. Emu figured out digital resonance, as well as Roland, with the D50.

[CfNorENa]

To go into the technical side a little, resonance is really just adding a feedback loop somewhere in the filter circuit, that feeds the output back into the input in some way.

Interesting. Never knew that this was what resonance was doing to the signal.

The reason you end up with a peak at the cutoff frequency is to do with both the phase response of filter networks, and the specifics of how you implement the feedback loop (you can play with stuff like inverting the feedback signal and so on).

Two more questions: what, specifically, is that "peak?" My understanding is that the resonance somehow "emphasizes" the frequency (as determined by the cutoff of the filter), but I'm not sure what that actually means. It's not simply a boost in volume per se, is it?

And why does increased resonance attenuate the signal in some filters (famously the Moog ladder, but other 24dB filters as well, like on my SH-2), but not in others (e.g. the SEM)?

[meatballfulton]

Two more questions: what, specifically, is that "peak?" My understanding is that the resonance somehow "emphasizes" the frequency (as determined by the cutoff of the filter), but I'm not sure what that actually means. It's not simply a boost in volume per se, is it?

Yes, it is a boost in volume at that frequency. If it's a low pass filter, the response is essentially flat below the cutoff frequency, boosted at the cutoff frequency and attenuated above the cutoff frequency. See below.



Take a sawtooth wave, run it through a LPF with zero resonance and sweep the filter. Now do it again with high resonance (but not self oscillating) and as you sweep you can hear the individual harmonics jumping out.

And why does increased resonance attenuate the signal in some filters (famously the Moog ladder, but other 24dB filters as well, like on my SH-2), but not in others (e.g. the SEM)?

It's just a flaw of that particular design. There are filters which copy the Moog LPF but add corrective circuitry (often switchable) to not attenuate the signal.

[piRoN]

Two more questions: what, specifically, is that "peak?" My understanding is that the resonance somehow "emphasizes" the frequency (as determined by the cutoff of the filter), but I'm not sure what that actually means. It's not simply a boost in volume per se, is it?

Yes, it's a boost in volume - the same as using a parametric equalizer to boost a frequency range.

And why does increased resonance attenuate the signal in some filters (famously the Moog ladder, but other 24dB filters as well, like on my SH-2), but not in others (e.g. the SEM)?

I'm not really too clear on the ladder filter, but I'd suspect that's to do with the specifics of how the feedback is implemented. If the resonance in a filter is feeding back an inverted version of the output for example, you'd expect to see phase cancellation below the cutoff point (where the filter's phase response is closer to zero degrees), and then increasingly less cancellation as you approach the cutoff frequency where the phase starts to get shifted significantly by the filter. So the more feedback you add, the more you attentuate the sub-cutoff frequencies.

However, this is all just basic theory stuff, and it's worth noting that the action of any real-world circuit that incorporates feedback can often be incredibly subtle and complex.

[briandc]

Here's a page with audio examples of a 12-dB state variable filter, in various modes:

SVF

Nice sounds!

Csound, a cross-platform sound synthesis engine (beast!) has a myriad of filter types that can be implemented in a synth instrument, including, yes, a state-variable filter! Details here:
SVF for Csound


brian

[garranimal]

I made a video about the state variable filter and how to get this behavior on multi-timbral polysynths.

[Bitexion]

Damn, the Matrix-12 is more like a spaceship than I imagined. It looks like Scotty configuring the Enterprise for warp speed in the 1960s.

[garranimal]

Damn, the Matrix-12 is more like a spaceship than I imagined. It looks like Scotty configuring the Enterprise for warp speed in the 1960s.

Scotty would be so lucky. It's more TNG technology. When I program it, I feel more like Data tweaking at android speed.