Self-oscillating filters?

[Hernan]

Hello,

I am trying to understand what a self-oscillating filter does. I read about them in a couple books/ magazine articles but I'm not sure I understand what they do. Generate a (sine) wave at the cutoff frequency?

Can anybody give a quick explanation/ point to a web page that explains these filters? Any links to sounds will also be useful.

Thanks!

[ChipCurtis]

I would just type "self-oscillating filter" into Google and see what you come up with.

Just a simplification: yes it is basically a sine wave produced by maxing out the Resonance while sweeping the Filter Cutoff. Not every synth can do it, it's a matter of how the trimpots were set internally, or if those circuits are robust enough to produce it. It is simply a kind of electronic feedback, for lack of a better term. Before the 90s synth/dance craze, it was considered largely an unwanted byproduct of those circuits. It was rediscovered as a 'serendipidous effect' on analog gear, and put to use in dance music of the 90s.

(I have been reading Keyboard Magazine and other electronic keyboard periodicals since the 70s. Not once did I ever come across the term 'self-oscillating filter' until about 1994, and that was on the internet.)

A lot of people misnomer it as "self-resonance".

[AdmiralQuality]

I too read Keyboard magazine (my subscription starting circa 1983 and lapsing around 87) and read much of filters that could be driven into self osscilation. Not sure how you missed it. It's certainly not a new thing.

The resonance control of the filter section of most synths, works through feeding back the difference between the output and the input, back into the input. This feedback is often referred to as "Q" by the analog designers who really understand this stuff, and is frequently labelled Q on the control panel as well (like the Q on a parametric EQ which controls the width of the band.)

Like any feedback, if you drive it too high, it'll cause oscillation. Some filters let you push it that high, some don't. That's all.

I don't believe it's ever been unwanted -- at least not in analog synthesizer design (cuz you can always turn it down and make it go away) Though we certainly don't want parametric EQs that self resonante. But synths have been doing it right back to the first Moog modules in the mid 60's.

Opening note of Tom Sawyer by Rush. There's a filter right at the edge of self-resonating. (It's a Minimoog) Its also got some cross mod from an oscillator to the filter's cutoff, making the warble/growl as opposed to sounding like a pure sine wave (which is the waveform you'll get out of any simple self oscillating system)

You need some impulse to start the process, but that's never a problem in an analog circuit that has inherent noise. I just wrote my first digital filter though, and, while it certainly oscillates when I max the resonance, I need to put a sound into it first to get it started. Thats because there's no noise to start it. (Before I can build a convincing softsynth with this filter, I'm going to have to provide an artificial noise source to get this happening.)

Think of it like a howling mic, you know how there's that point just before it starts to howl? How the sound gets colored and you can tell it's about to feedback. Well that's resonance.

Spectrum wise, resonance creates a peak around the cutoff frequency -- the point where the downward slope starts (in a low pass filter anyway) -- boosting frequencies in this narrow band, with the peak getting higher (louder) and narrower as you increase it. Eventually it reaches a point where it's so narrow it's a pure sine wave, and so loud it's drowning out almost everything that's coming into the filter.

Aside from spacey effects, it's a good way to determine what the filter's cutoff frequency is. (so you can tune it to an oscillator then back off a bit). Also comes in handy for making fake organ sounds on an analog synth (think jazz tone rather than pipe organ tone) You can use it to bring out a particular harmonic just before the cutoff point.

When the filter's cutoff is set to follow the keyboard, you can play this "whistling" sound just like it was another oscillator. On analog synths that didn't offer sine waves from the VCOs, this was often the only way to get a true sine out of them.

[ChipCurtis]

I too read Keyboard magazine (my subscription starting circa 1983 and lapsing around 87) and read much of filters that could be driven into self osscilation. Not sure how you missed it. It's certainly not a new thing.

I have a library of old issues dating from early 1985 to about 1991. No mention of “self-oscillation” in any article whatsoever. Can you show me the article you are referring to, and I can check on my copy? (if we both have the same issue in question).

This feedback is often referred to as "Q" by the analog designers who really understand this stuff, and is frequently labelled Q on the control panel as well (like the Q on a parametric EQ which controls the width of the band.)

Actually there may be like 3 or 4 synths ever that called it ‘Q’ on the control panel. Techs use the term freely in their jargon, whitepapers, whatever. But it's not a common thing to see on a synth panel, perhaps with exception of a lot of softsynth stuff.

I don't believe it's ever been unwanted -- at least not in analog synthesizer design (cuz you can always turn it down and make it go away).

I would make a distinction here between ‘capable’ and ‘desired’. During the time of the design/construction of the first modulars, not even the designers themselves knew all the kinds of sounds their machines were capable of. By far most synth sounds generated until about 1970 were basically bleeps and farts (almost without exception), until the first great players/programmers came around (K.Emerson, T.Banks, R.Wakeman, for instance) who were capable of creating sounds that were both fat and musical. Self-oscillation (useful mostly for electro/sci-fi FX) was certainly discovered back in the 60s, and perhaps used for something or other in sci-fi movies, Dr. Who, whatever. But the curiosity and infatuation with it (basically, “the current trend”) is for all practical purposes a phenomenon of the 90s.

Though we certainly don't want parametric EQs that self resonante.

Ooops. Refer to the bottom paragraph of my first post.

[AdmiralQuality]

Can't say I agree with all of that, but it's mostly pretty subjective stuff so I do respect your opinions.

And sorry, Mom threw out my old Keyboards. I learned it somewhere though, and Keyboard was a staple of my early learning. (I might have also read it in a really great book set that Roland had on synthesis theory. Can't remember, it was forever ago.)

BTW, I play self oscillating filters on my Juno-106 all the time. I'm lucky, my Juno is pretty stable, the filters are all in tune with each other, so this works fine for sine wave based sounds. About the only way to get a sound that isn't all static sounding due to the too -perfect clocks of the digitally controlled oscillators (though, sometimes that's a feature too. Expecially when it's phasing against itself in Pole 1+2 (Solo) mode!

Here's my first VST by the way, which just happens to be a resonant LPF -- it does self-oscillate but needs a kick to get it going as there's no noise in the digital circuit http://admiralquality.dynu.com/vstdev/?selfosc


Cheers!

[ChipCurtis]

Okay, cool.

I will check out your plug-in when I get a chance. From what you are saying, it just needs an oscillator.

BTW, have you checked out my MinimogueVA and Arppe2600va emulations? (Well, 'ours' not 'mine' hehehe!) They're pretty decent (and free!). Just look for them here at KVR.

[ZooTooK]

Opening note of Tom Sawyer by Rush. There's a filter right at the edge of self-resonating. (It's a Minimoog) Its also got some cross mod from an oscillator to the filter's cutoff, making the warble/growl as opposed to sounding like a pure sine wave (which is the waveform you'll get out of any simple self oscillating system)

OK, this is sort of off-topic but I believe the intro sound on Rush's Tom Sayer is made primarly by a Oberheim OB-X put into mono mode. Yes, near self oscillating, you can't make the OB-X go into self oscillating. The growl/warble is due to all the oscillators (16) and the 8 filter peaks are slightly detuned. You just can't make that massive sound with a Minimoog unless you do massive overlay recordings. I also suspect they used a Moog Taurus bass pedal in parallell.

The same OB-X sound is used by Tangerine Dream on their "Thief" sound track. Could it be a factory patch?

[kritikon]

I am trying to understand what a self-oscillating filter does. I read about them in a couple books/ magazine articles but I'm not sure I understand what they do. Generate a (sine) wave at the cutoff frequency?

Pretty much - the nearer to a sinewave it is, the worse the filter is. It's one of the reasons very few VSTi synths self-oscillate very well - they just sound like a nasty simple sinewave over the top of the patch. Nasty IMO. Although some devs have obviously put alot of hard work into emulating the little variations and inconsistencies that the better analogue filters had (such as GMedia). And it's not just VSTi - any digital h/w synths too - until about 2000 I'd never heard a digital filter I liked on any synth.

It's one of the few reasons analogue fans still collect them - for filter character - whether it's down to inherently noisy circuits, irregularities in the characteristics the filter circuits exhibit, and often analogue distortion/overdrive designed as part of the filter itself or a mixture of all of those (and more. So it's not just about the sine wave...it's about distortion and the harmonics that a filter feedback gives.

You find the real stand-out character filters are those on quite distorted and/or noisy synths(yet that doesn't mean the actual synth output is going to be noisy). OSCars - had an overdrive facility specifically designed for the filter. Korg MS series - lots of physical distortion when you overdrive the filter res, which you can then feed back into the input to make it distort even more, and it makes for a truly "screaming" filter - it's only as you really lay on the overdrive that the character comes out.


But don't assume that it's only analogue synths that have good resonant filters - there are some excellent VSTi ones...and there were plenty of analogue filters that were boring and dull. In fact someone mentioned the Juno106 - I have never used that filter in anger properly in any song, simply because it's another one of those nasty simple sine-wave irritating and un-musical filter self-oscillations that would be worthy of being a Synthedit filter self-oscillation.
That's why I tend not to use many Synthedit synths - some of them have great feature sets, and can be used to make very good sounds, but invariably they fall down completely flat when you self-oscillate the filter.
And many of the later Roland analogue filters were pretty weak too - even things like the 101/202 - although I like the general character of the filter, it's bloody useless when overdriven - it just whines. (Extremely loudly, I'll admit ).

A truly good filter will let you play the filter itself as a patch. You can turn off the oscillators in the onboard mixer and the filter will still make its own feedback sound. A bad filter, and that sound will just grate on your ears as a thin little whine.

Although many good filters still cannot fully self-oscillate to much good effect, they can sound very good at or only just over the point of self oscillation. Rolands are generally OK as long as the peak is not pushed to actual audible range. Oberheims - as mentioned - some of them couldn't truly self-oscillate (which was a shame) but they sounded their best at the fullest resonance or close to it.
I never much liked any Moog, and I feel they're pretty dull resonant filters, but again they can sound OK around the level of self-oscillation.


Alot of it is personal preference for style of sound. But to me, the noisier the filter, the better it is. Pure digital filters are pretty shite if left untweaked. But that could be one of the main benefits with coding digital filters - it can be possible to emulate all the little foibles of analogue circuits to get the character, and even more can be done to it - It's a shame more devs haven't got into actually carefully coding completely new filter inconsistencies and noises. There's no reason why we should see analogue filters as the be-all and end-all in filter sound. Why not design different types of overdrive? Different types of modulated noise to trigger the filter? There are all sorts of things that could be done to give filters completely new character.

I'd rather buy a VSTi that a dev has spent a year carefully coding a whooptastic filter and only has a saw,pulse and sine wave, than buy it half price with a simple dull digital filter but maybe 15,000 patches and 784 oscillators to choose from.

To my mind, and ears, a filter should be messed up as much as possible. Pristine clean self-oscillation is horrible.


And it's not a new 90s thing either - I can remember "self-oscillation" being a topic as far back as when I started out. I think the mistake was assumed that because self-oscillation was not a popularly used tool, that nobody knew about it or called it that. Remember that when the 303 1st came out, it wasn't at all succesful compared to the hype it gets (or used to get ony a handful of years ago). Hardly anybody used the filter res anywhere near full, it just wasn't a style that was around for whatever reason. People used it as a bass synth - it was marketed as a substiutute for real bass - and obviously that wasn't actually its real strong point - the bass on a 303 was never particularly good. However - those synth fanatics that were around at the time still knew exactly what self-oscillation was...they just got thrown out of the band if they used it.

[rick_jelliffe]

And not all self-oscillation are sine waves: an overdriven filter, or one which has high internal signal levels, may be driven past the saturation levels of the components. The result can be that the self-oscillation is not a clear, louder sine but a fuzzier soft-clipped note closer to the level of the input signal.

(For example of this, SynthEdit users can try out the 30 filter module on the Yahoo site. Most of the fitlers not named "f" exhibit this saturation behaviour.)

Rick

[AdmiralQuality]

If you guys don't like sine shaped self-oscillation, check out my Naive LPF v 0.2 http://admiralquality.dynu.com/vstdev/?selfosc2 It's my original 1-Pole LPF, now switchable to 2, 3 or 4 poles, as well as HPF mode. In the process of adding more stages, it's resonance went out of control, producing a lot of harmonics and aliasing frequencies when turned up too high.

I'm hoping to fix this next, but I think I'm going to keep this "dangerous" version for use in a phaaaat softsynth.

And kritikon, great post. Only thing I disagree about is that I always liked the sound of Moog and Roland filters, but that's all subjective of course.

Cheers!

[jasonsantiago]

It is simply a kind of electronic feedback, for lack of a better term. Before the 90s synth/dance craze, it was considered largely an unwanted byproduct of those circuits. It was rediscovered as a 'serendipidous effect' on analog gear, and put to use in dance music of the 90s.

actually there were major electronic acts, such as groups doug lynner was in in the late 70's and early 80s... when I went to his house he showed me how he did LIVE drums using an analog synthesizer 25 years ago. Pretty impressive. I think one of his bands sold a record of 500,000 copies so that I would say was fairly maintstream usage of self oscilating filters.

And the self oscilatting filter was in almost all of his patches... I tried really hard to find an equivelent in reaktor for him but alas I failed. Anyone know of a software equivelent in reaktor or tassman?

sorry to dig up such an old thread, but I have a huge interest in self oscilatting filters.

[Sascha Franck]

Nice post, kritikon.

[Debutante]

Can someone please refer a VSTi that does self oscilliation...I'd like to try it out to find out if it's something completely new to me or something I'm familiar with and just haven't heard in term yet.

[AdmiralQuality]

Can someone please refer a VSTi that does self oscilliation...I'd like to try it out to find out if it's something completely new to me or something I'm familiar with and just haven't heard in term yet.

I did, in my last post. Check it out, much improved since I originally posted in this thread. It's all pretty now too...


http://admiralquality.dynu.com/vstdev/?selfosc2

[jupiter8]

Can someone please refer a VSTi that does self oscilliation...I'd like to try it out to find out if it's something completely new to me or something I'm familiar with and just haven't heard in term yet.

More or less all of them does.