k y m a • t w e a k y — the kyma collective || view the current website || February 2015 Archive

/ WebHome / Sounds / Share.DiscussDispersivePhysicalModel


Share Section



All Sections


Login / Register 
Change password
Forgot password?

Symbolic Sound

Kyma Forum
Eighth Nerve

TWiki Links

TWiki Formatting FAQ
What is TWiki?
%SESSION_ENDIF% twiki.org
This sound uses waveguide synthesis along with a dispersive Allpass Filter in the feedback loop to produce anharmonic spectra. Similar to the case of metal bars, the resonant frequencies do not form a nice integer related sequence because the effective delay length depends on the frequency.

The allpass filter produces a group delay for frequencies in the vicinity of the filter frequency Frequency. The bandwidth of this dispersed group of frequencies is controlled by Q. The Delay parameter selects the duration of the delay line in series with the allpass filter. The amount of delay for those bent frequencies is directly proportional to Q, and inversely proportional to Frequency. You get the greatest bending of sounds by choosing a low value for Frequency.

Were it not for the presence of the all pass filter, the delay line along with positive or negative feedback models a simple resonator. But the presence of that all pass filter modifies the effective delay for frequencies around the filter tuning frequency. As Q is increased, so too is the effective delay for those frequencies.

As a result, you can get some very interesting sounds out of this little Sound. Unlike a metal bar, where the group delay goes as 1/Sqrt(F), this one has a more complicated and localized dispersion located around the filter frequency. So it won't ever sound like a bar. But it does have some features of rubber bands, coffee cans, and other non-musical instruments.

[Try setting delay to 0.005, and locking all but the Frequency and Q parameters in the VCS. Then roll the dice for some nice juicy sounds. Wooden Gamelons, anyone? ]

- DM

-- DavidMcClain - 16 Dec 2003

Hi David I just made a slight adjustment to your above thing. by simply moving the cut off filter so that it filters the source sound as well, makes the click at the beginning of the sound less offencive when you try to create a dull type of hit.

-- PeteJohnston - 18 Dec 2003

Now that is an interesting idea! Thanks for the tip, Pete. Ahh... but the lowpass filter in the feedback loop is still needed to avoid runaway higher "harmonics". My original intent was to darken the feedback gradually using a 1 pole LPF. Similar to what is done by many synthetic reverbs.

-- DavidMcClain - 18 Dec 2003

Yes David The low pass filter is still in the feedback path as well and it does still do the gradually darkening it just means that the first time round gets a bit of filtering also. On your Piecewide bar I put a copy of the Low pass filter in the source path only ( with seperate controls), I found you could get a darker sounding thud and a long ring. It seem to work quite well, ( I didn't post this one). I also tried to make one where the delay was controlled by the midi keyboard but I couldn't get the internation quite right. I think the delay caused by the filter is not linear (with respect to frequency) even when the Q is turned down.

-- PeteJohnston - 18 Dec 2003

Heh! Nonlinear indeed! By design! So yes, it will be difficult to do any kind of "tuning" to pitch. I did think of one other place where these kinds of Allpass filters are used in the studio. Probably not often by themselves, but when building a Phaser, you can hear the effects of changing Fc or Q. But it is generally recognized that the human ear cannot hear a static phase difference in sound. It is only sensitive to changes in phase with time. However, the bass enhancement using extreme amounts of group delay does add a noticeable but subtle element to the sound that can be heard by the human ear.

-- DavidMcClain - 19 Dec 2003

Here is an extension of the above idea, incorporating 4 of the Kyma AllPass Filters all in series. I tuned them roughly to a 1/Sqrt(F) dispersion sequence, setting the filters at octave intervals and increasing the Q by Sqrt(2) as we descend in frequency. This process could be carried further to get a more faithful approximation.

But equally as interesting is to just roll the dice on the four frequencies and their Q's. Now we truly have a bizarre and interesting dispersion curve with frequency!

[ in fact! you can actually remove the delay line element from this sound and it still behaves like a dispersive resonator. The delay comes entirely from the group delays in the filter bank. It sounds like a metal drum to me, something larger than a 55 gallon can, and smaller than a winery vat... ]

DavidMcClain - 16 Dec 2003

Here are some interesting applications of the multiple APF resonator. One is an expansion on the metal bar, using Pete's recommendation to lowpass filter the output. It runs at 300 bpm and feeds into a stereo ping pong delay line. There are a ton of presets available -- sounding anywhere from a garbage can to plastic tubing and a whole lot more...

The second example peels out that APF resonator for use as a filter on a sample. It makes Carla saying "Electron" sound like she's singing inside a can.

One of the characteristics of actual metal objects is the set of boundary conditions on the resonator that prevents certain frequencies from appearing. For example, a clamped metal bar cannot resonate at any frequencies other than those which would have zero amplitude along the bar at the clamp positions. We could simulate this by using a collection of notch filters to impose a sort of negative-formant on the sounds. It is difficult to impose boundary conditions directly on this APF resonator because there is no easy transformation from clamp positions to frequencies at which to notch the energy.

But listening to what happens when you roll the dice and the frequencies and their Q's get randomly assigned, the effect sounds very much like different kinds of objects. So perhaps by altering the APF frequencies and their Q's we get very much the same effect as imposing boundary conditions on the resonator. However, I don't know offhand how to deterministically alter the sound, say with the Mod Wheel to effect a varying clamp position.

So for now we just have to play with this and let serendipity take us where it may...

[ One thing to watch out for... I have noticed that under some conditions the APF filters get overloaded and break into wild ratty oscillation. Be careful of your ears when you roll the dice. Dropping the input amplitude to low values often helps. But even then it appears there are some signal conditions that just plain overload the APF's. The only thing to do then it to kill the Capy, or roll the dice again until the overload ceases. ]

-- DavidMcClain - 19 Dec 2003

Hi David,

I noticed there was a lot of Carla and not so much Can, so I made a set of identical all pass filters without the feedback and cancelled that from the output signal. I then thought why not give those second filters some feedback just not as much as the first, then you will only hear the effect and not the source signal. Also the effect will fade in instead of abrupt starts. Here it is.

-- PeteJohnston - 19 Dec 2003

Heh! That's pretty wicked! But I am confused by your statement that you would try to cancel out the feeback from the first filter chain? While it would be pretty neat to be able to undo some kinds of phase distortion, it is generally impossible to do so. IIR filters always progressively retard phase. Attempting to make phase corrections in the advance direction leads to unstable oscillators.

The math may seem a little complicated but it has to do with infinite responses outside of the unit circle in the complex Z-domain. Inverting an IIR filter is generally impossible to do. This makes sense when you think about the time domain response. An IIR filter rings into the future. In order to remove its effects you'd have to know about the future.

-- DavidMcClain - 19 Dec 2003

No Im not trying to cancel out the feedback. What I'm doing is making two similar curcuits with different decays and subtracting the fast decay from the slow , so as to make a slow attack as well as a slow decay. If the feedbacks and hence the decays were identical, the two circuits would cancel out compleatly.

-- PeteJohnston - 19 Dec 2003

© 2003-2014 by the contributing authors. / You are TWikiGuest