Cybernetic Music/Roland Kayn/Feedback Systems/AI

Thank you for these thoughts, and the videos of course.

You (and others on this thread) might already know it, but there was recently a seminar serie called Cybernetics for the 21st Century which reconsiders, reimagines and updates the tradition. It’s hosted by Media Lab of Guangdong Times Museum, and Research Network for Philosophy and Technology. All seminar recordings are you YouTube.

I have not yet heard all of them, but so far particularly enjoyed Mathieu Triclot’s return to the original sources of the first order.


Sorry, I forgot to mention the simplest way is some gain in the feedback loop. I’ve done this on my KARP 2600 via the preamp, and you get a nice chaotic source w/o using up other things.


Interesting. With Serge, direct feedback with gain through CV mixer still doesn’t result in chaos, just more extreme frequency and timbre shifts. I tried direct gain of 2 or 3, then potentially larger gains by feeding back the output of the CV mixer into one of its inputs. Tried all waveforms (sine/triangle/square) as well.

I did get chaos using VCA and/or Audio mixer for gain, at very high frequencies of the VCO. But these devices contain a DC blocking (Highpass) filter.

Possibly it is the DC blocking filter, or distortions in the 2600 preamp that do the trick (?)

Self-controlled VCO would be interesting to analyze mathematically, but unfortunately that’s beyond my knowledge or abilities.

I tried it on the Hordijk first and it worked there in the same way. To be clear VCO output to gain back into freq modulation with a good depth. Then raise the gain until you get a chaotic signal from a different output (mult) on the same oscillator. The base frequency of the oscillator should also not be too low if I recall.

I’d be surprised if exactly the same cannot accomplished on a Serge. I do have one handy and will try to verify if I can find time.

UPDATE: didn’t get a chance to try it on the Serge. But I tried with my Roland System 500 oscillators and I used the gain on the Make Noise case. Same result. Though I remembered incorrectly: with a sufficient amount of gain it’s not necessary to set the frequency very high, nor the depth to get an interesting chaotic source.


hi, curious where you’re at with this project?

I posted about where the project is at in the thread linked above. Short story is: its done, but no one can make another one without a teensy 3.6, which are not expected til later this year (maybe). I have parts for one more but I’d need to desolder a teensy 3.6 out of something to get it built… :frowning:

Probably more propper to keep further discussion over on that other thread tho. Happy someone asked about it. Sorry there’s not better news.


didn’t notice that…will jump over there, thanks!


Here’s a version of the Vink patch. It’s basically free running and being sampled and looped by a 1010 blackbox. I am using a Monome teletype module to basically recreate the HP tape machine that HAINBACH uses to construct pendulum loops. So I program the speed and direction of recording and the ‘tape machine’ is programmed to record at those speeds and switch directions. See this youtube video for an explanation of the looping system:


Jaap Vink has passed away.

Jaap Vink always tried to break out of the periodicity of the sounds so abundantly available in the electronic studio. Although his music was entirely produced with purely electronic sound material, its textures resemble the richness of orchestral sounds, or large natural sound-complexes, as a result of recursive processes. The density of their sound material increases and decreases by careful control of feedback networks with configurations of analogue tape recorders (delay lines), filters and modulators.


Has somebody here tried a frequency shifter in the feedback path of e.g. a Vink patch? I’m guessing it would work quite nicely but I’m a bit hesitant to buy a frequency shifter because they’re so expensive. But my guess would be they would sound glorious :slight_smile:

1 Like

Any examples of frequency shifters in 3U?

Horrible sequence through the two available frequency shifters, Xaoc is digital afaik, Doepfer is analog.

From a more detailed A-126-2 video

1 Like

technically the ring modulator in the vink patch is a frequency shifter - just providing both up and downshifted signals simultaneously

1 Like

Yes I know, but a frequency shifter that can separate the down- and upshift might provide very interesting results that are slightly ‘smoother’ than the ringmod. At least to my ears the resulting sounds are less metallic than standard ring modulation.

@ThanosF, you could look at the Doepfer A-126-2 (analog) and Xaoc Koszalin (digital). More rare ones are Modcan, Cwejman etc. It’s a very complicated module and thus quite expensive.

One thing I’m not clear on: for the feedback-taming usage of the freq shift, do you retain both side bands in the feedback path? Or only make use of the upper or lower sideband?

At sub-audio rates the behavior can be very different – Ring mod will just produce amplitude modulation, while the frequency shifter does a kind of “barber pole” phase shift, when mixed with the original signal it can sound like a phaser that is always ascending or descending, when unmixed (100% wet) it will sound similar to the dry signal.

If the frequency shift is on the order of the delay period, I’d think much larger feedback gains could be tolerated before instability, because there’s no frequency at which the echoes are consistently in phase [The phase is always moving at all frequencies]

Anyway it would be interesting, so would putting a phaser in the feedback path


I was finally able to try putting the loop routing through a frequency shifter set to a very sub-audio rate, then taking just the lower side out and back into the loop.

It did make for a much tamer system that persisted at a good level indefinitely. Still requires an amount of gain tweaking at various stages, but it made the whole thing a lot more pleasant to explore with different inputs and settings.

Still figuring out what results in ultimately homogenized frequencies / tones vs. what ends up in pleasant instabilities that continue to change, some.

1 Like

Re: upper lower band mix. There are several approaches and none are wrong per se but do have different characteristics. The post by @grayghost explains why.

I like the bands to be on a crossfade with offset control and then have an envelope follower with amount control also sent to the fade. It provides some dynamics that track an input (for example audio recordings). The other thing to check is whether your frequencies shifter cancels the reflection frequencies. That also can produce different outcomes. Less noticeable at lower shifted frequencies but definitely noticeable over time (for downward frequencies it is largely a function of the lowest input frequency, the processing within the feedback loop and feedback level, and shifted amount). The greater the shifted the amount the faster the onset.

Another live processing idea is to use an input to dynamically gate the triggers for the noise / sample & hold / fluctuating voltages that adds some instability while following along. If you also add the feedback to the gate, the threshold can help fine tune latching points.

Another potential is a frequency tracker on the input (though you’ll likely want to divide the frequency down quite a lot) that feeds to the frequency shifter.

1 Like

This is a good read,