Non linear feedback topologies

#1

I was thinking about distortions and how much i love feedback, so i came out with a somewhat interesting distortion signal path:

20170426_194555.jpg1024x283 76.9 KB

Basically.. The input signal passes through a linear gain and then it's summed with the feedback output. This passes through a high pass filter, in order to cancel dc. The signal goes through a waveshaper (soft clipper, tanh, hard clipper, sineshaper, antialiased or not.. you get the idea) and then outputted.
The output is low pass filtered.. This should in theory (in my mind, at least) correct high frequency instabilities.
This signal is then delayed by a fixed amount, even though variable delay could be implemented too, passes through a gain and is summed to the input signal.

At the moment i don't have my axoloti by hand.. and i won't for the next 3 weeks. But i'd like to hear what the community has to say about this.

I've already made similar distortion paths actually (sat ur hate objects), but there was only one filter (high pass OR low pass), and the delay was 1-sample fixed.
With a low pass filter the output was ultra beefy (too much, actually) and rich in low end. Noise tended to be canceled (not sure why, though).

So, i'm asking: what are your impressions on this? Will the delay line add some interesting character to the distortion? Will it be unstable? (i suppose it will, with negative feedback)

#2

Hi @Sputnki,

It's almost impossible to guess if a system will be stable when it is non linear with delays in the feedback.
Even very simple non linear fedback systems can be chaotic (see the famous Chua oscillator).

So, the right way to go is to experiment, and the Axoloti is a really good platform to do it :wink:

#3

Just a thought, instead of "correcting high frequency instabilities" could it be designed to control high freq instabilities, so it could be dialed in or out etc..
:grin:

#4

I've played with delays in feedback loops of phase modulation (with the modular synth of Caustic 3 on my old tablet... ), the results were very dependent of frequency and rather dirty (in a good way).

From a "mathematical complexity" point of view, all the values in the delay line are state variables.

#5

That's scary.

I was thinking to make a program in octave that tests impulse response over a range of parameters (say feedback amount and delay size).. Could it be used to determine a "map" of stable configurations?

@Gavin of course, the filter is just a classic low pass filter with variable cutoff. You have that degree of freedom indeed

#6

As scary as the Larsen effect :ghost:

A system S is linear when S(x+y) = S(x) + S(y).
That's why we can study linear systems with their impulse responses. (as any signal can be seen as a sum of such impulses).

That's cool, if you know the impulse response, you can derive accurately the response to any incoming signal.

For a slightly non linear system, as you said, it is almost possible to make a map, to find the dangerous zones.

With a Z-N delay, the stability issues will depend on N variables, the stability will depend on the "parameter" zone and of the path (as memorized in the N loop samples).

Anyway, it is worth experimenting with.

#7

Hi @Sputnki,

in fact, i think that feedback with non linearities and varying delay lines is a great experimental field.

What i've done is more or less similar to what you've drawn tonight and played with something like:

feedback.PNG1166x288 23.1 KB

There is filters, all pass filters, saturation, variable delays...
It self oscillates in a sort of "Forbidden planet" / 70s era Tangerine Dream way...

#8

@SmashedTransistors very nice! would you mind sharing the tear/fx/RanDelay?

#9

It's unfinished work - it was intended to be part of a rev/chorus effect - so i won't upload it yet to the library.

I post it here with the patch:

RanDelay.axo (2.0 KB)
A multitap delay with random taps xfading with various rates and random delays.

Ambiance.axp (15.5 KB)

Self oscillating stereo feedback.

#10

I updated and uploaded these in the library:
File -> library -> community -> tiar -> exp -> Ambiance

UPDATE: I added a pair of ring modulators à la Louis & Bebe Barron.