Hardware Physical Modelling


The KarplusRings engine for norns is quite nice. You can get a wide range of tone, from taps and plucks, to dulcimer and piano.


Isn’t the level of the incoming signal, or its variation, enough to make the tanh a more reactionary module? I would assume that the idea is to place it in the feedback path in order to set a wall/limit to how far your feedback goes. I’m unsure how it works, but I have to say I’m intrigued by it now. Reached out to Instruo to see if I can get one for myself - it doesn’t seem to be available through their distributors for some reason.


You are right, of course. It’s a limiter, not a compressor. My mistake. I assumed it responded in kind to whatever the level of the signal was. Still intrigued - I mean, if you can control the feedback, it’s rather nice to have.


yes, it reacts on different incoming levels…the louder the more waveshaping/limiting. i’m not sure what the voltage is where the waveshaping/limiting actually begins…
it acts as vca/loudness control in my feedback patches, a set and forget type module. but cv control over it would still be nice i think, so your feedback reached its limit, it stays more or less the same because of the tanh but now i would like to either attenuate or boost the signal for variation, i can do that by knob control but cv is always nice to have:)

in this muffs thread there are some examples and oscilloscope pictures:


Haha, read my mind (with reference to my last post on software thread)

any particular patching recipes to emphasise the ‘chaotic’ characteristics of feedback behaviour? I am aiming to get in this territory ‘in the box’ with my Faust programming!


those that are doing this with Eurorack,
what modules are you using? what ‘features’ do you look for? (e.g. open feedback loop on delay)

I love physical modelling, in both software and hardware and have been exploring this quite a bit, however, this topic has opened my eyes to perhaps adding some ‘basic components’, which I can then assemble in different ways, together and with what I already have.

Currently I have Rings, and Elements (love them both)
I also have a Bela Salt, running Supercollider , which has lots of physical modelling stuff, and as its located in the rack, I can mix and modulate that directly - fun hybrid experiments, still to explore there!

from the above, the ‘toolbox’ required seems to be modules covering the following:
Exciter - any, envelope noise source? (is colour interesting?)
Delay - for karplus, but with an open feedback , so it can be filtered/limited (suggestions?)
Filter on delay line - any thoughts on properties of this?
Body - hmm, what are you using as a resonator?

Ive got vcas/lfos/envelopes (though zadar is very interesting), but not much else…

Exciter: no dedicated noise, I often use Plaits as an exciter though, but a separate noise source might be useful.

Delay/Filter : I’ve no filters yet !! , Ive been a bit undecided what to add.
perhaps something flexible, then some smaller 2HP for variety? Delay, no idea on whats useful for this.

Resonator: again , Im not quite sure… I though a resonator tended to be delay lines + filtering? though perhaps you need more delay lines so you can diffuse sounds more.
(I guess a second set as the above delay/filter are being used for the karplus implementation)

Id love some suggestions on modules to look at to explore this area further…


Intellijel’s Plonk module is a collab with AAS.


I prefer tanh[3] more for shaping than feedback control specifically. Push it too much and you still get squarewaves, as opposed to just backing off. How well that works depends a lot on the content and how clean you want to keep it.


The goal is to achieve an infinite sustain while minimizing distortion. If the feedback gain is too small, the sound will decay. If the gain is too large, the sound will blow up, but will hit the saturation characteristic of the circuit and become distorted. The desired result is in between these conditions, but impossible to achieve exactly. It’s like balancing a pendulum on end.

Enter the compressor. It functions as an automatic control mechanism that keeps the gain balanced. It monitors the audio level somewhere in the loop and adjusts the gain accordingly. While it shouldn’t matter where the level is monitored, the smoothest sounding results seem to be achieved where the monitoring point is at the output of the compressor. Hence feedback compression. This is probably because the mixer (at the input of the delay) contributes the most in terms of saturation, and so directly controlling mixer input levels makes the most sense.

Also, while the primary push-pull mechanism is the amplitude of the loop signal, there’s a secondary push-pull with bandwidth. The filter (either lowpass or bandpass) tends to restrict the bandwidth of the loop signal. The saturation/distortion effect tends to expand it. You don’t want either a pure sine wave or a distorted mess, so the control mechanism tends to bring out the more interesting tones somewhere in between.


How about tanh, a smidge of gain over 1 and an expander? (In feedback loop)


well I think all these would drive the results towards the square-wave distortion. the more interesting chaotic effects are usually on this boundary between order and chaos and the compressor actually helps bring this out (because it balances the push-pull of the lowpass/bandpass filter and the implicit tanh of the mixer circuit).

I should add that if your entire signal path is digital, you’ll need an explicit tanh. I’m assuming analog circuitry for all except the delay, where a tanh-characteristic (soft saturation) is already built in to every module.


you’re absolutely right, a tanh isn’t the same as a compressor. i use it as a safety, the main control over these interesting chaotic spots come from playing, in my case usually pressure control into vca, this way i have a really fine control over the feedback behavior. the thing in my case is that it all is very frequency dependant, especially when i patch bandpass filters in the feedback path, some frequencies “behave” modest, others screech like there’s no tomorrow, that’s where the tanh comes in very handy and has a pleasant sound when driven hard (not like the d0 which takes the full -12 v to 12 v and can damage ears instantly).


Even though I’m in the same boat as @kilchhofer, as I control the feedback using a breath controller, I decided to get a Tanh[3] anyway. Just ordered one as I got intrigued by the description. So far I’ve been using the attenuators available at the inputs of the modules I use for feedback, to set the limits of where the feedback goes.

One nice workaround for modules that don’t contain feedback paths (feedback input/output) is to force them via patching through a mixer. We’ve tackled this before on this thread. Thankfully the Instruo module has three tracks, which allows for a lot of creative patching through the setup.

Unfortunately it seems to have replaced my plans of buying an Elby Designs Velocitizer. I’ve been curious about that as well, but between the Tanh[3], getting that Delayed LFO from Ladik I mentioned above and expanding the case I use for these things to 104hp, I think I have a lot of new ground to cover. Here’s hoping the Instruo module will increase the sweet spot of where the Resonant EQ breaks up like a woodwind.


You should be able to do quite a bit with what you have already.

Rings and Elements are both good resonators, which I’d say is the most difficult part to patch up manually. If you like the sound of them, there wouldn’t be much of a need to get an additional delay or filters.

If you do get a standalone delay, you’ll want one that does very short delay times–the shorter it can go, the higher pitched sounds you’ll be able to create. Mungo D0, CG Products Delay 1022, Dave Smiths’ feedback module, the shorter Doepfer BBD’s. As far as filters go, anything will probably work, but you’d have to try a few options to see what you like. I really like the 6dB channel of LxD as a filter on my resonators (but, again, Rings and Elements eliminate the need for an additional filter since they do all that internally).

Any sound will work as an exciter, though some better than others. Maybe try to make especially noisy sounds with Plaits and send them through a very tight envelope. Samples work really well, I have a bunch of samples of me hitting random objects together that I use as exciters. For a lot of patches you’ll also want to mix in a DC component, either a trigger pulse or an offset through a VCA with a very short envelope.

For a body filter you could use either a few bandpass filters, an EQ, or a convolution reverb. Three Sisters makes a great body filter. If I don’t use that, I use convolution in my DAW.

This site helped me a lot with visualizing patches and outlining the pieces you’ll need: https://www.frodebeats.com/physical-modelling


thanks @smbols - very useful info
yeah your right, I do have a lot to already play with… and given much of this is often done in the digital domain - I’ve got a lot to experiment with supercollider and Salt.

exciters, yeah, Ive been using noise blips mainly, but not tried adding a DC component nor samples, something to give a go.

so perhaps more experimenting to be done before more modules :slight_smile:

but very useful info, will help me, whilst im considering modules to ‘round out’ my rack.


Any er 301 users care to comment if it’s possible to build these types of connections using it alone?


Honestly I’m not sure how well suited the ER-301 is to physical modeling, yet. I haven’t experimented with it much myself.

I know crude Karplus-Strong has been done but people don’t seem to have had any luck getting it to track 1V/OCT, so far.

Internal feedback patching was recently introduced but has a 128 sample latency. There’s a waveguide delay on the development roadmap, not sure when that’s going to arrive though.

Bandpass filters were recent-ish third-party additions, but I don’t think building a modal resonator out of several bandpass filters is going to be practical. It might be possible in the DSP layer, but that’s not exposed to end users yet.


I’m doing some investigation on this to try and wrap my head around it and found these extremely useful charts and descriptions http://www.ron-berry.co.uk/page6.html


thanks, pretty cool site!

now someone should try and replicate this saxophone patch:


Industrial engineering calls for higher accuracy models of physical processes than music, and there is extensive research in this area. The mathematical term is partial differential equations, and models are available for fluid flow, excitation of 2D and 3D strings and membranes (the wave equation), heat diffusion, etc. etc.

From self funded research for my company Aether Machine, I have found that it is possible to create efficient, high accuracy analogs for these equations, using an extended form of the state variable equation. I am hoping to make a 3D model for a circular membrane, which can also solve fluid flow in 3 dimensions at the flip of a swtich available in the 6U synthesizer format within this year:

Interestingly I find the model to behave very differently to waveguide/Karplus Strong, not necessarily for the better. But the model is watertight, despite having a problem similar to my last design that no living person is available to consult for advice! I will try to post some examples over the next year or so once I have found what it is good at doing.

Please do not ask me for a schematic or to make a euro version, or try to explain why either of those things would benefit me. Someone always does.


…having said all this, the tools that are in the ER-301 are still ripe for experimentation. I just made a really cool drone with a trio of Schroeder allpass filters in a feedback loop, fed by Rings, fed by Plaits, where I tweaked the allpass filters with the 16n Faderbank :smiley: I personally prefer weird experimental resonant drone things to “modeling” as such.