Hi there,

I'd like to explore the sympathetic resonance concept.
The main idea would be to add to an incoming sound the sounds resulting of the sympathetic acoustical resonance of virtual objects implementend in a Pd patch.

Of course this would deal with physical modelling or "rule of thumb modelling" of such objects (guitar strings, sitar strings, rods, glasses ?,....), and fft in order to implement the resonance scheme.

For the sympathetic objects modelling I'll do a search on the forum, and will certainly find many interesting things (any idea welcome though).

Ideas about the resonance rule (how to convert "chunks" given by fft to excitation passed to my resonant objects ?, or other ways) would be welcome.

Thanx,

Nau

I've just recently been getting into physical modeling myself, so I only have so much to offer here, but your post is also raising some question for me. Mainly, what is rule of thumb modeling? And what makes you think you need an FFT for this? I mean, is there some FFT synthesis method your thinking of for modeling resonant systems?

If you go the physical modeling route, though, sympathetic vibrations or resonance is actually pretty simple as it is inherently part of the system. For example, to create the sympathetic vibrations of a string (as in a sitar or open strings of a guitar) you would simply feed the output of the plucked strings into a model of the open strings, possibly attenuated and filtered first. The model will only resonate with frequencies that are harmonic to it, as it should; other frequencies will quickly die out. If I remember right, the ideal place to tap the plucked string would be at the "bridge" of the model as that is where the coupling would be in a physical instrument, though I'm not certain it is a hard and fast rule (you can definitely get strings to resonate from energy propagating through air).

Sounds like Karplus-Strong algorithm would be a good choice. A quick search in the web led me to this:
http://lists.puredata.info/pipermail/pd-list/2006-04/037251.html
http://crca.ucsd.edu/~msp/techniques/v0.11/book-html/node120.html

http://lab.andre-michelle.com/karplus-strong-guitar
I don't know what has this been made with, but sounds pretty good.

There are plenty of sources of information. Good luck!

Sumidero

And the filter in that first link could be simplified as:

[rzero~ 1]
|
[*~ .5]

because [biquad~] is a two-zero two-pole filter and will calculate them all, but all you need is the one zero in the Karplus-Strong algorithm.

(In fact, the [biquad~] coefficients in that example are incorrect, as it actually implements y[n] = .5(x[n] + x[n-2]) instead of y[n] = .5(x[n] + x[n-1]).)

Dunno if it helps, but there's a rough KS sympathetic resonance model of a sitar inside "pd harmonium?" - http://radio.rumblesan.com/patches/Tablenge.tar.gz - not hugely convinced by the results, but perhaps it will serve as a starting point.

Hi there,

thank you for those quick answers.

@Maelstorm said:

I've just recently been getting into physical modeling myself, so I only have so much to offer here, but your post is also raising some question for me. Mainly, what is rule of thumb modeling? And what makes you think you need an FFT for this? I mean, is there some FFT synthesis method your thinking of for modeling resonant systems?

Excuse my approximate English. Writing "rule of thumb modelling", I was thinking about some kind of empirical recipe leading to interesting audio results but not using a bunch of (simplified and discretized) physical laws. An example would be the proposal of obiwannabe in this topic : http://puredata.hurleur.com/sujet-3299-creating-metallic-klangs-bullet-hits-doppler-effect .

When talking about fft, I was just thinking aloud.
Of course direct routing of (filtered and lowered, possibly integrated or derivated) incoming audio signal to a designated 'excitation point' of the model should work in (I hope) many cases, and would be a 'time-domain mechanical' approach. That way one can imagine make his virtual string playing a radio program sending the radio's audio to "the bridge", reproducing what some instrument makers do to "wake up" a new or "asleep" instrument.

But I was also wondering if fft wouldn't be useful for defining the incoming signal in the frequency domain and handle virtual resonant objects as a bunch of transfer functions... but "just thinking aloud freewheel", uh !

@Maelstorm said:

(you can definitely get strings to resonate from energy propagating through air).

That would be one of my goals.

Now it's definitively time to follow the links you brought to me, fellows, as this answer was written before reading those.

Thank you.

Nau

@lead said:

Dunno if it helps, but there's a rough KS sympathetic resonance model of a sitar inside "pd harmonium?" - http://radio.rumblesan.com/patches/Tablenge.tar.gz - not hugely convinced by the results, but perhaps it will serve as a starting point.

Let me thank you for highlighting this 'generative music webradio', wich discloses every Pd patch it plays ! This is kinda goldmine to me (as this very forum, of course).

Nau

Yop,

just read about PeRColate lib, dealing with physical modelling.

I can't find something on the web that directly suits my distro (TangoStudio based on Ubuntu Lucid). All I got is a .rpm package.

Did someone find those externals in 'simple archive' or .deb package ? Thank you.

As I think it could be possible to 'convert' rpm packages to deb packages, I think I could try, but would somebody here advise me wether this could be a 'dangerous' behaviour (for my beloved distro's stability)? As these are only text files I'd say it's not gonna make too much mess, but I prefer asking.

Thanks,

Nau