Hello all! This is my first time posting on this forum, so sorry if I did anything incorrectly.

I am trying to create a patch using modal synthesis, but I've noticed some odd behavior specifically with the sends and the bandpass filter I'm using. If I open up the clone and save it, the first time I send a bang I get a different sound from every other time I hit it. And oddly enough, I think that first time I hit it sounds better than all other times. I figure there's some under the hood processing that I'm not quite understanding. Could anyone explain to me how I can fix this issue?

Here's a quick explanation of the patch: Modal synthesis involves finding the most prominent harmonics, AKA modes, of an impulse and modeling how each one changes over time. I decided to model seven modes, so I set up three arrays with seven points. The first array sets which harmonic a given mode is on, between the first and 64th, the second sets the initial volume of each mode, and the third sets the amount of time it takes for each respective mode to decay in msec (following an attack/decay filter). The sound for each of the modes is currently a mix between an [osc~] and bandpassed noise with a very large Q.

Let me know what y'all think or if there's any other information needed.

Evelyn

modal synthesis_6.pd

mode.pd

not sure if this is your problem, but when you save a patcher which is used as object or in an object alrerady, those instances are loaded new and therefore inside everything defaults back to initial state.

btw. i am not too fond of the data rate triggered envelopes,

unfortunately there is no reson~ filter object for pd, but maybe there is something similar. if you use filters for the sinusoids, they will come with decay "built-in" and you can get rid of that envelope rubbish.

@Obineg Thank you for your reply! That's what I figured. I've been trying to determine what's going back to an initial state, and my only possible answer is the bandpass filter. Just did some troubleshooting and found that if I put a delay after the bang to set the bandpass filter back to its initial value of 440 hz, it gives me that more desirable sound consistently. So it's something about the bandpass filter moving from 440 hz to wherever the intended value pitch value is.

I'm not sure what you mean about data rate triggered envelopes. And how would I use a filter for the sinusoid? I've heard someone say that before, but I wasn't sure what it meant.

@Pathagas said:

I'm not sure what you mean about data rate triggered envelopes. And how would I use a filter for the sinusoid? I've heard someone say that before, but I wasn't sure what it meant.

Every filter can be described or understood in terms of its impulse response -- put a single sample impulse into it (the most neutral possible signal), and the result is 100% because of the filter's coloration.

The impulse response of a bandpass filter is a decaying sinewave -- the same as mechanical resonance. (I saw a video somewhere describing the movement of a weight on a spring using the classical physics formulas, and then reducing those down to a two-pole filter formula -- demonstrating that bandpass is literally a physical model of resonance.)

The catch is that the beginning of the impulse response is an instantaneous jump, producing a click.

I learned a neat trick from SuperCollider. It has a UGen called Formlet, which takes an input signal, frequency, and attack and decay times as input. It's implemented by subtracting two bandpass filters: in SC code, Ringz.ar(input, freq, decay) - Ringz.ar(input, freq, attack). These are constant skirt gain filters, meaning that energy at the center frequency is amplified, and the stopbands stay at the same volume -- meaning that, once subtracted, they cancel out, giving you 1/ a smoother sound and 2/ a natural envelope.

At some point, I found in Pd's ELSE library what appears to be an exact duplicate of Ringz: [resonant~]. It's easy to implement Formlet with it:

(Here, I'm using a standard exciter for modal synthesis -- a short burst of noise. Pink noise might sound better than white.)

ELSE is one of those external libraries I wish I'd known about at the beginning... a proper complement of standard DSP.

hjh

@ddw_music said:

(I saw a video somewhere describing the movement of a weight on a spring using the classical physics formulas, and then reducing those down to a two-pole filter formula -- demonstrating that bandpass is literally a physical model of resonance.)

The catch is that the beginning of the impulse response is an instantaneous jump, producing a click.

Wow! I can't explain how helpful this is! I've been trying to understand and recreate this concept for a while now. I just implemented this patch. I'm going to do some more research, and I'm especially interested in how mechanical resonance relates to a bandpass filter,

Thank you so much for your reply!

Evelyn