I need some Pro help! Is this possible with Synthmaker? !

oddson · 11 Feb 2010

From what I understand drumheads are much harder to model than strings using digital waveguides because they require a 2D mesh with all the displacements allowed to interact (and that's just the drumhead!).

Are you familiar with JOS?

Specifically https://ccrma.stanford.edu/~jos/wg.html And most especially: https://ccrma.stanford.edu/~jos/pdf/mesh.pdf Much of this last one might as well be in Greek for me... 😕

The summary section of that last document leads me to suspect the answer to "can Synthmaker handle this" is 'not really' as the amount of processing appears to be feasible only with a significant amount of parallel processing and I don't think four SSE channels will do it.

There appears to be another technique involving transfer fuction methods but I couldn't locate anything with sufficient detail to show it was feasible in SynthMaker or not but I'm pretty skeptical.

Apart from not supporting parallel processing as fully as might be required SynthMaker is lacking some array math components that are very likely to be needed with any technique that tries to physically model 2D services (or, worse yet, 3D resonating bodies).

But there's emulation rather than modelling. It may be feasible to produce an emulation of the sound by combining synthesis techniques to replicate the result of a tympani strike... (i.e. sample playback snippets for mallet impact noise, mixing wavetables across the note-envelope to mimic the harmonic dynamics... )

But anyway you approach it, it would not be an easy project and would depend heavily on your technical knowledge and level of commitment.

trogluddite · 11 Feb 2010

Check out the Synth Secrets articles at the Sound on Sound magazine website - there were a nice few articles describing the kind of waveforms you need to generate. I agree with oddson about the physical modelling though - drum heads are two dimensional (three dimensional if you include the air trapped inside) and so far harder to model than a one dimensional oscillator like a string or organ pipe - the articlles describe the kinds of oscillation modes involved and they are very complex. You can get very close using multiple oscillators and frequency modulation to generate the 'out of tune' enharmonic partials that you will need.

Tzarls · 11 Feb 2010

FWIW, you´ll get more "realistic" timpanis by using samples (velocity layered, keyswitched and the other tricks of the trade) than by using any kind of emulation/modeling. The problem comes when you actually want to sample the instrument: do you have a set of timpanis + timpani players at your disposal?

My opinion at least.

MegaHurtz · 11 Feb 2010

Psysical moddeling is very very good. The more woodish sound of a timpany can be accheived by using the resonance of filters (on a little impulse of noise) instead of a sine. I have to admit it works better on bells becouse of the big dent it makes when you smash a timpany. Maybe a little slow fm will help you out there.

NekroDisko · 27 Feb 2010

As far as I have learned, the trick to get any synth sound to be "realistic", an attempt in extensive additive synthesis is required. This technique may result in a mayor cpu-intensive effort (as it requires many oscillators at a time) but may be worth it if you only want to make one sound realistic. I suggest you make an additive synth with many alternatives to modulate the pitch (using both Envelopes and LFO:s). Noise-oscillators may be useful in case you want to use the synth for other rythm-sounds as well...

Disco_Steve · 27 Feb 2010

I think there is a timpani in csound which is done with physical modelling. so that is a great place to look as all that code is open source. and if you cair more about the sound than the making process. then just use the csound to get the sound. you can compile a csound instrument as a vst so you can use it as a synth. within your host.

happy synthing

Steve