Re: [motm] Re: JH thinks analogue modelling
2002-11-07 by jhaible@debitel.net
> First read this: > > http://www.sospubs.co.uk/sos/1997_articles/jul97/ronberry.html > > He has supposedly built some of his own devices Just checked this; I have seen it before, But I didn't remember the name. This goes into the right direction, but it has one drawback: The feedback loop is only around the delay line. For more natural emulations (or for more interesting new creations) you want all kinds of filters and nonlinear operations inside the loop. The problem with such extra devices inside the loop is that you cannot play "in tune" anymore. It still works with perfectly tracking filters inside the delay loop, but this doesn't help much. On a real instrument, it's a difference whether you just change the length of a tube, or if you have keyholes openend etc. So you want a mix of tracking and fixed filters inside the loop at least. You cannot move the formants (fixed stuff) out of the loop either, because they interact with nonlinear stuff inside the loop, such as air flow rate saturation. The direction in which I'm presently thinking is like this: To make a set of modules for the delay and for the nonlinearities, which could then be patched together with ordinary synth modules (like a LPF for the stiffness of a reed etc.). If you patch them together your tuning will be a mess. But then there must be an "overhead module" which performs an autotuning routine (think of a synth with very bad VCOs ...) and creates a large table to correct the pitch CV. You would go (1) patching whatever you find reasonable and what will oscillate, and which produces interesting higher modes when you change a "pressure" CV or something like that. (2) You set your "pressure" and other performance parameters to their nominal value and press "autotune" on th overhead module. (3) Afterwards, you can play this model you have created perfectly in tune (when your performance controls are at their nominal values), and you can go into the "wild other modes" just as your self-built model allows. What I like about this idea: You don't have to care about tuning, so you have *every* degree of freedom to patch in any existing module you like - everything that closes the loop and gives controlled oscillation. And all this with only 2 or three new modules (delay line + autotune unit could be in one module; and a good reed simulation would be a second new module.) Only thinking aloud at the moment. Spent a good part of last night with Yamaha patents and VL7 manuals - I finally begin to understand that stuff. JH.