MOTM

Jay writes:
>>Yes, but the question is does it have the playing dynamics a real string
instrument has? Would be interesting to hear a small "ensemble" so to
speak...<<

sucrosemusic writes:
>>Yes, I have to say, that it's very easy to catch all the harmonics of 
a sound, I can do that by ::gasp:: sampling the sound, or using my PC 
to figure out the harmonics from a recording and plug them into my 
additive synthesizer (Kawaii K5000s, digital, circa 1997)...
The real trick is in getting the expressive nature of the sound.  It 
can be done, of course, with an additive synth by mapping your 
velocity and aftertouch and whatnot to your various harmonics, and on 
a modular synth like you're describing by mapping that to your 
various section's VCAs, filters, etc.  Also, throw some vibrato in 
there for good measure...  and remember, when playing a violin, 
there's definitly more to it than one short sound clip.<<

The point of my MP3 was only to show a near dead on timbre match of a real 
violin's complex and moving harmonic structure, nothing more. I almost was 
about to splice it into the middle of the real violin mp3 to see if anybody 
could notice the change. Expressive playing dynamics will not be a problem 
later on, in fact I have a violin patch on my MOTM right now where I can vary 
the depth and speed of vibrato/tremelo with a breath controller, vary the 
loudness and brightness with one pedal, and vary the portamento with another 
pedal. It sounds very beautiful and expressive, but I'm still working on a 
really complex filter bank to process the sound in a more convincing way. This 
Nord Mod sound was done only as a proof of concept for a possible software 
synth I've been thinking of writing. I can't even find a decent synthesized 
violin from the latest research from any university. They're using waveguides 
and all the latest stuff but it sounds lousy. Haven't heard any decent 
violin/cello coming from any physical modeling synth either. That's why I'm 
thinking it's up to me to tackle the problem once and for all.

The fact is that you CAN'T just capture the harmonics from a real sound and 
plug them into a K5000 and expect to get a decent violin. That's probably why 
nobody has done it. And I've heard endless complaints about people trying to do 
just that same thing with other less complex instruments but only failed. Each 
harmonic has slow random changes over time plus fast noise-like jitter and is 
way more complex than the K5000 will handle with its envelopes. Not to mention 
that as soon as you play another note, the harmonic structure needs to 
drastically change. That's why the K5000 has a formant filter, but then you 
don't want the captured harmonics if you're going to use that, you need 
something like a sawtooth wave as input. But that formant filter isn't as 
effective as it should be (I've tried entering in the plot of a violin body 
with nothing convincing coming from it), and it's applied at the wrong time, 
before vibrato and pitch slides and therefore doesn't keep the formants fixed 
in those situations. And to add more complexity, it appears that those body 
resonances don't stay the same for every note either. It's a mess, but if I can 
get a string synth engine written, it will handle all of this with no problem.

Feel free to view the animated harmonic spectrum of the MP3 in something like 
Cool Edit Pro. You'll see that the real and fake timbre are virtually 
indistinguishable in every way. (except that my 2nd harmonic is slightly lower)

-Elhardt