[sdiy] transducers

[Richard Wentk]

At 19:01 09/11/2004 -0800, Metrophage wrote:
>--- Richard Wentk <richard at skydancer.com> wrote:
>
> > The good thing about Karplus Strong is that it's computationally
>cheap.
>
>Yes, that was the idea.
>
> > The bad thing is that no matter what you do to it, it always sounds
>like
> > a DX7 going 'plink.' So it was a good model for 1983,
>
>So it's a good thing that we all hate vintage synths here! lol I LOVE
>DX7 sounds.

I don't hate DX7 sounds. But I don't seem to use them much any more either.

>doing things that I'll never get my 13600 VCOs to do in
>centuries. My TX816 is one of my favorite synths for percussion, and
>many other musical applications.

That's true. But you should check out some of the FM tools in Reaktor if 
you want to go mad with FM.

They are *so* cool. Really.

>The KS algos don't, in my opinion, sound "realistic", but they don't 
>remind me of Yamaha FM. You can do a
>lot of interesting things with controlled feedback, depending on how
>you apply it.

KS always seems to have a very harsh, simple, in your face digital quality 
that reminds me of certain kinds of FM.

> > but the fact that no one has ever tried to sell a commercial KP synth
>says a
> > lot about how successful it is as a modern synthesis technique.
>
>Now that's just funny! I agree that the basic KS concept needs to be
>spiced up a lot to make an engaging instrument, but to say that a
>technique is only modern if someone has succeeded in marketing it is
>absurd.

Er, no, that's not what I said, or what I meant. What I meant, and perhaps 
also what I said, was that KS was never flexible or interesting enough to 
stand on its own as a synthesis technique that people would want to play or 
listen to on an extended basis. People bought DX7s because they liked the 
sound, not because they were interested in Bessel functions. I can't 
imagine a KS synth then or now having the same kind of appeal. The 
technique has  been in Csound since not long after it was invented, but 
even today very few virtual synths use it. Mostly because it sounds just 
plain boring, especially when compared to anything more recent than an OPL 
soundcard.

>Unless you are a marketer, I suppose! How many commercial
>additive synths have really taken off? Formant filters?

FS1R? K5000?

>If its cheap, it gets commercialized - that's why KS was proposed, and why 
>stuff like
>wavetable synths and DX7s became popular.

Bang per buck. FM has a lot. KS doesn't.

>Now that there is cheap,
>general-purpose DSP practically all hardware synths I find are "virtual 
>analog", which is the fuel of other tired rants.

Won't argue there.

>My point is that "success" not unlike "efficiency" and other such
>ideals, is entirely subjective to what your goals are. The goal of
>Karplus and Strong WAS to be computationally cheap, not to be an
>accurate model.

Well, sort of. The goal was to do the best that was possible at the time, 
and from that point of view it's a success of sorts.

But it has too much wrong with it to be more than an academic plaything. 
The Mk1 version is computationally cheap, but it also has very poor 
frequency resolution. You can improve on this by sampling and interpolating 
through the buffer, or by hiking up the sample rate. But then you've at 
least doubled the implementation cost, and quite possibly more than doubled 
it if you try to do anything cleverer than linear interpolation.

So it sort of works. And then you find it sort of doesn't if you try to do 
anything too obviously musical or expressive with it.

> > Some of the newer phys mod tools sound much more interesting. Check
> > out Tassman for some examples.
>
>Which is why I suggested checking out modal synthesis, which is the
>type of synthesis which Tassman is implemented with!

There you go then. :-)

All of which is a distraction from my basic point, which is that real 
physical modelling is extremely hard to do well. I only know of one model 
that comes really close to the real sound, and that's a grand piano model 
that needs to run on its own cluster.

Even modelling a reveb tank is a very complicated thing. You're going to 
get frequency dispersion, acoustic coupling between the springs, acoustic 
coupling between the springs and any surrounding noise (e.g. a monitor 
mix), nonlinearities in the spring extension, differences from temperature 
and air pressure variations, and - for all I know - moon phase too.

There's a place for making that kind of DSP effort, but if you just want to 
something that's going to go sproing in some vaguely sonic and musical way, 
it's a lot easier to wire up a spring in a drainpipe and leave it to get on 
with what it does best.  :-)

Richard