Casio CZ/ VZ/ FZ - Pro Series

On Mon, Sep 1, 2008 at 10:37 PM, Summa <flotorian@...> wrote:
> Hi Ezra,
>
> sorry for the late reply, was pretty busy last week...
>
> On 25 Aug 2008 at 17:33, ezra.buchla@... wrote:
>
>> well, maybe its worth noting that when i talk about "physical
>> modelling" synthesis i'm not necessarily talking about trying to
>> duplicate existing sounds or waveforms or whatever.
>
> I played around with the tassman that includes some physical
> modelling elements. I wasn't exactly satisfied with the results and
> the predictability of the results, one might getter results if he's
> into building real world instruments and having profound knowledge
> about their resonance bodies...

i don't know what that is but it seems smart for people making
physical modelling synths to have some experience with real objects. i
guess.

>> take the karplus-strong algortihm. this is a numerical model of a
>> system of masses and springs that can be set into oscillation. it is
>> quite computationally efficient. it sounds a little like a guitar
>> string or other kinds of strings, but based on the parameters and the
>> excitation methods, they can soundlike weird new stuff (you can't, for
>> example, sing into a string in the real world, or wrap the string in a
>> circle...)
>
> I always thought it's a tunable comb filter (that basicly is a delay)
> with some feedback that needs a short impulse burst to oscillate.

you're right, the term "Karplus-Strong" usually describes that very CS
implementation of the string model from physics (which itself has been
around for hundreds of years). the relevant sonic characteristic is
the way the various harmonics decay... with a direct  implementation
of the connected-mass system, you can change the spacing of the
harmonics with a stiffness coefficient, and their decay times with a
damping coefficient... this corresponds to delay time and feedback in
the comb filter implementation. (i did, after all, carefully read the
Karplus-Strong patent application that i posted earlier...)

but in the direct implementation, you can also change the order or
topology of the force equation, which produces enharmonic as well as
harmonic components. you can't get this from a comb filter (but you
could get them, i suppose, from some kind of really really weird
complicated APF phaser implementation... hm)

> at least get some marvelous plucked string sounds using formants in
> conjunction with comb filters....

sure... BTW, already when you say "formant" in this context you are
invoking a physical modeling concept (the sets of frequencies produced
by the vocal tract when forming different vowels...)

> Here some Comb-Filter examples I did with the microQ some years ago.

> http://www.summasounds.de/files/mQcomb.mp3

very nice... sounds like there's an awful lot of stuff on there...
delays... multiband compression...  already by introducing formant
filtering (by any numerical or analog means you care to choose), this
becomes a whole new ballgame (compared to the humble comb filter), and
indeed a very chaotic one if you're filtering or waveshaping the
feedback path (as it sounds like you are...)... you're going to end up
hearing a lot of artifacts resulting from "limitations" or "flaws" in
the actual hardware or software... totally "chaotic" stuff.

>> where this gets interesting to me is with the addition of nonlinear
>> terms and other sources of chaos to the equations. now you have a
>> situation where it is very difficult to analytically predict what
>> orbits (read: waveforms) will be produced, and the best way to observe
>> the system is to compute it. this exact problem (or very near) was the
>> subject of the historical work by fermi, pasta, and ulam at los alamos
>> in the 1950's, which formed the basis for modern chaos theory and
>> numerical analysis...
>
> Even so chaos theory is pretty fascinating, especially when it come
> to weather or big bang prediction, when it comes to sounds I'm rather
> interested in results than in theory. It simply doesn't need that
> much chaos to trick the ear/brain combination, it can trace a very
> limited amount of parameters at once only.
> I must admit that I'm more interested in practical, emotional
> playable, in multiple aspect changeable sounds, than in unpredictable
> coincidental generated stuff.

i guess i wouldn't argue with that. i spend part of my time building
modular synthesizers that employ a relatively "traditional" sound
design paradigm, but even in the guts of an analog waveshaper there is
much beautiful folded strangeness to be found. hard to understand
anything in circuit design without some nonlinear analysis tools.

i happen to enjoy designing synthesis algorithms in my "spare time,"
and i happen to have a numerical modelling background, so it makes
sense to use that paradigm when it's useful. most things having to do
with sound are "chaotic," including (for example) the spectra
resulting from the cz-series phase-reset synthesis algorithm that i
was talking about earlier.

(interested in the chaotic dynamics of coupled oscillators? no? how
bout cross-modulated FM oscillators? how bout that
formant-combfilter-feedback patch? these can all be studied with
"chaos theory," and indeed coupled-oscillator problems form one of the
most fruitful areas of recent research...)

anyway... here's a single "pluck" from a KS synth i made a few years
ago, with nonlinear force terms and arbitrary correlation between
masses; it sounds more like a gong than a string:

http://music.calarts.edu/~ebuchla/mp3/losalamos.mp3

as you say, that's not a playable instrument, its the raw output of a
numerical system, an array of 8 floats and some very very simple DE's.
as i mentioned before, ableton's "tensor" is a playable, commercial
implementation of a similar algorithm (or so i surmise.)... i like it
ok... anyway, it's easy to get the demo and play with it, and decide
for yourself...

anyways, i wouldn't know how to make that sliding enharmonic decay
stuff without some very intricate programming of sinusoids... or a
super, SUPER weird phaser (that does not yet exist)... anyway it would
take a lot more "effort" than 8 floats and two DE's...

>
> I still think our perception is to limited, so it's like throwing
> pearls before the swine and at the end (to us/our minds) it might do
> nothing more than adding a nice texture to the sound ;) I'm not sure
> if this would be worth all the effort, at least when it comes to
> sound generation...

ok. i guess i don't agree. our ears and minds are fucking amazing,
capable of incredible feats of acoustic analysis: pitch
discrimintation, rhythmic discrimination, spectral feature extraction,
masking...

i grew up on a diet of experimental electronic music and close
listening. i think "texture" is very deep. you're a sound designer so
you probably do too, i'd imagine...

in any case, i don't understand your distaste for researching novel
synthesis architectures. they don't have to be more complicated or
chaotic. they could be more flexible and computationally efficient,
like wave terrain synthesis, or the original KS algorithm in its
day... in my observations, it really looks like most useful audio
technologies being life as unplayable academic toys...

i guess i keep thinking about if my father and bob moog hadn't thought
to apply FM (a telecom engineering method) to musical synthesis in the
1960's, on the grounds that the resultant sounds were too
"unpredictable," and unnecessary for the ingenious patchmaker (hey,
you got sines, saws, squares and AM, what more do you need?)...
chowning formulated the "musical" characteristics of FM in the late
'70's...

i would love it if someone would formulate the "musical"
characterstics of nonlinear mass-spring systems (what are the
frequencies of the enharmonic orbits? how fast do they decay?). i
tried once, but it's very very very tricky...

>> there are people who want to use powerful dsp's to emulate vacuum
>> tubes or something. that's not so innteresting to me. what is
>> interesting is virtual toroidal gongs and pianos made out of rubbery
>> non-wood that you can bend into a horseshoe....
>
> I simply want good sounding (high audio bandwidth) tools that are
> powerfull enough to make my sonic fantasies come true, giving me the
> most easy and direct access to certain sonic parameters...

sure... fortunately you have that in abundance, these days. might as
well move forward.

>> i really would check out perry cook's work and his "synthesis toolkit"
>> c/c++ library if you'e even the slightest bit interested in realtime (not
>> lookup) synthesis informed by physics.
>
> I'm pretty sure that simulation isn't the way I want to go, even so
> it might be a shorcut for ppl. that aren't really into deep synthesis
> stuff or don't have a feeling for sound, to simply giving them a tool
> to reach something complex/pleasing sounding...

hm. well, thanks for that... jeez...

anyway, if you actually read any of cook's work (as i again recommend
to anyone interested), you'll see that a great deal of it is concerned
with isolating the parametric handles on the numerical models that
have the greatest psychoacoustic effect (and hence are the most
"playable....")

in a lot of ways i totally agree with your exhortation to use simpler
tools with more imagination. if i understood that right. but i also
like making tools, and it seems stupid to make the same ones over and
over again (software is cheap), and i can still think of some stuff
that i want that doesn't exist yet.

but, you know, i don't care much.

i just realized that i've never read anything i needed to read on this
list. i subscribed cause i wanted a cz-101 service manual.

now unsubscribing.

take care,

- eb.