20 Feb
2004
20 Feb
'04
7:01 p.m.
[David Olofson]
On Friday 20 February 2004 20.41, Joern Nettingsmeier
wrote:
[...]
Well, there's a big difference between a fake crunchy guitar sound,
and a crunchy synth sound - but basically, I agree; use the Real
Thing(TM) instead. Easier, and it feels and sounds better. :-)
However, in this case, I was rather thinking about getting this sort
of sound into a "typical" Audiality song, which means a total file
size around 10 kB or so. (Sort of like IXS, except Free/Open Source
and based on MIDI + modular synthesis rather than a tracker format +
FM synthesis.)
What I want to do is push the limits of extremely compact music
formats, and proving that virtual analog style synthesis doesn't
*have* to sound like minimalistic german synth music. I may fail in
doing this, but I hope to at least learn a few useful things about
creating good synth sounds in the process.
the most promising concept if you want realism is probably physical
modeling, i.e. a waveguide. used with a filter modeling the response
(resonance) of the guitar body, you'll already have a good start. a
bit more troublesome will be feedback between strings, but doable.
as has been observed, the real trouble starts when you want to model
the non-linear aspects, like palm muting, finger tip interaction with
the string during oscillation, when releasing or tapping, and most
importantly the plektron or finger exciting the string.
there's some very interesting work by someone at hut.fi (sorry i'm so
vague on this, it's some time ago i read it) on modeling the piano.
the biggest part of it is dedicated to the hammer striking the string.
iirc, it needs oversampling. and the interaction of the hammer with
the piano string is much better defined than that of the plektron (let
alone finger tip or nail) with the guitar string.
for a cheap and quick solution, you'll probably have to reduce the
complexity by a fair amount. i would imagine sampling different
string-exciting techniques, analyzing and reducing them to a simple
data set and blending with a waveguide or even simpler oscillator
model could be fruitful. it won't be easy to faithfully emulate the
transients and their decay over the whole range though. for
palm-muting you just might get away with a rigid low-pass feeding back
to the oscillator.
that being said, i concur with jörn. :)
cheers,
tim
ps: thinking this over, i seem to remember those guys at hut.fi doing
some quite nice work modeling an acoustic guitar, too.
Tricky.
To get crunchy hard-rock guitar sounds like Pete's
(nice track pete!), you'll have to realistically emulate
palm-muting, which I've never heard in a synth. And how would you
control the amount of muting? Map it to a CC and play a slider?
<p class="heretic">
why spend precious coding time faking electric guitars when there
are so may excellent and highly trained guitarists around?
nothing against electronic sounds where they are appropriate. but
when you need a crunchy guitar, simulators strike me as the wrong
tool for the job. :-D
</p>