Archive of the former Yahoo!Groups mailing list: MOTM

I would like to add my own thoughts to Paul's comments on this subject.

The challenge in creating "new" waveforms for a modular synth (i.e., not
sine, triangle, square/pulse, sawtooth) isn't so much in the technology for
making it happen, but rather in selecting useful waveforms. What is needed
is an 'orthogonal' set of basic sound 'primitives,' and strangely enough,
although the 'big four' probably exist mostly because they're easy to
generate using basic circuitry, they do tend to cover some basics nicely,
which is why they have endured.

Sine is the purest component (ala Fourier), triangle contains odd partials,
square also contains odd-only [with more upper-spectrum energy] but by
altering duty cycle produces useful spectral modulations, and sawtooth
contains a robust set of odd & even partials, ripe for subtractive
synthesis.

The idea of creating some new basic colors is very appealing, but in
practice, I've been pretty disappointed with most things I've heard in this
regard. As Paul indicated, they sound like buzzes & mush for the most part.
Like random slices from a sampled waveform, looped.

If one were to try to add new waveforms to this palette, I think the thing
to do is worry about the sound first and the technology (analog vs. d∗g∗t∗l
wavetable) last. In looking at the existing set (the "Big 4"), the first
thing that probably comes to mind for most people is the possibility of an
"even-partial-only" complement to the tri/square. I believe (could be wrong
here) that this wave might look a lot like a full-wave-rectified sine.

On it's own, not a real pleasant sound. But probably as we search for new
waveforms, we will find that they may be of the kind that aren't as 'useful'
on their own as the Big 4, but are purely designed to be 'additive' elements
to a fundamental provided by a sine, etc.

Probably not a bad way to think. Consider Carlos' comment on 'Secrets of
Synthesis' that analog synthesis tends to be bold, but as it gets layered,
starts to sound mushy and thick. My feeling is that this is largely due to
the fact that the fundamental and lower-order partials are heavy in each of
the existing waveforms, being duplicated in a layered tone more than they
would in a naturally-produced sound. (In sound design, as in orchestration,
one of the first things you learn is that it's easier to 'fillout' a sound
in higher registers than in lower ones, where too much activity starts to
sound mushy quickly.) To brighten a timbre, we may add a triangle, etc., an
octave or more up, but this is only a crude simulation of what happens in a
naturally-generated sound because the spacing of the partials is different
as you move further up from the fundamental.

The ESQ-1/M has a large set of wavetable components ("synthetic waveforms")
that were created digitally with the idea that they would complement the
synth's basic waveforms. These include peak formant waves, waves with only
prime-numbered partials, variants on the Big 4 with portions of their
spectra removed, etc. On paper, they sound very exciting, but in practice,
while it's good to have them and they can be used to brighten up a sound in
interesting waves, when push comes to shove, none of them is as useful as a
sine, triangle, pulse, or sawtooth. They are almost more academic exercises
rather than fundamental (no pun intended) additions to the "four elements"
of the traditional waveforms.

Anyway, I'm just rambling, but I think it's interesting ground for
investigation. I do think that what needs to happen in creating an extended
orthoganal waveform set is to think about what might complement the current
waveforms, then experiment with those sounds to prove them out, and lastly
decide how to create them. I have a feeling, sad as it is for purists, that
truly useful waveforms may be subtle combinations of partials that can
really only be created and stored as d∗g∗t∗l wavetables, rather than be made
with discharging capacitors, etc. (the full-wave-rectified sine may be an
exception). I hope I'm wrong!!

The other road to go down would be to worry less about static spectra and be
more concerned with modulation possibilities (I'm not talking about
subtractive filtering). For example, any static spectrum of a particular
pulse duty cycle may be only 'so' interesting in a "forest for the trees"
kind of way, but pulse width modulation is an effect all of its own. Maybe
there are analog-domain ways of doing waveform morphing & animation along
those lines that no one has yet implemented.

Enough for now; apologies for all the hot air.

(and sorry for mispellings... I'm using Outlook Web Access so this letter
was edited in plain text, revealing one of my weaker traits...)


--KT