[By the way, it's "Vocoder" for "Vocal Encoder / Decoder," not "Voco∗r∗der."
I hear a lot of people say it that way thinking that it's related to a
'recorder' but that's not the case.]
Think of vocal processing as one way to use a vocoder, not the only way
(although Homer Dudley did invent it for voice use, ostensibly as an
experiment in telephone communications, like so much of the equipment we
use).
A vocoder analyzes the frequency/amplitude spectrum of one signal and
'imposes' it on another. The human voice is very complex, and difficult to
synthesize (Tomita's little "voices" doing countdowns and such on his "The
Planets" album are about as good as you typically get without a vocoder).
The problem is controlling all of the subtle aspects of voice... generally
too much to do with two hands. A vocoder lets you simply speak into the
"modulator" input and impose that spectrum onto a "carrier." Some vocoders
have built-in vocal samples as carriers (like the Korg DVP-1, which I have),
but this is so limited-if you can't choose the sound that you want to
actually ∗shape∗, what good is it?
A vocoder works by inputting a signal into the Analysis section (the
Modulator signal). It is then sliced into frequency bands--the more the
better for best resolution---via a fixed filter bank (think graphic
equalizer). Each slice of the frequency band is passed to an envelope
follower. So now we have a control voltage that tracks each slice of the
frequency spectrum.
You also have a Carrier input that feeds the Synthesis section. This goes to
an identical fixed filter bank, which slices up the frequency spectrum the
same way, but in this case the slices go to VCAs. The VCAs from this section
are controlled with the CV outputs of the EFs from the previous section. The
outputs of the VCAs are then mixed together to an output (but it would be
nice if you could tap them before they did that... hint-hint...). In this
way, the spectrum of the second signal "follows" that of the first.
Because the subtle non-vowel plosives of speech ("shh" and the pop from a
"p" and so on) don't 'analyze' well with vocoders, especially those with a
small number of frequency bands, most vocoders have a bypass chain that goes
through a highpass filter to allow these sibilant sounds to bleed into the
carrier output to increase intelligibility.
Wendy Carlos made a vocoder out of two Moog Fixed Filter Banks, ten EFs and
ten VCAs for use in Timesteps and the Beethoven's Ninth parts of the
Clockwork Orange soundtrack. There are photos of it in the booklet from her
new S-O-Bach boxed set.
Okay, theory behind us... does it need a vocal input? Absolutely not! Like I
say, that just shows off the device because changes to the voice are easily
recognized by the human brain, and because it's so hard to synthesize any
other way.
But there are loads of cool things you can do with a vocoder. Feeding
percussion into the Mod input is very neat---each drum, etc., in a drum set
"lights up" a different part of the spectrum, so if you impose this onto
another sound---even with voice as the ∗Carrier∗ this time!---you get wild
results. I've heard weird metallic drum kits created this way by drumming
into the modulator and as a carrier using a ring-modulated sound, where one
of the VCOs was also controlled by a separate EF tracking the modulator
input. Very expressive... sounded like a drum kit made up of car fenders,
gongs, etc.
I've always been interested in being able to screw around with the
relationships between the bands so that the voice could be used to
formant-filter another sound, but not in the 1-to-1 way that produces
recognizable vowel sounds on the other side. When I was in high school, some
friends and I had an idea of creating a weird band called the "Rigelian War
Monks" where each of us would sing into a similarly cross-wired vocoder to
create a 'foreign language' of these War Monk performers (who would be
anonymously hooded in black and red). Kind of a Residents from Space thing
before I'd ever heard of the Residents. We were young, and never got our
hands of vocoders that allowed us to do that.
Even simply being able to shift the control relationship between the two
halves would allow you to impose a formant from, say, a low male voice onto
the upper part of another spectrum for a female-ish voice. Having the bands
locked 1-to-1, as it is with 99% of vocoders, is very limiting, it seems to
me. It almost demands that your carrier & modulator signals occupy the same
portions of the frequency spectrum, which cancels out a whole boatload of
possibilities.
I've also heard instruments with very recognizable formants---oboes,
banjoes, etc.---used as modulators to warp synthesizer pads, string
sections, etc. to very unique effect.
Wendy Carlos, on "Secrets of Synthesis" (highly recommended album)
demonstrates a vocoder used with a Pitchtracker to create fairly realistic
voices that are not her own. Perfect for demanding ransom money over the
telephone.
Oh well, enough typing. Back to my siesta...
-----Original Message-----
From: Nathan Hunsicker [mailto:
nate@...]
Sent:Thursday, 16 March, 2000 11:00 AM
To:
motm@onelist.comSubject:Re: [motm] Vocoder Wish List / Daydreams
From: Nathan Hunsicker <
nate@...>
Since the only vocorder I ever used was an old korg vocorder, what can been
done besides vocal processing? What makes a vocorder great. As always, if a
vocorder can be an asset in unique sound creation i'm all for it, but the
only use I've even known one for was the "robot voice" effect. Does a
vocorder need a vocal input to function? will it work with other instuments
/ sound sources? Any feedback would be helpful -Nate