Understanding Filter Banks And Vocoders; Filter Banks; Vocoders - Moog SPECTRAVOX User Manual

UNDERSTANDING FILTER BANKS AND VOCODERS

FILTER BANKS

Electronic sounds, upon their widespread introduction into mainstream culture through the 1960's,
were initially perceived as cold, strange, and alien. The early pioneers of synthesizers, Bob Moog
among them, wanted synthesizers to not only create new sounds but also model existing acoustic
sounds. Pure electronic tones emerging from a speaker, however, were a completely new experience
to listeners – no one had ever heard sounds that were not acoustic in nature. The electronic sounds
lacked two aspects that always accompanied any acoustic sound – the tone of the room the sound is
made in (reverb) and the tone of the resonant acoustic body itself.
The resonant fixed filter bank was created to solve the second problem – to shape the sound of an
electronic tone and simulate a resonant body, like the wooden frame of a guitar, the skin and shell of a
drum, or the physical body of a human singer. The Moog 907 Fixed Filter Bank was the essential tool in
the early Moog modular systems to shape the spectral signature of a tone and thus give an electronic
sound a simulated resonant, acoustic body. Spectravox, like the 907, contains ten filters capable of
sculpting the spectrum of sound, but its filters have variable resonance and can be jointly shifted
through the frequency domain – greatly increasing the tonal possibilities.
12
UNDERSTANDING FILTER BANKS AND VOCODERS

VOCODERS

When we speak, our vocal cords begin to vibrate and create a tone. In order to form words, we move
muscles in our mouth to filter the sound—to articulate certain vowel or consonant sounds. Open your
mouth and make a long open "aaaaaaaah" sound. Keep your vocal cords vibrating and try to change
that "ah" sound to an "iiiii" sound. Play around with your mouth like Homer Dudley did back in 1928—
move to an open 'e' sound (like in the word bet) and then to an "oooooh" sound. Notice how all the
work is done by the muscles in your mouth—your vocal cords are always oscillating at the same pitch!
The human mouth is in fact an extremely expressive resonant filter. A vocoder seeks to capture the
movements of the mouth and impart the tonal characteristics of the mouth shape onto a second
sound (called the "carrier"). In essence, by using a vocoder you are swapping your vocal cords with the
carrier. You can take any sound (a synthesizer, a drum machine, the sound of an aircraft taking off) and
shape it with the dynamics of your mouth!
The Wendy Carlos/Bob Moog vocoder which is at the heart of Spectravox was constructed out of a
core of a few elements: two ten-band fixed filter banks (with matching frequency bands), ten envelope
followers, and ten voltage-controlled amplifiers. The sound of a human voice (called the PROGRAM) is
sent to the first bank of ten filters (the "analysis filter bank"). The output of each individual filter band is
sent to an envelope follower, which tracks the amplitude of the voice signal in that particular frequency
band. The output of each individual envelope follower essentially encodes the motion of a muscle in
the mouth—the way that muscle is filtering the sound at any given moment.
A second input—the CARRIER—is sent to the second bank (the "synthesis filter bank"). In this filter
bank, however, the level of each individual filter is controlled by the envelopes from the analysis filter
bank. The PROGRAM signal creates ten different filter envelopes with the motions of the mouth, and
those motions are used to control the shapes of the synthesis filter bank which is filtering the CARRIER.
This is how a vocoder maps the motions of the mouth onto a completely different sound!
ENVELOPE
PROGRAM
FOLLOWERS
1 1
2 2
3 3
4 4
ANALYSIS
5 5
FILTER
6 6
BANK
7 7
8 8
9 9
10 10
FIGURE: The PROGRAM is connected to the ANALYSIS filter bank, which breaks the PROGRAM sound
up into ten frequency bands. Each band is sent to an ENVELOPE FOLLOWER which converts the audio
in each PROGRAM band into a slow moving control voltage. These control voltages are then used to
control the level of each filter band in a SYNTHESIS filter bank. A different sound—the CARRIER—is fed
through the SYNTHESIS filter bank—thus mapping the timbral characteristics of the PROGRAM onto
the CARRIER.
(Continued)
CARRIER
1
2
3
4
SYNTHESIS
5
FILTER OUTPUT
6
BANK
7
8
9
10
13