8

150 Chapter 14 EVOC 20 PolySynth
Vocoder Basics
If you are new to vocoders you should read this section. It provides you with basic
knowledge about vocoders and their functionality. You will also find tips on using
vocoders, and achieving good speech intelligibility.
What Is a Vocoder?
The word vocoder is an abbreviation for VOice enCODER. A vocoder analyses and
transfers the sonic character of the audio signal arriving at its analysis input to the
audio signal present at its synthesis input. The result of this process is heard at the
output of the vocoder.
The classic vocoder sound uses speech as the analysis signal and a synthesizer sound as
the synthesis signal. This classic sound was popularized in the late 70s and early 80s.
You’ll probably know it from tracks such as “O Superman” by Laurie Anderson, “Funky
Town” by Lipps Inc. and numerous Kraftwerk pieces—from “Autobahn” and “Europe
Endless” up to “The Robots” and “Computer World.”
Away from these “singing robot” sounds, vocoding has also been used in many films. As
examples: the Cylons in Battlestar Galactica, and most famously, on the voice of Darth
Vader from the Star Wars saga.
Vocoding, as a process, is not strictly limited to vocal performances. You could use a
drum loop as the analysis signal to shape a string ensemble sound arriving at the
synthesis input.
How Does a Vocoder Work?
The speech analyzer and synthesizer referred to above are actually two filter banks of
bandpass filters. Bandpass filters allow a frequency band (a slice) in the overall
frequency spectrum to pass through unchanged, and cut the frequencies which fall
outside of the band’s range.
In the EVOC 20 plug-ins, these filter banks are named the Analysis and Synthesis
sections. These filter banks have a matching number of corresponding bands—if the
analysis filter bank has five bands (1, 2, 3, 4, and 5), there will be a corresponding set of
five bands in the synthesis filter bank. Band 1 in the analysis bank is matched to band 1
in the synthesis bank, band 2 to band 2, and so on.
The audio signal arriving at the analysis input passes through the analysis filter bank
where it is divided into up to 20 bands.