There's more to vocoders than 'Mister Blue Sky' — and with the vocoders supplied on some current multi‑effects processors, creative vocoding is now available to almost anyone.
Vocoders were very popular in the '70s, but their 'talking keyboard' sound soon became clichéd, and from then on, their popularity steadily declined. By the time MIDI started to take off, vocoders were all but extinct, with only a couple of manufacturers continuing to make them — which was a pity, because a vocoder really comes into its own when used as part of a MIDI system. Fortunately, a few multi‑effects units now include a vocoder as part of their repertoire, and with a little ingenuity, they can be used to modify sounds in a number of creative ways — other than producing the classic 'asthmatic who's swallowed a harmonica' vocal effect.
But before exploring some of the processing tricks made possible by this unique device, it's useful to take a look inside to see how it works.
A vocoder enables the tonal character of one sound to be imposed on another, quite different sound; the classic talking keyboard effect is produced by using the changing characteristics of the human voice to shape a sustained synth sound. What really happens is this: the vocal signal, which we shall call the modulator, is analysed by a bank of filters that continually measure the signal envelope in each part of the spectrum in exactly the same way as a spectrum analyser does. The more filters in the bank, the more accurate the analysis.
The signal to be modified, known as the carrier, is also fed to a bank of filters, but this time the level of signal passing through each filter band is modulated according to the output from the spectrum analyser section. In other words, the spectral characteristics of the modulating signal are duplicated in the filter bank processing the carrier. Figure 1 shows a simplified block diagram of what's actually going on. If the modulating signal is continually changing in character, as is true of the human voice, these dynamic changes will be passed onto the carrier, giving the synth sound a recognisable vocal quality. So effective is this process that it is possible to pick out intelligible words, even when none of the original vocal signal is present. And because we're analysing the spectral content and not the absolute pitch of the modulating signal, it doesn't even matter if the words are sung out of tune, or even spoken.
Apart from the obvious spectral variations generated by the vocal chords, human speech also includes 'fricatives' — short, high‑frequency sounds present in syllables such as 'S' and 'T' which are formed in the mouth. If these are separated out from the main vocal signal by means of a high pass filter, they can be added to the output to increase the intelligibility of the sound, and because they don't relate to the musical pitch of the vocal, they can be added to any musical output without compromising the tuning. A simple system for adding fricatives is also shown in Figure 1.