If you thought vocoders were just an obsolete way of making a singer into a Dalek, you've been missing out. Tom McLaughlin explains music technology's most overlooked innovation.
If you've always associated vocoders with tacky sci-fi robot voices rather than music you could be missing out on one of the most creative musical devices technology has yet produced.
THE VOCODER IS a most misunderstood beast. A hangover from the "prehistoric" analogue days, its mere mention conjures up images of alien and robot voice effects of dubious intelligibility. In a nutshell, the vocoder analyses the tonal qualities and loudness contour of one signal (the modulator) and superimposes them upon another signal (the carrier). But if you thought that the vocoder's use in life was restricted to making strange voice effects you've been missing the point of creative sound manipulation. Read on.
The earliest mention of the vocoder (from voice coder) I can find is 1948, preceding the voltage controlled synthesiser by almost two decades. In an attempt to minimise the amount of information needed to transmit speech down electric lines, the Bell Telephone Laboratory realised that speech could be broken up into discrete frequency bands relating to vocal formants using filters.
Unlike modem analysis, synthesis and re-synthesis techniques, vocoder technology relies solely on detecting the amount of energy present in these predetermined frequency bands with good old analogue filters and envelope followers. The music and recording world actually owe quite a lot to communications technology, for in addition to the vocoder, equalisers came out of the concentrated analogue research in the '40s '50s and '60s.
The vocoder synthesises its speech by employing two identical sets of band-pass filters with centre frequencies matching key vocal formant bands. A voice signal is analysed and broken down into its constituent formant bands with the first set of band filters and re-constructed at the receiving end with the other set. A vocoder greatly reduces the amount of information required to transmit speech down telephone lines. Thus the vocoder was born. It may come as a surprise, but the voice you hear on the other end of the telephone is often an analogue or digital real-time reconstruction.
Although a standard vocoder is limited to dealing with frequency bands important to speech (the EMS 2000 has two sets of 16 filters: low-pass 14OHz, band-pass 185, 270, 367, 444, 539, 653, 791, 958, 1161, 1406, 1703, 2064, 2700, 4000Hz and a high-pass starting at 5388 for sibilant and un-voiced sounds), it still has to be considered one of the most intelligent filter banks on the market. There are filter banks and programmable equalisers available with better technical specifications, especially in the signal-to-noise ratio department, and some can even string together and play back settings via MIDI, but none of them can match the sheer versatility of the vocoder's ability to continuously track an incoming signal's frequency spectrum and superimpose that upon another signal in real time.
There are several types of vocoders still being made. EMS seem to be the most readily available and reckon they'll be producing their excellent vocoders (designed by Tim Orr) for many years to come. If not available immediately from stock, they'll manufacture any of their existing models, to specification if need be, in as little as a few weeks. I believe that Synton and AKG still make their vocoders but as far as I know these are only to order. The vocoder is far from dead, in fact Robin Wood of EMS UK assures me that extensive use of the vocoder was made in the recent Falklands war for scrambling top secret communications.
Roland, Korg, Moog and Bode made vocoders which you may still find floating around in secondhand columns, hire companies and universities. In its heyday, vocoder research led to some pretty hairy designs. The most complex vocoder EMS currently manufacture is their Vocoder 5000 with its two sets of 22 filters. In the late '60s David Cockerell and Peter Zinovieff came up with a system involving a 60-channel analysis/resynthesis section. Some, like the Roland Vocoder-Plus, were geared to the masses, having a simplified filter circuit and string synth intrinsic to its design. Roland couldn't have made it easier to get ethereal choirs, backing or harmony vocals: simply plug a microphone or recorded vocal in and follow the vocal line on the keyboard. Fine if that's all you wanted, but a true vocoder is capable of so very much more.
Recently two digital processors with vocoding modes have appeared, namely the Korg DVP1 and Roland VP90 neither of which allow the flexibility of a true vocoder, and although fun to use and perfectly valid as specialised signal processors it is questionable whether these can actually be described as vocoders. It all depends what sort of effect you're looking for, the possibilities and permutations are vast, but vocoder use seems to fall into the following categories: creating synthetic speech effects, creating choirs and harmony vocals, making instruments sing, enhancing natural speech, altering a speaker's identity, correcting formants of sampled vocals, formant cross-modulation (voice scrambling), instrument cross-modulation, "impossible" vocal effects.
EARLY USE OF the vocoder on tracks such as the Electric Light Orchestra's 'Mr Blue Sky' leaves a lot to be desired in the intelligibility department. This is understandable as the vocoder was probably hired in and, with hire and studio rates being what they are, insufficient time given to getting the most out of this unique form of signal processor.
Vocoders like working with sounds that have been optimised dynamically and spectrally. Feed them mediocre modulator signals and no matter how good the carrier signal is, the end result will be mediocre. Medium to heavily compressed modulator vocals are in order here, with equalisation tweaked to make the voice as intelligible as possible (a favourite trick is to tweak a vocal, or mix, then listen to it from the next room. Success can be judged on the important elements still being present after making it through the walls).
Careful microphone technique is also important. Vocalists shouldn't be much closer than 6" from the microphone and should speak across, rather than into, the mic to prevent overloading the vocoder with plosive sounds. A pop-shield wouldn't go amiss, either. If there is considerable noise or ambience, judicious use of a noise gate will be of assistance.
Aural exciters may increase the intelligibility of non-vocoded vocals but tend to confuse vocoders when used to the extreme on modulator signals. With the exciters added top end high formants get shifted even higher and things like "V"s start sounding like "F"s. Best to use aural exciters in moderation on the modulator or, better still, on the carrier or vocoded output.
The standard procedure with vocoders has been to superimpose speech upon a wide spectrum carrier signal, typically an unfiltered or lightly filtered synthesiser tone or oscillator. The easiest effect to produce is the droning robot voice. Little work needs to be done to the carrier signal and it's as simple as finding the pitch, register and synthesiser tone colour suitable for the character of your "robot", putting that into the carrier input of a vocoder and speaking into a microphone sent to the modulator input. The result is the familiar monotone robot or alien voice.
But the carrier doesn't have to be a simple synth tone. As a matter of fact, your alien voice will be considerably more interesting if you use a combination of synth waveforms. Modular analogue synthesisers give a lot of scope for electronic effect source material. Complex LFO pitch and amplitude modulations, waveform cross-modulations and ring modulations can be set up so that what was once speech becomes nothing less than "out of this world". To help make fricatives and non-voices sounds (h-p-t-k-f-s) more intelligible, it is often necessary to mix in a small amount of high energy/high frequency sound material - traditionally white noise.
You'll find that your vocoded character will sound more natural if you vary the pitch of the character to mimic the inflections of human speech. Very few of us speak in a monotone (robots, aliens, and cyborgs excepted). Using a pitch-bend wheel is preferable to using a keyboard for varying the pitch of the carrier, the frequency animation characteristics of speech are rarely heard as discrete semitone steps. Take a good listen to yourself reciting a short poem and experiment with pitch-bending your synth to follow its natural fluctuations in pitch. In addition to allowing external carrier signals to be processed, EMS vocoders have an onboard oscillator that may be used for the "voiced" portion of vocoded speech, and a circuit that may be used to vary the pitch of that oscillator according to the amplitude of the modulator in a most natural manner - speak softly and the oscillator will pitch itself lower, louder and its pitch will raise. Good thinking, Batman.
EVEN THE SIMPLEST of vocoders makes giving the impression of a lead vocal backed by harmonies, backing vocals and choirs a breeze. The modulator may be a pre-recorded vocal track or a live source, and the carrier can be anything from a synth string or choir patch, sample library voices or even a sample of the actual lead vocalist.
Pulling the effect off requires painstaking attention to playing the carrier in time with the line to be vocoded and sensitive use of pitch-bend and vibrato. As with keyboard string sounds, slightly anticipating the entry of each note helps the illusion. Don't worry about hearing the carrier too soon. The brilliant thing about working with vocoders is that the carrier is gated by the modulator - even though it may still be playing, you won't hear the carrier if there is no signal present at the modulator input.
It's tempting to speed up the recording process by playing chords into your vocoder, but the realism of choirs and backing vocal ensembles will be increased if you overdub each chord's note as a separate line and track, each with a slightly different vibrato, tone colour and pitch bend. Careful EQ'ing and balancing of vocoded vocals in a mix completes the illusion.
"At its peak the vocoder was an industry buzz-word - like AMS, SSL, and DX - everybody was using them and, inevitably, they suffered from overkill."
To realistically vocode speech, it has been necessary to use either a very bright synth patch or mix in a noise source to give the vocoder something to work with for un-voiced or breathy speech components. With sampling now as readily available as synthesis there's no reason not to use sampled voices as carriers. Breathy vocal samples like 'Fairvoice' give the added benefit of already having a good amount of high-frequency material present. Using synthesisers as a carrier's raw material may have a lot to do with vocoders sounding overtly synthetic in the past. This need not be the case; any signal may be used as a carrier or modulator, as we'll see.
THIS IS AN effect that hasn't been used in quite a while. You have to remember that at its peak the vocoder was an industry buzz-word, much like the LinnDrum, AMS, sampling, SSL, and DX have been. Everybody and his brother were using them (not always tastefully) at one point and, inevitably, they suffered from overkill.
Instruments may be made to sing by playing the instrument and singing into the vocoder, singing live to a prerecorded instrument track or playing to a taped vocal track. Musical skill is required: the vocoder is an instrument and to get the most out of it you have to practice using it. With a sync code on tape, modulators or carriers may be sequenced - this takes just as much skill as playing live. Emotive vocals require spot-on timing to follow convincingly.
MORE COMPLEX VOCODERS like the Roland SVC 350 (11-band) and EMS 3000 (16-band) allow the user to adjust the amount by which specific formant bands are excited by the incoming modulator. Using the same voice signal as both modulator and carrier, deficiencies in the original signal can often be made up for.
I had occasion to use this technique on a series of samples made from a trans-Atlantic telephone messages left on an answering machine - you could just about make out what was being said. On top of that the production required that "the phone message be placed in an AMS reverb approximating a large airport". Trying to make things more intelligible with an aural exciter only added to the already uncertain top end and, although equalisation helped, it wasn't the answer. Dusting off my trusty vocoder I found that compressing and EQ'ing the voice samples, and tweaking the formant bandpass filters on the vocoder helped them to sound more like the caller was phoning from around the corner rather than from across the sea.
The vocoder can also be used in a similar fashion to a de-esser, but with much more flexibility. Vocals suffering from over-sibilance can modulate themselves while rolling off a little bit of sensitivity on the upper formant filters to cut down on the offending "sssss". Similarly, poppy "P"s and "B"s can be minimised by doing the same with the lower filters. Questionable vocal takes may be given a new lease of life. With attention to a good balance between the original and vocoder-corrected voice, you can often conceal the fact that any signal processor is being used at all.
THIS IS FUN. Apart from thinking in terms of the vocoder as a speech synthesiser or specialised vocal equaliser, it may also be used to alter or mask the speaker's identity. Different formant bands can be accentuated, giving a voice another quality. The sex of the speaker can convincingly be changed by harmonising a split of the vocal to the desired register, then cross-modulating that with the original - anyone remember Will Powers?
THERE ARE MANY instances where use of the "palaeolithic" vocoder can help make vocal samples a bit more believable. As the major reason for munchkinisation is the shifting of a vocal sample's formants up the keyboard along with its pitch, we can use the vocoder to superimpose a fixed set of formants on our vocal samples, adding considerably to their realism.
This is no more difficult than playing a vocal sample (the modulator) in a natural register, to sound the same note every time the sampled voices (the carrier) are played. You'll need to plan a keyboard map that sends the modulator to one output of your sampler and the carrier(s) to another. With sequencing, an extremely tight marriage of modulator and carrier signals is possible.
Some vocoders have Hold or Freeze facilities allowing their analysis and resynthesis circuits to "hold onto" a sound at a given point in time. This has been useful in the past to superimpose "oo" and "aah" vocal sounds onto carriers but, like looping around single wave cycles of samples, one tone colour can become rather monotonous to listen to for any length of time. Now that we have the technology, it makes far more sense to use a sampled vowel sound as the modulator - offering all the animation of sampled sounds - along with proper formant registers and ratios.
ONE OF THE earliest uses of the vocoder seems to have been in the postwar period by the US Navy. Messages could be sent down communications lines without the enemy having the slightest clue as to what was being said, by taking the values arrived at by the speech analysis section of a "modular" vocoder (such as Moog's and EMS' 5000) and swapping them around at the resynthesis input. This process is known as formant crossmodulation or Voice Scrambling and really does produce quite unidentifiable synthetic speech. I wonder what the result would sound like after locking someone away with one of these to use on an acid house mix?
I'VE SPENT MANY gleeful hours experimenting with sampled as well as synthesised sound sources. Believe me, dealing with even the most boring synth patches yields surprisingly musical results, giving them an almost digital, sampled quality far more interesting than layering the sounds upon one another. I'm not a great lover of standard FM sounds (except when layered with digital, analogue or sampled sounds), but the effect of modulating simple FM and analogue sounds by one another produces acoustic events that border on what you'd expect to hear coming from Synclaviers or Fairlights - and at a fraction of their prices.
Frequency spectrum and loudness envelopes of sampled acoustic events can be magically transferred to any synth or sampled sound you can dream of with a vocoder. The 'Let's Dance' snare (thanks David) envelope might control how we hear a steam burst, complete with sizzling ambience. A more realistic tone colour and bow scrape can be added to synthesised strings using sampled strings as the modulator. Something as simple as a pitch-swept sine wave will impose amazing harmonic sweeps on your carrier - nifty means of adding animation to boring sounds or drawing one's attention away from tell-tale sample loops.
But a vocoder isn't limited to processing simple speech or sustained notes. Depending on whether its slew rate is fast enough, a vocoder can be used to superimpose or cross-modulate almost any acoustic event, even complex rhythmic patterns. Hip hop, house and high energy sampling fanatics take note that no-one has yet modulated a James Brown groove with, say, a Stevie Wonder riff. This could be a creative way around the use of copyrighted material.
Unexpected poly-rhythms and accent shifts crop up when modulating one rhythmic pattern with another, which may just prove musically useful to you. Modulating a very ambient, though rhythmically simple, drum pattern with a more complex pattern will have the effect of placing previously non-existent emphasis on the ambience for a kind of ghost or phantom drum sound. Spooky.
Mega-tight gated reverb effects are created as simply as vocoding the reverb return with the untreated drum and combining the two through a mixer. Try this with a slow slew rate to exaggerate the effect.
A most unusual effect can be obtained by modulating the reverbed, delayed or effected signal with the original or vice versa. This is virtually uncharted territory; if you're in search of new effects, make a point of checking this one out. Do somebody's brain in by vocoding a reversed copy of a sound or better still, a reverse reverb.
ANYTHING YOU CAN do on a keyboard, instrument or sequencer can be superimposed upon a vocal, including any conceivable variation of trill, arpeggio, glissando, portamento or intervals technically difficult to sing live.
Vocoders might not be an answer to all your problems (or lack of inspiration), and you'll get far more use out of a reverb unit, but there are definitely some amazing sounds and effects hidden away inside them - certainly enough to warrant a closer inspection now that multitrack recording gear, quality instruments, synthesisers and samplers are within many a muso's financial grasp. With a little bit of imagination and experimentation who knows what sort of sonic mind-benders may emerge from someone's bedroom?
Feature by Tom McLaughlin
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!