In the first part of a two-part series, Gordon Reid charts the rise of EMS and their creation of the world's first self-contained portable synthesizer, the VCS3, and its velocity-sensitive keyboard, the DK1.
Here's a puzzle for you. Study the following list of names: ARP, EMS, Moog, and Sequential Circuits. Which is the odd one out? Well... they all manufactured analogue synthesizers, they were each world leaders in their fields at some point during the '70s, and they all went bust and disappeared in the late '70s or '80s. All but one, that is. EMS became bankrupt like all the others but still exists, albeit in modified form, and continues to sell classic analogue synthesizers.
Do you want another difference? OK, on an ARP, Moog, or Sequential Circuits synthesizer, you can set up and play a patch with reasonable confidence that — if you know what you are doing — it will perform as expected. You can also be sure that the patch will sound the same if you recreate it sometime later. Unfortunately, this isn't true for the EMS range of synthesizers. The company's most famous instruments are undoubtedly classics, and they remain sound effects machines of almost unsurpassed quality and flexibility. However, unless you're very brave (or you've paid for a whole bunch of modifications) you won't even attempt to play a tune on one. This is their story.
Peter Zinovieff was born in London in 1933. His parents were wealthy Russians who had fled to the UK following the 1917 revolution. A geologist who filled his home with samples (rocks, not audio) he was fascinated by electronic music, and used his wealth to develop a huge voltage-controlled studio that occupied an entire room in his Putney home.
When this became too unwieldy, he enlisted the help of engineer Dave Cockerell and programmer Peter Grogno. They helped him to design an enhanced system which used two DEC PDP8 minicomputers to control the voltage-controlled modules of Zinovieff's early synthesizers. Unlike previous electronic studios, their 'MUSYS' system proved reasonably user friendly, with a standard QWERTY keyboard and a velocity-sensitive piano-style keyboard, rather like today's computer-based studios.
Bearing in mind that Bob Moog had only recently published his seminal papers regarding voltage control, Zinovieff's ideas and instruments were incredible. It's worth remembering that, at that time, the closest most people had come to a computer was watching the flashing monstrosities featured in episodes of Star Trek and Lost In Space. But twenty years before affordable computing and sequencing packages, Zinovieff's PDP8s could store and replay compositions, complete with sound shaping parameters such as loudness and envelopes. His software was even capable of twisting the music into bizarre new sounds and effects. All this, from a program that loaded from a long strip of punched paper tape! In 1968, Zinovieff and Cockerell also invented a form of computer-controlled spectral (or 'additive') synthesis, using a system of 60 resonant filters that could analyse sounds and resynthesise them.
In 1969, when MUSYS became too expensive for Zinovieff alone, he decided to 'offer it to the nation' as a free resource for the Arts. To this end, he placed an advertisement in The Times newspaper but just £50 was forthcoming. The benefactor, a gentleman named Don Banks had, it seems, misunderstood Zinovieff's offer because, in return for his cheque, he asked Zinovieff to make him a synthesizer. Zinovieff and Cockerell therefore decided to embark upon a substantially different money-making scheme.
Together with Tristram Cary, a composer of electronic music for TV series such as Dr Who, Zinovieff and Cockerell created a new company, Electronic Music Studios Ltd. Given their mutual interest in synthesis, and the existing design of Cockerell's — the Voltage Controlled Studio No. 1 — it was hardly surprising that their first product should be a synthesizer.
The VCS No.1 was a hand-built rackmount unit with two oscillators, one filter and one envelope. In an era when a synthesizer was, almost by default, a huge Moog modular, this was not thought to be adequate, so the partners set out to enhance Cockerell's ideas. They designed an instrument that was small, portable, but very powerful and flexible. It was the Voltage Controlled Studio No. 3 or, as it is now known, the EMS VCS3.
In addition to his musical and technical accomplishments, Zinovieff was also interested in language. For Harrison Birtwhistle's opera entitled The Mask Of Orpheus, he chopped up the sounds in the words 'Eurice' and 'Orpheus' to create a language of 151 new words. He remained active in his MUSYS studio until 1979, producing works including Birtwhistle's Chronometer and Hans Werner Hense's Glass Music.
In 1975, the long-forgotten techno pioneers Zorch became the only 'rock' band ever to record in Peter Zinovieff's studio. Long-time fans and users of EMS synthesisers, they recorded a soundtrack for an early music video called Mother Earth. Zinovieff moved to a remote Scottish island in 1979 where, I understand, he remains to this day.
Since this article was first published in November 2000, Peter moved to Cambridge where he was interviewed on video by this article's author in 2016.
The VCS3 is, essentially, a modular synthesizer. Given the prevailing ideas about music synthesis in 1969, this is not surprising. After all, the first integrated synth — the Minimoog — was not to appear for another year, and all previous synthesizers had allowed you to connect the various sound creating sections as you wished.
The VCS3 comes in two parts. The synthesizer — nicknamed 'The Putney' because EMS was located in that part of London — contains the bulk of the audio modules. It also incorporates two power amplifiers and two speakers that make it a complete, self-contained sound-effects generator.
Oscillators one and two are the primary sound sources, and these produce a remarkable range of frequencies, from below 1Hz to around 10kHz. Oscillator one produces sine and sawtooth waveforms with a form of rectifying waveshaping for the sine wave. Independent level controls allow you to select the amounts of each waveform in the oscillator's output. The second oscillator also produces two simultaneous waveforms, and again it offers independent level controls for each. This time, the waveforms are pulse and triangle waves, with simultaneous waveshaping from zero percent to 100 percent on the former, and from sawtooth to ramp wave on the latter. It's a shame that, on an unmodified VCS3, none of the waveshapers can be voltage controlled, because this would introduce many forms of pulse width modulation (PWM) and dramatically increase the range of sounds available. Once selected, a waveform is static.
A third oscillator is similar to oscillator two, with pulse and triangular waveforms, but its frequency range is concentrated much further down the spectrum, lying between 0.025Hz (one cycle every 40 seconds) to around 500Hz. Oscillator three also offers a waveshaper, and this is identical in concept to oscillator two's. Using this, you can easily demonstrate just how precise the ramp/triangle/sawtooth waves can be. The same is not true, however, for the pulse waves!
An independent section on the panel contains a noise generator, with a level control and a 'colour' control that varies from predominantly low frequencies (red) to 'white' noise, and up to predominantly high frequency (blue) noise. Another section contains the ring modulator which, as you would expect, offers just an output level control.
Many players and writers have described the VCS3's filter as a conventional low-pass filter with an 18dB/octave slope, but they are, to some extent, wrong. The VCS3 filter exhibits a 'knee' in its cutoff profile; the first octave above the cutoff frequency rolls off at 12dB/octave, but the slope increases to 18dB/octave at frequencies above that. Furthermore, any amount of filter resonance significantly depresses the low frequency gain, so EMS described it as a combined low-pass/band-pass device. At high 'response' (the EMS term for resonance or 'emphasis') the filter self-oscillates, so it becomes an oscillator. This is not strange to a generation of players brought up on multi-mode filters and today's complex virtual analogue synthesizers, but it was mind-boggling stuff in the late '60s.
If the filter is slightly unusual, the envelope generator (which EMS called a 'shaper') and its associated VCA are positively arcane. It has six controls. The first, the Attack, is straightforward enough and has a range of about two milliseconds to one second. So far, so good. The next control is labelled 'On', but nowadays we would call this a sustain level 'Hold' because it determines the length of time that the envelope stays 'high' after you release the gate. (The Korg MS20 also has this on its HADSR envelope generator.) EMS claimed that 'On' has a range of zero to two and a half seconds, but mine holds for a full five seconds. Control number three is more recognisable — it's a Decay Rate, with a claimed range of three milliseconds to around 15 seconds. Mine lasts for 60 seconds. The fourth knob is labelled 'Off' and it determines the delay before auto-retriggering of the envelope cycle. Until you understand that this must be in the '10' position (called 'Manual') to play the VCS3 conventionally, it can be very confusing. Indeed, the envelope will auto-repeat at frequencies of up to 60Hz, which is well inside the audio range, so the 'Shaper' can also act as an LFO or even as a deep bass oscillator. Again, this must have confused many players dreadfully when they first tried to get a sensible sound out of a VCS3.
The envelope has two outputs with independent level controls. The first (and the fifth in the 'shaper section') is the 'Trapezoid' level — the one that confuses most people. To understand this, just picture the envelope shape produced by an AHD (attack/hold/decay) contour generator. The Trapezoid level simply determines the level of the envelope CV. The second level control (the sixth shaper control) is the signal level, and this controls the loudness of any signal passing through the Shaper. There is also a large, red Attack button, which we would nowadays describe as a Manual Gate.
The VCS3 also provided a spring reverb with Mix and Level controls. This is a simple dual-spring device with a maximum reverberation time of approximately two seconds. Unfortunately, when using the VCS3's internal speakers, the reverb howls uncontrollably before the mix gets very dense, and you can only use it to its full potential with external amplification and speakers.
It may not be obvious at first sight, but the VCS3 is a stereo synthesizer with independent output channels A and B that drive the left and right speakers respectively. These have independent level controls, panning controls, and output filters that, depending upon position, attenuate the bass or treble, or provide a flat response.
Performance controls are limited to the enormous X/Y joystick. This has two controls that govern the X and Y ranges but, unfortunately, its maximum range is about plus/minus two Volts, so it's not often that you can plumb the extremes of any parameters it controls. In addition, just to place the VCS3 firmly in '50s sci-fi territory, a voltmeter allows you to measure any control voltages (which are close to DC) or signal levels (which are AC) within your patches. You can even connect an oscilloscope to a dedicated quarter inch output on the rear. Wonderful!
On any synthesizer — whether vintage or modern — you expect a patch to sound the same, no matter how many times you return to it, but this is not the case with the VCS3. This is not due to its famous oscillator instabilities, but rather a consequence of the inconsistencies of the patch pins.
On a conventional modular synth, the impedances of the short patch cables are virtually zero, and are too small to affect the sound in any significant fashion. However, a white VCS3 pin may vary between 2,565Ω and 2,835Ω, and this is easily enough to change things. The most obvious way to demonstrate this is to change between two white pins in position I8 or J8... you may hear the keyboard tracking change significantly. Likewise, filter tracking, filter cutoff, audio signal levels between modules, other CVs between modules all change slightly when you replace one pin with another, ostensibly identical one.
To recreate a patch, therefore, you must not only record all the knob positions and the matrix patching map, but you must record which pin went in which position within the matrix. Fortunately, since you can now buy resistors with tolerances as good as 0.1 percent, you can alleviate this problem by replacing the resistors within all your existing pins. It's laborious work, but it pays in the end.
The separate DK1 keyboard — known as 'The Cricklewood', because that was where Cockerell worked — was as radical as the VCS3 it controlled. It was, of course, monophonic (there weren't any polyphonic synthesizers in 1969) but it was velocity sensitive, allowing players to add expression in a way that had previously been impossible.
The DK1 is connected to the VCS3 using a dedicated eight-way cable that provides two power rails, two CVs, and a Gate pulse for the envelope shaper. To the left of the keyboard, two switches control the two output CVs (called 'Channels') produced by the DK1. The first of these has 'Signal' and 'CV1' positions — CV1 was what we would now call the pitch CV. In contrast, switch two had a 'Dynamic Voltage' position and a 'CV2' position. The former of these produces a CV proportional to the velocity with which you hit a key, while the latter produces a pitch CV. But CV1, and therefore channel one, produce the same thing, so there's no point in having both switches set to 'CV'.
Returning to the 'Signal' position... the DK1 has a built-in sawtooth oscillator and an associated VCA with frequency, 'spread', level and dynamic range controls. This is a godsend because, with the spread set to '10' the oscillator tracks the keyboard in a conventional 1:1 relationship. In other words, you can play the keyboard and, with everything else set up appropriately, you'll hear the notes that you would expect. This is not necessarily the case when you use the keyboard CV channels. The reason is because the keyboard CV channels enter the VCS3 through two input level controls marked, sensibly enough, Channel one and Channel two. The problem arises because 1:1 key tracking occurs somewhere between 'six' and 'seven' on the knobs, and the exact position can fluctuate wildly with the oscillators' temperature, the time of day, and the FTSE100 index. This makes it very tricky to use the VCS3's internal oscillators for correctly pitched melodies. Every time you play the thing, and even after an hour of 'warming up', you are constantly trimming the tuning and scaling the Channels, much as you might expect to do on a Minimoog once every few years.
Furthermore, the VCS3 doesn't conform to either the one Volt per octave or Hz/V standards used by every other manufacturer before and after. It uses internal voltages of 0.32V/octave for oscillators one and two, 0.26V/octave for oscillator three, and 0.20 V/octave for the self-oscillating filter. However, because there are CV amplifiers on the internal module inputs, you need to double these figures to 0.64V/octave, 0.52V/octave and 0.40V/octave respectively for external CV sources!
Likewise, the usual 10V peak-to-peak signal levels are eschewed in favour of three Volts, four Volts and six Volts for the oscillators (depending upon waveform) five Volts for the filter, three Volts for the noise generator... and so on. There was nothing conventional about the VCS3.
Tristram Ogilvie Cary OAM, MA (Oxon), LMus TCL, HonRCM, I Eng (to give him his full title) was born in Oxford in 1925. His education was as posh as it gets, with scholarships galore, but the latter part of the Second World War interrupted this, and he spent four years in the Navy as a specialist working on the RADAR technology developed a few years earlier. It was here that he received his training in electronics, and developed his first ideas regarding electronic music recording.
He studied piano, horn, viola and conducting, and spent a number of years teaching music, working in a gramophone shop, and developing the ideas for his first electronic music studio. From 1954 onwards, he became known as a composer, gave up his other jobs, and began living off his commissions for theatre, radio, film, and television.
In 1967, Cary founded the electronic music studio at the Royal College of Music. In 1973, he was invited to become the senior visiting lecturer at the University of Melbourne in Australia, later moving to Adelaide, and eventually becoming the Dean of Music at Adelaide in 1982. In 1986 he left the University and now operates as Tristram Cary Creative Music Services, where he continues to produce specialised recordings for film, TV, theatre, and radio. His book, The Illustrated Compendium Of Musical Technology, was published in 1992.
We haven't yet discussed the VCS3's most notable characteristic: the patch matrix. The most important thing to note here is that the VCS3 will remain forever silent unless you stick some pins into the matrix. This is because none of the devices described are connected to each other unless you use the matrix to determine which signal goes where. Fortunately, the 16 x 16 matrix allows you to connect any of the VCS3's modules to each other.
Let's say that you want to direct the output of oscillator one to output channel one. Since the signal generated by oscillator one emerges from the list of sources in row three, and the input to channel one is column A, you simply stick a patch pin in position A3, and the connection is made. Of course, this doesn't preclude you from sticking more pins in row three, and yet more in column A, so patches can become very complex, very quickly. Indeed, you can stick 256 pins into all 256 available sockets, but I doubt that it would create a sound. (I've never tried it — pins are too expensive.) Also, you must remember that, at this point, you have only made a set of connections between modules. Whether you hear a sound, or whether it's a usable one, still depends on the positions of the front panel controls.
It may not be obvious from my description, but the VCS3's patch matrix is much more powerful than the cable patching systems of other synthesizers. This is because, whereas a Moog, ARP, or even a modern VCO may have one or two outputs, the VCS3's have, in effect, 16 outputs that you can direct to every input on the front panel. Likewise, each input on a conventional modular synthesizer is replaced by up to 16 inputs on the patch matrix. You would need sixteen large mixers and sixteen 16-way multiples to imitate this, using a conventional patching architecture, and I can't think of any synthesizer that offers this. So, despite its limited number of oscillators, and its limited filter and envelope architecture, the VCS3 is nonetheless an incredibly powerful synthesizer.
Unfortunately, there are three problems with the matrix. The first two are simple to avoid: if mistreated it can become unreliable; and it's very expensive to replace. The third is more fundamental. The matrix is not 'buffered', and this means that, every time you insert a pin into an existing patch, the actions of other patch connections will change to some degree. Let's suppose that you have spent an hour creating a complex patch and getting every knob exactly as you want it. You then decide that you want to add, say, oscillator two to the filter input. You insert the appropriate pin and ... everything else changes. As you can imagine, this is infuriating.
Now let's turn to the patch pins themselves. These are not simple metal connectors that short between the row and column rails. They are resistors, and there are three types of these in common use. White ones (with a resistance of 2.7kΩ) are the most common, and you can use them for almost anything. However, because the resistors in the pins have a wide (five percent) tolerance, they are not suitable for some jobs. In particular, two white pins inserted into I8 and J8 (CV Channel A connected to the pitch CV inputs of oscillators one and two) will often be sufficiently different to make the oscillators track differently. To overcome this, EMS supplied red pins, also 2.7kΩ, but with two percent tolerances. The third of the common pin colours is green. These pins have a higher resistance than the others, thus reducing the amplitude of a signal considerably. Most often, you use these when you want to attenuate a control signal, such as applying a delicate amount of modulation to a pitch CV input.
Typical patches require up to around 15 pins, but you will often see VCS3s with fewer than this. Indeed, the number of usable pins may be close to zero because mishandling and age damage the delicate internal connections, making patching haphazard at best. Fortunately, you can still obtain these from EMS.
Nobody knows how many VCS3s were built and, since EMS's serial numbering was neither consistent nor sequential, it's unlikely that we ever will. The most reliable estimate — given to me some years ago by Robin Cook — is that fewer than 800 ever saw the light. Of these, many have been destroyed, dumped into skips by owners who found them unusable and worthless in the mid-80s. Given that today they're worth £1,800, it makes you want to scour the rubbish tips and dumps around the country... maybe.
If you read some of the conversations flying around on the Internet, you might be forgiven for thinking that the VCS3 is no more than a glorified effects unit. In part, this is because few casual users have the knowledge or the patience to squeeze conventional musical signals from the instrument. But perhaps more significantly, it's because the VCS3 has four quarter inch inputs on the rear panel — two for microphones, two for line level signals — routed to the Channel one and Channel two rows on the patch matrix. Because the VCS3 is modular, this is a far more powerful arrangement than the signal inputs on pre-patched monosynths, allowing you to use an external signal as an extra module, maybe as an audio source, a CV source, or even a Gate.
There's another reason why the VCS3 is often regarded as a sound mangler. Because its internal oscillators are so unstable, using external signals (such as that generated by the DK1) is often the only way that you can play conventional melodies. So, in many ways, the VCS3's status as 'effects generator extraordinaire' is a classic case of making a virtue out of a necessity.
At £330 for the synthesizer itself and a further £150 for the keyboard, the VCS3+DK1 cost a small fortune in 1969. On the other hand, your £480 bought you the world's first self-contained, portable synthesizer and velocity-sensitive synthesizer keyboard. The VCS3 was difficult to program but, a year before the appearance of the Minimoog and ARP2600, it brought synthesis within the reach of the public. (How many players do you think ever had access to the esoteric music computers installed in studios such as MUSYS, or the BBC Radiophonic Workshop where Cary had worked earlier in the decade?)
As a result, success came quickly, and EMS soon grew, employing more engineers and more sales staff, including a young chap named Robin Wood (whose role in this story will become much clearer next month). Several breadboard prototypes appeared. Perhaps the most interesting was the VCS4; a five-octave keyboard married to a pair of VCS3s, a mixer and signal-processing unit. The 1969 prototype still exists and I understand that it's still functional, so why did EMS decline to produce it? Another was the Synthi KB1, a fantastic looking re-hash of the VCS3 with a 29-note mini-keyboard in a single case. Prototyped in 1970, EMS sold the only one of these to prog-rock band Yes and I'm not aware that it has never been seen again.
The VCS4 and KB1 were intriguing designs, and they may well have been commercially successful had they ever made it into production. However, in marked contrast to these compact synths, the next EMS instrument to appear was the monstrous Synthi 100, which we'll take a brief look at in Part 2, as we continue to chart the further rise and dramatic fall of EMS in the '70s, and its Phoenix-like rebirth in the '90s. Until then...
As you will have gathered, you can't treat a VCS3 like a conventional synthesizer, and you certainly can't sit down at a DK1 keyboard and just start to play. Even if you are adept at patching and programming the VCS3, its quirks and instabilities make conventional melodies all but impossible. Fortunately, there are numerous modifications possible, some of which tame the VCS3's worst excesses, and some of which considerably extend its usefulness. The following list describes a range of these, all available from EMS. It's by no means an exhaustive list, but it'll get you started.
- Oscillator stabilisation modification (per osc) £18
- Metal-can dual transistors for oscillators (per osc) £10
- Oscillator sync — variable via potentiometer (per osc) £20
- Voltage controlled shape (per osc) £20
- Hi/Lo switchable frequency range (per osc) £25
- Portamento/glide £45
- Attack time extension to five seconds £15
- 10-turn pots on input channels (each) £15
- Centre-zero trapezoid for inverted bi-polar output £30
- External gate input socket £5
- Patchable voltage inverter £35
- Extra input channel £30
Of these, I would urge anybody with a VCS3 to pay for the stabilisation modification for all three oscillators. Once these are completed, the synthesizer becomes very stable, and many of its perceived deficiencies simply vanish. You should consider the voltage controlled shape mod, because this adds all the Pulse Wave Modulation possibilities that you could ever desire. I also suggest the voltage inverter (very useful) plus the 10-turn pots, which make it much simpler to tune the input CVs for conventional playing. A full range of accessories, spares, books and manuals is also available from EMS. These include a special jack lead that permits conventional MIDI/CV operation, and the original user manuals, service manuals, and patch books first released from 1969 onwards.
EMS, Trendeal Vean Barn, Ladock, Truro, Cornwall TR2 4NW.
Tel +44 (0)1726 883265.