Although FM synthesis has its roots in the sixties, the instruments that popularised it were to dominate the synth scene of the eighties. Gordon Reid uncovers the origins of FM and charts its rise to fame from its unlikely beginings in academic research in the USA.
For once, one of my retros does not begin with the sentence, "I remember the first time I saw a..." That's because it starts at the Bell Telephone Laboratories in the 1950s, some considerable time before I was born. This was where a gentleman by the name of Max Matthews began experimenting with digital computers, to see whether they could become a viable means for generating audio signals. Matthews was far ahead of his time, if only because he realised that — unlike the primitive analogue signal generators of the time — computer‑generated digital audio could be consistent and controllable. In 1957 he wrote a program called MUSIC I, programmed in assembly code for an IBM 704 mainframe computer. You may think this unremarkable, but you should realise that the 704 was a vacuum‑tube (valve) computer utterly incapable of executing the program in real time.
Understandably, MUSIC I was only capable of generating very basic sounds (it had a single triangle‑wave digital oscillator) but Matthews continued his developments and, in 1958, wrote MUSIC II. This had four triangle oscillators and was capable of much more interesting sounds. Inevitably, MUSIC III followed. Completed in 1960, Matthews wrote the program for a more advanced (transistor) mainframe called the IBM 7094. Then there was MUSIC IV (1962), MUSIC IVF (written in 1965 by a man named Arthur Roberts) and MUSIC IV BF (1966/67).
Enter John Chowning
At this point we must change scene, to Stanford University's computer department, and observe two researchers named John Chowning and Leland Smith, without whom the music industry would be a very different place today. They were working on a new version of the programs developed by Matthews and his associates, which they called MUSIC V. At the same time, Chowning was experimenting with huge amounts of vibrato (amounts that you could never obtain without the use of electronics), applying this to the audio–frequency signals generated by the digital oscillators within the program. Apocrypha has it that he accidentally programmed a vibrato that was larger than even he intended, discovering that the result was not vibrato, but a unique tone unlike anything he had heard before.
Chowning was apparently unaware that he had stumbled across a common technique used to broadcast radio transmissions: Frequency Modulation of a carrier signal (an 'operator') by a modulator (another 'operator'). This is understandable: FM radio signals exist at frequencies around 100Mhz, and even exceptional human hearing is limited to about 25kHz, so nobody had fully appreciated the audible effect that the modulator could have. By accidentally modulating a signal in the audio band, Chowning became the first person to hear what we now call FM synthesis, and he discovered that it was a powerful way to create new sounds. Consequently, in 1966, he was the first person to compose and perform a piece of music using FM as the sound generator. Called Sabelithe, it predated commercial digital synthesis by more than a decade.
Meanwhile, development of the various MUSIC programs continued apace, and in 1968 Princeton released a much faster version for the IBM 360 mainframe. Even MIT got in on the act and, in 1973, a compact version, called MUSIC 11, was developed for the PDP 11 minicomputer. MUSIC 11 was a huge step forward. For one thing, the PDP 11 was much more compact than the huge IBM mainframes required by previous versions. Secondly, MUSIC 11 was the first 'mini' digital synthesis program to use a QWERTY keyboard and VDU.
Throughout the latter half of the sixties, Chowning continued to develop FM synthesis, adding functions that allowed him to control the evolution of the sounds he created. Then, in 1971, Matthews suggeste d that he should create a range of recognisable sounds, such as organs or brass, to demonstrate that FM was musically useful and, therefore, a possible basis for a commercially viable product. Chowning did this and, with a view to licensing the technology, persuaded Stanford's Office of Technology Licensing to approach companies for him.
Yamaha Step In
Do you know the story of the A&R man at Decca who turned down The Beatles? Well, how do you think Hammond and Wurlitzer feel knowing that they turned down FM? As did all the other American manufacturers that the university approached. In desperation, Stanford contacted the Californian office of a well‑known manufacturer of motorbikes, powerboat engines, and construction equipment. Yamaha duly despatched a young engineer named Ichimura who, after a brief evaluation, recommended that the company take out a licence on Chowning's system.
With understandable caution, Yamaha negotiated a one‑year license that it believed would be sufficient to enable it to investigate FM and decide whether the technique was commercially viable. So it was that, in 1973, Yamaha's organ division (which, in 1974, would be responsible for the mighty GX1) began development of a prototype FM monosynth.
Despite the commercial development taking place elsewhere, Chowning was still an academic, and he continued working at Stanford on MUSIC 10 (yet another version, this time for the PDP 10). Unfortunately for him, the university failed to see the value of his work so, after a somewhat acrimonious parting of the ways, he moved to Europe to continue his research. This later proved to be a significant embarrassment for Stanford because, when Yamaha approached it to negotiate an exclusive commercial license for FM, Chowning was no longer a member of the faculty. The best that can be said of the university is that it knew when to eat humble pie, reinstating Chowning as Research Associate of its Center for Computer Research and Musical Acoustics (CCRMA). Chowning then assigned the copyright of FM to Stanford, which duly assigned the license to Yamaha.
Chowning was, of course, no fool, and he later received a royalty on the sale of all Yamaha's FM synthesizers. And, while his personal deal with Stanford has remained highly confidential, the university is rumoured to have collected more than $20 million in license fees. Whatever the exact figures, it's no coincidence that CCRMA was later re–housed in its own purpose‑built facility.
Anyway, with the legal technicalities sorted out, it was time to begin serious commercial development of an FM synthesizer. So, in 1975, Yamaha built a prototype polyphonic FM synth. But first...
The First FM Synths
Surprisingly, neither Yamaha's commercial work nor Chowning's academic efforts led directly to the world's first commercially produced digital synthesizer. Called the Synclavier, this was developed by NED (the New England Digital Corporation) and it was a polyphonic digital synth based upon 8‑bit FM and additive synthesis.
Despite its famous name, it's highly unlikely that you have ever seen an original Synclavier. This is because they are incredibly rare. Released in 1978 or thereabouts, only 20 or so were made. The early Synclavier that some people recognise — the one with the incredibly sexy five‑octave keyboard spangled with little red buttons — was, in fact, the second version. First appearing in 1980, NED called this, with remarkable logic, the Synclavier II. The instrument that most people recognise — the huge black, weighted keyboard — was the final version of the Synclavier. This soon became a mainstay of top professionals in the film and video industries, later spawning a powerful range of synth/sampling workstations, and nearly 20 years later it remains a favourite in post‑production studios all over the world.
Yamaha, meanwhile, were in the synthesizer doldrums. Having scaled all the peaks with the GX1 and CS80, in 1975 and 1976 respectively, they had experienced a slow but inexorable slide from favour during the later years of the decade. The Prophet 5 and Oberheim OBX were cheaper and much lighter than the CS80, and the CS70M — designed to replace the CS80 and cure some of its perceived deficiencies — was a turkey of the first order. Similarly, the CS range of monosynths and the SK multi‑keyboards failed to capture our hearts, souls, and wallets. Yamaha was floating into the backwaters of the professional keyboard world. Something had to be done.
It was not until 1981 that Yamaha unveiled their first commercial FM synthesizers, trailing NED by three years. Named the GS1 and the GS2, these abandoned recognisable synth facilities such as oscillators and filters, in favour of frightening new things such as multi‑operator equation generators, 30kHz data rates, and digital‑to‑analogue converters. What's more, despite offering just two‑operator algorithms, the operation of these monstrously expensive machines (the GS1 retailed for £12,000) was a complete mystery to all but the most mathematically orientated. Not that this mattered, as neither machine offered editing, each being (more or less) a preset instrument, capable of playing only the 500‑odd voices supplied by Yamaha on little magnetic 'lollipop' sticks. Yet the GSs were, for a couple of years, reasonably successful. Toto, for example, layered nearly a dozen tracks of GS1 on million‑selling hits such as 'Rosanna' and 'Africa', and used two of them in their live shows.
There was a good reason for this success. The sound quality and playability of the GS1 was exceptional (and the GS2 wasn't too shoddy either). But, in an otherwise pre‑digital world, for most players these were hyper‑expensive oddities. It's thus no surprise that Yamaha sold only around 100 or so GS1s and, in all likelihood, few more GS2s.
Then, in 1982, Yamaha demonstrated a keyboard with 'six‑operator equation generators' that you could edit. It was never released in its original form, but this instrument was to be the progenitor of one of the most important synthesizers ever created. It would cost a fraction of the price of its predecessors. Moreover, it would weigh far less, would offer an unprecedented 16‑note polyphony, and would incorporate velocity and pressure sensitivity. Furthermore, using a radical new technology called 'storage cartridges', it would be able to store and access thousands of new sounds. But before that, there were the CE20 and CE25 Combo Ensembles.
The First Affordable FM Synths
I remember seeing a second‑hand CE20 for sale in the late '80s. It looked like one of those 'diddly‑bop' four‑octave home keyboards that Yamaha and Casio seemed to churn out with frightening frequency, and I duly ignored it. Had I known its significance, I'm sure I would have forked out the £60 or so that the seller wanted.
With 14 preset monophonic voices and six preset, eight‑note polyphonic voices, this was undoubtedly the first affordable FM keyboard on the planet. Yet, like the heavyweight GS1, it offered vibrato, 'symphonic' (ensemble chorus), and programmable velocity and pressure sensitivity, at the time unknown on cheap polyphonic keyboards. That the aftertouch offered control over tone, volume and vibrato depth was just astounding. Another neat trick was offered by the portamento provided on the monophonic voices, which only created its glide effect when both the initial and the destination notes were depressed simultaneously. Later copied by a handful of esoteric synths such as the Crumar Spirit, and more recently the Alesis Andromeda, this feature allowed you to use playing technique (rather than knobs or wheels) to create portamento and pitch‑bend effects.
Despite its groundbreaking price/performance ratio, the CE20 was clearly aimed squarely at the home‑organ market. Its solo voices comprised five woodwind, three brass, four strings, and two electric‑bass imitations, while the polyphonic voices included such 'classics' as polyphonic brass, horn, organ, electric piano, harpsichord and strings. Of course, none of these — by today's standards — sounded much like the real thing, but when you consider the alternatives available in 1982, they were remarkable.
The CE25 was a somewhat different beast, with 20 polyphonic voices — five brass, three strings, two organs, two poly‑leads, two poly‑bass, and six percussive — so it lacked the CE20's monophonic portamento. By way of compensation, it offered a slider that controlled the depth of the ensemble effect. This allowed players to add just a hint of depth to voices that would have suffered had 100 percent of the effect been applied to them.
Whichever way you look at them, the CE20 and CE25 were revolutionary keyboards that should have created far more of a stir at the time, and should have carved a far deeper niche in synthesizer history than they have. The reason that they did not is, perhaps, that they were completely overshadowed by what was to come...
The World's Most Popular Synth
In retrospect, it seems clear that, just as the GX1 had been used to prototype technology for the CS50, CS60 and CS80, the GS1 and GS2 were test‑beds for the DX series. Likewise, I suspect that the CE20 and CE25 were little more than market‑testers for Yamaha, designed to ascertain whether the public was interested in the sound of FM. Yet, despite their decade‑long gestation, finalising the first batch of DX7s proved to be a bit of a problem because, just as they entered production, Sequential Circuits launched the Prophet 600 — the world's first MIDI synth. The prototype DX7 had no MIDI, so Yamaha quickly upgraded the electronics to allow the instruments to transmit and receive on MIDI Channel 1. Unfortunately, the cases had already been manufactured, so you can recognise the earliest DX7s by the silk‑screening of the word 'MIDI' on their control panels. The word was added later and, apparently, it's a slightly different colour from the other legends.
Despite these teething troubles, the DX7 was an instant sensation. Its six‑operator, 32‑algorithm FM synthesis allowed programmers to create sounds of unprecedented complexity and/or subtlety. Its 16–voice polyphony made a mockery of the five‑note Prophet 5 and the eight‑note Oberheims and Jupiter 8. It offered 32 patch memories, and you could use ROM and RAM cartridges to expand the number of sounds available at any given moment. Furthermore, breath control augmented the velocity and aftertouch sensitivity. This was an unparalleled specification for the time.
Then there was the price... at under £1500, the DX7 was not only the most highly specified of professional synthesizers, it was by far the most affordable. It lacked the sonic depth of its analogue competitors, but it more than made up for this by providing a breathtaking new palette of sounds, many of which remain classics to this day. Of these, the now clichéd DX7 piano patch is probably the most (in)famous, and 18 years later it still shows no sign of disappearing. Indeed, so successful was this sound that it later became impossible to buy a synthesizer — analogue or digital — without an imitative patch named 'DX Piano', or something similar. Even sampled pianos and Roland's SAS (modelled) pianos offered DX imitations alongside their emulations of acoustic and electric pianos.
It soon became clear that the DX7's forté was percussive instruments, and anything that sounded like two bits of metal being banged against each other. Furthermore, it excelled at orchestral imitations such as brass and woodwind. Taking the untreated sound from its single output and adding a little chorus and reverb proved highly successful, adding a realism that no analogue synth had ever approached. I even became a surprise beneficiary of this when, in 1984, Robert John Godfrey of the Enid (a band then famous for its huge rig of analogue synths) bought a DX7 and gave me his 'redundant' ARP Odyssey!
In stark contrast to its wide‑ranging capabilities, the DX7 looked surprisingly conservative. At a time when many American instruments sported multi–coloured control panels, graphic affectations such as blue pin‑stripes, and large blocks of wood on either end, Yamaha had given the DX7 a slim, sober, and professional appearance. This proved to be a master stroke. In 1983, the world was poised on the brink of the digital revolution. Large turntables with granite bases were out, and compact disc players were in; huge, mainframe computers were out, and tiny 'home' computers were in; large, knobby analogue synthesizers were out, and sleek digital synthesizers were very much in.
But just as the DX7 had great strengths, it had huge weaknesses. As we now know, it proved completely unable to create the warm pads that are meat and drink to any self‑respecting analogue polysynth. It was also incapable of the characteristic filter sweeps that dominate so much of popular music in the 21st century. Likewise, its achievements in the field of lead synthesis are (to be polite) less than memorable.
It also had functional disabilities. It wasn't long before players started to notice the hiss generated by the 14‑bit DACs used in the early models. In addition, it was a MIDI imbecile, and suffered from the most obnoxious operating system ever devised for a commercial keyboard. And, even ignoring the woefully inadequate screen and horribly abbreviated parameter names, FM synthesis was simply too much science for players brought up on VCOs, VCFs and VCAs. Far from being, as Yamaha claimed, "an easier synthesiser to program than ever before", it precluded most players from all but the most serendipitous twiddling. Indeed, the vast majority of players never reached first base in programming the DX7, electing instead to use it as a preset instrument. Consequently, a whole industry was built on players' needs for more sounds on ROM, more RAM packs (both internal and external) and even computer‑based patch editors for the brave.
Of the third‑party products, perhaps the most important was the Grey Matter E! board, which increased the number of on–board memories from 32 to 320 (each with independent MIDI parameters) and added patch layering, plus limited EQ. Two French developers named Dan Armandy and Alain Seghir produced the less well‑known but superior SuperMax. Like E!, this added more memories, layering, and enhanced MIDI controller capabilities, but also offered arpeggios, MIDI echo and delay, velocity cross‑switching, and more.
But perhaps the strangest add‑on (being about as large as the DX7 itself) was the Jellinghaus DX Programmer, a large blue control surface that offered dedicated knobs for all the important FM parameters. I only ever saw one of these, at Rogue Music in New York, and was sorely tempted to buy it. Unfortunately, I dithered an hour too long, and by the time I went back for it, it was gone.
If the DX7 took the world by storm (and it did), the same can not be said of its less‑than‑impressive little brother, the DX9. This lacked the DX7's velocity and pressure sensitivity and offered an emasculated eight‑algorithm, four‑operator‑per‑voice version of FM with just 20 patch memories. It was a strange sibling for the DX7. Whereas big brother almost single‑handedly redefined the synthesizer world, little brother was — to be blunt — a complete turkey. Of course, Yamaha may be a lot sneakier than people realise. If ever there was an incentive to spend a bit more to get a DX7, the DX9 was it!
More successful was the TX7, a desktop MIDI module that contained the guts of a DX7, but offered no on‑board editing. Then, in 1984, the monstrous TXn16 appeared. This was a 19–inch rackmount frame that housed up to eight TF1 modules, each of which was, in essence, a DX7. These came configured as the TX216 (with two TF1s) to the truly monstrous TX816 — a 128‑note polyphonic, eight‑part multitimbral device equivalent to no fewer than eight DX7s in a single cabinet.
But none of these modules captured the public's imagination as had the original DX7. That honour was left to another of Yamaha's monster synths with the number '1' in its name. It was time for the DX1.
FM Giant: The GS1 In More Detail
Even today, the GS1 is a remarkable instrument, weighing 90kg (a hair under 200lbs) and designed like a miniature grand piano. It's clear that Yamaha intended it for studio use. But don't let the 19th century looks deceive you. Its beautifully weighted 88‑note keyboard is both velocity and poly‑pressure sensitive, and — with the exception of on‑board editing — it is everything you would expect from a top‑of‑the‑range synth dating back to 1981.
Setting one up is no mean feat. You have to tip the GS1 forward onto its front surface, so that you can bolt on the legs and the pedal unit. Then you have to get it upright again. I've managed to do this alone on two occasions, and my back still complains about it.
Once you've assembled the GS1, you can turn your attention to the sparse control panel situated immediately above and behind the keys themselves. At first this looks horribly limited. For example, there are only nine performance controls: vibrato and tremolo sections (each with speed and depth knobs), a 3‑band EQ, a detune knob, and a master volume control. But these hide some superb performance capabilities. For example, the 'soft' pedal is actually a pressure‑sensitive vibrato controller, while the centre 'sostenuto' pedal is a pressure‑sensitive tremolo controller. (The damper, or 'sustain' pedal is exactly what it appears to be.) Likewise, the detune knob, far from being a simple detune, offers four settings: shallow and deep 'static' detune, and shallow and deep 'random' detune. It's remarkable how much life is injected into the sound by the random variations of the latter settings.
The only other control, aside from those that select and store voices, is the Ensemble On/Off switch. This introduces a real analogue delay line into the signal path, and makes the GS1 as swooshy and noisy as any analogue ensemble keyboard.
Finally, we come to the memories themselves. There are 16 of these, which you fill using Yamaha's preset voice library. To load a memory, you insert side 'A' of the small magnetic 'lollipop stick' in the card reader slot, then press one of the 16 location buttons to start the reading process. Once this is complete, you turn the magnetic stick over and load the information on side 'B'.
It's not obvious, but you've just loaded different voice data into two otherwise identical dual‑operator/dual‑carrier 16‑voice synthesizers. This means that you can mix and match 'A' sides and 'B' sides from different patches, to create new, composite sounds in memory. Yamaha actually encouraged this, and the GS1 provides a 'Store' capability that allows you to record your new combinations onto single magnetic sticks.
You would probably be amazed at the expressive nature of the GS1. Even ignoring the polyphonic aftertouch control of both loudness and brightness, its sounds can be both digital — in terms of clarity and brightness — and very analogue in their warmth and depth. Indeed, had it weighed 80kg less, I would have been delighted to use one as my string ensemble of choice in the early 1980s. You won't hear me say that of any other FM synth! Nonetheless, the GS1 also pointed the way forward to the digital sounds of the '80s, with many of the FM pianos, harpsichords, and other plucked sounds that Yamaha would (almost) perfect in the DX7.
In 1984, Yamaha released a MIDI upgrade for the GS1. This fits inside the body of the instrument and has an interface/control panel that replaces one of the two radiator grilles on the underside. (Radiator grilles! I kid you not... The GS1 consumed 95W of power.) By modern standards, the MIDI specification added by the upgrade is incredibly primitive. The GS1 remains monotimbral, and communicates only on MIDI channel 1, 2 or 3. Furthermore, it only transmits and receives Note On/Off and Velocity messages. Oh yes, and it can't transmit and receive simultaneously. There is a switch to determine which it is doing at any given time!
Despite all the GS1's shortcomings, it was — and remains — a remarkable instrument, and a surprisingly satisfying one to play. Hmmm... maybe that's not so surprising, given that its 1981 price of £12,000 is equivalent to around £50,000 today.
The GS2 In Brief
The GS2 offers only one of the 2x2‑operator synthesizers found in its bigger brother, and this limits you to simpler sounds comprising just side 'A' of a voice card. This may seem like a huge restriction, but given how radical FM synthesis was in 1981, and how bright and percussive FM voices seemed to a generation brought up on fat, splodgy analogue synths, it wasn't a problem.
Unfortunately, there are a couple of other deficiencies in the GS2's specification. On the electronic side, the GS2 loses the random modes of detune that add so much to some GS1 sounds. More significantly, it also loses the beautiful 88‑note velocity and poly‑pressure sensitive keyboard, which is replaced with an equally beautiful (but only velocity‑sensitive) 73‑note keyboard. This is the real difference. Whereas the GS1 is a wonderful instrument to play, the GS2 is merely streets ahead of anything else that was available back in 1981. Once you had played the GS1, nothing else would do.
Surprisingly, the GS2 is only 18kg lighter than its big brother, so you gain little benefit from its lesser specification. I suspect this is why — despite its much lower retail price — the GS2 sold little better than the GS1.
Sine Of The Times: A Bit About FM Synthesis
FM synthesizers create complex sounds by arranging and combining the outputs of digital sine-wave generators ('Operators') in different ways ('Algorithms').
The interactions of the Operators within the Algorithm determine the nature of the sound. Each Operator can act in two ways: as a Carrier, or as a Modulator. In simple terms, a Carrier is acted upon, whereas a Modulator acts upon. To be precise, a Modulator can also be 'acted upon' by another Modulator, but this is not the place to discuss FM in that much depth. Simply remember this: (i) the more Operators that can interact, the more complex the sound can be, and (ii) the more Algorithms there are, the more varied the range of sounds can be.
The GS1 has eight operators per voice (which is a lot) but these are arranged as four modulator/carrier pairs (which is extremely primitive). The GS2 is even worse: it sports just two modulator/carrier pairs per voice. In contrast, the DX7 has six operators per voice, but offers 32 algorithms that allow programmers to design sounds that take advantage of far more complex interactions between the operators. In this light, it's amazing that the two GS keyboards sounded as rich and engaging as they did.
For a full explanation of FM synthesis, take a look back at Parts 12 and 13 of our epic synthesis series, Synth Secrets (see SOS April 2000 and May 2000; also available on the SOS web site at www.soundonsound.com/series/synth-secrets).