Between the extremes of the broad brushstrokes of subtractive synthesis and the painstaking detail of additive, there have existed many hybrid styles of synthesis combining the speed of the former with the precision of the latter. Paul Wiffen traces the development of this middle ground through its successes and heroic failures. This is the seventh article in a 12‑part series.
Throughout much of the last 20 years, there has been a strain of synthesis which, although it has never challenged the dominant variety at any point, has always provided a worthy alternative for the synthesist looking for that little bit extra control over the timbre of the source waveform without having to go to all the effort of specifying the shifting level of each harmonic individually, as in additive synthesis. While individual manufacturers have coined many terms for their variation on the theme — Wavetable Synthesis, Vector Synthesis, Wave Sequencing, and so on — the overall term which seems to best fit this broad category is Transitional synthesis, because the sound, broadly speaking, starts with the specific harmonic content of one or more waveforms, and evolves, through various means, to end with a different harmonic spectrum (rather than decaying to fundamentals, as with a closing analogue‑style filter). How this is achieved varies from one implementation to another, but what all the forms of this type of synthesis have in common is that they offer the user greater control of the harmonic content of the sound as time passes, by allowing him/her to specify the waveform at given moments in the sound's development.
In this they are very different from analogue synthesis, where the fundamental nature of the timbre throughout its development is determined by the basic waveform selected. All that can happen is that some or most of the frequencies this waveform contains can be removed by the filter cutoff or exaggerated by resonance; no radical shift in harmonic content can be achieved. In PCM‑based synthesis, the harmonic content of the sound is dictated by the frequencies present when the recording was made. Although these can also be modified by cutoff and resonance, new frequencies, again, cannot be introduced.
Perhaps the earliest manifestation of the kind of Transitional synthesis I'm talking about this month was on the original CMI (Computer Music Instrument) from the innovative Australian company Fairlight. Press and media coverage of the instrument made much of its light pen and the facility to draw single‑cycle waveforms that it offered. Those who tried this method, however, soon found that, without analogue filters to run through the harmonic content of waveforms, picking out and exaggerating their differing compositions, most hand‑drawn waveforms sounded rather ordinary and often bland, despite the revolutionary way in which they were created. The simple fact of the matter is that the human ear is sensitive to change in harmonic content, and tends to be unimpressed by a static harmonic content, however complex. The secret of the success of the enveloped filter, as a mainstay of synthesis over the years, is that it's an exceptionally quick and easy way to vary this harmonic content.
Lacking any such filtering capability, the Fairlight engineers had to look for another way to make harmonic content change. Obviously, the samples the Fairlight made could contain timbral changes but only if they were present in the source being recorded. Introducing timbral change on the machine itself would be a tougher job. The system they eventually came up with was perhaps the only function on the CMI which really made use of its computational power, all its other facilities being simple RAM storage and replay tasks, whether of sample data or sequences.
...if you can get your hands on a PPG, Prophet VS or Yamaha SY22, you'll discover a style of synthesis which is perhaps the most powerful of all the non‑imitative styles.
Having created two waveforms, the user could place one at the beginning of the available sound memory and the other at the end. The computer would then calculate a waveform for every other memory location in between, by interpolating between all the corresponding points on the two waveforms (this process was known, a little inaccurately, as a Merge). As a result, each waveform played back in the course of a sound made in this way was similar to the one which preceded it but with subtle changes as the waveform was slowly altered to evolve towards the final result. The important thing was that these changes were entirely different to those which a filter would give, as they were produced by a mathematical method which was not in any way restricted by how sound behaves in the real world. Exciting new timbres emerged which had never been heard before. These could be radical (if two completely dissimilar waveforms were specified as start and end points) or subtle (if the two waveforms were closer together in appearance). There were changes for the ear to pick up on, and these changes were also unpredictable and different.
Unfortunately, although the CMI's method might sound like a sound designer's dream, the technology of the time had some major limitations which restricted its usefulness in mainstream musical applications. Firstly, memory size was limited, so the transition from one waveform to the other happened fairly quickly at the nominal original pitch. This meant that it was no good for sustaining sounds, where a gradual change in timbre works wonders; the sound was always of finite length. Secondly, the memory into which the transitional sound was loaded achieved pitch changes in exactly the same way as a sampler — by replaying at different rates. So the higher up the keyboard you triggered the sound, the shorter it became, and the quicker the timbral change happened. A sound triggered lower in the keyboard would last longer, and its timbral change would take longer to happen.
Of course, there were times when this unalterable relationship was fortuitous. Occasionally the short high notes would have enough punch and character to stand out well over a pad sound, which made up for their brevity. More often, the longer, slower notes worked well as sustained low‑end sounds, with the timbral change accentuated because the higher harmonics were more audible.
But most Fairlight users lost patience with these limitations, at least in part because the CMI had nowhere near enough computional power to perform its operations in real time (which is why the computations had to be stored in memory and played back just like samples). As a result, they would stick to the (at the time) unique sampling and rhythm‑sequencing capabilities of the Fairlight.
Very little use of the CMI's Merge facility was recorded for posterity, and I have never heard a sound on record which I could positively identify as having been created in this way. I live in hope that, since today's computational power could create the interpolated waveforms needed on the fly, without even breathing hard, someone will do a real‑time implementation of this exciting feature of the grandaddy of all digital systems, since, if the calculations were done in real time, the speed change needed to vary the pitch would disappear. The bright, metallic sounds I found the CMI created would suit current musical styles, like techno, down to the ground.
The next digital instruments to venture into the territory of harmonic transition were the PPG Wave series. Happily, this system, invented by Wolfgang Palm, did not rely on computation in real time, so the Wave synthesizers did not suffer the problem of the evolution of the sound being linked to its replay pitch. As a result, our old friend the envelope generator could be used to control the speed and direction of the movement between waveforms.
This was possible because the waveforms were created at the factory and loaded, in 'family' groupings, into so‑called 'wavetables', sets of digital memory locations exactly like single‑cycle sampled waveforms. These wavetables allowed a style of harmonic transition which was very similar to the Merge facility on the Fairlight, in that each waveform was only slightly different to the one on either side of it — but over the 32 locations within each wavetable, wide timbral changes were possible.
Of course, those wavetable groupings were decided by the manufacturer, removing the element of serendipity available on the user‑defined Fairlight implementation. But this was more than made up for by the fact that the results were usable in a real‑time mainstream format.
The synthesist was able to specify the wavetable used by each oscillator, and the starting waveform. However, there was then no obligation to make use of the wavetable's harmonic flexibility. The specified waveform could be used through the sound's duration, complete with normal amplifier and filter enveloping, exactly as on an analogue synth (although the waveform was generated digitally). In fact, one of the PPG's wavetables contained the standard sine, sawtooth, square and pulse waveforms, so that you could make sounds in exactly the same way as with analogue synths, although they never sounded quite the same.
However, nobody bought PPG Wave synths for their ability to duplicate the analogue synthesis process, but rather for the fact that they could supersede it. Once you had specified the initial harmonic content with the starting waveform, an envelope or LFO could be used to change that harmonic content, by moving around inside the wavetable in much the same way that envelopes and LFOs can change the filter cutoff from its initial harmonic‑content setting in analogue synthesis. The greater the envelope depth or LFO amount, the further away it was possible to move from the original waveform in the wavetable. The speed of that movement was determined by the attack, decay and release times of the envelope, or the frequency of the LFO.
Despite the fact that the wavetables were factory‑preset, this gave the PPG Wave synthesizers a much broader timbral range than standard analogue synths, especially as enveloped analogue filters could also be brought to bear on the sound after the wavetable synthesis had done its unique job. The closest analogy for those of you who have only heard analogue synthesizers is Pulse Width Modulation (PWM): the timbre changes without any movement on the part of the filter as the waveform moves between different variations of the basic waveshape. It was when I first heard Pulse Width Modulation that synthesis came alive for me, and the PPG system offered this same kind of movement, but with a host of different timbral groups in the various wavetables.
Just as with PWM, you could choose to set a constant timbral motion, with an LFO moving the wave readout evenly on each side of the starter waveform, or set up a more tailored single harmonic movement using the attack, decay, sustain and release phases of an envelope. You could use the attack to move quickly from the initial waveform to another further along the wavetable, move back a portion of that distance using the decay, hold on one particular waveform for the sustain segment, and then move slowly back to the original waveform during the release phase.
The beauty of Vector Synthesis was that it was very 'hands on' (to use the modern jargon) and simple to grasp (figuratively and literally).
As stated earlier, the PPG system was considerably more musically useful than the Merge function of the Fairlight, but was restricted to the waveforms provided by PPG in the Wave synthesizer. PPG's Waveterm changed all this, by providing the computational power for users to create their own waveforms (and, incidentally, make samples) and download these into the Wave 2.2 synth for use just like the factory‑preset wavetables. On the Wave 2.3, the whole memory could be used to download and play back 12‑bit samples linearly.
There were two versions of the Waveterm (A & B), easily distinguished externally by the fact that most 'A's had 8‑inch floppy drives, while the 'B's used the newer 5.25‑inch disks. More importantly, the 'B's were improved internally by 16‑bit resolution and better A/D conversion for the sampling side. Of course, sample playback through the analogue filters of the Wave 2.3 drew most attention (not surprisingly, as it pre‑dated PCM‑based synthesis by seven or eight years) but more creative users latched onto the fact that with the Waveterm they could build their own wavetables and turn them into custom sounds on the 2.3 synth.
Of course, time eventually catches up with any technological innovation, and PPG's fortunes faltered with the arrival of cheap samplers from Ensoniq, Sequential and Akai. Ironically, these never attempted to cover wavetable synthesis, but nevertheless the writing was on the wall for the Wave system. Despite ground‑breaking new product designs, which were the first attempts anywhere in the world at stand‑alone hard disk recording and virtual synthesis (called the HDR and the Realizer), PPG finally went bankrupt in 1987 (see the 'Thoroughly Modern Wave' box for what happened next).
The next company to go in for a system which allowed you to change the harmonic content of the source sound in real time, before the filter section, was Sequential Circuits. However, instead of changing the waveform that an oscillator was generating, their system allowed you to set up four different waveforms on four different oscillators and then mix between them by means of a joystick. This was clearly a much cheaper system: the Prophet VS, the synth which used this technology, was released with a price tag of around £2000 instead of the £3‑4000 price tag the PPG Waves had carried. Of course, the resulting sound was not quite as smooth as that produced by the PPG, where the harmonic content of every waveform was closely related to that of the one either side in the wavetable. On the VS you could choose to mix between waveforms with vastly different harmonic contents, which made many of the resulting sounds a little harsh to the average ear.
Sequential dubbed this technology Vector Synthesis, which was perhaps a bit of a misnomer. A vector is a straight line between two points, but the VS's joystick allowed you to take any indirect path between the starting and end positions of the oscillator mix. No doubt Sequential thought that Vector Synthesis sounded better than Cartesian Synthesis, or any other more accurate name.
Apart from the waveforms supplied as standard (which included sine, sawtooth, square and various widths of pulse, so you could produce standard analogue timbres through the filter), the VS also allowed you to create your own waveforms, through a basic form of additive synthesis (which had only been available in the PPG system if you could afford a Waveterm to go with your keyboard). On the VS, this was done by stepping through the various harmonics and specifying a level for each. What's more, you could actually hear the resulting change in real time (unlike with the Waveterm, which had to compute when all harmonic levels had been set).
Once you'd created your waveforms, or just selected the factory‑preset ones you wanted to use, you could place two pairs of them on the X and Y axes of the joystick. This meant that left/right movement would control the mix between one pair and up/down movement would simultaneously do the same for the other pair. Any position of the joystick thus gave a unique mix of the four oscillators, and as a result, extremely complex timbral changes could be produced as part of a real‑time performance. This was something the PPG could only achieve through programming. Of course, there were envelopes on the VS, to allow this mix to be altered automatically during the playback of a note, but to make life even easier the VS could record a manual joystick movement, and use this as the model for automatic change in the mix.
The beauty of Vector Synthesis was that it was very 'hands on' (to use the modern jargon) and simple to grasp (figuratively and literally). There were no difficult concepts to get your head round. Everyone understands the concept of 'mixing', and a couple of minutes with the joystick made it very easy to understand the possibilities for unique sound creation.
So if it was such a great idea, why was the Prophet VS the last synth Sequential made, before going bankrupt and being taken over by Yamaha? Well, it was the usual combination of poor mechanical reliability and other developments in the industry with more mass‑market appeal. Sequential had their problems with quality control: one particular quirk with the case design made aftertouch stop working if you put the keyboard at an angle on an A‑frame stand. The joystick on the first VS I used to reacquaint myself with Vector Synthesis for this piece had partially dropped inside the case and was held in place by tape.
The other reason the VS remained a specialist taste was what I refer to as the 'Piano, Strings and Brass Effect'. The VS couldn't do any of these at all authentically. Unfortunately, it came out at around the same time as the D50, which had the rudiments of PCM‑based synthesis, so it could get close, and it had built‑in digital effects to boot. The fact that it was also cheaper than the VS was the final clincher. Sequential returned to sampling technology in the Prophet 3000, the first 16‑bit sampler to hit the market, but this came to market too late to save them.
Fortunately, the design talents at Sequential were not scattered to the four winds, as Yamaha stepped in and kept Dave Smith and his team together. They took over Sequential's building in San José, and although little was seen from the team in the year after the takeover, they were put to work on a more commercial implementation of Vector Synthesis. This eventually emerged a couple of years later as the Yamaha SY22, its ancestry clear from the joystick on the front panel. The synth included some PCM source waveforms (to take care of the Piano, Strings and Brass Effect) but lost out on the ability to build custom waveforms through additive synthesis, as this was felt to be too marginal.
But the main advantage the SY22 had was that it was built by Yamaha. The case design was much more solid and the reliability a thousand times better. In addition, Japanese manufacturing techniques had brought the price down to well under £1000. As a result, the SY22 sold in much greater numbers, and if you find one on the second‑hand market, the chances are that it will be in much better condition than a VS and will continue to work properly for many years to come (even if there are those, like myself, who would argue that the Yamaha version misses out on much of the uniqueness and character of the original VS).
Curiously, by the time the SY22 hit the market, Yamaha had already been without the Sequential design team for almost a year.
Curiously, by the time the SY22 hit the market, Yamaha had already been without the Sequential team for almost a year — inscrutably, they had parted company with the ex‑Sequential personnel almost as quickly as they had moved to keep them together. However, another Japanese manufacturer, Korg, stepped in to preserve the unity of the design team, and Korg have continued to use their talents as an R&D facility ever since (two current Korg products which owe their existence to this facility are the 1212 I/O PCI card and the Z1 synth). Ironically, the first product they presented to Korg, another implementation of some of the concepts first introduced in the Prophet VS, was developed so quickly that it was launched at the same NAMM show which saw the introduction of the Yamaha SY22. We'll look at this instrument, the Korg Wavestation (perhaps the most successful of all the transitional synthesizers), in the next instalment of Synth School, as well as the most powerful implementation yet from Emu Systems, in the form of the Morpheus.
In the meantime, if you can get your hands on a PPG, Prophet VS or Yamaha SY22, you'll discover a style of synthesis which is perhaps the most powerful of all the non‑imitative styles — no use at all if you want authentic piano, strings and brass sounds, but all the better for that if you want to come up with truly unique and personalised synth timbres.
Fortunately, when PPG ceased to be, its Wave technology was not lost forever. Wolfgang Düren, who had masterminded worldwide sales for PPG in their heyday, decided, at the end of the '80s, to recruit designer Wolfgang Palm. The aim was to use the new LSI (Large Scale Integrated) circuit technology to produce a MIDI‑controllable rackmount version of the Wave system. In a inspired moment they decided to call it the Microwave, and this instrument is still available in an updated form today — the Microwave and the Microwave II boast the original wavetables from the PPG instruments. Rumour has it that the new company spent over a year trying to make digital filtering sound as good as the original analogue filters of the 2, 2.2 and 2.3, but, in a move reflecting the original Wave keyboard's design years before, they were forced to go with analogue filters to keep the sound authentic. Distinguished visually by a large, bright‑red parameter value dial (reminiscent of Comic Relief's Red Nose), these instruments have brought the price of wavetable technology down to around the £1000 mark, without sonic compromise, thanks to the modern economy of single‑parameter access.
A few years later the same team followed the Microwave with the impressive Waldorf Wave keyboard, boasting a front panel which can be raised up, Minimoog style, for ease of use when programming. In addition to the trademark big red dial, there are scores of smaller red knobs and switches to make programming as quick and easy as possible. Unfortunately, all this instantaneous parameter access has its downside: the price. The Waldorf Wave is one of the most expensive synthesizers on the market, but this hasn't stopped the production being pre‑sold for years in advance.
As a result of Waldorf's efforts, if this piece has whetted your appetite for wavetable synthesis, you're not obliged to brave the second‑ (and third‑) hand marketplace. You can purchase a current Wave synth in either a very affordable (Microwave) or very expensive (Waldorf Wave) form depending on your budget, but either way you will have perhaps the most authentic recreation of a vintage technology on the market. If you do go, instead, for an original PPG, make sure you know a good service engineer.