You are here

Synth School: Part 6

Building On PCM — The Next Generations By Paul Wiffen
Published February 1998

Ensoniq's VFX‑SD workstation.Ensoniq's VFX‑SD workstation.

The ball of S&S synthesis had been thrown, and most of the big names in synthesis caught it and ran with it, scoring some notable goals in the process. Paul Wiffen continues his chronicle of modern synthesiswith a look at the state of play from the late '80s to the present day. This is the sixth article in a 12‑part series.

When we left the PCM‑based synth story in the last instalment of this series, Korg's M1 workstation had superseded the Roland D50 and was in a dominant market position. But technology marches on, and there were many new developments in sampled‑based synthesis and hybrid systems still to come from Ensoniq and Yamaha (amongst others), but also from Korg themselves.

An American Tale

Sampling and FM synthesis technology in one box — Yamaha's SY99.Sampling and FM synthesis technology in one box — Yamaha's SY99.

The next company to develop PCM‑based synthesis was the American manufacturer Ensoniq. In fairness to Ensoniq, they were actually the first company to put sample‑based waveforms in synthesizers. Back in 1985, the ESQ1 had a few small PCM samples built in to allow drums and strings to sound more authentic. In fact, these samples, combined with a built‑in sequencer, made the ESQ1 a candidate for the title of first workstation synth.

However, the VFX was really the first machine from Ensoniq which could be compared with the Korg M1, in that it had quality samples and effects, multi‑stage envelopes and multitimbrality. Introduced in late 1988, the VFX lacked only one thing to qualify it as a workstation — a sequencer. This was added in the VFX‑SD the next year. As the suffix implies, Ensoniq added not only a Sequencer but a floppy Disk drive to this model, to enable saving and loading of sequences and programs.

The VFX architecture is well worth examining, because for the first time it made a PCM‑based synth as easy to set up for live use as the split/layer keyboards of the late '70s and early '80s. Combining M1 Programs into Combis had always been a bit of an effort, and certainly not the sort of thing you tackled on stage halfway through a gig. For the VFX, though, a system was developed which made the process easy enough to contemplate in front of an audience.

Live Or Programmed?

Emu's Proteus 1.Emu's Proteus 1.

Ensoniq achieved the task by adding a different section to the VFX, from which the combining sounds could be played live. This was separate from the setup used for multitimbral access by internal or external sequencer. The arrangement made perfect sense, because the parameters you need to change quickly when layering two or three sounds together live are very different to the parameters you might need to adjust during the playback of a multitimbral sequence. The VFX allowed the user to quickly select three different patches (the second and third by double‑clicking) and then see and adjust their respective volumes, pan, keyboard range, effect amounts and other important parameters related to live presentation. This meant that VFX users could very quickly assemble a complex split/layer setup, with maybe two sounds under the right hand and a bass sound under the left, balance them, and position them in the stereo mix, without the minutes of parameter adjustment that would have been required to do the same with an M1 Combi, for example.

As a result, a whole new breed of players was encouraged to start programming, because the surface layer of the VFX gave easy access to the combining of programs into performances without the need to develop an in‑depth knowledge of how the machine worked. Whether any of them were encouraged by this to delve deeper into the machine's archictecture is open to debate, as the story of synthesis seems to be one of more and more user‑accessible parameters being accessed by fewer and fewer users. It seems to have taken the return to popularity of simple analogue synths to encourage people back into knob‑tweaking for themselves.

Inside The VFX

Korg's T2 workstation.Korg's T2 workstation.

Those intrepid users who did venture into the structure of the individual VFX patch were rewarded with a voice structure laden with possibilities. The VFX's voice architecture was actually more reminiscent of that of the D50 than that of the M1 (which had tended to sound OK with one or two base samples, so Korg's engineers hadn't needed to develop so complex a voice architecture). The VFX, however, now allowed up to six voice components instead of the four components (Partials) of the Roland machine. Not all of these had to be used, and many sounds used only one or two source waveforms, especially as the VFX did not split individual sounds into attack and loop segments. Using all six components meant that you could create some of the most complex, evolving sounds ever possible on a synthesizer, especially as the multi‑stage envelopes available on the VFX could be used to control the level of each component individually. As a result, the VFX could produce sounds of such complexity that they made the D50 sound like an old two‑oscillator synth. If three patches were layered together, up to 18 oscillators could be triggered from a single key. However, that many oscillators piled together can be rather overpowering in everyday sounds (not to mention the fact that they exhausted the synth's polyphony very quickly).

Yamaha's SY99 took things a stage further, by adding a disk drive and the ability to load samples into RAM. This meant that users could actually take their own samples and combine them with FM sounds.

Within each component of a sound the possibilities were even more complex. In another development reminiscent of the D50, a series of waveforms could be strung together to create a loop, and these were no longer fixed by the manufacturer as on the D50; users could now specify the starting and finishing source waveforms for their loops. They were, however, restricted to using them in an order defined by Ensoniq, so the best results still came from the serendipitous sequence of waveforms at the design stage. However, it was possible to add or take away waveforms from the beginning or the end of a sequence, or even to move to another part of the waveform ROM completely, giving a lot more control to the user than the D50's fixed loops did. Playback rate was, of course, still fixed, but at least the rhythmic patterns created could be changed to some considerable extent.

As Ensoniq had also taken multi‑stage envelopes further than before, with multiple rates, levels and loop points, the potential for creating sounds of unprecedented complexity was great (as were the chances of getting completely confused and giving up). However, in the hands of Ensoniq's creative team of developers astounding results were achieved, some almost qualifying as pieces of music in their own right. A selection of breathtaking programs shipped with the VFX, some still evolving and bringing in new components a couple of minutes after being triggered. This only led to one problem: how should you use them in a track? It was back to the old story of intros and quiet middle sections (where the synthesist in a band has too often been banished before). Problems with sync'ing rhythmic elements to the tempo of the song still existed, so keyboard players stuck to the basic pads and imitative instruments unless they were doing film, TV or ambient music where such restraints are less common. It would be left to Korg, a few years later, to solve this problem and allow complex changing timbres with rhythmic elements to be sync'ed to the tempo of the song.

Performance Controls

To make the VFX's six available components within the voice more versatile for performance (and to prevent enthusiastic programmers exhausting the polyphony too quickly), Ensoniq used an expanded set of real‑time controllers to bring different voices in and out of play. So, instead of combining together voices with radically different elements, it was possible to group together voices which were very similar but with slight variations. Additional switches provided the means to mute and un‑mute different components. For example, a flute patch might have a straight flute sound as one of its components, one with extra breath noise as its second, a third with a 'flutter‑tongue' effect, and so on. Switching these in and out in different combinations meant that the user could circumvent one of the biggest problems with sample‑based synthesizers, the fact that the source PCM waveform cannot be altered in real‑time.

This was an important development in PCM‑based synthesizers. It has been taken on by many other manufacturers, whether in keyboards which feature expanded performance controllers, like Korg's most recent PCM‑based machine, the Trinity, or Emu's range of modules featuring extensive modulation routings, which allow the user to make the most of the standard MIDI Continuous Controller inputs. A raw PCM sample can only ever be a 'snapshot' of an instrument at one moment in time, played in one way. Only by mixing between different snapshots in real time can any sense of the motion and change that is part of the nature of real instruments be conveyed by a PCM‑based synth.

The VFX also improved on the amount of multi‑channel MIDI access that a user could have to PCM‑based synthesis. Twelve patches could be set up very quickly to be controlled on any MIDI channels, with similar key range, transposition and controller setups to the performance mode. As computer‑based sequencing became more and more important to the majority of users, it was a major plus that 12 of the VFX's programs could be used simultaneously on different MIDI channels. Hence, the Multi was born (as opposed to the Combi, a Korg invention). In general, these days, PCM‑based keyboards tend to feature both a Combi‑derived mode for performance (it may even be called Performance Mode) and a Multi‑derived mode for internal or external sequencing (usually accessed by a switch marked 'Seq' or 'Multi'). The jargon may vary from manufacturer but the two are usually easily distinguished, as one gives quick access to a lesser number of sounds, while the other gives much more complete access to at least 16 different timbres (several manufacturers now have schemes, especially on PCM‑based modules, for 32 different timbres to be used simultaneously).

For under £1000, Emu's Proteus gave the user access to many of the sounds which had only previously been attainable from a £2000 sampler.

Ensoniq & Effects

Of course, the old problem of different effects being an integral part of sounds and not being available simultaneously reared its head when MIDI was used to sequence multiple programs. Whilst Ensoniq didn't have the solution to this (which can only be the expense of the extra hardware to provide a separate effect circuit for each multitimbral voice, as on the Korg Trinity), its Copy routines did allow the most important effect in any combination of programs to be quickly copied to the effects buss of the Multi section. This meant that the procedure described last month when I talked about the M1's effects (for copying the effect of the sound most reliant on it and assigning some of the other sounds to it as appropriate) could not only be carried out on the VFX, but it could also be achieved more quickly. It is noticeable that nowadays most PCM‑based synths allow this procedure; the main difference between synths is whether it is made easier or more difficult by the architecture of the machine in question.

This capability was equally useful on the VFX‑SD, which had the necessary on‑board sequencer and floppy disk drive to qualify it as a workstation. Now the chances were increased that the user might be trying to create his/her entire piece on the one instrument, it was even more important that the ability to share the two simultaneously available effects between up to 12 musical parts was as easy and flexible as possible. In addition, the sequencer could record any changes to this shared effects capability.

Although they have not made such a giant leap forward again since the release of the VFX, Ensoniq have continued to include the originally unique aspects of the VFX design in their subsequent synthesizers. After the VFX‑SD came the TS10 and its 76‑note keyboard variant, the TS12. These had even more complete sequencers and so took over the 'top‑of‑the‑range' workstation mantle. They also added the ability to load samples from the Ensoniq library as source samples, allowing the user to customise the basic set of sounds in the machine. (One of the principal problems with PCM‑based synths is that if you don't have a kazoo sample in the basic waveform ROM, you are unlikely ever to get a kazoo sound out of the machine.) For the smaller budget, Ensoniq also produced the SQ1 and SQ2, which came with a less powerful sequencer and synthesis architecture. Those who are interested in the combination of multiple sound sources within a single program, however, will still find the same ability in these machines.

EMU Exploit Their Sound Library

Around this time there was a major change in direction from Emu Systems, who, until this point, had concentrated almost exclusively on samplers. Although Emu's samplers had much of the subtractive synthesis architecture of the machines we have already looked at here (in fact, the Emulator II was the first sampler which included a full complement of filters, envelopes and other subtractive standards), they could not really be called synthesizers because all source waveforms had to be loaded from disk into RAM. It should be noted, though, that Emu samplers come much closer to the traditional synthesizer than those from Akai or even Roland.

Emu had been making a big investment in sound sampling for their machines for more than 10 years by this stage, and they suddenly realised they had a marketable asset (outside the library disks which they been selling or bundling for their end users). They decided that if they were to design a small rackmount module which had the synthesis capabilities of their samplers, but with lots of short samples pre‑loaded into masked ROM, they too could join the PCM synthesizer revolution.

Thus was Proteus born, named after the Greek god who could change his shape at will (a reference to the number of different instrument multisamples in ROM, which the user could instantly switch between). For under £1000, it gave the user access to many of the sounds (albeit in much shorter samples and loops) which had only previously been attainable from a £2000 sampler, and then only after waiting minutes for the sample data to load from floppy or hard disk. The result was an overnight success, and thousands of these modules are still in use.

But the real appeal of Proteus for those interested in synthesis was not the instantaneous availablility of quality Emu sounds but the very real synthesizer architecture of the machine. It inherited all the standard Emu synthesis capabilities, with proper filtering, enveloping, and a modulation routing system to die for.

As computer‑based sequencing became more and more important to the majority of users, it was a major plus that 12 of the Ensoniq VFX's programs could be used simultaneously on different MIDI channels.

Not only could Proteus change its shape from program to program, but also from machine to machine. Emu hit on the idea of selling their extensive library piecemeal, divided into categories for the needs of different users — so whereas the original Proteus gave you a wide sweep of sounds for general use, subsequent machines became more targeted to specific music styles. The first of these was Proteus 2 Orchestral, a big hit with film and TV composers who, even if they didn't use it in their final mixes, found it invaluable for composing and trying out arrangements and orchestrations. Proteus 3 World satisfied a growing demand for ethnic samples after the influence of Peter Gabriel's Real World label started to make itself felt in the crossover markets, and the Procussion gave drummers and synthesists the same editable access to a huge drum library. The real payoff, though, came with the Vintage and Classic Keys models (by now the numbering system had been abandoned). These offered sampled synthesizer waveforms from classic synths of yore, which could be properly filtered, modulated and enveloped through a real synthesizer voice circuit. If you want an example of how Samples & Synthesis can be a really creative tool, take a look at these two modules.

The Proteus heritage still continues today in the UltraProteus (a sort of greatest hits with some extra filtering capabilities from Emu's own Morpheus, which we will look at next time) not to mention the increasingly bizarrely named modules (Orbit, Planet Phatt and Carnaval) which court the dance market. Other manufacturers have clearly learnt the lessons of marketing specific sound sets at different target groups (Roland and Akai in particular), but for me the joy of Proteus remains in the synthesis rather than the sample side, which is why those machines loaded with synth waveform samples give the synthesist the greatest creative potential!

Korg's Response

After the success of the M1 and its rack relations, the M1R and M3R, the next series of synths from Korg, the T1, T2 and T3, addressed the growing criticism the M1 had started to suffer for the compressed nature of its more percussive instruments, such as piano and guitar. Korg did this by taking advantage of the continually dropping price of ROM and allocating twice as much memory to the storage of the PCM samples. This meant that the sample could be longer before the loop needed to start on sounds which decayed, and the loops could also be longer, if required, on sustained timbres. As a result, the piano and guitar sounds gained much more natural decays and so could be used for a wider range of music styles, rather than just the fast repetitive triggering of dance music, where subtlety was not required. The T‑series appealed much more to the performer, and this was why the flagship T1 had a full‑range weighted keyboard, to allow traditional pianists to feel more at home with it.

Of course, as Winston Churchill remarked, you can't please all of the people all of the time, and many people in the emerging field of dance music complained that the T‑series didn't sound like the M1, and so didn't use them. The reason was simple. The compressed nature of the sounds in the M1 made them ideal for the no‑holds barred sound of dance music, where everything needed to cut through and be louder than everything else. Pre‑compressed sounds such as the M1 pianos and guitars were ideal, especially if the user didn't have the money — or the awareness of the need — for a separate compressor. As a result, Korg have made the orginal M1 samples available several times in more recent products to court the dance market (the X5DR and Trinity PBS options, as we shall see later).

Another key feature of the T‑series was that the user could load his or her own samples into the machine for processing through the instrument's filtering and enveloping. Like the Ensoniq TS‑series, this gave the user a way around the main limitation of PCM‑based synthesis — that if the waveform ROM does not contain a multisample approximating to the sound you need, you'll be hard‑pressed to drag said sound out of the machine. Now Korg users could at least expand and customise the source waveforms with samples to take care of their less mainstream needs. On the T1 the sample RAM to do this came as standard, whereas on the T2 and T3 it was an optional extra.

The 0/W series added more refinements; the major step forward was the doubling of polyphony to 32 notes. The module version, the successful 05R/W, also helped develop the PCM‑based multitimbral module into a commonplace item in any setup. One of the most important improvements that arrived with the 05R/W was the implementation of effects via a send amount system. This allowed you to remove effects from sounds like the bassline (which often didn't need them, especially if the effect was reverb) by setting the effects send amount on your bass sound to zero.

The story of synthesis seems to be one of more and more user‑accessible parameters being accessed by fewer and fewer users.

Next came the X‑series, which made the concept more accessible at the low‑budget end and gave more user control. The module versions, the X5D and X5DR, added some of the original sounds from the M1 because these had become so important for certain styles of dance music that they became a major selling point (despite being less authentic than the more recent versions of pianos and organs which used more generous allocations of RAM).

The M1 tradition culminated at the beginning of 1996 with the Korg Trinity, still using PCM samples as its main source of sounds (now at 48kHz sample rate, allowing this synth to be used in an all digital systems with the addition of S/PDIF or ADAT interfaces), but with the addition of separate DSP circuits to finally get around the problem of changing effects when it was being used as a multitimbral instrument. The Trinity comes in various versions, but nevertheless still uses the same fundamental technology that Korg introduced back in 1988 with the M1.

Apart From PCM...

Clearly, other companies have made extremely good use of PCM technology in their synthesizers (including newcomers to synthesis Alesis, as well as Kawai, Roland, and Akai), but it is the quality and type of the source samples rather than the innovation of their synthesis archictecture which makes them useful. However, the mid‑'90s saw several developments in synthesis, including Emu's Morpheus and Korg's Wavestation, and we'll focus on these transitive types of synthesis next time, as well as looking at some of the precedecessors, such as the PPG Wave and the Prophet VS, which inspired them.

Yamaha Join The PCM Party

So great was the success of Yamaha's FM (Frequency Modulation) synthesis in the early '80s that the company spent most of that decade 'trickling down' the technology into cheaper and cheaper synthesizers. As a result, they were the last of the 'big names' to introduce PCM‑based technology into their synthesizers, and it was initially a supplement to their FM technology, not a replacement for it. This meant that the sounds which FM had proved very good at producing — electric pianos, tuned percussion and woodwind — could still be provided by the FM circuitry, but the sounds which were better produced from PCM samples, like acoustic piano, strings and other fuller sounds, could be generated using the more recent technology.

However, the reason why the SY77 (the first machine to combine FM and S&S) proved so popular with its more professional users was not the ability to produce sounds with one or other of the two complementary technologies, but the ability to produce hybrid sounds from the combination of the two. PCM's biggest weakness was still the difficulty of adding expression to the performance. The basic sample would sound very authentic but somewhat static. FM was the perfect antidote for this, as it has always been very responsive and expressive, although not the most authentic way of reproducing the fundamental timbre of sounds.

Of course, the two technologies did not necessarily always sit together well in the mix. Fortunately, the SY77 also had the necessary DSP hardware to produce effects. This meant that the same 'smearing' techniques as used on the Roland D50 (see December '97's Synth School) could be used to bind the two sounds together. The only difference was that, instead of using the effects to join together an attack with the sustain portion of the sound, both the FM and PCM parts would sound simultaneously but, being of different characters, they would stand out from each other. The effects would be used to blend the two sounds together.

Of course, many of the programs would only use one sound or the other, so in this case the effects would be used just to add reverb or chorus, in the normal way. The SY77 also had the ability to sequence multitimbrally, but the effects had to be shared between all the different programs being triggered. It therefore allowed you to use the effect/s from whichever program seemed to need it most and then assign the other sounds to those effects where appropriate.

The SY99 took things a stage further, by adding a disk drive and the ability to load samples into RAM. This meant that users could actually take their own samples and combine them with the FM sounds. At the time, however, there was not much general cross‑platform support for reading other manufacturers' disks, so unless you wanted to contend with MIDI sample dump, your main option was sample disks for Yamaha's own TX16W sampler (a 12‑bit machine which had been able to do stereo samples but only at 33kHz, whereas mono ones were available at 48kHz). One of the interesting side‑effects of this was that some use was finally made of a TX16W sample library which the old Yamaha R&D Centre in London's Conduit Street had given me a splendid budget to create. Yamaha distributors all over the world finally had a use for the piles of beautifully bound disk sets for the rather overlooked sampler, which had been in stock for years. For me personally, it meant the ability to load all these great sounds, which no‑one else had ever given me a budget to record and edit properly, eaxctly as I had set them up, but in a synthesizer (without two hours wasted on MIDI sample dump and basic editing, which tends to de‑motivate me seriously!). This isn't the only reason why the SY99 is my favourite Yamaha synth of all time, but it certainly goes a long way towards it. The SY99 certainly is unique in allowing you to combine FM, subtractive synthesis, and your own user samples all within a single machine.

Even when Yamaha eventually dropped the FM capability and produced their first purely PCM‑based synthesizer, the SY85, they still kept the ability to load user samples into RAM, but using the much cheaper method of SIMMs (Single In‑Line Memory Modules), just like modern samplers. As a result, a colleague (now the manager) at a certain retail store I used to manage was able to use his library of dance loops and drum samples to cheaply customise that machine for the evolving dance market. The rack version of this synth, the TG500, actually went a stage further, using Flash ROM to store samples so that they were retained in memory after power‑down (I suspect because it had no disk drive to quickly reload from, so MIDI Sample Dump was the only way to get them in). Korg are now offering this as an option for the Trinity in the form of the PBS‑TRI option, and it really expands the usefulness of a PCM‑based synth, especially for live applications, where loading samples even from hard disk is altogether too long a procedure. Nothing beats turning a machine on and finding your own personalised samples already to go.

Yamaha's most recent PCM‑based synths unfortunately no longer have the ability to load samples into RAM or Flash ROM. Despite the fact that they retain the DSP effects capabilities of all the previous Yamaha workstations and the ability to sequence multitimbrally, they're not as exciting, for me, as the glorious hybrids of the late '80s and early '90s. The really notable and innovative products from Yamaha at the moment are in the physical modelling arena.