You are here

Multitimbrality: A Brief History

Exploration By Paul Wiffen
Published June 1994

Although much more widely used than ever before, the multitimbral ability of electronic instruments to produce several different sounds at the same time is still not always fully exploited. Paul Wiffen traces the evolution of this powerful feature.

When, as a Sequential sales rep back in 1984, I opened the manual for the about‑to‑be‑released SixTrak synth and my eyes alighted on the word multitimbral for the very first time, little did I realise what a major concept I had encountered. Sure, my schoolboy Latin enabled me to work out fairly quickly what the actual word meant — 'capable of producing several different sounds at the same time' — but I can't claim to have had a prophetic vision of just what it would mean to the development of MIDI, which was still very much in its infancy at the time. But it didn't take more than a couple of minutes messing about with the prototype to realise that, in terms of sequencing, the multitimbral synth was probably the most powerful development in electronic musical instruments since the polyphonic keyboard. More than any other feature, multitimbrality has been central to the development of MIDI sequencing, permitting multiple tracks with different voicings from relatively modest setups.

But before we go on to discuss the role of multitimbrality in MIDI sequencing, exactly what does it mean and how did it come about?

What Is 'Multitimbrality'?

Traditional instruments produce one basic sound, regardless of how many notes are sounded simultaneously; a piano produces the sound of a piano, a violin that of a violin, and so on. There are obviously major variations within this sound, depending on how forcefully the instrument is played or what additional factors are in use (pedals, mutes, etc), but in general these can still be regarded as different shades of the same basic tonal colour. The word timbre (pronounced tam‑brer) is often used to describe this overall character of an instrument; the word 'sound' is used in so many contexts that it becomes confusing.

A few traditional instruments, like the Church Organ, can change their timbre so dramatically that they might be regarded as an exception to the above rule, but for the most part a conventional musical instrument produces only one timbre. Just for the time being, let us refer to this as monotimbral

The first polyphonic synthesizers (preset or programmable) made a major advance on this in that they were able to completely change timbre at the touch of a button (or more accurately, a couple of dozen knobs). From one moment to the next the instrument could change character completely, to the point where each of these timbres could be thought of as different instruments. However, within our terms of reference, nothing has changed; these devices were still monotimbral, for even though they could make chameleon‑like changes from one instrument sound to another, at any given moment they could still only produce one timbre at a time.

Keyboard players began to imagine the possibility of having two different timbres sounding simultaneously, either layered together or perhaps a different one under each hand on the keyboard. This was not a huge conceptual leap, as most keyboard players were already treating their two hands as separate instruments, one filling the same role as a bass player, the other as a chordal instrument or soloist. This was also fairly easy to achieve electronically, as all that was required was the ability to split the available number of voices within the synth into two and assign a different sound program or timbre to each group of voices. This facility, often referred to as Split/Layer, soon became fairly common, especially in the more expensive professional instruments. With the benefit of hindsight, we might refer to such instruments as bitimbral or having dual timbrality.

The Sequencer Influence

Since a Split/Layer combination was the most that any but the nimblest keyboard player could hope to control when playing live, things did not advance any further until the newly evolving digital technology brought in sequencers which were capable of producing more than 16‑note monophonic cycles. Once it was possible to run more than two parts at a time from a sequencer, then people began to think how useful it would be if all these parts could have different timbres. Of course, before MIDI came along, it was only instruments with built‑in (or specifically designed add‑on) sequencers which were able to do this.

The first of these truly multitimbral instruments (although the term had yet to be coined) was the now legendary Fairlight CMI. Interestingly, the CMI's (now) standard‑type polyphonic sequencer was not able to work multitimbrally, but its phenomenally successful Page R was. Just as many people used the Fairlight for the compositional tool that Page R was as did for its sampling capabilities, and it continued to be heavily used long after the CMI Series II's sampling fidelity was surpassed. Each of Page R's eight monophonic voices could be loaded with a different sound (normally samples and often percussion sounds — Page R was a Rhythm Sequencer, after all) and then sequenced independently.

The easiest way to comprehend how a multitimbral instrument works is to actually think of it as a number of different instruments all packaged in the same box. Each of these instruments is free to operate in the same way as a traditional instrument, with its own sound playing its own set of notes (ie. voices). When you visualise a multitimbral instrument in this way, you can see why they represent such excellent value for money (and how just one can really flesh out a MIDI sequencing setup). In the case of instruments with no multitimbral capability, voices often lie unused much of the time. With a multitimbral instrument there is very little chance of voices being unused most of the time, since multiple parts increase the demand for voices in direct proportion to their number. In this way, the full potential of an instrument is exploited at all times.

The only problem with multitimbral instruments (apart from the fact that you soon discover that there is no such thing as 'too much polyphony') is making sure that each 'virtual' instrument, to use the current hip (or do I mean hype) jargon, knows which notes it is supposed to play with its individually‑assigned timbre. As we saw in the case of the Fairlight CMI above, the built‑in sequencer was the first solution to this, as manufacturers could develop their own transparent assignment systems which did not tax the player's comprehension too much. We will see later how MIDI came to address this problem and thrived as a result.

'Thank You' Sequential

The real breakthrough (as well as the term multitimbral) came with the launch of the Sequential SixTrak, which not only opened up the world of synthesis to multitimbrality, but also reduced the entry price into the sub‑£1000 category. Ironically, the multitimbral capability of the SixTrak came about more as a combination of chance circumstance than as a deliberate strategy. Sequential had been commissioned by an American fast food chain (rejoicing in the name of Chukky Cheese) to produce a complete synthesizer on a chip for their own brand video games division (presumably you would play a Chukky Cheese video game whilst waiting for your Chukky Cheeseburger or whatever it was).

Sequential spent a fortune developing this chip, but Chukky Cheese went bust without paying anything, so Sequential looked around for something they could make with the thousands of single‑oscillator chips which they were now stuck with. When they totted up the cost of putting six of them together in a box with a keyboard, they found that they could produce a 6‑voice polyphonic synth for less than any of their current Prophet range. The fact that each of the SixTrak's voices was actually a self‑contained synthesizer (remember our way of looking at a multitimbral device earlier), and so could be set to play with a different timbre, was almost a side issue — until someone in Sequential's marketing department coined the term 'multitimbral' and decided this was the way to promote the new synth. An internal real‑time sequencer was added (MIDI sequencers were not yet available in any numbers) which allowed the SixTrak to demonstrate its own multitimbrality for all the world to hear (pre‑empting today's built‑in demo tunes by several years).

Although a major step forward in the world of multitimbrality, the SixTrak and its descendants (the non‑programmable MaxTrak and the velocity‑sensitive MultiTrak) failed to make a huge impression on the market, as the basic 'sound' of the synth chip was a little thin compared to previous Sequential synths (although you could get a big, fat, lead sound by 'stacking' all six voices with different patches in a special multitimbral version of Unison mode) and the onboard real‑time sequencer was a little on the primitive side.

Mass Acceptance

The first multitimbral synth to really take the market by storm was the Casio CZ101 (and the full‑sized key version, the CZ1000), although in truth this was more to do with the £249 price tag for a digital synth than with its multitimbral capabilities. Indeed, many owners must have used the CZ101 for years without knowing of its multitimbral potential. One reason for this, of course, was that the CZ101 had no internal sequencer (it was the first multitimbral instrument without one). This meant that you had to have a MIDI sequencer and know what 'Mono Mode' was and how to activate it on the CZ. All this conspired to hide the CZ's multitimbral light under a bushel.

Nevertheless, certain more clued‑up users really went to town with this feature. The most well‑known case was Vince Clarke, who bought half a dozen Casios to use with his UMI sequencer on the BBC Micro. Each CZ101 was 4‑part multitimbral, so this gave him 24 monophonic parts (the monophonic nature of each part was actually ideal for Vince's style of pop music).

Monophonic Limitations

There were two reasons why all early multitimbral devices operated monophonically. The first was that in analogue synths, or digital instruments that still used analogue filtering, it was difficult to implement dynamic allocation of sounds to voices. When using the instrument multitimbrally, this meant that a voice had to be permanently assigned to each timbre ('hard‑wired' to use old‑fashioned terminology). As a result, to get the maximum number of multitimbral parts, each part was restricted to being monophonic (ie. playing only one note per part at a time). This was how the Oberheim Xpander and Matrix 12 worked, but don't let that make you think any the less of them. Anyone who ever heard the powerful voices of these beasts in multitimbral operation will tell you there was nothing missing from that sound (they certainly didn't suffer from the SixTrak's thinness). Indeed, good analogue sounds fill so much of the audible frequency range that it is no bad thing to limit the number of voices each part is playing.

The second reason for early monophonic operation was to do with the original MIDI Specification. Because the aforementioned hardware limitation was still very prevalent when MIDI was first proposed, Mono Mode — ie. the original MIDI Mode, designed to allow multitimbral operation (and, incidentally, the use of individual pitch bend and other controllers for each voice) — was, as its name suggests, specified to only expect monophonic operation on each incoming MIDI channel. As a result, even those machines which could have dynamically allocated voices to the different multitimbral parts didn't do so, simply because the MIDI Spec specified monophonic operation.

The Sequential Prophet 2000 sampler, the next important multitimbral instrument, was a case in point. Samplers are ideal for multitimbral operation as they are already dynamically allocating voices to the different zones (or keygroups) within a multi‑sample. There was no reason why the 2000 couldn't have responded polyphonically to multiple incoming MIDI channels (it was already covering 16 channels with eight voices, so it had to use dynamic allocation), but the MIDI Spec restricted multi‑channel response to monophonic operation, so that was how Sequential made it work.

It is worth remembering that, at this time, there was a lot of paranoia about manufacturers departing from the official MIDI Spec and so Sequential (being one of the founding fathers of MIDI) were obviously intent on setting a good example and following the Spec to the letter.

However, this didn't stop the Prophet 2000 from being the first sampler that really could be used multitimbrally to good effect in conjunction with a MIDI sequencer. Nor did the fact that each channel could only have one sample on it, and that there was only one pitch bend and LFO amount for the entire machine. At that time, multitimbral operation with a sequencer was such a breakthrough anyway that those in the know were prepared to accept an incomplete implementation of it.

The Akai S900, following hard on the heels of the Prophet 2000, allowed each keygroup to be individually assigned to a MIDI channel, instead of switching between two different operational modes. This meant that the sampler could respond polyphonically to each MIDI channel and that each MIDI channel could have several keygroups (and therefore multi‑samples) on it — this represented two major advances over the 2000. The major drawback was that all this had to be set up within one program, which made it time‑consuming and difficult to keep track of. As a result, only a small percentage of the many people who bought an S900 ever got around to using its multitimbral capabilities!

Major Leap Forward

Manufacturers eventually realised that it was not enough to simply make multitimbral operation possible via MIDI, it also had to be quick and simple to use. The first sampler to achieve this was the Akai S1000, which allowed previously created programs to be simultaneously active and quickly set to respond to different MIDI channels. This was a major leap forward as setting up the machine to respond to multiple incoming MIDI channels took seconds, not minutes or even hours as previously. Again, as in the case of the S900, the implementation took a non‑standard form, in that the machine was not switched between Poly Mode (where only one channel is responded to) and Mono Mode (where each voice responds to its own MIDI channel).

Amazingly, it took much longer for a similarly user‑friendly multitimbral setup to appear on non‑sampling instruments. Although the Roland MT32 was a versatile multitimbral module, it was not the easiest device to set up to work in this way. Perhaps the most elegant multitimbral solution to date was that which arrived with the Emu Proteus...

Despite a minimum of user controls, Proteus makes it very simple to scroll through each MIDI channel, assigning the timbre you want to each channel as you go. Provided the Proteus is switched to Multi Mode (which is buried about half‑way through the Master Menu parameters), then its design actually encourages multitimbral operation, unlike previous synth and sample players which seemed to set it as a challenge for the expert user. It's interesting to note that when Emu produced the keyboard version of Proteus, the MPS, they allocated Multi Mode a button and LED all to itself on the front panel.

Multi VS Mono Mode

At this stage it is worth differentiating between Multi and Mono Modes (the latter originally being referred to as MIDI Mode 4). As we noted earlier, the original MIDI Spec only allowed for monophonic operation in multitimbral working, but this was more to do with hardware limitations at the time of MIDI's introduction than anything else. It didn't take manufacturers very long to realise that they could simply have their dynamically allocating instruments respond to as many notes on each MIDI channel as they had voices free at any given moment. Although not strictly Mono Mode, several manufacturers followed this scheme for quite a while.

Then companies like Ensoniq and Emu started adding a Multi Mode to the normal MIDI mode options of Omni, Poly and Mono. In effect Multi Mode set the machine up to respond exactly like Mono Mode, except that polyphony was fully supported on each channel. Eventually, this concept of Multi Mode was fully absorbed into the official MIDI Spec, leaving Mono Mode to return to its original one‑voice‑per‑channel specification for use with MIDI guitars (which, for the most part, need the same timbre assigned to each string but independent pitch and expression control for each). These days most multitimbral devices use Multi to denote multitimbral operation (with the exception of high‑end samplers from the likes of Akai and Roland, which are permanently in a Multi‑style mode of operation).

16 Channels And Beyond...

It also took a long while before program changes on different MIDI channels could be accepted. Most modern instruments now support operation on all 16 MIDI channels, although some manufacturers stuck with only allowing six or eight parts until quite recently. For instance, it is only with the introduction of the JV family in the last year that Roland mid‑priced keyboards and modules have finally been able to respond to all 16 MIDI channels at once.

Of course, nowadays it is not unusual to find that 16 MIDI channels is simply not enough! Clearly, a multitimbral instrument can only respond to more than 16 MIDI channels if it has two MIDI Ins (a feature I believe we will see increasingly appearing on the new generation of 64‑voice instruments, which are ideally suited to control by 32 MIDI channels), but there are often ways to get around this. On the Proteus modules, for example, if you are only using the lower register of the MIDI range for a bass line on one channel, you can link another preset to the bass sound and split the keyboard range between them. Similar tricks are available on other manufacturers' machines.

It is difficult to offer general advice on how to set up an instrument for multitimbral operation (because virtually every manufacturer implements it differently on different products!), but keep your eyes open for new or secondhand instruments with multitimbral ability, as you may be getting up to 16 instruments for the price of one!

"It didn't take more than a couple of minutes to realise that, in terms of sequencing, the multitimbral synth was probably the most powerful development in electronic musical instruments since the polyphonic keyboard.""Manufacturers eventually realised that it was not enough to simply make multitimbral operation possible via MIDI, it also had to be quick and simple to use."

Multitimbral Shopping Tips

If the economy of producing lots of different sounds from one keyboard or module appeals to you (and it should!), here are some brief guidelines to finding a unit, new or secondhand, which will give you the maximum for your money.

  • Find out how many part multitimbral it is, ie. how many MIDI channels it can be set to respond to simultaneously. Chances are with new products that they will respond on all 16 MIDI channels, but if the person trying to sell it to you (be they shop salesman or private vendor) does not know, or gives a glib 'all MIDI channels' answer which they can't illustrate, make a point of finding out from another source (for example, ringing the manufacturer), otherwise you may find your creative juices drying up when you get the instrument home and try adding that seventh or ninth part to your sequence.
  • Check that the unit has sufficient diversity of timbres to warrant multitimbral operation. If you don't have a wide enough palette of sounds then there may be little point in having full 16 MIDI channel operation. (This doesn't mean that a limited multitimbral operation within the context of a larger setup might not be wonderful.) If you have percussion sounds, so much the better as you can produce entire pieces from one module.
  • Check that the unit has sufficient voices to play back the number of tracks you want it to, especially if the allocation of voices to MIDI channels is fixed. Obviously, if the voices are dynamically allocated instead, they can be stretched much further — provided you don't want multiple voices on all channels all the time.
  • Check that the unit sounds good. Multitimbral operation is all very well, but multiplying timbres which sound thin or unpleasant on their own probably won't produce a very exciting result, unless you have them all play in unison (like the SixTrak's Stack mode).

Making The Most Of Multitimbrality

Once you have acquired a multitimbral instrument, there are many ways to ensure you exploit its full potential. Here are a few pointers:

  • If, when you have 'finished' adding tracks to your musical masterpiece, you have voices or unused channels left over, try doubling timbres by copying or 'ghosting' tracks on your sequencer or by setting two parts on your module to respond to the same MIDI channel.
  • Make sure that 'pad' parts (strings, organ chords, ambient sounds etc) are not using up too many voices unnecessarily and thus starving your more economic, rhythmic elements (which release voices much more quickly for reallocation to other parts).
  • Use MIDI program changes to extend the number of apparent parts. For instance, the percussive figure in a verse could alternate with a chordal part in the chorus by judicious insertion of program changes.
  • Use linking of programs (ie. keyboard splits) to run a bass line together with a middle and/or high register track on the same MIDI channel, leaving other MIDI channels free for parts with less restricted ranges.
  • On aesthetic as well as practical grounds, make sure each part is adding something to the whole, rather than just adding tracks because you can.

Some Multitimbral Milestones

The approximate original launch price of each instrument is shown, with estimated current secondhand values given in brackets.

  • 1982 Fairlight CMI 8 fixed monophonic parts £25,000 (£1500)
  • 1984 Sequential SixTrak 6 fixed monophonic parts £799 (£150)
  • 1985 Casio CZ101 4 fixed monophonic parts £250 (£50)
  • 1985 Sequential Prophet 2000 16 dynamically‑assigned mono parts £2000 (£400)
  • 1986 Oberheim Xpander/Matrix 12 6/12 fixed (huge‑sounding) analogue parts £2000/£4000 (£1000/£2000)
  • 1986 Akai S900 16 dynamically‑assigned polyphonic parts £1899 (£500)
  • 1989 Emu Proteus 16 dynamically‑assigned polyphonic parts £999 (£500)
  • 1990 Cheetah MS6 6 analogue parts £299 (£200)

Sampling And Multitimbrality

Sampling lends itself to multitimbral operation, as each sample within a multisampled instrument is in itself a different timbre.

Multisampling was originally developed to allow different pitches of the same instrument to be sampled, but it wasn't long before people realised that there was nothing preventing them from sampling completely different sounds and placing them at different points on the keyboard.

The first examples of this were probably multisamples containing the different instruments that typically make up a drum kit (the strictest definition of 'multitimbral' would of course encompass all drum machines right back to the original LinnDrum), but examples of sample disks for instruments like the Emulator II started to appear, featuring all sorts of combinations of different instruments set across the keyboard. The internal sequencer of the EII, primitive though it was, allowed full multitimbral operation of the machine by assigning different presets to each track, but combining sounds into a single preset was the only way to operate the EII multitimbrally via MIDI.