MIDI instruments first appeared in early 1983 and represented the product of a collaboration unique in an industry best known for its jealous isolationism and protectionism.The germ of the idea came at a meeting in late 1981 between representatives from Roland, Yamaha, Korg, Kawai, Oberheim and Sequential. The intention was to sever once and for all the dependence on analogue control systems and create a ubiquitous digital interface between synthesisers and other instruments from all the major manufacturers. While digital interfaces had existed previously, they had almost invariably been specific to a particular manufacturer if not to a specific instrument.
It had become apparent that if the market for electronic instruments were to be successfully exploited, there was a need for the adoption of common standards across the entire industry. The last days of voltage control has produced a broad if inexact consensus of a voltage range of 1 volt per octave but there were by then too many exceptions to this “rule” to regard it as being even the basis for a new standard.
The emergence of digital control and memory, however, made the creation of a standard far simpler. Components or at least families of components used by the various manufacturers were often of the same family if not actually alike and hence would accept much the same control commands. These commands took the form of digital messages not unlike those used to convey information within and between computers. The computer industry had long since evolved a set of common standards and communications protocols so the intention was to adapt these for use within and between instruments.
Two basic approaches were possible: serial or parallel. Parallel interfaces were desirable because they offered huge data capacity and high transfer rates but, unfortunately, they were known to be considerably more expensive to implement than the relatively slower serial systems. Early proposals envisaged two standards - fast parallel system for professional users and a slower serial one for “domestic” applications - but this approach was quickly dismissed as being unacceptably divisive. Shortly thereafter, Sequential proposed a compromise system using a (relatively) high speed serial interface called USI (Universal Synthesiser Interface). Early in 1982, Roland responded with a similar system called UMII (Universal Musical Instrument Interface) and by the end of the year, the consortium had agreed the basic specification for what was to be called MIDI (Musical Instrument Digital Interface). Remarkably, this specification remains in use (albeit with certain additions) to the present time.
Not everyone, however was convinced:
“The only thing I won’t give in on is this stupid 5-pin plug thing. I can’t stand it. To force people to go out and buy a piece of shit connector that they can’t use for anything else in their whole rig is insane. It’s just totally insane.”
Carmine Bonnano President of Octave-Plateau Electronics.
“I think that MIDI has been brought on because of market pressures, not because the de facto standard is technically exquisite”
Tom Rhea Director of Marketing, Moog Music
Notwithstanding the remarkable consensus that had led to the evolution of the MIDI standard, concerns remained about its quality and, in particular, its hardware implementation which, being based on domestic connectors, was felt by many to be insufficiently robust for professional use. Other doubts were raised about cable lengths, signal distortions and so on but, nonetheless, the standard demonstrably worked, even between machines from different manufacturers.
In the summer of 1983, nearly two years before the formal publication of the detailed MIDI Standard, Yamaha were able to launch a seminal product which heralded the advent of the age of digital music making: the DX-7. Although not the first MIDI-equipped synthesiser, it was the first fully to embrace digital systems, using a new synthesis system - frequency modulation - to provide radically new sounds and textures and combining this with a computer-style user interface that featured a small LCD display and a single assignable data entry slider.
The DX-7 proved immensely successful despite the difficulty involved in programming sounds using its minimalist interface: most users simply used the 32 presets preprogrammed into its memory or purchased additional sounds on plugin data cartridges. So difficult did the DX-7 prove to be to programme that the vast majority of machines returned for repair proved, on examination, to have the original 32 presets still intact !
Fortunately, it was possible to control virtually all parameters of the DX-7 remotely via the MIDI interface and this, in turn had been derived fairly directly from computer interface technology. Hence it proved to be relatively easy to interface a MIDI instrument directly with a computer via a simple adaptor and the idea of a more user-friendly on-screen editor was born. At much the same time, a number of software developers began to look at the possibility of using MIDI information stored on computers as the basis for software sequencers. Until the advent of MIDI, sequencers had been dedicated hardware systems that could only be used in one way. The advent of the (relatively) affordable personal computer meant that potential sequencer users could elect to work with any one of a number of sequencers or editors but, here again, lack of compatibility - this time between computer operating systems - proved to be problematic.
Since there seemed little possibility of a short-term resolution to the computer wars, expansion of MIDI software was restricted until the emergence of the computer that became the de facto standard for music work for many years: the Atari 520ST. Succeeded by the 1040ST, the Atari appeared in 1986 and to the delight of music programmers and users, it was fitted with its own inbuilt MIDI interface and furthermore, since it was originally intended for games use (the name “Atari” apparently translates as “I am going to attack you”), it featured a small, robust operating system and, best of all, a user-friendly graphic interface.
The universality and success of the MIDI standard was viewed as surprising in the industry since the specification finally adopted was relatively slow (by virtue of its serial nature) and used componentry that was specified for its cheapness rather than suitability. Thus it was guaranteed entry to the fast expanding home keyboard market but, more significantly, the possibility of fairly direct interaction with true computers (as opposed to dedicated hardware sequencers) began to be considered. Where the relatively clumsy closed system offered by the Fairlight Page R had led, others soon followed.
In 1983, Karl Steinberg and Manfred Rurup developed a computer based sequencer that ran on a Commodore 64 microcomputer and communicated with instruments via the new MIDI system, using a simple serial interface. The uniquely attractive aspect of their programme was its one-page graphic interface. Joel Chadabe quotes Rurup thus:
“The funny thing is that because he (Steinberg) wasn’t a trained programmer, he couldn’t do the multiple page approach that was common so he put it all on one page. And the one-page graphical interface which nobody else had at the time was a breakthrough”
By contrast to the impenetrably difficult interfaces offered by other sequencer designers, the Steinberg approach proved immediately attractive to users and with the advent of affordable microcomputers such as the AtariST, the prospect of creating a virtual studio from a relatively small number of relatively cheap devices became a reality.
Oddly enough, the Atari had a fundamental component in common with its ostensible polar opposite in the music technology hierarchy. It shared the same processor - the 6800 - as the monstrously expensive and prestigious Fairlight CMI (Computer Musical Instrument) although it must be admitted that the CMI actually boasted two such processors. With the emergence of the Atari, user friendly software such as Steinberg’s Pro-24 sequencer and the universal adoption of the MIDI Standard by virtually all hardware and software developers and manufacturers, the possibility of creating a “virtual” studio which had little if any dependence on tape recorders became a reality. Exactly how the whole thing could be made to work in practice was, however, quite a different matter.
The basic idea of the tapeless studio represented a substantial conceptual leap for the recording industry. Hitherto, the choice had been to record “live” or to overdub successive instruments or voices. Both approaches had their pros and cons but one factor was inescapable: however good the quality of the tape and tape machine, the act of recording something engendered a degradation of audio quality, usually too slight to matter but sometimes leading to unacceptable levels of background noise and/or distortions. This problem was addressed in part by the introduction of noise reduction systems but even these were not without their particular set of demerits (eg need for constant and accurate alignment with consequent difficulties in moving material from one system to another). With the development of MIDI systems and synthesisers capable of producing high quality instrumental sounds, it became possible to at least partially circumvent the conventional programme chain of source to multitrack tape followed by multitrack tape to stereo master and hence to reduce the adverse effects of the recording process. The idea was simply to use a MIDI sequencer as if it were a multitrack recorder, playing in the various instrumental parts one at a time onto separate “tracks” and then, when all were completed, running the sequencer in playback mode with the various synthesisers, drum machines etc feeding into the mixing desk and then to the master stereo recorder thereby avoiding one generation of recording with its attendant problems.
This approach, while convenient and, above all, cost effective had two main limitations: it did not allow for other than live vocal or acoustic instrument additions (during the mixdown process) and the original MIDI specification did not allow for synchronisation to external devices: indeed it no way of recognising or responding to any temporal information at all. It was, however, already possible to synchronise tape machines and mixers by means of SMPTE timecode, a system borrowed from the film and television industries but MIDI devices had no way to respond to what they perceived as a wholly foreign language. It was therefore necessary to devise a means of converting SMPTE values to a form that sequencers etc could recognise. An addition was therefore made to the original MIDI specification in the form of MIDI Time Code, a format derived from incoming SMPTE code that could drive a sequencer directly rather than using the internal clock of the host computer.
The only requirements to make MTC a workable system were a relatively simple converter and minor modifications to sequencing software. In the case of the Atari, the converter was a matchbox-sized unit connected to a spare serial port and other manufacturers were quick to follow with the development of freestanding units. With this simple development, it became possible to synchronise a MIDI system with a tape machine and, by implication with any similarly controlled studio equipment such as automated mixing desks.
The ideal of the tapeless studio had been somewhat compromised but at least multiple tracks of vocals and “real” instruments could now be added to the outputs of synthesisers and drum machines despite the fact that the sound quality of the former remained compromised by the shortcomings of the recording process. MIDI continued to develop and diversify at a steady rate: for example it became possible to employ computers as sample editors by downloading actual audio over a MIDI connection and using on-screen editors not dissimilar to those created for the more intransigent synthesisers such as the DX-7. The MIDI Standard as first written has been little changed in the intervening years although there have been numerous additions to enhance its diversity and usefulness. These additions have enabled it to remain one of the few universal standards in use in the audio industry despite the fact that it was originally based on a series of compromises that many thought unacceptable.
That there is a connection between MIDI and computer systems is indisputable: both use much the same form of communication and both make heavy use of hexadecimal arithmetic. In MIDI, the entire message content is hexadecimal and the majority of early computer forays into audio and music required that they be programmed in assembler languages which often make use of similar arithmetic. There is, however, a fundamental difference: MIDI is bounded and constrained by the MIDI Standard to be a control and communications protocol. It cannot, by itself, process information “creatively” whereas, for an appropriately programmed computer, the world is very much its oyster.
The manifestations of the computer in audio and music have been many and diverse. It has appeared as composer, synthesiser, sampler, instrumentalist, improviser, singer, conductor, copyist, mixing engineer, recordist, editor, audio processor, environment creator, spatialiser and even studio manager (early models of the Fairlight CMI included software applications that would administer client databases and studio bookings). In the diversity of its employment, one can perhaps detect a parallell with its influence in the visual media where it has come to be regarded as far more than a convenient tool. Rather it has become an integral part of the creative process, exploiting its unique qualities to create images that would otherwise be unimaginable. So it has been in audio too and, just as computer graphics was initially a painfully slow, low-resolution activity, computer audio as it is now, has developed from the laboured pontifications of the corporate mainframe.
It is perhaps difficult to imagine the clumsiness of such systems from a turn-of-the-century perspective where high speed computation can exist cheaply on any desktop. The first stored programme computers evolved during World War Two based upon the seminal work of mathematician John von Neumann. In much the same way that the development of the tape recorder was spurred on by its application as a propaganda tool, the computer made possible the (relatively) rapid computation of artillery ballistics and, later, the operational parameters of the atomic bomb.
The major problem that dogged the computer was interfacing and interaction. The entire power of a large corporate mainframe system would not have been anywhere near sufficient to maintain the graphic user display on a modern laptop machine so the idea of point-and-click was clearly not viable. Instead, the user would submit the data to be computed in conventional form. From this data, operators would create a stack of punched cards or a roll of perforated paper tape. This could then be read into the computer, processed and output in similar form. The outputted cards or tapes would then be processed in reverse, creating text or numerical printout on an electric typewriter. This would then be returned to the client. On a typical system, the turnround time would a matter of days for a straightforward task, stretching to weeks if specialist processing was required.
The early involvement of computers in audio required just such processing since the machines available at the time had neither audio inputs nor outputs so the conversion from numerical output to audio output on tape required a second, dedicated computer to process the raw data from the first.
Programming too was problematic since few high-level languages suitable for audio work existed and most computers used manufacturer-specific low level languages. The implication of this was that programming had perforce to be undertaken in assembly languages or at the even lower level of machine code, making the whole process horribly laborious.
These difficulties were perhaps not fully anticipated when the first forays into computer music were made in the late 1950s. Working for the Bell Telephone Company, Max Matthews became interested in the possibility of using computers to produce musical sounds. Chadabe quotes him thus:
“It was immediately apparent that once we could get sound out of a computer, we could write programs to play music on the computer. That interested me a great deal. The computer was an unlimited instrument and every sound that could be heard could be made this way. And the other thing was that I liked music. I had played the violin for a long time.”
This interest led Matthews to begin the development of a long series of programmes designed to make computers produce sound, starting with Music 1 in 1957 and culminating in Music 11 in 1973. Matthews was by no means the sole author of these programmes. Major contributors included Jean-Claude Risset, Richard Moore, Joan Miller, Barry Vercoe and later, John Chowning whose work led to the development of FM synthesis, later licensed to Yamaha for their DX series of synthesisers. Part way through this period, the move was made to higher level languages such as FORTRAN which made the programmes portable inasmuch as they could now run on a range of mainframe systems. This, together with the availability of new “minicomputers” such as the DEC PDP series helped dramatically to increase the numbers of computer music practitioners which came to include Peter Zinovieff of EMS. Although improvements were continually made, the system remained relatively clumsy: erstwhile users of Music V, which was released in 1968 would receive the programme on 3,500 punched cards with a letter saying, “Good luck”.
Development work continued at Bell in collaboration with Stanford and Princeton Universities but by the early 1970s competition had begun to emerge from Europe in the form of the French government-sponsored Institut de Recherche et Coordination Acoustique Musique (IRCAM). Working initially with the Music series of programmes, IRCAM researchers went on to develop their own approaches, often characterised by their identification with vocal sounds and processes, the best known being CHANT, released in 1978.
Such programmes still required the power (by then not inconsiderable) of mainframes and minicomputers. Such microcomputers as existed were hopelessly underpowered for audio work and even when expanded (as in the Alpha Syntauri system for the Apple 2 micro) remained extremely limited and of relatively poor audio quality. A need emerged to either develop dedicated hardware or to find a way of improving the performance of existing systems. This became a significant fork on the road with the evolution of the large, predominantly hardware-based system (eg the Fairlight or its main rival, the NED Synclavier) on the one hand and the largely software based approach (eg Steinberg, the Composers Desktop Project) on the other. This latter approach was initially predominantly based on the newly released Atari computer which was later supplanted by the Apple Macintosh (released in 1984) and clones of the ubiquitous IBM PC.
Meanwhile, the commercial audio industry had embraced computers with generally less enthusiasm. The expansion of tape track numbers and mixing desk complexity had made a degree of automation almost a necessity. Many desk manufacturers developed their own automation systems, perhaps the most famous early system being NECAM produced by Neve. The fashion for working in many different studios made compatibility between systems an important issue. Unfortunately this was never resolved (and remains so to this day) with the consequence that major studios use desks from one of a very small number of manufacturers and, in so doing, are forced to accept that their client base is often restricted as a result: a client who has begun work on a project in an SSL-equipped studio is likely to continue it in other SSL-equipped studios since a Neve (or, for that matter, any other) system cannot read the data that it generates.
In recent years, the relatively recent developments of affordable, high quality analogue/digital and digital/analogue interfaces and the cheapness of memory chips and disk drives have triggered a massive explosion of so-called “audio workstations”. Stemming from the relatively crude 2 track editing systems of the 1980s, these systems - usually based on an off-the-shelf computer - have now reached a level of sophistication that allows them to subsume virtually all the major studio operations: tracklaying, overdubbing, effects processing, mixdown and final mastering. Advantageously, all these processes take place in the digital domain and are largely undoable if circumstances demand. Furthermore, software designers have shown themselves able accurately to reproduce the characteristics of established hardware processors, bringing a whole range of “traditional” qualities to what is otherwise often a very clinical approach. Thus it is possible, for example, to process a sound in such a way as to simulate the effect of valve circuitry, creating the “warmth” so prized by fashion-conscious clients.
With some notable exceptions, the split between dedicated hardware and software-based systems has remained. Heavy industrial users such as broadcasting organisations have traditionally opted for hardware systems, citing operating speed and reliability as their main reasons whereas the more creative and experimental end of the industry has embraced the PC-based system for its flexibility and cheapness. Since software is relatively cheap when purchased legitimately and readily available free of charge when pirated, it is a simple matter to expand and enhance any such system on a regular basis whereas upgrades to dedicated systems tend to be expensive and relatively infrequent, being tied to a single supplier and system.
One may fairly assert that there is now almost no commercially created and released recording that does not make at least some use of computers and their related systems: computer music has become a mainstream activity.
As Chadabe puts it:
“(Indeed), one could say that by the late 1980s, the age of computer music was over because everything was computer music”