The Future of Synthesis
With the summer shows out of the way and the forthcoming new products all unveiled, Martin Russ looks into the crystal ball and plots out his version of where the world of synthesis might be heading...
Having crossed his palm with silver, MIDI consultant Martin Russ was more than happy to indulge in a spot of crystal ball gazing for us...
The release in early 1987 of the DX7 MkII by Yamaha has only part fulfilled the promise of the rumours about the likely successor to the justly famous DX7. Lots of other very interesting developments in the field of synthesis are also at the speculative stage. The next few years promise yet another turn around of some well established ideas. In this article, I intend to present some of my thoughts and ideas on the future of music synthesis, with particular emphasis on keyboard instruments.
Digital synthesis has rapidly taken over from analogue synthesis as the common method of producing sound. Yamaha were the first to take the plunge into completely digital synthesis by exploiting John Chowning's FM patents, resulting in the famous X-series synthesizers. (Casio's Phase Distortion (PD) system is really only FM by another name...) This left the other manufacturers (Roland, Korg, Sequential, Ensoniq, etc) with only the option of using hybrid digital and analogue technology. Digital oscillators, initially producing the simple 'classic' waveforms (square, sawtooth, etc), were teamed with analogue filters to produce products like the Roland Juno and JX series, and the more advanced Korg DW range.
Purely digital synthesis has been slow in arriving, mainly because of the cost in both RAM and software of implementing any other digital synthesis method other than the non-linear modulation methods of FM and PD. (Phase Distortion is essentially FM, but with non-sine waves available as sources in addition to the standard sine waves and a different method of controlling the parameters. Additionally, Casio's choice of 8-bit quantisation, although cheap, fast and easy to implement in their LSI technology, is rather noisy and coarse in application and in stark contrast to the 16-bit clean, quiet and professional sound of Yamaha's FM - especially in the DX7 MkII). The new Roland method of Linear Arithmetic synthesis is an example of a different approach to synthesis - it uses software simulation of a conventional analogue monosynth combined with PCM waveshape playback and, therefore, exploits VLSI technology and software to produce a very complicated product at very low cost.
The recent lapse of the FM patents has resulted in a rush of new releases using this technology. The Elka EK44 was the first multitimbral FM synth to offer six operators, and the new Korg DS8 uses simplified controls over FM and is the first result of the sharing of research facilities by Yamaha and Korg. Significantly, Yamaha have made no attempt to simplify the programming of FM in the DX7 Mk II, although several third party software packages are being evolved to simplify the user interface to FM.
Sampling does not really fall into the synthesis arena at the moment, since it is really nothing more than the modern day replacement for the Mellotron, but using solid state memory rather than magnetic tape. The original sampling machines like the Fairlight Series I and Emulator I have been overtaken by units like the Synclavier, the Ensoniq Mirage and Sequential Prophet 2000. Apart from Akai, who nevertheless had to rely on the design skills of David Cockerell (former EMS designer) for their S900 sampler, the Japanese manufacturers have been more reticent in their response to sampling. Roland's S-10 and S-50 (in their original form) are pretty unadventurous in their use or control of sampling, but offer useful sampling capability at a reasonable price. Also noteworthy is the Korg DSS-1, which seems to go over the top in its approach, resulting in a large, complex and expensive unit which may overwhelm the prospective user.
The latest US offerings in the sampling arena are the Emax - an adventurous addition in the low-to-mid price region from E-mu Systems, the manufacturers of the pioneering Emulator I - and the new Ensoniq Mirage replacement, the DSK-1. Casio have recently entered the market with the FZ-1, a 16-bit machine that is slightly amazing in its specifications and its price.
Noticeably absent from sampling keyboards have been Oberheim (with currently only a very clever replay-only machine in their portfolio - the DPX1) and Yamaha, with the SPX90 being their only means of playing back sampled ('digitally frozen') material. Sampling really is fashionable at the moment, with the result that this article will probably be well out of date by the time you read it - but the overall trends will be the same.
The future of sampling is relatively easy to predict: longer, higher quality samples will follow the price of RAM. In other words, as RAM gets cheaper so you will get more multi-samples for longer times, with better signal bandwidths and more bits. The natural limit to sampling will probably be at about CD quality, so you can expect things to bottom out at 16-bit, 44.1 kHz sampling in a year or two.
At this sort of sampling rate, storing information becomes a major problem - already floppy disks are beginning to look too small for some of the multi-sampling instruments. Both E-mu and Akai have recently introduced a hard disk option to their Emax and S900 samplers and this is probably the way forward in the short term. Unfortunately, you need to backup the contents of the hard disk onto another storage medium - usually floppy disk. The long term solution is probably hiding in the wings, of which more later...
None of these sampling keyboards have anything significant in the way of digital signal processing on-board, preferring to provide analogue filters and envelope shapers for post-processing. Limited digital signal processing is available on some products by using an external personal computer, notably the Apple Macintosh for use with the Prophet 2000/2002. The Atari ST seems to be rapidly gaining ground as the standard musician's computer, and some very powerful sampling packages are rumoured to be imminent...
Roland's LA synthesis technique lends itself to signal processing in the digital domain and by replacing the PCM wavetable ROMs with RAM, a formidable combination could be formed. The real future of sampling probably lies with a combination of sampling and processing, to enable editing and re-synthesis of sounds.
Since the clever parts of LA synthesis exist in software anyway, the processing possibilities which could be offered would probably only be limited by the need for real-time or rapid processing - no half-hour FFT analysis allowed here! Most people's awareness of computers does not generally include any knowledge of the limits of real-time processing, and so any digital signal processing will need to be applicable in real time, or very impressive in its aural effect if it is slower than real time. This tends to limit the sort of processing you can carry out on standard samples but, as is often the case with synthesis, it is the limitations which spur on greater creativity - so expect some 'off the wall' sampling data formats which facilitate signal processing.
A simple starting point would be to store some parameters based on the spectrum of the sound rather than its time waveform, which would aid processing no end. Digital Signal Processing chips are rapidly accelerating in their speed and processing power and the very fast FFTs (Fast Fourier Transforms) and inverse FFTs which will be needed may just be around the corner.
Returning for a moment to Yamaha's current 'no sampling' situation - why no sampler yet? Perhaps the RX5 and HX20 give some hints as to their ultimate intentions, although both currently use ROM-based samples.
At the time of writing, there were still only vague hints of the RX5 sampling add-on, perhaps using the MIDI Sample Transfer format to move external samples into its editing RAM. Since Yamaha currently have no sampler, it would be probable that such provision will not be announced (perhaps as a ROM update for the RX5) until a sampler is announced.
You would then use the sampler to sample sounds for use in the drum machine. It seems sensible to only pay out for one expensive sampling unit rather than buy several disguised as sampling keyboards, drum machines, etc...
I would not be surprised to see addons from other manufacturers appearing in the next year or so, to enable user-sampling for non-sampling drum machines... Akai's new MPC60 'Akai by Roger Linn' drum machine offers an RX5 look-alike (spec-wise) but with built-in sampling, much as the Sequential Studio 440 offers sampling and drums with a sequencer thrown in as well!
The editing capability of the RAM-stored sounds in the Yamaha RX5 shows the sort of control which will become increasingly important in the future - wide ranges of pitch variation and user definable envelopes should soon be the order of the day in all digital drum machines. More complicated processing should help to remove some of the need for ever larger libraries of 'new and better' sounds and should enable the user to tweak the sounds to his or her own taste.
AWM has elements of sampling and Roland SAS-type processing implied in its title, and this probably means that postprocessing of the AWM sounds should be versatile and relatively easy. The processing involved in the DMP7 digital mixer points the way to the current Yamaha expertise in signal processing, as well as their knowledge of analogue-to-digital conversion. Korg's DS8 synth uses clever software to make FM voice creation and editing easier, showing the way forward.
There is a great deal of interest in using digital signal processing technology in musical products and the next few years should see many instruments whose performance is considerably enhanced by internal software and processing. As with the RX5, simple envelope and pitch editing should be the first facilities to appear, followed by more complicated time-based processing and equalisation. Eventually, you should be able to effect major changes to the overall timbre and its evolution.
Increasing activity by the major Japanese companies in the field of DAT (Digital Audio Tape) recorders could provide another pointer to future products. Yamaha, Casio and Technics, as well as the more traditional hi-fi manufacturers like Sony, Aiwa, Sharp and JVC, all have working DAT players. DAT uses a similar technology to the latest generation of video recorders, especially the Video 8 system, and uses a cassette which is slightly smaller than the conventional audio casssette (see photo).
DAT recorders are already on sale in Japan [HHB Hire & Sales plan to be selling the Sony model in the UK by September/ October we are told - Ed], DAT is suitable not only for hi-fi audio use but, more interestingly, also as a mass data storage medium, with much cheaper potential costings and higher reliability than traditional Winchester hard disks (used on the Synclavier).
Samplers currently use the 3.5-inch or 5¼-inch floppy disk format as their main storage medium, though E-mu and Akai have announced a hard disk add-on for their Emax and S900 machines, respectively. The problem with floppy disks is that they can only store somewhere in the region of 1 Mbyte of data (more bits can be obtained, but their integrity over longterm use is very suspect). As I mentioned before, multi-sampling, increasing sample rates and larger numbers of bits will all increase the requirement for data storage beyond the capacity of a floppy disk. DAT would enable rapid storage and retrieval of very large amounts of sample data, so it could be that Yamaha intend to leapfrog the current 'limited by disk capacity' samplers by releasing a DAT-based sampler, with no intermediate disk-based system. Makes sense, doesn't it?
DAT cassette tapes have the capacity to store many gigabytes of information, and so long sample times, extensive multisampling and other techniques currently considered impracticable, could be easily implemented in such a DAT sampler.
DAT also has interesting implications for AWM and LA. Both synthesis methods use ROM-based samples of real sounds, perhaps with some pre-processing and some time-based parameters, and both currently require large amounts of ROM. ROM is historically cheaper than RAM, and is likely to stay so in the future, and so changing sounds stored as AWM samples by using RAM to replace the ROM is likely to be very difficult using floppy disks.
DAT's massive storage potential, however, offers an excellent means of storing and loading AWM or LA samples. Once in RAM, rather than the current ROM, the samples can then be processed as well as played back. Depending on the amount of pre-processing of the samples, the possibilities of further processing will either be similar to those practicable on conventional samples, or a new set of operations may arise.
Digital Signal Processing uses lots of computer power and expertise in DSP technologies. It would be surprising if Yamaha were not applying their computer skills to producing an IBM PC compatible computer, as a replacement for the MSX computers. The CX5M seems to be fading in popularity for new purchases, but a fair number are still being used in homes and schools and will probably be used for some years to come. More likely, the IBM PC compatible would be an addition aimed at the higher end of the business market and professional music area. Roland already produce PC-based software systems designed for composition, transcription and control of MIDI-based equipment, although they do not currently produce or even 'badge-engineer' a PC look-alike. Thus, future Roland and Yamaha products in the software field can be expected to be PC compatible, and interfaces to MIDI and other means of controlling Yamaha products can be expected too. The combination of an FM/AWM synthesizer, perhaps with sampling options for the AWM, combined with a PC for signal processing would be a powerful product indeed. Perhaps this is the real replacement for the DX7 (and the DX7 II) and we are seeing only the result of DAT's current 'just appearing' situation.
Casio started out as a calculator manufacturer and their range has extended as far as small BASIC computers/calculators with a few K of RAM. It would be quite a leap for Casio to jump to a PC lookalike, but it pays to expect the unexpected from the Japanese. Akai, meanwhile, seem to have abandoned their music computer plans for the moment.
The world dominance of the IBM PC and its compatibles (about 80% of personal computers are PC compatible, with about 10% Apple Macintosh and the rest sharing the other 10% between them) means that you can expect to see only PC compatibles being introduced in the near future. Strangely enough, the professional and hobbyist music markets both currently seem to be showing a strong preference for the Apple Macintosh and Atari ST computers, probably because of their undeniable strengths in high resolution graphics and user-friendly operating systems.
I would be very surprised to see any major manufacturers releasing software for the Mac or the ST, mainly because the third party companies like Mark Of The Unicorn, Steinberg, Dr.T, Hybrid Arts and Opcode are doing such a good job. There is currently a flood of MIDI-based software for both the Macintosh and the Atari, with the Atari range very quickly coming up to the sort of quality which has, until now, been the exclusive province of the Mac.
This article has presented some of my personal views on what is happening and what will happen in the hi-tech music market, together with some thoughts on what some companies are probably planning. I have also included a few suggestions as to possible future products and approaches to improving existing product lines. As with all such 'crystal ball gazing', I am probably wrong in quite a few respects. The best bet is to expect the reality to be a lot better, and expect it here a lot quicker than I have described.