Zipi: Is This The End Of MIDI?

Is MIDI about to be superceded? Martin Russ investigates proposals for a new standard for interfacing and controlling electronic musical instruments.

In a text file posted to the MIT Press ftp server on September 8, 1994, the Editor of Computer Music Journal (CMJ), Steven Travis Pope, mentioned a folder containing six articles which were to be included in the Winter 1994 issue (18:4) of CMJ. Text and Postscript versions of the articles were provided, with the text files occupying about 170 kilobytes, and the Postscript versions containing diagrams too. Eager Internet‑surfing musicians, myself included, downloaded the six articles and proceeded to wade through the 27,000 words of text and diagrams, which amounted to almost 80 A4 pages on my printer. Something major was occurring here!

So what is all this about? You may already have heard rumours about ZIPI being the replacement for MIDI, or that it finally makes guitar synthesizers usable, or that it has something to do with resynthesis and physical modelling. Whilst I would recommend that you read the Winter 1994 issue of the CMJ, not everyone is going to be able to, so this article aims to provide an overview on information which was current in September 1994, and which may be superceded by any revisions in the published CMJ articles.

Zipi

Nowhere in the CMJ documentation does it state what ZIPI stands for, only that it has resulted from a collaboration between Zeta Music Systems (makers of MIDI violins, guitar synths etc) and the Centre for New Music and Audio Technology (CNMAT), based at the University of California, Berkeley. The 'Z' probably refers to Zeta — the rest is unknown. Perhaps it is being played down because of the strong connection with one manufacturer's name?

Tech Spec

Physically, ZIPI uses either a 7‑pin DIN plug or an 8‑pin mini‑DIN (for use with laptop computers or perhaps PC plug‑in cards). It employs current loop interfaces, but information travels in both directions within the same cable; thus only one ZIPI socket is needed on a piece of equipment. The minimum transmission rate is 250 kBaud, although it can run faster than that. ZIPI is designed around the OSI layered model, and so can be described in a series of layers: the lower layers deal with the nitty‑gritty of hardware and bit transport, whilst the middle layers describe various application‑level protocols; one of these is the Music Parameter Description Language (MPDL) — the rough equivalent of the MIDI messages format. MPDL describes how musical information is packaged up for sending around the ZIPI network. Unlike MIDI, ZIPI instruments are connected in a ring, although in practice this will be implemented as a central 'hub' with the cables to the instruments radiating out from it.

MPDL provides a rich source of control for musical purposes. For example, it allows you to directly 'address' individual notes, within 'instruments', within families, and it breaks the link between a note and its pitch. Whereas in MIDI it can be awkward to deal with individual notes which change pitch with time, ZIPI's MPDL allows you to have two different 'note numbers' — the note number identifies the note, not its pitch. ZIPI's roots lie in the frustration that many 'alternative controller' manufacturers have with MIDI's keyboard‑centric design. Guitar and wind controllers can generate huge amounts of information, and MIDI is really not well suited to coping with it. ZIPI is designed with small overheads for controller information (in fact, for almost everything) and so can have large numbers of parameters devoted to providing real‑time control.

When MIDI was first proposed over 12 years ago, it was at the limits of available technology in terms of UART speed and processor capability; ZIPI is designed with future proofing in terms of the speed (only a minimum speed is defined), and it can be expanded to suit advances in networking technology. For example, it should be possible to carry ZIPI MPDL data packets over networks using FireWire (1394 serial bus), Ethernet, ATM, Iso‑Ethernet and many others. The complexity of the MPDL and the other application‑layer protocols is greater than MIDI, but should not present too many problems to today's fast and sophisticated processors. With the benefit of hindsight learned from MIDI's example, ZIPI has a strong base which should provide a firm footing for future enhancement.

Replacing MIDI?

So with a simplified socket, similar connector, and at least 10 times faster transmission speed, is ZIPI a replacement for MIDI? In my opinion, this seems to be true for some applications but not for others. The obvious use is with guitar synthesizers, where the individual pitch and volume parameters from the six strings can generate large amounts of controller information. For monophonic instruments, ZIPI provides extra control over the triggering of notes so you can change pitch without starting a new envelope (to give legato phrasing, for example), something which is almost impossible to achieve easily with MIDI. For physical modelling instruments, like Yamaha's VL1/VL7, ZIPI would provide even more control over the sound, and this is probably how the G‑WIZ resynthesis work described in SOS January 1995 will be controlled.

For non‑keyboard applications, ZIPI seems to offer exactly what MIDI has never really delivered: enough bandwidth for expressive and effective control of the sound. For keyboard applications, there is more of an uphill struggle. Whereas guitar controllers have never really been popular (because of the problems of control via MIDI), keyboard controllers have become the major source of control for synthesizers — something which many synthesists consider has reduced synthesizers to a role of replacing conventional keyboard instruments, rather than opening up their true sonic potential.

ZIPI is thus faced with the prospect of being successful in finally providing guitarists, wind players, and other non‑keyboard instrumentalists with an excellent way of interfacing to a synthesizer module. But at the same time, ZIPI may have to overcome the preconceptions of a large established user base who are probably well satisfied with what MIDI offers for the majority of their work. The people who are dissatisfied with MIDI may well turn out to be the serious academic researchers and professional musicians, and their relatively small numbers (compared to the General MIDI, 'fun' keyboard and computer musicians) may mean that ZIPI for keyboard use becomes a niche market.

The scenario therefore might go something like this: ZIPI becomes a runaway success for 'alternative controller' applications like guitar and wind controllers, perhaps even drums — and deservedly so. It is also adopted by the leading edge academic musical researchers for producing music without the limitations resulting from MIDI. But for mainstream interfacing between electronic musical instruments and computers, MIDI may well retain a strong hold for a long time to come, meaning that ZIPI and MIDI will have to co‑exist.

ZIPI's roots lie in the frustration that many 'alternative controller' manufacturers have with MIDI's keyboard‑centric design.

Overcoming the limitations of MIDI may well become the battle cry of equipment designers — especially if physical modelling and resynthesis techniques become more prevalent. Luckily the transition from MIDI to ZIPI is relatively straightforward, since ZIPI offers more bandwidth and facilities than MIDI, although converting from ZIPI back to MIDI may be accompanied by a loss of information due to the mismatch of bandwidth and control. Even so, I do not see MIDI losing its dominance very quickly, since it is perfectly adequate for most musicians and dabblers, and as the recent resurgence of interest in classic 'analogue' instruments has shown, for some purposes lack of precision and unsophisticated control are still useful. So the next few years should see a gradual mixing together of MIDI and ZIPI‑equipped instruments, with the problems of inter‑working slowly being resolved.

Adoption

For ZIPI to succeed at all, of course, it needs to be implemented by manufacturers and adopted by musicians. So far, only G‑WIZ's FAR resynthesis system [see SOS January 1995, p.22] and some computer‑based sample playback systems are offering ZIPI interfaces. Since one of the major influences on manufacturers is their customers, if you and lots of other people ask for a ZIPI interface, then it may well eventually materialise. Certainly ZIPI seems like a well thought out solution to the guitar and wind instrument control problem, and is appearing at just the right time — with several new guitar and instrument synthesis approaches in the pipeline.

The ZIPI Development Group have already presented their proposals at the 1993 and 1994 NAMM shows. The articles in Computer Music Journal represent a well thought out and powerful set of ideas for a far‑reaching new electronic musical instrument interface. Expect a flurry of developments in the next year or so — the key indicators will be the response of the Japanese companies to ZIPI. Personally, I wish the ZIPI group every success; I just wonder why they never contacted me for advice on what's wrong with MIDI.