MIDI: The Protocol That Never Broke
Abstract
MIDI, the Musical Instrument Digital Interface, was worked out in 1981 and 1982 by Dave Smith of the Californian synthesizer maker Sequential Circuits and Ikutaro Kakehashi of Roland in Japan, with engineers from Yamaha, Korg, and Kawai. It was first shown in public in January 1983, when a Sequential Prophet-600 and a Roland Jupiter-6 were connected by a cable and one played the other. The specification stayed at version 1.0 for thirty-seven years. MIDI 2.0, adopted in 2020, begins every connection by asking whether the other device understands it, and speaks 1.0 if there is no answer.
Islands
A synthesizer of 1980 could usually be played from its own keyboard and nothing else. The older analog instruments could be linked by control voltages, one cable per note and parameter, but manufacturers used different voltage scales. The new microprocessor-controlled instruments had digital interfaces, and every company designed its own: Roland had its Digital Control Bus, Oberheim a parallel bus. A musician who owned a Roland sequencer and a Sequential synthesizer owned two machines that could not be connected.
The makers of electronic instruments were small companies competing for a small market. Kakehashi, the founder and president of Roland, thought the incompatibility was holding the whole market back.
Kakehashi, Oberheim, and Smith
In June 1981 Kakehashi raised the idea of a common standard with Tom Oberheim, whose own interface he considered too cumbersome. Oberheim brought in Dave Smith, the founder of Sequential Circuits and designer of the Prophet-5. Smith talked to the American manufacturers; Kakehashi talked to Yamaha, Korg, and Kawai. Representatives met in October 1981. That autumn Smith and the Sequential engineer Chet Wood presented a paper at the Audio Engineering Society convention proposing a “Universal Synthesizer Interface”, a simple serial link that drew on Roland’s bus.
Most American companies stayed out. The standard was finished between Sequential and the four Japanese firms, by fax and on trips across the Pacific. Kakehashi wanted to call it the Universal Musical Interface. Smith thought that “a little corny” and proposed Musical Instrument Digital Interface.
Robert Moog, the best-known name in synthesizers, announced the coming standard in his column in the October 1982 issue of Keyboard magazine. He was not among its designers.
What It Is
MIDI carries no sound. It carries a description of a performance: this key went down this hard, this key came up, this pedal was pressed, switch to program 12. A message is one to three bytes. The receiving instrument turns the messages into sound with whatever sound it has, which is why the same recording of a performance can be played back as a piano, an organ, or a drum kit.
The hardware was chosen to be cheap. The link is a one-way serial connection at 31,250 bits per second, a rate that falls out of dividing a one-megahertz clock by 32, over a five-pin DIN plug of the kind already used in European hi-fi equipment. Each input is electrically isolated by an optocoupler, so that connecting two instruments cannot create a ground loop and a hum. There are sixteen channels on a cable, so one keyboard can address sixteen instruments. Adding the interface to a synthesizer that already had a microprocessor cost a few dollars in parts.
There was no licence fee and no royalty. Smith’s reasoning was that MIDI was worth nothing unless every manufacturer adopted it: “we had to give it away.”
January 1983
At the Winter NAMM trade show in January 1983, Smith connected a Sequential Prophet-600 to a Roland Jupiter-6 with a MIDI cable. Notes played on one sounded on the other, across two instruments built by different companies on different continents. The specification was published in August 1983.
Within two years nearly every new synthesizer, drum machine, and sequencer had the sockets. Yamaha’s DX7, launched in 1983, had them (see John Chowning and FM Synthesis). The Atari ST of 1985 had MIDI ports built into the computer and became the standard studio sequencer for a decade. A keyboard player could stack the sounds of several instruments under one set of keys; a composer could record a performance into a computer, correct the wrong notes, and change the instrument afterwards. General MIDI, in 1991, fixed a common list of 128 instrument sounds, and the Standard MIDI File became the format of early web-page music, mobile ringtones, and karaoke machines.
Why It Lasted
The specification of 1983 was not elegant, and its limits were known within a few years. Values such as key velocity have 7 bits, 128 steps. The data rate is low enough that dense passages can audibly smear. Pitch is assumed to come in twelve equal semitones. There were proposals to replace it throughout the 1990s and 2000s, and none got anywhere, for the reason that had made MIDI succeed in the first place: every instrument already had it, and a studio’s equipment is bought over decades. A 1983 Prophet-600 will still respond to a controller built this year.
The physical transport changed underneath. MIDI messages now travel mostly over USB and Bluetooth, not DIN cables, and inside a computer between programs, with the same byte format.
Smith and Kakehashi shared a Technical Grammy Award in 2013. Kakehashi died in 2017 and Smith in 2022.
MIDI 2.0
Work on a successor was announced in January 2019 and adopted at the Winter NAMM show in January 2020, thirty-seven years after the Prophet and the Jupiter. MIDI 2.0 has finer resolution for every controller, control of individual notes, and, for the first time, two-way communication: devices can ask each other what they are and what they can do.
The designers made translation to and from MIDI 1.0 a first priority. A MIDI 2.0 device opens with a capability inquiry. If the other end answers, the two agree on the new protocol. If nothing comes back, the device assumes it is talking to an instrument from the first thirty-seven years and uses MIDI 1.0. The new standard is built on top of the old one and replaces none of it.
📚 Sources
- The MIDI Association: “Details about MIDI 2.0, MIDI-CI, Profiles and Property Exchange” (MIDI 2.0 layered on MIDI 1.0, capability inquiry, translation as a design priority, two-way communication, per-note control)
- Smith, D., and Wood, C.: “The ‘USI’, or Universal Synthesizer Interface”, AES Convention 70, paper 1845 (October 1981) (record only; the paper is behind the AES paywall)
- Wikipedia: MIDI (finding aid: control voltages and the proprietary buses, Kakehashi’s approach to Oberheim in June 1981, the October 1981 meeting, the naming, Moog’s announcement in Keyboard of October 1982, Winter NAMM 1983, specification in August 1983, General MIDI 1991, Technical Grammy 2013, January 2019 and 2020, 31,250 baud, the DIN connector, “we had to give it away”)