Skip to content

John Chowning and FM Synthesis

Abstract

John Chowning (born 1934) is a composer who, in 1967, while experimenting with vibrato on the computer of the Stanford Artificial Intelligence Laboratory, found that modulating the frequency of one sine wave with another at audio rates produces rich, controllable timbres from almost no computation. Stanford patented the method and licensed it to Yamaha, which turned it into the DX7 synthesizer of 1983 and the sound chips of the first PC sound cards. The licence earned Stanford more than $20 million and paid for its computer music centre. In the middle of all this, Stanford let Chowning go.

A Percussionist at the AI Lab

Chowning was born on 22 August 1934 in Salem, New Jersey. He was a percussionist, took a music degree at Wittenberg University in 1959, studied composition for two years in Paris with Nadia Boulanger, and came to Stanford, where he received his doctorate in 1966 under the composer Leland Smith.

He had no training in mathematics or electronics. What he had was an article by Max Mathews of Bell Labs, who had written the first programs that made a computer compute sound sample by sample (see the first computer music). Chowning obtained Mathews’s program and, with the help of the programmer David Poole, got it running on the time-shared computer of John McCarthy’s Stanford Artificial Intelligence Laboratory. The music department’s first computer music was made in the off hours of an AI lab.

His own interest was in space: how to make a sound appear to move around a listener, using loudspeakers and a computer. That work led to the four-channel piece Turenas in 1972.

1967

Computed sound in the 1960s was expensive and dull. Building a natural-sounding tone by adding up sine waves, one per harmonic, each with its own changing loudness, took minutes of machine time per second of music, and simpler methods sounded like test equipment.

In 1967 Chowning was experimenting with vibrato, the slight periodic wobble in pitch that singers and string players use. In the program it was produced by letting one oscillator vary the frequency of another a few times a second. He pushed the rate and depth of the wobble far beyond anything a musician would do, into the range of audible frequencies. The wobble stopped being heard as a wobble. Instead the tone changed colour, and became a complex spectrum with many harmonics.

This is frequency modulation, the principle of FM radio, applied at audio rates. Two oscillators, a carrier and a modulator, are enough. The ratio of their frequencies decides which overtones appear, and the depth of modulation decides how strong they are. Change the depth over the course of a note and the tone brightens and dulls as a struck bell or a blown trumpet does. Chowning spent the following years working out which settings gave brass, woodwind, bells, and drums, and published the results in 1973 in the Journal of the Audio Engineering Society, in a paper titled “The Synthesis of Complex Audio Spectra by Means of Frequency Modulation”.

What mattered commercially was the arithmetic. An FM voice needs a few table lookups, additions, and multiplications per sample. It could be built into a chip.

Yamaha

Chowning took the invention to Stanford’s Office of Technology Licensing, which offered it to American organ manufacturers, Hammond and Wurlitzer among them. Their engineers worked in analog electronics, and they declined. Yamaha, the Japanese instrument maker, sent a chief engineer who was visiting its American branch; “In ten minutes he understood”, Chowning recalled. Stanford granted Yamaha a one-year trial licence in the early 1970s and an exclusive one in 1975. The patent was issued in 1977.

It took Yamaha most of a decade to get the method into affordable hardware. The first FM instrument, the GS1 of 1980, needed about fifty chips and cost as much as a car. The DX7, released in 1983 at $1,995, needed two. It had sixteen voices, six oscillators per voice, MIDI sockets, and a glassy electric piano preset, E. Piano 1, that by one count was heard on 40 percent of the American number-one singles of 1986. Yamaha had hoped for about 20,000 sales; it sold some 200,000 in three years (Mix gives about 180,000). Few owners ever programmed one. FM is unintuitive, the DX7 offered a small display and a membrane keypad to do it with, and most players used the presets it came with.

Yamaha also sold FM as bare chips. A cut-down DX7 voice went into the Sega Genesis console. Simpler chips went into the AdLib and Sound Blaster cards, and from there into nearly every PC of the early 1990s, so that the sound of DOS games is the sound of Chowning’s vibrato experiment with two oscillators per voice.

Let Go and Rehired

While Yamaha was negotiating its exclusive licence, Chowning was in Berlin on an artist’s residency, because he no longer had a job. Stanford had declined to keep him on the music faculty. As he told the magazine Mix: “It was the only time they had ever let a junior professor go.” The University of California then offered him a full professorship, and Stanford rehired him with tenure.

In 1975 he founded the Center for Computer Research in Music and Acoustics (CCRMA, pronounced “karma”) at Stanford and directed it for many years. The royalties sustained it. By June 1994, as the patent ran out, Stanford counted more than $20 million from the Yamaha licence and called it the second most lucrative licence in the university’s history; a later figure, cited by Mix, is $22.9 million. Mix called it, in 2005, the third most lucrative patent the university had ever licensed.

His own pieces are few and carefully made: Sabelithe (1966, revised 1971), Turenas (1972), Stria (1977), commissioned by Pierre Boulez’s institute IRCAM in Paris and built entirely on ratios of the golden mean, and Phoné (1981), which uses FM to imitate the singing voice. The Computer History Museum made him a Fellow in 2026.

📚 Sources