“A Greater Achievement than the Telephone”: Alexander Graham Bell’s Synthesizer (1884)

On November 14, 1884, Alexander Graham Bell wrote several pages’ worth of notes spelling out the rudiments of an optical sound synthesizer.  As with many of the ideas he toyed with only in passing, this proposal seems to have escaped the attention of past historians,* which may be understandable; after all, it seems to have gone nowhere and influenced nobody.  However, it uniquely anticipated some important future developments in sound synthesis, and its quirky technical details are interesting in their own right.  The same set of notes also reveals Bell’s own sense of the sound synthesizer’s importance: if such an instrument could be perfected, he wrote, “it will be a greater achievement than the telephone, photophone, or spectrophone.”  The telephone probably requires no explanation, but the photophone could transmit sound wirelessly using a beam of modulated light, and the spectrophone used the same principle to identify materials by ear based on their absorption spectra.  Bell thought that the sound synthesizer would eclipse all three of these inventions if it could be made to work as he imagined.

Histories of electronic musical instruments typically jump from Elisha Gray’s musical telegraph (1874) to Thaddeus Cahill’s telharmonium (ca. 1895) and associate the transition between them with one crucial development.  In Gray’s musical telegraph, each musical note had corresponded to a simple electrical signal at a single frequency.  In Cahill’s telharmonium, by contrast, signals generated by multiple tonewheels were combined to produce complex timbres through additive synthesis, bringing electricity to bear on sound quality.  The synthesizer Bell proposed in 1884, midway in time between those two other inventions, was supposed to generate sounds electrically and to control their quality, which was in fact the whole point of it.  He never built the contraption he’d imagined, unlike Gray and Cahill, but conceptually his idea appears to represent some rather significant firsts.

Bell had already written back on January 16, 1879: “Why not try the synthesis of sound as suggested by Sir W. Thomson.  Sir William seemed much struck by the idea. Could I not have construct a curve theoretically on a large scale and then have it photographed or reduced and applied to a phonograph.”  Whatever Bell may have discussed with Sir William Thomson—later 1st Baron Kelvin—his thoughts at that time had apparently been limited to forming a lateral waveform as a ridge or indentation on the surface of a cylinder, presumably for use in some non-electrical method of sound generation.  By 1884, however, he and his associates in the Volta Laboratory were exploring some new approaches to recording and reproducing sounds using light and electricity.  On November 14th, they had just begun a new series of experiments which we know today mainly from surviving artifacts at the National Museum of American History, described in my Discography of Volta Laboratory Recordings.

The course of experiments as a whole centered on recording sound photographically, and the specific method used on November 14th was to cause the vibrations of a stylus attached to a diaphragm to vary the length of a slit through which light was passing, with these variations being perpendicular to the direction of rotation of a photosensitive surface. This method formed an exposed band of varying width, as seen above.  Experiments I and II were both made on November 14th on smaller rectangular glass plates that could only accommodate part of the spiral, while Experiment III, conducted on November 17th, was made on a full-sized glass plate and has been played back digitally using IRENE:

Another disc prepared on the same day (NMAH 287679) featured the word “phonography.”  At the time, the plan had been to shine a beam of light through the glass after the emulsion had been developed with a selenium cell linked into a telephone circuit on the other side that would transduce the light modulations back into sound, thereby adapting the principle which the photophone used for real-time sound transmission to the problem of playback.  The whole process, which closely anticipated the optical film sound track, was described along with some others in U. S. Patent 341,213, filed in November 1885 and issued in May 1886.

It was the sound-recording experiments of November 14th, and more specifically the “beautiful curves produced photographically” on that date, that suggested to Bell a new and more flexible means of synthesizing sounds.

The jottings in question appear in Volume 20 of the Laboratory Notes, in which Bell had made a number of entries about his work on the metal detector (or “induction balance”) back in 1882, but which he had then set aside with many pages remaining blank until he suddenly found he needed something handy to write on two years later.  They’re somewhat messy—after all, we’re dealing here with the raw vestiges of dynamic brainstorming—but here’s my attempt at an exact transcription, complete with as many of Bell’s corrections and strikethroughs as I could untangle:

{75} Nov. 14^th 1884
1500 R I. Av. Wash.

Not having a note-book at hand here—I make use of this old book to jot down some ideas.  The beautiful curves produced photographically today—have led my thoughts back to the old idea of the utilization of radiant energy for the reproduction of sounds—and tonight I have obtained a glimpse of the possibility of realizing, by means of radiant energy, an Instrument that has long been a kind of dream to me.

An instrument which we may consider as a new kind of musical instrument—or even a kind of talking machine.  As a musical instrument it should be superior to any or all musical instruments.

The old idea, which has hitherto existed as a sort of dream without any substantial reality, and without my being able to see any practical way of realizing it—is this.

Given an instrument by means of which we could at will control the form of a vibration—then we could control the character of the sound emitted from it—and reproduce produce from it by the manipulation of its keys or moveable parts—any sound or sounds whatever.  All that would be necessary would be to know the necessary form of vibration.  Suppose for instance—that a musical chord produced by a full orchestra—should produce as the resultant of all the vibrations of all the musical instruments employed—a vibration of the air in the ear passage having a form like the following:—

&c.— Then if which is supposed to be a periodical curve.  Then if we could by manipulation of an instrument adjust an instrument to produce into a shape like one of these curves

(a complete vibration) and cause this to agitate the air producing an air vibration of its {77} own shape which should be repeated & repeated a definite number of times per second—the frequency of repetition corresponding to the frequency of the fundamental note of the chord—we would have—not a single sound—but a full chord—in fact the effect would be as if the orchestra were playing.

If we can contrive an instrument that will control the for by which we can control (a) the form of a vibration & (b) the frequency with which that form comes is repeated—we [would, corrected to:] shall have an instrument from which we can produce any sound or sounds at present whatever.  It will at once be a violin, a violoncello,—a piano an orchestra—&c.—and even a singer with all the who sings & articulates at the same time.

I now see that it is possible to construct an apparatus capable of realizing this conception.  It may even be a comparatively simple apparatus if radiant energy is employed.

Let me take the first crude conception as it exists to-day—which, while unsatisfactory as a final apparatus—points the way to a solution.

Let 1, 2, 3, 4, &c. 10.

[“1 / 2 / 3 / 4 / 5 / 6 / 7 / 8 / 9 / 10”]

be upright slips of metal or opaque substances on the end of levers l, l, &c. which could by means of keys, k,

[“k / k / l / l / axis / 2 / 1”]

be depressed to a greater or less extent by the ten fingers.  Then it would be possible to cause the upright slips to {78} take a position like this

or this

or in fact any [shape, corrected to:] other shape that ten slips could assume.  It is evidently a mechanical problem simply to multiply the effect shape.  Each lever for instance could adjust 10 or twenty slips to the same height—so as to produce by a multitude of slips an appearance like this:—

&c.

Having adjusted our slips as shown—we have a periodic curve on a large & gross scale.

Now it is evidently possible to throw an image of this periodic curve upon a screen, and by suitable adjustment of our light &c—we can make this image larger or smaller than the original as we please.  We can reduce the image until it is of the size of a phonautograph tracing produced by the voice—or m and in this case the imperfections of the curve due to the separate pieces will also be reduced.  If we enlarge the curve—the faults will be magnified.

Let us reduce the image to a suitable size and cause it to fall upon an opaque screen containing a narrow slit—so that the image would fall as shown here

[“(opaque screen) / light / slit / shadow”]

Now place behind the opaque screen a selenium cell having a flat surface [of corrected to:] at least as wide as the length of the slit and of considerable length—a cell the surface of which might be somewhat of this shape

[“surface of selenium / slit in opaque screen”]

{79} Connect selenium cell with a telephone & battery—and listen at telephone while the opaque screen with slit is moved in the direction of the arrow head.

Query—Should we not hear a sound?  And should not the quality or character of the sound be controlled by the shape of the curve-image?

Suppose we cause the slit to move at a uniform rate of speed in the direction shown by arrow head (or reverse)—we should obtain a sounds [of a corrected to:] that would always be of the same pitch—but the quality or character of the sound would be changed by every adjustment of the levers.  An  One adjustment would yield a vowel sound as if sung to a certain pitch; another adjustment might cause the sensation of a musical chord in which more than one musical tone would be perceived clearly perceived—but the fundamental of the chord would be of the same pitch as the pitch of the vowel formerly heard &c.

The pitch also could be controlled in various ways: For instance let the source of light be brought nearer the upright slips—then [their corrected to:] the image on the screen would grow larger—and though the slit is supposed to move with the same velocity as before—the variations the frequency of the light-fluctuations on the selenium will be reduced & the pitch of the resulting sound will be lower.  If the source of light  The enlargement of the image will reduce the intensity of the light passing through the slit—(for the same amount of light is now spread over a larger surface)—but I would not anticipate that the loudness of the sound heard from the telephone would be thereby affected—for the amplitude of the curve would be magnified in exactly the same proportion as the light is light is diminished.  While the diminution in the intensity of the light might weaken the sound—the increase in the variation would increase it—so that the resultant loudness would be the same as {80} at first.

A. If the slit is caused to move at a uniform speed then the an enlargement of the curve-image will lower the pitch—and its reduct & the reduction and a diminution raise it—without changing the form of the vibration or affecting the loudness of the sound.

B. The manipulation of the keys will affect the quality or character of the sound & control the form of the curve & its amplitude—& hence control the character or quality of the sound perceived & its loudness but will not affect the pitch of the fundamental sound perceived.

C. Given the power of moulding the form of the curve-vibration & controlling the size of the projected image—we have the power of producing at will any kind of sound or sounds whatever,—by the

It should be possible for instance—so to manipulate the instrument—as to cause in the listeners ear—the sensation of listening to an opera—with full brass band—& violin accompaniment— We could make him perceive a multitude of sounds from the one instrument.  The possibilities include the production of articulate speech—& the production of a number of voices—a crowd of speakers.

I feel that the crude conception described above—contains a valuable germ.  I already perceive a multitude of ways in which the vi the curve-image could be controlled—all of which appear to me more practical than those described above.

If the above invention can be perfected it will be a greater achievement than the telephone, photophone, or spectrophone.

AGB
Nov. 14^th 1884.

To summarize Bell’s idea as it stood in 1884:

• Ten keys would control the heights of opaque “slips” so that they would make up multiple iterations of a desired curve.
• A light source behind the “slips” would project this curve onto an opaque screen as a static bright band of varying height.
• A slit in the opaque screen would move horizontally across the projected curve at a constant rate of speed, causing it to be exposed to varying amounts of light depending on the height of the curve at each point along its path.
• A selenium cell mounted behind the slit would receive the modulated light passing through said slit and transduce it into an electrical signal.
• An ordinary telephone would in turn transduce that electrical signal into sound.
• The positions of the “slips” would control sound quality, and their variable distance from the screen would control pitch.

The proposed synthesizer was programmable insofar as the keyboard could make its curves assume any desired shape, corresponding to any desired timbre, although Bell may have expected the performer to adjust timbre continuously and dynamically rather than leaving it set to one consistent “voice.”  It was also digital to the limited degree that its curves were to be made up of discrete samples embodied in the “slips,” although the height of each “slip” would have been continuously adjustable as an analog value.  Moreover, it was based not on additive synthesis, as the telharmonium was, but on sample-based synthesis (invoking another use of the word “sample”).  Its optical approach would thus have shared more in common with the Optophonic Piano (1916), the Cellulophone (1927), the Variophone (1930), and later instruments conceived in a similar spirit up to and including the Optigan; and also with the tradition of hand-drawn optical sound tracks on motion picture film pioneered in the 1930s by Arseny Avraamov, Rudolf Pfenninger, Oskar Fischinger, and others.

These notes present a remarkable and unique vision, but it’s a vision that contains a number of gaps.  For example, if all ten of the performer’s fingers were occupied by the keyboard that controlled sound quality, how would pitch be adjusted?  With foot pedals?  By a second performer?  To what extent would chords or the sounds of multiple instruments be produced all at once by a single artificial wave shape, and to what extent (if any) would the component elements be produced separately and combined?  And when the slit had made one full pass across the screen, how was it supposed to get back again to the other side?  Could the screen have taken the form of an endless loop on rollers with a pair of slits, leaving one slit exposed to the light while the other passed behind?  Bell seems not to have thought things quite this far through.  In his own words, he’s presenting only a “crude conception” of a “valuable germ.”

He was to revisit the idea of electrical sound synthesis four and a half years later, in a set of notes dated March 1, 1889, found in Volume 28 of the Laboratory Notes.  (The pagination of this volume was twice corrected at much later date, but the initial crossed-out page numbers appear to be the ones used for internal references, including the reference to “page 112.”)

{111 / 123 / 113}

Friday March 1^st 1889
Old Idea—Electrical musical instrument

Contact point p. passed from c to d rapidly with uniform velocity will cause over the conductors e f g h i j &c will yield a musical tone from telephone.

If contact point p. be passed with same velocity from a to b—the musical tone will be lower in pitch—because the same number of [interrupters, corrected to:] contacts will be made in a longer period of time.  Might not a musical instrument be based on this principle

Again Let A be a conducting disk which is rapidly whirled round and let B be a small wheel or roller arranged so as to have alternate conducting & insulating portions so that its rotation against a conducting surface will make a musical tone in teleph.  Then the nearer to the axle of the whirling disk B is pressed the slower will the rotation of B be—& the lower the tone from the telephone.  Think an undulatory current could be produced as easily as an intermittent one—and even the form of the electrical undulations be moulded by the use of imperfect conductors having considerable resistance.

{112 / 124 / 114}

[“Good Conductor / Carbon or imperfect conductor”]

Let metallic point p. be moved along line a b—Will not the resistance of circuit be least when it is [near, corrected to:] passing c and greatest when passing d.  If edge e c d f is sinusoidal curve—will not a pure musical tone be produced from telephone—a simple tone like that produced (theoretically) from a tuning fork?

Would similar effect but different in pitch (higher) and feebler be produced by causing point p. to traverse line g h with same velocity as before?

[“Good conductor / Poor conductor”]

Would not the form of vibration curves of metall conducting edge affect form of electrical vibration—& thus affect the character or quality of resulting sound?

Would not [same, corrected to:] similar effect be produced by making the permanent contact with the poor conductor & the moving contact on the good conductor.  An experiment could easily decide the character of the effect and settle whether it is worth carrying out.

{113 / 125 / 115}

If feasible to produce different qualities of sounds in this way—could make electrical apparatus to mould by manipulation—the form of a sound vibration—in a more simple way than the Radiant Energy plan.

[“Good conductor / Poor conductor / Good conductor”]

If above will work it has [an, corrected to:] this advantage as over converse that the sound should not be feebler along line c d—for the distance from the poor conductor would not sensibly affect resistance.  It is problematic what effect would be.  The potential at any point on surface of good conductor (lower one) would be a resultant effect—It is conceivable that it would vary from point to point.  The variations would probably be greater in converse arrangement shown page 112.

Bell starts out here with the “old” idea of synthesizing tones via simple, intermittent electrical signals, like the ones which Elisha Gray had used in his musical telephone and which Bell himself had used for less “musical” purposes in some of his own experiments.  His first diagram shows six evenly-spaced wires radiating out from a single point in a V-shape; the plan in that case is to drag a contact point across the wires so that a circuit will be repeatedly made and broken, and at a faster rate the further down the V one goes.  The second diagram shows another arrangement with an electrical breakwheel pressed up against another wheel such that it will spin less rapidly, and thereby interrupt the circuit less frequently, as it moves towards the other wheel’s axle.  The remaining diagrams then show arrangements in which moving the contact point across the V is instead supposed to produce a continuous variation in resistance based on the physical shape of a curve.  It’s hard to tell exactly what Bell has in mind here, but he may be trying to illustrate a three-dimensional structure something like the one I’ve shown below, where variations in brightness correspond to variations in the relative thickness of good and bad conductors.

Bell speculates that some approach along these lines could provide a “more simple way” of molding sound waves “by manipulation” than the “radiant energy plan” he had outlined back in 1884.  He doesn’t elaborate, and I can’t think of any obvious mechanism for dynamically manipulating the shape of such a curve either.  Indeed, I don’t think this new idea is nearly as interesting as his earlier one.  But the problem of sound synthesis was evidently still on his mind in the spring of 1889, and he hadn’t forgotten his “radiant energy plan,” even if he still hadn’t tried to develop it.

I’m going to assume that Bell never actually built a musical instrument based on electrical sound synthesis; certainly he never exhibited one.  But I hate to leave this subject without giving you any relevant audio to listen to, so I took one of his synthetic waveform sketches from 1884—

—did some graphical processing to it; looped it—

—converted it into a WAV file using Picture Kymophone, transducing its brightness into amplitude while running it from left to right (Bell wrote that it could have gone either way, but it doesn’t sound significantly different backwards); and then harnessed the result as a sample for control by MIDI.

By way of conclusion, then, here are a couple musical selections played with an actual synthesizer voice designed by Alexander Graham Bell, even though he never had an opportunity to hear it in action himself.  He speculated that his apparatus could “at once be a violin, a violoncello,—a piano an orchestra—&c.,” but the specific curve he sketched turns out, unsurprisingly, not to resemble that of any preexisting instrument.  Instead, it produces a new sound that’s entirely its own, albeit not unlike the sound of other early experiments in optical synthesis.  I wonder what he’d have thought of it.

---

Bach, Prelude No. 14 from The Well-Tempered Clavier, Book I.

---

Chopin, “Grande Valse Brillante.”

---

*Postscript (October 27, 2017): Shortly after publishing this piece, I did spot one brief mention of Bell’s 1884 proposal in Robert Bruce’s Bell: Alexander Graham Bell and the Conquest of Solitude, at page 355.  However, it’s wrongly dated there to 1885, so any historians of electronic music who’d wanted to track down the reference would have scoured Bell’s notebooks for the latter year in vain.