Rediscovering the Ballet Mecanique

What was the first sequencer? That depends on your definition of sequencer. Does it have to be electronic and involve synthesizers, or is it any device
Publish date:
Updated on

What was the first sequencer? That depends on your definition ofsequencer. Does it have to be electronic and involve synthesizers, oris it any device that plays music according to a "program"?

Preprogrammed mechanical instruments have been around forcenturies-think calliopes, hurdy-gurdies, and music boxes. Mozart evenwrote a couple of pieces for a mechanical organ.

The technology reached its zenith at the end of the 19th centurywith the invention of the player piano. Programmed music in the homesoon became wildly popular: within a few years, automatic music-producing hardware and its attendant software (piano rolls) becamefixtures in the upper-middle-class parlors of America and WesternEurope.

One of the biggest drawbacks of player-piano technology was that youcouldn't sync two or more instruments with each other. To createmechanical ensembles, inventors grafted additional instruments-violins, bells, organ pipes, and drums-onto player-piano mechanisms,which resulted in monstrosities like the Orchestrion. Still, the musicwas coming from one instrument.

George Antheil, a young American composer living in Paris in theearly 1920s, refused to accept the technological limitations. Heenvisioned the music of the future coming from ensembles of mechanicalinstruments, linked together like looms in a factory. His Ballet pourinstruments mecaniques et percussion (1924), popularly known as Balletmecanique, embraced industrial technology. It calls for 16 synchronizedplayer pianos (referred to as pianolas in the score) playing fourdifferent parts (four pianos play each part), along with a percussionensemble (see Fig. 1). The orchestration is for two conventionalpianos, three xylophones, four bass drums, a gong, a siren, sevenelectric bells, and three airplane propellers. By most definitions,Antheil created the world's first multitrack sequence.

Antheil was relying on Parisian piano manufacturer Pleyel to help himachieve his goal. Pleyel had patented a method of slaving any number ofplayer pianos to one master instrument by means of a complex system ofelectrical pulses and self-adjusting paper-roll mechanisms. (Pleyel'screation bears a striking resemblance to today's time-code-basedsynchronization systems.) Unfortunately, either the system was neverbuilt, or if it was, it proved impractical for public performance.

Only after he had finished composing Ballet mecanique did Antheilrealize that it couldn't be played. He then composed a reorchestratedversion for multiple conventional pianos and only one player piano.

Today, Antheil's dream of a gigantic machine-controlled performancecan at last be realized. Locking together 16 player pianos is simple,thanks to MIDI-compatible player pianos from manufacturers such asPianoDisc, QRS, and Yamaha. With the assistance of publisher G.Schirmer, Yamaha, and a host of consultants, technicians, andmusicians, I helped make the dream a reality.

Antheil, the son of a shoe salesman, was born in 1900 in Trenton, NewJersey. A piano prodigy, he toured Europe at the age of 21. He fell inlove with the expatriate scene in Paris, which was brimming withartists such as Ernest Hemingway, James Joyce, Pablo Picasso, ErikSatie, and Igor Stravinsky. Antheil planted himself at the center ofthis remarkable assemblage and became Paris's "Bad Boy of Music" (helater used the nickname as the title of his autobiography). In thisheady atmosphere, he wrote a number of pieces celebrating the machineage. But none was as "mechanistic" as Ballet.

Ballet mecanique is quite long-1,240 measures-and can take 26 to 36minutes to perform (depending on the tempo). The piece is unremittinglycacophonous and brutally rhythmic, incorporating more than 600time-signature changes (see Fig. 2). The xylophones play in parallelmajor sevenths, the pianos boom out huge repeated chromatic clusters(which at times require the players to use their forearms), the sirenrises and falls above the din, and the bells and airplane propellersprovide "pedal points" to keep the noise floor high (see Fig. 3).Snatches of jazzy melodies occasionally bubble up, only to be subsumedby the sheer wall of sound.

Ballet mecanique also introduced a new element to musicalcomposition: silence. The piece contains several sections, some as longas 20 seconds, in which absolutely nothing happens.

The reorchestrated version of Ballet mecanique (1926) was a roaringsuccess (that is, if the amount of controversy a work generates is ameasure of its success, which was the Parisian attitude at the time).The public greeted the first performance with boos, catcalls, andwhistles; fistfights between Antheil's supporters and detractors brokeout in the auditorium and in the street. The event inspired thegreatest artistic riot since the premiere of Stravinsky's Rite ofSpring 13 years earlier, and the young Antheil became the toast of thetown. "Antheil had Paris by the ear," composer Aaron Copland wrote.

An American promoter tried to re-create the spectacle the followingyear in New York City's Carnegie Hall, but the effort fell flat. NewYork audiences were apparently too jaded to appreciate Balletmecanique, and technical difficulties induced laughter and derision.Antheil's reputation as a serious composer never fully recovered.

Antheil enjoyed a few successful years in Germany as an operacomposer before the Nazis banned his music because he was non-Aryan.Antheil then returned to the United States and settled in Hollywood,where he built a solid, if undistinguished, career as a film composerand orchestrator. In 1952, he again reworked Ballet mecanique, thistime into a much shorter, tighter version that excluded the playerpiano entirely. Antheil died of a heart attack in 1959.

The earlier versions of Ballet mecanique languished in a closet in thehome of Antheil's widow until the late '70s, when she gave them toCharles Amirkhanian, a composer and historian in San Francisco.Amirkhanian subsequently published many other Antheil pieces under thename Antheil Press. In the early '90s, he sold the publishing rights toAntheil's catalog to G. Schirmer, the New York-based music- publishinggiant.

Ambitious percussion ensembles often played the 1952 version ofBallet mecanique (and today it still remains the most popular choice),but until November 1999 neither of the older versions had been heardfor many years. In 1989, conductor Maurice Peress revived thesingle-player-piano version in a concert at Carnegie Hall, but no onehas performed it since. Early in 1999, the German Ensemble Moderneperformed the multiple-player-piano version for the very first time,using 2 player pianos instead of 16. Musicologist and scientist JurgenHocker modified the 1920s pneumatic instruments used in thisperformance, customizing them to respond to MIDI.

Bill Holab, G. Schirmer's publications director, saw the original1924 version as a great opportunity, and he began working on a versionof Ballet mecanique that any contemporary performing group (withsufficient resources) would be able to play. The necessary technology,of course, was MIDI. Aware of Yamaha's Disklavier line ofMIDI-compatible player pianos, Holab approached Yamaha and persuadedthe company to supply Disklaviers for the piece's first performance.Then Holab contacted me.

Holab was familiar with my work for EM and Mix, as well as with thebook MIDI for the Professional, which I cowrote with former EM editorTim Tully. When he asked me if I was interested in helping with theproject, my jaw dropped.

What Holab didn't realize was that I already knew about Balletmecanique. Thirty years earlier, at a music camp in Vermont, apercussion teacher had given me a reel-to-reel tape of the 1952version, which I listened to dozens of times. The earlier versions,however, were unknown to me. Needless to say, I was thrilled to havethe opportunity to help Antheil's impossible dream come true.

Holab asked me to create a set of MIDI sequences that he could sendalong with the printed score and parts to groups who wanted to performthe piece. He knew that the groups would have to deal with manytechnical issues. My specific assignment was to research these issuesand then write a booklet to accompany the score and MIDI files. Ioffered to do even more: I figured that many ensembles would have ahard time procuring sirens, bells, and airplane propellers, so Iproposed to create samples of those sounds and include them with theother materials. Holab readily agreed.

Freelance editor George McGuire had used Leland Smith's Score, a PCnotation program, to enter Antheil's original score into a computer forpublishing purposes. Score doesn't output MIDI files, but somethird-party add-ons do. Holab and I tried to create a sequence fromMcGuire's file, but the unusual time signatures (11/16, 7/32, and soon), the huge number of meter changes, and the sheer volume of notes(some measures contain clusters of 23 notes in each of the player-pianoparts) made the MIDI files a useless jumble, and no sequencer I triedcould even open them.

I then contacted inventor and composer Trimpin (see "PerformingMusician: Taming the Elements with MIDI" in the December 1999 issue ofEM). Trimpin, who is an expert on player pianos, owns a copy of therolls for the 1926 single-player-piano version of Ballet mecanique. Hehad invented a scanning device capable of converting the paper rollsinto MIDI data, and he had created Mark of the Unicorn Performer filesfrom his rolls.

But three problems prevented me from taking advantage of his work.First, there was no clean way to separate Antheil's original four partsfrom this "combined" file. Second, the barlines in Trimpin's filesdidn't correspond to those in Antheil's score, so they would have hadto be entered by hand-an extremely complicated operation. Finally,Antheil himself had never been particularly happy with any of therolls, so I would have had to check every note against the score foraccuracy.

Ultimately, the most sensible option was to create the files fromscratch-all 1,240 measures. I loaded the notes into Opcode's Vision 3.6but soon realized the program couldn't handle sequences longer than 999bars. Fortunately, version 4.01 handled the 1,240 bars without protest.But Vision doesn't like bars with more than 32 beats, so I had toconvert some of Antheil's more expansive time signatures into a metricequivalent that the application could handle; for example, I changed64/8 into 32/4. Vision didn't blink at ridiculous time signatures like11/16 and 7/32.

Since nuance of performance is not a high priority with theplayer-piano parts, I took advantage of a sequencer feature that Iusually avoid-step time. Step time allowed me to build the impossiblycomplex chords one note at a time.

Unfortunately, the copy and paste functions were of limited use:although the score has plenty of duplication, Antheil often made subtlechanges to repeated phrases. For example, a phrase in a bar of 4/4 willbe slightly modified (perhaps by the removal of an eighth note) in thefollowing bar of 7/8. These two bars would themselves be repeated, butthe fourth bar would be slightly different from the second-a differenteighth note would be missing. So after cutting and pasting a section, Ihad to make additional changes by hand.

A player-piano roll is made of paper. If you cut a hole in the paperthat corresponds in length to exactly one eighth note and then youpunch another eighth note right after it, the second note won't play.You must leave some paper between the holes to allow the hammer torecover and strike the second note. MIDI needs much less time between aNote Off and the next Note On-one sequencer "tick." But even aMIDI-driven player piano, which uses solenoids to move the hammers andkeys, needs time to recover.

Because Ballet mecanique is full of repeated chords (at one point,the same 16th-note chord repeats 300 times), I had to decide how longeach note should sound. I chose an arbitrary duration of 80 percent,which worked fine (most of the time).

Additionally, Antheil's score has no dynamics markings. Elsewhere hewrote that Ballet mecanique should be "played as loud as possible."Setting all the note Velocities to 127, however, would break thepianos; they aren't designed to take that kind of sustained punishment.So I set the note Velocities to 80, which gave me plenty of room toexperiment with the dynamics. This choice also worked well (most of thetime).

So I could hear what I was doing, I borrowed four MicroPiano modulesfrom Kurzweil, installed them in my home studio, and rewired the room,adding a second pair of speakers behind me. The four player-piano partssurrounded me as I sequenced them. The setup was terrific for proofingthe parts, because despite the incredible atonal cacophony of thepiece, if anything was the least bit wrong with the texture at anypoint (say, a wrong note) the problem stuck out spatially even moreclearly than it did musically. Entering all of the notes took about sixweeks.

Next, it was time to test the sequences on real instruments. Modernplayer-piano mechanisms are very different from the pneumatic systemsof the early 20th century, and I wasn't sure that today's pianos couldhandle Antheil's demands, even with the reduced Velocities andshortened note durations. In addition, modern pianos are designedprimarily to be used with their own internal computers. They all haveMIDI inputs, but they behave somewhat unpredictably when driven byexternal MIDI signals. I wanted to make sure that the files would workwith the three major brands of MIDI player pianos: Yamaha, QRS, andPianoDisc (the latter two are sold under a variety of brand names).

My first test was with an upright Disklavier at the MIT MediaLab. Iinstalled Vision on one of the lab's Macs and ran my sequence. Troublesset in immediately: notes were dropped, rhythms were jerky, and thesustain pedal wouldn't release. MIT's Disklavier is one of the firstthat Yamaha made, and the MIDI implementation in the ancient onboardCPU was, to put it politely, a little flaky.

George Litterst, a software developer, teacher, and consultant forYamaha, invited me to his basement studio, where he had a Disklavierconservatory grand with more recent firmware. On this piano thesequences ran much better, but I still encountered problems.

The first complication had to do with polyphony. Antheil's scorecalls for chords of up to 23 notes, but the Disklavier's polyphony islimited to 16. Over that limit, the solenoids become too much of adrain on the power supply, and a last-note-priority voice-stealingalgorithm comes into play.

The Disklavier's method of handling incoming MIDI data posed anotherinteresting problem: there's a built-in 500-millisecond delay betweenits responses to Note commands. Why? Think of an electronic synthsimultaneously receiving two notes at different Velocity levels.Although the amplitudes of the two notes will be different, they willsound at exactly the same time. With an acoustic piano, however,velocity refers to the speed of the keystroke. If you simultaneouslysend an acoustic piano two Note commands with different Velocities, thesolenoid-controlled keys will start to move at the same time, but thekey with the higher Velocity will strike the string before the key withthe lower Velocity does.

Yamaha has compensated for the time difference with its 500 ms"predelay," which is reduced according to the individual noteVelocities. High-Velocity notes end up more delayed than low-Velocitynotes. This delay can be switched off, but without it the MIDI responseis extremely "squishy" and unpredictable because other crucialMIDI-data parsing takes place during the delay interval. So whereas thedelay feature solved Yamaha's problem, it complicated my workconsiderably.

Even with the latest firmware, Litterst's Disklavier couldn't quitehandle Antheil's score. Large-chord tremolos and extended fast trillscaused it to miss notes. I experimented with note durations andVelocities and found that the trills behaved better if I increased theVelocity to 99-Velocities greater than 99 made the Disklavier sound asif the hammers would fly off the keys. In some cases, setting thedurations to 60 percent or shorter helped, as it seemed to give thekeys more time to recover. Randomizing durations so that all thesolenoids didn't release at once also helped. Occasionally, however,setting the durations to 120 percent did the job better: having theNote Off occur after the next Note On invoked some kind of fast-recovery feature in the Disklavier, allowing it to play the passageswithout falling over.

However, no amount of tweaking could make the really large chordsand tremolos sound right, so I simply thinned them out. In almost everycase, the part in question was doubled across at least two and usuallyall of the player-piano parts; I stole different notes from thedifferent parts, and the effect wasn't audible.

Finally, I tested my files on other MIDI player pianos. The Piano Mill,a dealer in Newton, Massachusetts, had both PianoDisc and QRS systemson its huge showroom floor. I brought a computer to the store early inthe morning, when I was less likely to disturb any customers.

QRS's system behaved similarly to Yamaha's, with 16-note polyphony(last-note priority), the same fixed 500 ms delay, and the sameproblems with repeated chords. The PianoDisc systems were a differentstory-they had a polyphony limit of 32 notes and could play most ofAntheil's chords without glitches. Their delay time was much shorterthan that of the Yamaha instruments: anywhere between 50 and 150 ms,depending on how the system was configured. With fast repeating chords,however, PianoDiscs had the same problems as the Yamahas. I includedall of this information in my documentation for G. Schirmer so thatperforming groups tackling Ballet mecanique would understand what theywere dealing with.

While investigating the player-piano situation, I was also working onthe samples that I had promised to create for G. Schirmer. I auditioneddozens of airplane sounds from a variety of effects libraries on discand online. Antheil's score, which stipulates small wood, large wood,and metal propellers, requires them to sound for varying lengths oftime, from one eighth note to several dozen measures. To cover therange, the samples had to be loopable; that is, their pitch couldn'tchange. But the propellers in all of the library sounds were active insome way-flying by, starting up, or winding down-so they wereunsuitable for my purposes.

Tim Tully solved this dilemma by going to a private airfield with aportable DAT machine and taping stationary airplanes. His recordingswere terrific, and I got three very different prop samples. I loopedthem in BIAS's Peak and loaded them into my Kurzweil K2500 sampler. Imapped them over three sections of the keyboard and set the pitch-bendsensitivity at 12 semitones so that others could add some Dopplereffect in real time.

I recorded a fire-engine siren directly to DAT, which yielded asample that looped easily and had the characteristic falling toneduring the release. For the bell samples, I used an old alarm bellaugmented by bells that I purchased at hardware stores. I recorded themdirectly to disk in my studio.

By then, I had almost all the materials necessary to produce a completeMIDI-controlled performance of Ballet mecanique right in my own studio.All I had to do was sequence the live piano and percussion parts.Compared with what I had gone through to get the player-piano parts,sequencing was a snap. The note entry took about two days.

Employing a combination of Roland, Kurzweil, and E-mu synthesizersalong with the four MicroPianos, I created the first performance of theoriginal Ballet mecanique, for which I was the sole audience. It was agreat moment.

I burned a couple of CDs of the MIDI version and sent them off toBill Holab and Charles Amirkhanian. My job for G. Schirmer was prettymuch done. But I wanted to take this project through to its logicalconclusion.

The University of Massachusetts at Lowell, where I had been on thefaculty since 1988, has a first-class percussion ensemble led by JeffFischer, one of the Boston area's top freelance percussionists. Lowelloffers an excellent program in sound-recording technology and boasts awealth of technical expertise among the faculty and students, as wellas top-notch recording and production facilities. In the same buildingas the studios is a 1,000-seat concert hall with all the amenities of aprofessional theater, including lighting, sound systems, and tie linesto the recording rooms. Since the hall regularly books professionaltheatrical performances, it has a full-time box- office manager and apublicity office. With its unique resources, Lowell seemed the perfectplace to present the world-premiere performance of Balletmecanique.

Michael Bates, director of academic relations for Yamaha, hadpromised G. Schirmer that the company would provide 16 Disklaviers tothe first ensemble that performed Ballet mecanique. I told Bates myidea, and his only question was, "When do you need them?"

Then I approached Fischer. Not only did he know about the 1952version of Ballet mecanique (the one without player pianos), but he hadactually played it. I showed him the 1924 score, and although herealized immediately that it was very different from the piece he knew,he decided that his ensemble could play it.

In the weeks that followed, I recruited the two live pianists thatthe piece requires. One was Juanita Tsu, head of Lowell's pianodepartment; the other was John McDonald, a composer and pianist on thefaculty of Tufts University. He, too, had played the 1952 version.

But Tsu and McDonald were not to play conventional grand pianos.Their parts would never be heard amid the din of a percussion ensemble,samples, and 16 player pianos. Miking them seemed more effort than itwas worth. Instead, the pianists would play 88-note MIDI keyboardscontrolling two Kurzweil MicroPianos, which would be fed throughspeaker wedges placed beside the two performers on the stage floor (seeFig. 4).

So it came to pass that a computer played acoustic pianos whilehuman beings played electronic pianos. The irony would have delightedGeorge Antheil.

As the concert date grew nearer, it occurred to me that real bellsmight work better than samples. On the Web, I came across the EdwardsSignals company and found more bells than I could possibly haveimagined. I ordered four with similar voltage and currentrequirements.

I also contacted MIDI Solutions, a company located in Vancouver,British Columbia. MIDI Solutions makes a MIDI-to-relay converter, adevice that opens and closes eight different relays in response to MIDIcommands. The converter worked great, and because a little SysExdoesn't scare me, programming it was straightforward.

Unfortunately, the converter's relay contacts were not rated highenough in current capacity to handle the substantial amperage that thebells draw. So I recruited Coleman Rogers, an engineer and fellowLowell faculty member, to help design and build a system whereby theMIDI Solutions relays triggered other relays to control the bells.

Ballet mecanique is played at a constant tempo, but when an ensembleand computer have to stay in sync through 600 time-signature changesand several long silences, who is leading whom? Should the playerpianos follow the conductor, speeding up and slowing down according tohis direction? Or should the conductor follow the computer? And how doyou get a conductor to follow a machine?

There are plenty of ways to make a sequencer follow a conductor, butit seemed very unlikely that any existing software-based systems couldhandle Ballet mecanique. Most autofollowing programs are designed towork with relatively simple meters, but Ballet mecanique clearly has alarge number of meter changes. A single mistake by the conductor couldthrow off the whole system.

A musician would find it very difficult to control the sequencer'stempo in real time, especially if the response of the player pianos wasdelayed by half a second (as with the Disklaviers). When the results ofa tempo change aren't heard instantaneously, the performers are likelyto overshoot it.

We therefore decided that the computer would be the master and JeffFischer, the conductor, would be the slave. I wrote a click track intothe sequence using kick drum and sidestick sounds-the kick on strongbeats and the stick on weak beats-which Fischer would listen to throughheadphones. He and I went over the score in minute detail, deciding howto subdivide bars with odd meters such as 5/8 and 11/16. We first useda drum machine for the click track, but that didn't work. The snare andkick drum got lost in the sound of the four bass drums and threexylophones (later four, because three people alone couldn't play theparts) flailing away on stage. So we then used a cowbell and loosesnare with a long decay for the click track. The sounds were sodifferent from everything else that there was no way Fischer could missthem.

I put vocal cues for rehearsal letters into the sequence so thatFischer could find his place again if he got lost. I also added"countdowns" to certain bars to prepare him for important transitionsthat would otherwise be difficult to hear. The most important countdownwas at the beginning. Because of the Disklaviers' built-in delay, theyneeded a head start. I inserted three extra beats at the top of thesequence, during which my voice said "3, 2, 1," and the first chord onthe player pianos entered half a second before the fourth beat. Knowingthe piece's tempo, we easily calculated how far in advance, in beatsand ticks, the pianos had to enter.

We eventually replaced the headphones with a Shure PSM600 in-earmonitoring system. It worked well, and it prevented the audience fromhearing the click during those long silences (as they might have withconventional headphones). In fact, it occurred to me that the piececould quite possibly be performed without a conductor if every memberof the ensemble wore a personal monitor.

Because of the density of the MIDI data in Ballet mecanique, a singleMIDI cable can't handle all of the information. A MOTU MIDI Time Pieceallowed me to assign (using OMS) a MIDI output to each group of pianos.(I set aside a fifth output for the bells.)

But when I started working with the Disklaviers, I made a startlingdiscovery. Remember that 500 ms delay? It shows up not only at thekeyboard but also at the piano's MIDI-Thru jacks! So in a daisy chainof four Disklaviers, each responds later-by half a second-than the onebefore it.

The solution was to use MIDI splitter boxes. We used a Roland 1 5 4MIDI-Thru box for one group of pianos, a MOTU MIDI Express in 1 5 6mode for another group, and a Kawai 4 5 8 manually switched MIDI patchbay for the remaining groups (see Fig. 5).

To record this important event, I recruited Jonathan Wyner, a localclassical engineer and owner of M- Works Mastering. Together we decidednot only to record the concert but also to schedule a second sessionduring which we could do a studio-quality recording. Along withtechnical director Bill Carman, Wyner and I outlined two recordingsetups. The concert setup was relatively simple: a pair of cardioidmicrophones in ORTF configuration on the stage, just behind theconductor's podium; and a pair of spaced omnis and a coincident pair offigure-8 mics in the first row of seats. The hall is very live, sosetting additional mics farther from the stage was unnecessary.

For the postconcert recording sessions, we added spot mics on thexylophones, gong, bass drums (a pair in mid-side configuration), thefour groups of Disklaviers, and the speaker array at the back. We alsotook direct feeds from the samplers and synths. To offset the boominessof the hall and approximate the sound of a full house, we covered thebalcony and the rear of the orchestra seats with 144 blankets.

Thanks to the generosity of the professional audio-manufacturingcommunity, we were able to produce a top-notch recording. EAW providedthe speakers; from Shure came the personal monitor; Apogee Digitalloaned us three AD-8000 analog-to-digital converters and donated DTRStape; Tascam came through with three DA-88s; Millennia Mediacontributed 16 channels of mic preamps; and Redco Audio donatedcables.

As I listened to the final dress rehearsal, I was overwhelmed by thepower of the piece. Although listening to the MIDI re-creation ofBallet mecanique in my studio had been exciting, the sheer volume ofsound that filled the concert hall was stupendous. The ensemble'splaying was considerably slower than my sequenced version (achievingAntheil's specified tempo isn't humanly possible in some parts), butthe live musicians brought drama, shape, and musicality to thepiece-elements that were far less evident in my MIDI version. It wasquite a moment.

The performance went much more smoothly than any of us had a rightto expect. The ensemble played brilliantly, and nothing significantwent wrong. The concert (see the sidebar "Filling Out the Program") ranalmost three hours, but none of the 1,000 people in the audience left.Instead they gave the ensemble a well-earned standing ovation.

The most remarkable thing about the Ballet mecanique project is that itnever stopped being fun. The technical challenges; the musicalresearch; the "Aha!" factor (when I figured something out); the "Wow!"factor (as I heard the piece develop); and meeting people, drawing theminto the project, and seeing them become as enthusiastic as I was-everyaspect was a tremendous rush.

I feel very lucky to have had the opportunity to work on such afantastic project and to have poured so many of my passions and skillsinto it. It took 75 years and many generations of technicaldevelopment, but here, at the close of the 20th century, we havefinally made George Antheil's youthful dream come true.

Paul D. Lehrman is a composer and author, the "Insider Audio"columnist for Mix, and the editorial director of Mix Online. Visit hisBallet mecanique Web site at

Filling out a high-tech evening

The chance to do a concert with 16 player pianos doesn't come upvery often. To make the most of it, I searched for other composers andpieces that might benefit from this unique opportunity. RichardGrayson, chairman of the music department at Occidental College in LosAngeles, sent me MIDI files for three pieces using various combinationsof MIDI synths and Disklaviers, and I saw immediately that they wouldfit into our concern perpectly.

Two of the compositions used a fascinating shareware program calledMIDIGraphy, which graphically diplays notes and other commandsonscreen. It's similar to the piano-roll notation of many sequencers,but it has more options and jazzier graphics. One piece, in fact,Fantasy on Broadway Boogie-Woogie, used as its raw material a paintingby Mondrian. Grayson broke down the artwork into individual elementsand then created a sequence that reconstructed the elements aurally andvisually, so that as the piece played, the painting assembled itself onthe screen. We displayed the graphics using Macintosh-compatible videoprojector connected to Grayson's PowerBook.

The morning of the concert, however, we made a disturbing discovery:while MIDIGraphy is OMS compatible, it can address only one OMS cableat a time. It seemed as though we would have to take apart ourmulticable MIDI network to play Grayson's pieces. Fortunately mystudent assistant realized that we could attach a second MIDI interfaceto Grayson's computer and simply run a MIDI cable from it to an unusedinput on my Kawai switcher. All we have to do between pieces was changethe source selectors on the switchers.

From Germany, Jurgen Hocker sent me MIDI file for two of the Studiesfor Two Player Pianos by the late American expariate composer ConlonNancarrow. Nancarrow developed his own musical language using playerpianos almost exclusively. But because he didn't have access to MIDIuntil late in his life, he was able to hear only a few of hismultiple-player-piano pieces played in proper sync.

I worked out an arrangement of the saltarello-presto movement fromMendelssohn's Symphony no. 4 (commonly known as the Itahan symphony).Very fast and contrapuntal, it lends itself to multiple pianos; myversion used eight different parts, each played by two Disklaviers.

To round out the program, Jeff Fischer chose three compositions fropercussion ensemble. Double music (1941), a piece by John Cage and LouHarrison, uses brake drums, water buffalo bells, sistrums, and a gongthat's lowered into a tub of water as it's played. Ritmicas nos. 5 and6 (1930), by Cuban composer Amadeo Roldan, are probably the first worksever written for percussion ensemble. They use the whole vocabulary ofwhat we now call Latin percussion, integrating African, Caribbean, andjazz rhythms in a way that still sounds fresh.