The first weekend of April, no joke, I got to take part in a concert in one of the great performing spaces on Earth, New York's Carnegie Hall. I wasn't exactly performing, but I wasn't exactly crew either. Nor was I exactly conducting: The American Composers Orchestra, a group devoted to presenting large-scale works of 20th (and soon, presumably, 21st) century composers, had chosen to present the New York premiere of the new version of George Antheil's Ballet Mecanique, which I had helped prepare for publisher G. Schirmer. Although I had done my best to make the piece playable by any ambitious performing group, the ACO still wanted my help. So I had the distinct pleasure of sitting on the stage and overseeing a computer, a bunch of mechanical pianos, and an ensemble of amazing musicians raise the roof in one of the loudest "classical" pieces those walls had ever contained.
Now, I know some of you are probably sick of me writing about this project (I talked about it in the March "Insider Audio," which is still available online at www.antheil.org, and also at great length in an upcoming article in Electronic Musician), and I have promised many people (like my editors) that I will soon shut up on the subject and get on with my life. But the story of what happened at Carnegie Hall is simply too good, and too relevant to Mix readers, to let pass. So indulge me one more time.
As you may recall, the piece calls for seven or eight percussionists, two pianists, four to 16 player pianos (in this concert, the parts were handled by eight Yamaha Disklaviers-which were plenty loud), seven bells, a siren and three airplane propellers. A computer running a MIDI sequencer controls the player pianos, and at the same time, triggers the sound effects, and also cues the conductor through a fiendishly complicated click track. At the school where I was teaching last fall, a student ensemble premiered the piece, and it was a tremendously exciting event. After that initial "shakedown" performance (and recording, which is now available on CD from the Electronic Music Foundation at www.cdemusic.org), the piece was ready to send out to the real world. The ACO was the first group to take up the challenge. The ensemble handled it beautifully-but letting my baby take its first steps, without me holding both its hands, turned out to have some interesting problems.
My job in New York was to supervise the MIDI and sound setups, answer any questions the conductor might have and sit on a preconcert panel discussing the piece. The orchestra hired Miles Green, a New York MIDI/audio expert who has worked with luminaries like Philip Glass and Laurie Anderson, to handle the hardware side of things. A couple of weeks before the performance, I sent Miles my sequence files (converted from Studio Vision Pro to MOTU Performer) and my samples (converted from SampleCell to the Kurzweil format), which he loaded into his Powerbook G3 and his K2000 rack unit. He hooked the computer up to a MOTU Micro Express XT USB interface and connected the interface to a bunch of synth modules.
Miles called immediately; there was trouble. The player-piano parts consist largely of huge, thick chords playing very fast, in precise rhythms. There are four parts, and in order to make them work at all, each part needs its own dedicated MIDI cable, or the note data would be way too thick-but this is no problem, one assumes, when you have a multiport interface.
But coming out of Miles' synths, the piano parts sounded alarmingly sloppy, and the click track was all over the place: sometimes early, sometimes late, sometimes disappearing entirely. It sounded as if he were trying to shove all of the parts down a single MIDI cable, and the dreaded MIDI choke was rearing its ugly head. The computer was certainly fast enough-heck, we had played the piece at school on a pokey, old 100MHz Macintosh 8100 with no problems at all, and in some rehearsals on an antique Quadra 650-so that wasn't the issue. USB is supposed to be a much faster protocol than Apple's ancient serial driver, so that shouldn't have caused any problems. So what the heck was going on?
A few frantic calls and e-mails to Mark of the Unicorn revealed the cause. Most multiport MIDI interfaces that communicate with the host computer at a speed faster than MIDI itself have a "throttle-back" feature. This is necessary when you want to send a long string of data to a single device, something that commonly happens when you are bulk-dumping a synth's program banks to the synth using MIDI system exclusive. If the data coming out of the interface is too fast, the synth's input buffer will overflow and the message may get garbled or lost. To prevent this, when the driver software controlling the interface detects this kind of rapid-fire MIDI stream, it throttles back the speed of the interface to the actual speed of MIDI, and sometimes even slower. In the case of the MOTU interfaces, my huge, lumbering piano chords were being misinterpreted as a sysex bulk dump, and the interface was deliberately slowing them down.
MOTU offered two solutions: We could use two interfaces, thereby lowering the data density through each and hopefully avoiding the throttle-back, or they could rewrite the drivers so that the throttle-back feature would kick in at a higher data rate. Miles opted for both solutions: MOTU sent him another interface and the updated drivers. There were no more problems.
But we weren't finished yet. A different kind of problem arose when I was told by the orchestra administration that the elaborate new sound system at Carnegie Hall was mono. Although none of the instruments need to be miked, the Ballet Mecanique does require a sound system with as many as seven individual speakers; the exact number depends on how many of the soundmakers in the piece are provided by a sampler and how many are played "live." The orchestra planned to use a live siren and bells, which left the three propeller sounds coming from Miles' Kurzweil. To keep the propellers from sounding like total mud, and to give them a proper stage perspective, the sounds needed to be physically isolated from each other; a mono system wasn't going to allow that.
Fortunately, in the March issue of Mix, there happened to be a great article by Mark Frink about this very subject: the new Meyer Sound system in Carnegie Hall. The article described the systems as consisting of two stacks on the sides, and an impressive cluster that was flown above center stage-and although it didn't say so specifically, it seemed obvious to me that, yes, the whole thing was mono. But the article revealed something else, which the orchestra hadn't told me: The hall also had a 6-channel monitor system, with six Meyer Sound wedges. Since we weren't going for huge SPLs on the propellers, three of the wedges, pointed at the audience, would seem to do the job admirably. I posed this question to my New York colleagues, and their cautious reply was that if we could work it out with the hall sound crew, maybe we could use the monitors instead.
Now, I had been warned about this crew. Union crews in New York are notorious for being difficult to work with (Joke flying around the Javits Center whenever AES is in town: How many union guys does it take to screw in a light bulb? Ten! You gotta problem widdat?), and Carnegie Hall has a reputation in some circles for being one of the hardest venues of all. But when I arrived at the hall the morning of the performance (which was scheduled for 3 o'clock that afternoon), I found the absolute opposite to be true. The stage, sound and electrical crews were cooperative, knowledgeable, professional and eager to help with what they realized right away was going to be a fun gig. It didn't hurt that the sound crew are all avid Mix readers, especially when they realized that the piece I had just written about in my March column was the one that was about to be on their stage. And some are even Jean Shepherd fans.
It turned out my request that we use the monitor system had already been relayed to the crew, and so when we came in that morning, the wedges were already out and wired up. (Using the monitors also meant they wouldn't have to fly the center cluster, for which I imagine they were grateful.) It took us an amazingly short time to get everything in place and running. It simply couldn't have been handled better.
At the same time, the eight percussionists, two pianists, and the siren and bell players arrived and set up their hardware among the eight Disklaviers Yamaha supplied. A violist had been dragooned into the siren job, which she took on with great fervor-in concert, the sight of her formal black gown flying as she cranked furiously added immeasurably to the performance. A trombonist played the bells, which were connected to a MIDI-to-contact-closure converter, using an old Yamaha synth.
Miles and I set up our MIDI paraphernalia on a table onstage, next to the door leading backstage, that the crew had supplied. We were originally told by one of the orchestra managers that we should set up offstage, so the audience wouldn't see the "machinery," but we convinced him that that was just too risky.
The group had had only one rehearsal-the day before in a different hall-and this was going to be their only chance to run down the piece with the Disklaviers and the full sound system. I had to admit I had trepidations. At school, we had put four months of rehearsal into it, while these folks were going to do it in two days. But I knew from the first downbeat that it was going to be fine. Percussionists are a class of players I have always admired for their skill in navigating seemingly impossible scores, and these cats were among the best New York had to offer. They were good, and so was the conductor, Dennis Russell Davies, who threaded his way through the more than 600 time-signature changes without hesitation.
The only glitch was that Davies couldn't hear the click track, which was the only way the Disklaviers and the players were going to be able to stay together. I had specified a Shure wireless in-ear monitor system (we used one at my school, and it worked great), but the orchestra had gotten something else-and it just wasn't loud enough. No matter how we gain-staged the inputs and outputs, the thing would go into hard clipping whenever we got anywhere near a usable level. So we got rid of it, and the crew quickly ran a cable from the Mackie 1202 mixer handling the click (which itself was coming from an isolated output on the Kurzweil), across the stage to a single headphone for Davies to wear. Now he had plenty of volume.
Two run-throughs later, the ensemble had the piece up to the tempo they wanted, while I jumped on and off the stage and ran around the hall checking the propeller levels, and signaling tweaks to the sound crew. It went like clockwork, and everyone was happy.
And this is where the story should end, with a line like, "The performance was great, all our preparation paid off, the audience loved it, and we got a rave review in the New York Times." But it didn't quite turn out that way.
The preconcert panel I was speaking on started at 1:30, 90 minutes before show time. A table with four mics was set up on stage, and the main house system (the side stacks) was turned on. We panelists talked and answered questions for about 45 minutes, and then final preparations for the concert began. Miles and I asked if we could do a soundcheck, and the orchestra management said no: Since the house was open, too many people would hear the noises, and they wanted to "surprise" them. So Miles walked around the stage doing an "idiot check," making sure none of the MIDI cables were pulled out or splitter switches inadvertently thrown, and I double-checked the MIDI-to-bells hookup and listened to the click track.
The house lights went down, conductor Davies took the stage, put on his cans, nodded to Miles, and we were off. After about 30 seconds, Miles and I looked at each other: Where the hell were the propellers? We waited for another place where the propellers came in, watching the score closely, and sure enough, no propellers. Fortunately, the table that had been set up for us was right near the exit door, and so, as nonchalantly as I could, I got up from my seat, left the stage, and announced to the crew that the propellers were not coming through the sound system. All hell broke loose backstage, so I, still nonchalantly, closed the door and walked back to my seat. After about a minute, the door opened a crack, so I got up again, and a sheepish crew member told me what had happened: During the preconcert panel, one of the crew had powered down the monitor system and then forgot to bring it back up again.
Would it be okay, I was asked, if an electrician were to walk across the stage, while the performance was going on, to where the power amps were, and turn them on? No problem, I said. Cool as a cucumber, the electrician did exactly that, and at the next propeller cue, all was well. All the audience knew was that something had to be adjusted onstage. The piece finished in grand style, the audience did love it, and we did get a very positive review in the Times.
At a reception after the concert, I asked Davies if he knew the propellers were missing at the beginning. No, he said, he was too busy conducting the live musicians. But toward the end of the piece, there are long sections in which the propellers sound by themselves, and had they remained turned off for the entire performance, there would have been huge inexplicable holes. Since I had the power to stop the performance by stopping the sequencer, I asked him if, in the event we couldn't fix the propellers right away and I had stopped the piece, would he have had me shot. He replied, "No, that would have been the right thing to do." He then told me a story of a piece he was conducting in Europe that starts with the entire orchestra playing an F major chord, only one of the violinists played an F sharp. He stopped the piece, turned to the audience, and said, "We can do better than that." And they did. A class act, to be sure.
So there are a few lessons to be learned here. One, if you're doing a high-profile live performance, don't use the very latest technology, or even Revision x.0 of anything. Go back a generation or two and use tried-and-true stuff. If you have to use brand-new tools, try to get some inside contacts with the company that makes them, so you can get problems fixed fast. Two, try everything out way ahead of time, in a situation as close to the real performance as possible, to give you room to maneuver in case things don't work the way they're supposed to. The interface problems that we were able to solve very nicely would have been fatal had we discovered them the day before the show. Three, even the best, most experienced people make mistakes. And four, which is a direct corollary to three, whenever you have the opportunity to check your equipment just before a gig, take it. And then check it again. Because even when you make it to Carnegie Hall, you still have to practice.