SAN MATEO, Calif. As the towers of the World Trade Center crumbled on that September day two years ago, the plume of debris they spewed claimed a victim that, although not human, was nonetheless irreplaceable to those who loved it: the mighty pipe organ in Trinity Episcopal Church at Broadway and Wall Street. Dust and detritus from the collapse rendered the church's 80-year-old Aeolian-Skinner organ unusable.
But that lesser tragedy opened the door to a unique story of musical skill, love of an instrument, insight into signal processing and innovation in electronic systems design that this year has brought about a resurrection of sorts. Today a new mighty organ plays at Trinity not a pipe organ, but perhaps the most innovative electronic instrument ever to fill a sacred space.
For Douglas Marshall and David Ogletree, principals of the Marshall & Ogletree LLC organ company in Needham Heights, Mass.), the loss at Trinity Church would transform what had been a hobby, perhaps an obsession, into a product development. It would be a unique chance to prove that the contrarian ideas they had formed over the course of a decade were correct, and that conventional wisdom about sampled-data electronic instruments was at best incomplete.
Vanished gloryThe story began long before terrorists boarded airplanes that late-summer morning in 2001. Marshall and Ogletree, who grew up 15 years apart in Westwood, Mass., and who both went on to careers as concert organists, in the early 1990s formed a company to represent major organ builders to the church market.
Along the way the two developed a love of the great Aeolian-Skinner pipe organs that are arguably the finest organs ever built in North America. With the Aeolian-Skinner Organ Co. gone, Marshall and Ogletree speculated about how the sounds and, more important, the playing experience of such instruments could be preserved.
They hit upon a project to record samples of existing instruments around the United States and experiment with them. "We had hoped that we could build some sort of addition to an electronic organ that would convey the real sound of these instruments," Ogletree explained.
The theory of sampling musical instruments is relatively straightforward. Depending on its physics, an instrument has an attack, a steady-state and a release, each of which has a different tonal signature. This complexity can be reduced substantially by a few approximations for instance, using the steady-state tone and simply modifying it with attack and release envelopes, perhaps throwing in some transients known to occur at the opening and closing of the "voice."
In this sort of approach, it's only necessary to get a good recording of a moment of the instrument's steady-state voice, and the rest is signal processing.
But Marshall and Ogletree ran into a problem: That approach didn't end up sounding like the real thing. "In 1997 or so, we were doing experiments with the digital sampling software that was just coming on the market," Ogletree said. "They were just starting to stream high-fidelity recording onto hard disks, and we were playing with it. We began to realize that the length of the recording made far more difference than we had believed."
It had been assumed, Ogletree explained, that the sonic content of an organ voice matched what most people reported hearing. There is an initial, transient period of a second or so, after which the human ear can identify a particular note and stop, and then the voice remains constant until the release.
But the data Marshall and Ogletree were collecting contradicted that. "The tone actually become steady a very long time after it seems to have settled," Ogletree said. "Recorded samples of the attack have to be as long as 15 to 20 seconds to capture the actual voice, because there are things going on during that entire period it's not at all a steady tone."
In fact, he said, the actual voice of a pipe organ stop is a tangled web of transients, from the sound of the mechanical action to the pressurizing and venting of air chambers to the voices of the pipe itself. Sorting out which transients contributed to a qualitative sense of realness to a master organist was a job that only an experienced player could hope to achieve. Late nights and many samples led to a collection of proprietary techniques for combing the transients out of a recording and ordering them for reproduction.
But there were more surprises in store. For one, there was the matter of microphone placement. Organ pipes don't stand in the middle of a room and sing they stimulate a pipe cabinet that brings its own important contributions to the sound. This fact would demand exhaustive work on microphone placement, multichannel recordings and even more nights over the editing software teasing out just what mix of channels would represent the truest voice of the pipe. Again, the ear of the organist and a commitment to the task were essential.
Not least was the problem of authenticity. There is a characteristic sound to the works of any great organ builder. But like any other kind of public art, a completed organ is subject to the attentions well meant or otherwise of many people, often with sad consequences. Cabinets get rebuilt or moved, pipes modified or damaged, pressure and flow specifications changed. "It became a significant project just to identify which pipes in which organs were in fact authentic," Ogletree said.
Maybe an obsession
The hobby began to take on the proportions of an obsession. A big Skinner organ can have 80 or more stops, as the individual voices are called. Each stop in turn can be applied to each key on the manual (the organ keyboard), resulting in a unique pipe or grouping of pipes for each key/ stop combination. Using the painstaking long-sample recordings for each note for each stop generated nearly 20 Gbytes of sampled data. "The cost of gathering all this data is almost incalculable," Ogletree said. "Without a group of people who were dedicated to the point of being a little bit possessed, it could never have been done." But gather they did.
The next problem in creating an instrument was to design an electronic platform that could make use of the samples, and to develop tone-generation and postprocessing applications that could turn the samples back into voices, singing out in response to an organist's fingers and feet.
Since "neither of us was a hardware person," Ogletree said, the partners and their growing team of developers turned to off-the-shelf hardware rather than custom DSP design. Upon investigation, the group concluded that a personal computer or perhaps a linked network of PCs with existing professional-grade sound cards could do the work.
So the team set out to develop tone-generation software running on a stock PC. Windows was quickly rejected as unreliable, and development was done on the Linux operating system. Following the death of James Murray the key algorithm and software developer, himself an organist new software designers joined the project, which by now was taking on a life of its own within the company. And then came the 9/11 attacks.
To build an organ
The organ destroyed with the World Trade Center at Trinity Church, situated just 600 feet from ground zero, was originally built by Ernest Skinner in 1923 and substantially revised in 1958 and again in 1968 by the Aeolian-Skinner company, making it a classic of its breed. Marshall and Ogletree approached the organist and choirmaster of Trinity, Owen Burdick, with a plan. What if they produced not a box to add to an organ, but an actual organ, built with the scope of the original Aeolian-Skinner?
With Burdick's help, the project was approved, and after extensive site surveys, an implementation plan emerged. The new organ would be an interim installation, in place for about five years, until Trinity either rebuilds its original pipe organ now dismantled and sitting in storage or acquires a new one.
Like its predecessor, the electronic organ would sing out from two spots in the church: the chancel and the gallery. To handle the massive amount of polyphony necessary to blend the samples into a precise re-creation of a pipe organ, about 80 channels would be needed between the two locations, each with its own digital output, amplifier and speaker. As in the original installation, there would be two consoles, one at each site.
Marshall & Ogletree determined that 10 PCs would be necessary to drive all the channels. The 74 speakers would be placed inside the existing organ chambers, now hollowed out by removal of the destroyed Skinner pipes. The long cable that runs from the tone-generating PCs to the amplifier/speaker banks would necessitate balanced audio lines, which in turn would need studio-quality sound card technology.
Definitive Technology (Owings Mills, Md.) was chosen for the speakers that would give voice to the organ manuals. Massive subwoofers would be used in tandem with the Definitives the signal split by low-pass and high-pass filters-for the deep notes played from the organ pedals. Each speaker would be driven by its own Carver power amplifier a total of 15,000 watts.
"It was very important that the Definitive speakers were bipolar they radiate both forward and to the rear," Ogletree said. "Real organ pipes don't just radiate forward-they fill the cabinet in all directions, and turn the space around them into part of the instrument. To reproduce that spatial depth at Trinity, the speakers would have to do the same."
Consoles were ordered a relatively modest one for the gallery from Klann Organ Supply in Waynesboro, Va., and a much more ornate one for the chancel from Fratelli Ruffatti in Padua, Italy. It is scheduled for installation by Christmas. Each console would house a custom control system developed jointly with Classic Organ Works (Markham, Ontario).
Remarkably, after an official project start in December 2002, the organ was ready to play at a commemorative service on Sept. 11 of this year. The new instrument even had a name: Epiphany.
With the masses of hardware and cabling installed, the tone-generation, postprocessing and control software debugged and the instrument tuned to the space, Ogletree is pleased with the results. "Of course I have a biased opinion," he said. "But I would say that it's one of the two or three best instruments on which I've played. The experience is what a great organ should be: It is humbling, in a way staggering, to sit at the console and hear."
And what is next for the company? Interest is stirring among churches and concert halls, Ogletree said, as word of the Epiphany spreads through the small but close-knit community of concert organists. The company believes there will almost certainly be more Epiphany installations.
And the development will go on. The present samples were collected with what was then state-of-the-art equipment.
But today, said Ogletree, "the digital technology has advanced to the point that it is not a limiting factor. The limits are microphone and speaker technology, and those are really lagging behind." New microphone technology might very well trigger a new round of sample collection, he said.