Carl Haber and Earl Cornell, Berkeley Lab researchers, digitally recovered a 128-year-old recording of Alexander Graham Bell’s voice.
Carl Haber (left) and Earl Cornell, Berkeley Lab researchers, digitally recovered a 128-year-old recording of Alexander Graham Bell’s voice. Photo: Berkeley Lab

What does winning sound like? The definition of winning for MacArthur Genius Award recipient and Lawrence Berkeley National Laboratory physicist Carl Haber is surprising.

“Winning” is not the unrestricted $625,000 the Berkeley scientist will receive over the next five years from the John D. and Catherine T. MacArthur Foundation for his revelatory IRENE/3D Project. And Haber’s “winning sound” is not even the miraculous audible recording of Alexander Graham Bell’s voice he and his colleagues, Earl Cornell and Peter Alyea, pulled from a 128-year-old, wax-and-cardboard disc.

What rings Haber’s bell, is bounding out of his office chair to inscribe arcs, arrows, “x’s” and wavy, snakelike lines on a nearby whiteboard in an all-out-effort to explain his project to a visitor.

“Particles collide and you can discover things from how they scatter,” he says, scribbling a collision point and surrounding it with “sensor” marks representing energy-absorbing silicon. “The shower of particles, they happen 40 million times a second, so you need a fast mechanism to capture them.”

IRENE (Image, Reconstruct, Erase Noise, Etc.) photographs the grooves in fragile or decayed recordings, stitches the “sounds” together with software into an unblemished image file, and reconstructs the “untouchable” recording by converting the images into an audio file.

Haber’s “fast mechanism” is a high-resolution digital camera capable of measuring the movements of particles smaller than a human hair. The image analysis produces mountains of data, but Haber says, “It’s not so far flung.”

The physical science, on its own, may not be outrageous to a man who spends his days measuring subatomic patterns. Stuck in a traffic jam in 2000, Haber heard a KQED broadcast about Library of Congress audio recordings so fragile they could not bear the ravages of a stylus. “If you played them back with a needle, they’d be ruined,” he says. “We were measuring silicon, why couldn’t we measure the surface of a record? The grooves at every point and amplitude on a cylinder or disc could be mapped with our digital imaging suite, then converted to sound.”

One frustrated driver’s daydream became revolutionary science. IRENE holds enormous implications for scholars, historians, curators, audio collectors and populations made “mute” by moldy wax cylinders or acid-leaching lacquer discs.

At the Northeast Document Conservation Center in Andover, MA, a pilot project using the system will digitize “instantaneous recordings” — one-of-a-kind, live recordings of native languages, musical performances, political speeches and radio broadcasts. Many of them have been held in archival collections and never played for fear that a contact approach would ruin them.

Large institutions like the Library of Congress can afford to purchase Haber’s machine for their own use, but for smaller organizations, a central server like NEDCC is vital. A Woody Guthrie Archive the NEDCC will digitize has held the discs for years, waiting for technology to release them from their fragile state. Carnegie Hall has a small collection of damaged radio transcript recordings that, until now, were unrecoverable. Recent improvements to IRENE’s software prevents the cameras from going out of focus or “jumping” when reading over a crack (resulting in “lost” data). Instead, technology enables the virtual stylus to ramp smoothly from a record’s one damaged level to the next.

Carl Haber and Earl Cornell, Berkeley Lab researchers, digitally recovered a 128-year-old recording of Alexander Graham Bell’s voice.
IRENE, the machine used by Carl Haber and Earl Cornell, Berkeley Lab researchers, to digitally recover a 128-year-old recording of Alexander Graham Bell’s voice. Photo: Berkeley Lab

NEDCC Executive Director Bill Veillette says IRENE teaches the value of being patient — just like the ground-penetrating radar used before shovels in archeological digs, and the x-ray technology applied to artifacts instead of dismantling them to understand furniture joinery.

“It’s human nature to be impatient, but if it means potentially damaging an historic artifact in the process, it might be better to wait until a technology inevitably comes along to make the discovery possible,” he said. “It’s ironic that as we put more time between us and the history we are exploring, technology allows us to learn more than if we had acted earlier.”

Kathryn Bunn-Marcuse, Assistant Director at the University of Washington, Seattle’s Burke Museum, says the IRENE project recently digitized 24 cylinder recordings of Kwakwaka’wakw songs made in conjunction with 1930s films by anthropologist Franz Boas. She says Haber and his staff donated “critical time and expertise” to preserving their “significant archival holdings.”

Working with descendants of Boas’ original films to create a DVD, Bunn-Marcuse is combining the digitized audio files with archival photos and film of native communities’ First Nations dances. “The project aims to redress colonial histories by ensuring the community of origin has access to all archival material as well as an active part in shaping the final publication,” she said. “Close collaboration will ensure that the new presentation will not re-inscribe outdated approaches of salvage anthropology.”

Closer to home, Haber says 60 of the 3,000 native Californian recordings made by Alfred Kroeber and held at the Phoebe A. Hearst Museum of Anthropology are an “extended world of traditional cultures.” Sound recordings were a novelty in the early 1900s and ethnographers feverishly created ten’s of thousands of recordings. “They’re at risk in playback, but if there was enough of this equipment,” Haber says, gesturing at the nearby turntable, mandrel, and black, bridge-like, “big stage” arm holding the 2D or 3D camera, “you could imagine an international program for things otherwise lost.”

Hearst Head of Collections Victoria Bradshaw says many California indigenous languages are seldom or no longer spoken. Generations of transfer — from cylinder to open reel to cassette tape — have compromised the recordings. IRENE will allow the Hearst to return to original source material to recapture the cultural information. And the new, clearer recordings will be emotionally significant for descendants who may hear the voice of an ancestor for the first time.

“In the past few decades there has been a resurgence in cultural revitalization for California Indian tribes. The audio recordings are a top priority. I look forward to the day when we are able to provide high-quality recordings,” Bradshaw said.

Haber is making gradual improvements, transitioning IRENE from an expert orientation, to a robust system that will work in the hands of a trained technician. Separate scanning systems for cylinders and discs were designed and the blocky “big stage” eliminates the need to recalibrate when switching from 2D to 3D.

“I’ve had the privilege of being able to look over the shoulder of Bell and Edison and see how they invented information technology,” he said. “Since then, we’ve just refined what they captured in real time. In a scientist’s lifetime, if you get just one idea, you’re feeling pretty good.”

Watch Carl Haber discuss his work in the Berkeley Lab video below:

Would you like a digest of the day’s Berkeley news in your inbox at the end of your working day? Click here to subscribe to Berkeleyside’s free Daily Briefing.