On the Wednesday before Halloween, physicist Carl Haber has gathered with several colleagues in an underground complex at the Library of Congress in Washington, D.C., for an unusual experiment: conjuring voices from the dead.
The resurrection involves a collection of audio recordings held by the National Museum of American History
, produced by the long-defunct Volta Laboratory. Founded in the 1880s, in part by Alexander Graham Bell, the lab was once a nerve center for scientists experimenting with ways to record and play sound. The recordings they produced—encrypted in objects like disks and cylinders—are part of a trove of audio relics from around the world dating back to the late 19th century. Some are too delicate to play or are damaged, even broken, while others remain trapped in obscure formats that modern devices can’t read. Containing folklore, music, and political speeches, as well as early experiments with sound recording such as number counting and poetry recitation, the recordings represent a muted history. But Haber, a senior scientist in the Physics Division at Lawrence Berkeley National Laboratory, and colleagues have devised a way to listen in on the past—by adapting digital photography and image processing techniques originally used to facilitate the search for subatomic particles.
On this autumn day, Haber and his teammates are working on an unusual disc from the Volta collection, made of bookbinder’s board topped with wax—it looks like a primitive record. Wearing purple surgical gloves to protect the media from the griminess of human hands, a curator takes the recording and places it on a clear plastic turntable connected to specialized optical equipment. She must be careful—nobody really knows what is on these files. While some recordings might contain nothing more than a simple story or a trill, others could hold more precious sounds. In 2011, for instance, Haber’s team extracted the voice of Alexander Graham Bell
himself, reciting a random list of numbers—an achievement that helped Haber win a MacArthur Genius grant
Once the disc is situated, Haber begins photographing the channels cut into its waxy surface using a high-resolution camera that creates thousands of magnified, 3D images of the surface. A computer will then mathematically interpret the images to reconstruct the route a record player’s stylus would have taken through the grooves. This virtual path gets converted into a digital audio file that can be played from any laptop or even uploaded onto YouTube. The whole process, which can take from minutes to days, is known as IRENE, for “Image, Reconstruct, Erase Noise, Etcetera.” Haber talks about the equipment as if it were another member of the team.
IRENE is the unexpected byproduct of Haber’s 30 years of experience working in the field of high-energy physics. In the 1980s, Haber and colleagues were among the first to use extremely precise silicon detectors that characterize the spray of particles in high-energy particle collisions. The world’s two most powerful particle accelerators—Fermilab’s Tevatron, outside of Chicago, and CERN’s Large Hadron Collider (LHC), outside of Geneva, Switzerland—have used these detectors to prove the existence of fundamental particles—the top quark and the Higgs boson, respectively.
“He’s really one of the big wizards in the field of silicon detectors for particle physics,” says Beate Heinemann, deputy spokesperson for the ATLAS experiment at the LHC and a colleague of Haber’s at Lawrence Berkeley.
Silicon detectors are useful because they supply scientists with precise measurements of a particle’s position in space, says Haber. Each detector is a small unit—just a few inches—and many hundreds are placed together into a large array. During assembly, researchers use high-resolution cameras to determine the position of each detector using tiny marks, called fiducials, etched into its surface. Knowing the detectors' arrangement is helpful in later shaping a consistent picture of all the particles created at the collider.
In the early 2000s, when Haber was working on ATLAS, he began to wonder, “What else could we do with these kinds of high-resolution tools?” Coincidentally, he happened to catch a radio report about how the Library of Congress’s historic audio collections were too fragile or degraded to be played and transferred to other formats. That’s when inspiration struck—why not photograph the audio discs with equipment used to survey the particle detectors?
For several years, Haber kicked around the idea with his physicist colleagues. Eventually, he suggested to a post-doc at Lawrence Berkeley named Vitaliy Fadeyev that he take a crack at recreating sound from a beat up copy of “Goodnight Irene,”
a rendition by The Weavers, a folk group led by the iconic Pete Seeger
. Using the camera system reserved for silicon detectors, they captured the surface data from the record and analyzed it to produce a sound clip.
“It actually sounded like a phonograph record,” Haber says.
Journal of the Audio Engineering Society
and sent it out as a cold call to anybody they thought might be interested. At the Library of Congress, digital conversion specialist Peter Alyea had already taken notice, thanks to an earlier draft that he had received. Haber and Fadeyev’s audio snippet, while crude, lacked certain distortions typical of other attempts to revive old recordings, says Alyea. He contacted the pair, and the three soon came together in 2003 to develop a more targeted optical system. They named it IRENE, after the song Pete Seeger’s group had sung. A 2008 New York Times
article about an early success
drew Carlene Stephens, a curator at the National Museum of American History’s Division of Work and Industry, and the Volta collection into the collaboration.
Haber and Fadeyev published a paper about the experiment in the
Haber’s team has since scanned hundreds of items. “It’s a revival,” says Stephens. “We had the stuff for decades with the thought they would be mute artifacts.”
Perhaps the biggest impact their work can have, Haber says, is to digitize early anthropological field recordings containing oral traditions from groups like Native Americans and Jews in the former Soviet Union. “That would be a great contribution to make, and I’d like to see that happen,” he says. Placing more machines around the world and establishing a community of preservationists could turn his hope into a reality, he adds. IRENE is already set up in Berkeley, Washington; Andover, Massachusetts; and Chennai, India.
Back at the Library of Congress, Haber and his Washington colleagues end up spending all afternoon on the waxy disc, managing to extract a sample of sound—coarse but accessible. The data have not been released yet, but once they are, perhaps another voice will find its way back into the chorus of history.