With a vintage Edison Standard Phonograph cranking out tinny Hawaiian hula songs in the background, he put the moldy cylinder under a microscope and pointed out the grooves that were still visible underneath the mold and other damage.
"We might be able to fix that," said Haber, a researcher for the University of California's Lawrence Berkeley National Laboratory. "There is still information there."
Haber and Berkeley Lab colleague Vitaliy Fadeyev are working on a breakthrough way of digitizing and archiving old recordings, such as wax cylinders and traditional flat records, that are too far gone for a standard stylus. If successful, the pair may be able to help archivists at The Library of Congress and elsewhere rescue swaths of recorded musical and audio history that are today in danger of being lost.
They're using technology that is more traditionally applied to their "day jobs" in high-energy particle physics, creatively reapplied to extract the information contained on the old recordings without touching the needle of a record player to the fragile surfaces.
The Library of Congress has given them a grant to explore the technique. They've presented it to archivists at the New York Public Library for the Performing Arts and at Stanford University, each of which has large collections of old recordings. The response has been promising.
"This is exactly the kind of innovation that sound archivists need," said Hannah Frost, a media preservation librarian at Stanford University who is familiar with the pair's work. "If we don't have to physically touch the surface of the fragile materials but can find an alternative way to read them, it has great implications for preservation."
Getting into the grooves
Haber's team creates sensors that are sensitive enough to detect the passage of a single subatomic particle released inside a high-energy particle accelerator, Haber said. The sensors, which look like a little like shiny aluminum plates just a few inches across, are connected to custom microchips that record where and when the particle was detected, and send this information to central computers.
Haber's team constructs these sensors at the Berkeley Lab facility, using some tools similar to what large chipmaking companies like Intel might use for their own quality-checking process. One of these is an optical scanner, used to take a digital picture of the sensors at a microscopic resolution and check quality of the production.
Four years ago, the researchers heard a story on the radio that piqued their interest in the music realm.
"We heard on NPR that there was this issue with audio and old recordings, and how they're preserved," Haber said. "We thought we could do this same analysis with records."
The optical equipment used in the labs already took finely defined pictures of a flat surface. They tried turning it on old vinyl records instead of on their own sensors and came up with high-resolution pictures of the way the grooves on the records wiggled back and forth--changes that are turned into sound by the needle of a record player.
Over time, they customized their own software to analyze those wiggles and turn them into usable digital data--and from that into digital audio.
"The groove has a well-defined geometry," Haber said. "If there is dust or a scratch, (the analysis) just skips that place."
The next step was to try a three-dimensional image. Old flat records were comparatively easy, since the back-and-forth wiggle of the grooves posed few challenges to the $75,000 optical scanners used for two-dimensional images.
Older Edison wax cylinders, by contrast, have straight grooves that create sound by varying their depth. To capture that information, a scanner needed to be set up to take a large number of "slices" across each spot on the cylinder, capturing information at various depths. These individual pictures could then be combined to create a 3D composite picture of the grooves and again turned into digital music.
The result, they found, were digital files that actually sounded considerably better than the original records. (Audio comparisons of the optically scanned files to digital files recorded with a traditional turntable and stylus can be found here.)
The process is far from being ready for commercial development. The Library of Congress grant money, which is aimed at improving the 3D version of the technique, will run roughly through the end of the year and is aimed, in part, at figuring out whether the process could be feasible for an archive like the Library's.
Certainly, the Library, which has the largest public collection of sound recordings in the United States, is potentially the biggest beneficiary from the technology under development. According to the Library's Web site, its collection includes 47,000 wax cylinders, 650,000 78rpm records and nearly a million other records in various formats.
That collection includes music and other historical documents, including a wax cylinder recording of Germany's Kaiser Wilhelm II acquired in 1904, about 2,500 records made between 1943 and 1945 by U.S. servicemen serving in the Pacific, and a collection of rare operatic recordings made in Russia before the revolution.
"I think the tools that they're developing will really help the sound-archiving community turn a corner into the 21st century," said Mark Roosa, director for preservation at the Library of Congress. "We know the concept works. Now, it's a question of refining the tools and the software, and then developing a prototype."In theory, the process could be used much more widely--perhaps even on the vast collections kept by record companies of original masters and outtakes from recording sessions. But Haber said he and his colleagues are sticking with the archives for now, not wanting to get involved with copyright or ownership issues.
"With all the stuff going on with record industry, I'm not going there," he said.