How do you play a broken record? “Take a picture of it,” says audio engineer David Ackerman, who heads the Audio Preservation Services (APS) unit in the Loeb Music Library. Ackerman explains that new technology (developed for IRENE—the Image, Reconstruct, Erase, Noise, Etc. project of the Library of Congress, based at the Lawrence Berkeley National Laboratory at the University of California, Berkeley) can optically scan a record groove. Then, with high-resolution digital maps of the shapes of the grooves, computers build digital files that reproduce the audio data on the original record, in many cases rescuing sounds from the realm of seemingly permanent silence.
Ackerman monitors such emerging technologies, which might someday come to Harvard, as part of his mission within the Harvard College Library—preserving rare, even unique, audio materials and making them accessible. His little-known group of four audio engineers quietly does for sound recordings what art conservationists do for paintings: preserve, clean, repair, restore, and ensure that they live on.
But sound preservation involves computer data files, because the work nearly always means transforming an analog recording into a digital one. Although none of Thomas Alva Edison’s wax cylinders has yet presented itself, the analog materials do come in widely diverse formats, including disk records that are made of aluminum, shellac, acetate, and vinyl; reel-to-reel tapes in both acetate and polyester-based formulas; cassette tapes, digital audiotapes, Dictaphone databelts, minidiscs, and even eight-track cassettes. (The music library does own a couple of boxes of wax cylinders, but APS does not own a wax-cylinder player.)
The output formats are varied, too. The engineers can burn CDs and CD-ROMs, as well as audio and video DVDs. They can make analog tapes and create files in three Internet-friendly modes: Real Audio, MP3, and MP4. (The last two are playable on iPod devices; the music library keeps several, so patrons can borrow one, go into a listening room, and work directly with MP4 files that APS has reformatted and placed on a library server.)
Recordings show up in a range of conditions. “The Harvard College Library has enormous holdings of audio and visual material,” says Ackerman. “It is scattered throughout many units and has been cared for differently in different places. Storage conditions vary widely, from good to very bad. We’ve seen tapes that were stored in a hot, humid attic. Some recordings are unplayable.”
Take certain audiotapes manufactured during a particular period in the late 1970s and early 1980s that tend to suffer from “sticky shed syndrome.” With age, the tape’s backing gets sticky and starts to peel off, shedding the ferric oxide on the recording side of the tape, where the magnetic signal resides. “You can do things like ‘baking’ those tapes in a laboratory convection oven,” Ackerman says. “If done properly, you may be able to recover all or some of the magnetic signal.”
Photograph by Jim Harrison
“Sticky shed syndrome” causes inadvertent loss of data, but deliberate loss of data, as in the compressed-file formats used to save memory space, can pose even more vexing problems. “Compressed file formats are a nightmare for us,” says Ackerman. “It’s about throwing the data away.” He explains that closed proprietary file formats like those used by Apple’s iTunes build in copy protection—which, of course, makes copying and preserving the file difficult. “Studios have a vested interest in seeing those things go out of print and become unavailable,” says Ackerman. “Then demand for them can build again. But suppose, say, they are bought out by a company that decides to blow off preservation activities. Twenty years out, that could also blow off a chunk of our cultural heritage. The next generation, which might only have heard compressed, auditorily diminished recordings, wouldn’t have a benchmark of how some song originally sounded, and might even ask, ‘Why are we spending this money to preserve?’ ”
That question doesn’t arise with regard to the rare, obscure, and often irreplaceable sound recordings that APS handles. For example, the Woodberry Poetry Room’s collection of spoken poetry recordings, one of the world’s largest, includes readings by Jack Kerouac, Marianne Moore, Ezra Pound, the young Robert Frost, and even Alfred Lord Tennyson. (Some readings are available at http://hcl.harvard.edu/libraries/houghton/collections/poetry_room.html.) The Harvard Iranian Oral History Project has about 900 tapes of interviews with 134 eyewitnesses, including political and military leaders, to important historical events in Iran from the 1920s to the 1980s. APS digitized these recordings and deposited 100 gigabytes of data into Harvard’s on-line Digital Repository Service, managed by the Office for Information Systems. APS is also at work on special collections, like the James Rubin collection of South Indian classical music and the Laura Boulton collection of Eastern Orthodox liturgical chants. Any recorded sound is fair game, including the voices of birds and tree frogs.
APS can protect original recordings by making “listening copies.” For example, a request to hear some part of the Rubin collection may arrive at the music library, but “You don’t want to give someone a reel-to-reel tape and have the tape machine snap it or stretch it,” Ackerman says. Instead, APS reformats the music for patron access.
One fascinating project is the Milman Parry Collection, recorded in the 1930s in various Slavic regions of eastern Europe. Milman Parry (1902-35) and folklorist Albert Bates Lord, the late Porter professor of Slavic and comparative literature, laid down these tracks with two machines that made direct, instantaneous recordings onto aluminum records. (Each disk could store only about four minutes of sound, after which the researchers threw a switch to continue recording on the other machine.) The sung tales they preserved, which form the basis of Lord’s 1960 classic, The Singer of Tales, can go on for 45 minutes to two hours, so “a single tale can take up to 30 to 40 aluminum disks,” Ackerman explains. From these aluminum disks, APS creates very large digital preservation files, and then makes a lower-resolution copy for Internet distribution.
“We have people on our staff with a lot of experience working with antiquated audio formats,” says Ackerman. To play back the old materials, they also need some vintage high-end equipment like the versatile, studio-quality Studer and Ampex reel-to-reel tape recorders that date from the early 1980s and late 1970s, respectively, which can handle any tape speed from 1 7/8 to 30 inches per second, and read signals with full-, half-, or quarter-track tape heads. (At 8 Story Street, APS works in studios formerly used by the National Public Radio program Living on Earth, hosted by Steve Curwood ’69.) A technician in Boston comes in every few months to keep the machines properly tuned up, but people with the skill to care for such technology are becoming scarce.
Although abrasion causes analog tapes to lose information each time they are played, backing up in the digital world is easy: it simply means copying computer files. But the files are typically huge. APS owns a RAID (Redundant Array of Independent Disks)—45 hard drives capable of storing four terabytes of data (a terabyte equals 1,000 gigabytes) that provide a safety net for each other if one drive fails. Those hard drives work hard: each night it takes hours to back up the mountain of data generated by two shifts of engineers who work day and night, five days a week.
Furthermore, the audio data “need to be digital in a way that you can characterize,” says Ackerman. APS typically migrates its digital data into a standard personal-computer format, Broadcast WAV (Waveform Audio Format) files, but “in 100 years it may be impossible to find equipment to play them, so we try not to lock ourselves into a physical format,” he notes. “Anybody with a sound card can take analog information and turn it into something digital. We want to make the best-quality digital copy that an analog-to-digital converter can produce. And we want to maintain the integrity of the process, to preserve the relationship of the digital output to the original object. Partly we can do that with metadata, ‘data about data.’ That’s information that tells you something about your primary audio data—the sampling rate, where the tracks are, what kind of signal processing was used. We keep track of everything we do, so you can trace the relationship to the original.
“It’s one thing to take a spoken text and make it sound good, but did you edit it or leave something out?” he continues. “Did you move something to a different place, or insert or swap some things around? Did you merge two different takes of the same thing to get a better-sounding version? We don’t do those things; we maintain the authenticity of the content through the transfer process, and that sets us apart from everybody else. If something is not playable from beginning to end, we reassemble it in a way that you can hear it, [but we] do that in such a way that you can take it apart and edit it yourself—you can get your hands on the original. We also encourage our clients to keep the analog stuff after the transfer.”
APS’s painstaking work qualified it to participate, along with Harvard’s Archive of World Music, as a secondary partner on a National Endowment for the Humanities research and development grant, “Sound Directions: Digital Preservation and Access for Global Audio Heritage,” led by the Indiana University Archives of Traditional Music. “We’ve written a suite of software tools for managing the audio preservation process,” says Ackerman. Their report will appear this spring.
Much of what is preserved digitally goes onto websites, where hundreds of recordings are available now at the click of a mouse, raising intellectual-property questions. “A lot of recordings can’t be made available due to copyright issues,” Ackerman explains—and for religious reasons, too. “It may not be ethical,” he notes, “to put the Native American creation chants up on the Web”; such a decision calls for curatorial judgment. But in the studio itself, a simpler kind of judgment prevails; as Ackerman says, “The ear is the final arbiter.”