Playing Old Records (No Needle Required)
By ANNE EISENBERG
Published: May 6, 2004
THE traditional way to preserve old sound recordings is to play them,
typically with a stylus, and then convert the sound into a file that can be
stored digitally. But two physicists at the Lawrence Berkeley National
Laboratory in California have developed a new way to preserve the contents
of old discs and wax cylinders: they take pictures of the groove instead of
dropping a needle into it.
The team shoots thousands of precise sequential images of the groove and
then stitches the images together, measuring the shape of each undulation
and calculating the route a stylus would take along the path.
"We grab the image and let the computer model what the stylus would have
done if it had run through the surface," said Carl Haber, a senior
scientist at the lab who led the research team in collaboration with
Vitaliy Fadeyev, a postdoctoral researcher there.
The new method may be particularly important for recovering the contents of
recordings that are too fragile or too damaged to be played in traditional
ways. It can work on disks or cylinders that have been scratched, cracked
or even shattered.
"The real excitement for me is that the method has the potential to rescue
recordings," said Daniel P. Sbardella, a sound engineer at the Rodgers and
Hammerstein archives of recorded sound of the New York Public Library for
the Performing Arts. A recording could even be scanned in bits and pieces,
Mr. Sbardella said, and then converted to audio files that can be edited to
reconstruct the whole recording.
"The mission of a sound archive is to preserve as much sound as possible,"
he said, "so even if that means rescuing a few seconds that are one of a
kind, it's really worthwhile."
The Library of Congress is financing research in the new method, now in the
early stage of development. The library holds some of the earliest sound
recordings, including many wax cylinders, said Mark S. Roosa, the director
for preservation at the library. The cylinders were made of inexpensive
materials and have not held up well, he said. Some of the cylinders and
early Edison discs are cracked or in pieces.
"This method is going to have far-reaching impact on sound archives," he
Mr. Roosa predicts that the technology will one day make a dent in the
enormous preservation tasks that libraries face. "We have thousands and
thousands of cylinders and hundreds of thousands of discs, and we are just
one library," he said.
The method involves no contact with the recording surface. After the camera
does its work, image-processing algorithms take over, detecting scratches
or spots of dust and deleting them. Then software simulates the stylus
motion, and the results are converted to a digital sound format.
"The advantage of the method is that it is completely noncontact,"
extracting information from the groove by mapping the surface, Dr. Haber
said. "You take these pictures and it's purely a software issue of how the
recording is processed after that," he said.
Both Dr. Haber and Dr. Fadeyev came to music preservation accidentally:
they are particle physicists, not music archivists. But at work they
routinely use precision optical techniques to align arrays of
One day a few years ago, a radio program that caught their attention
prompted them to consider a new application. "We heard a show on National
Public Radio on the problems of preserving delicate recordings of the
past," Dr. Haber said. He wondered whether the precision methods the group
used for particle detectors might be of use. "Why not just measure the
shape of the grooves on the surface?" Dr. Haber said, and then pose the
question to a software program: what would a needle do?
The physicists began with old 78 r.p.m. discs, on which grooves run
laterally, undulating in the plane of the record parallel to its surface.
"So from the top down, you can see the groove profile," Dr. Haber said.
The team used a commercially available electronic camera and zoom
microscope to acquire images of the grooves. But it was a slow process. It
took 40 minutes to scan one second of audio, primarily because the optical
tools were not optimized for the task. "It will run much faster when people
use a machine built solely for scanning records," Dr. Haber said.
The reconstruction of a snippet of a 1950 recording of the folk song
"Goodnight Irene" by the Weavers and Marion Anderson's 1947 rendering of
"Nobody Knows the Trouble I've Seen," both released on shellac discs, can
be heard at www-cdf.lbl.gov/~av. For comparison, the same music can be
heard there drawn from a commercial CD remastered from the original studio
tape, as well as in a playback by stylus of the original shellac disc.
Recently the team reconstructed music stored on a 1909 wax cylinder. In
cylinders the information is stored vertically, perpendicular to the
surface. "You can't tell height when looking at it from above," Dr. Haber
said, so the two-dimensional techniques used with discs would not work.
Instead, the group used a scanning microscope able to measure the height
for any given point on the surface of the groove.
The three-dimensional method is even slower than the two-dimensional one,
and much work lies ahead to develop special machines configured to match
the requirements of record scanning.
But Dr. Haber is confident that such machines will emerge, partly in
response to demands for precision measurement in many fields. "We just
wanted to prove in principle that optical methods could do this job," he