Historical story

Historic sound can be heard in a detour

Old sound recordings have been ready to be listened to for over a hundred years, but the records and cylinders are too fragile to play. Researchers found a solution:create a virtual version and play it. For example, recordings by Alexander Graham Bell can now be heard.

More than a hundred years ago, scientists and inventors experimented with recording sound. Since the magnetic tape did not yet exist, they used all kinds of other materials, such as glass, shellac and stannool (metal foil). Unfortunately, these sound carriers are very delicate, which means that the recordings can no longer be played.

At least not in the usual way. Researchers from Berkeley (University of California, USA), the National Museum of American History and the Library of Congress succeeded in making the sound audible via a 3D scan.

Record, but not play

The first sound recordings were made by pressing a needle into a cylinder of wax:the wax roll turned under the needle and the louder the sound, the more the needle depressed the wax. Thus, the depth of the groove depends on the volume. The rate at which the depth changes (the number of bumps and dips per centimeter) corresponds to the tone.

The cylinder was later replaced by a flat record, the gramophone record. The grooves are also made with a needle, only now the needle does not go deeper/shallower based on the sound, but to the left or right. You can see this very well from the bends in the groove (see below).

To play back the recorded sound, you normally have to run a stylus along the groove and convert the movements back into sound. But very old cylinders or the gramophone records can be damaged by this. They are made of such a delicate material that the needle can deform the grooves.

Based on photos

The American research team therefore came up with something different to listen to old recordings:they do not play the real cylinder or record, but a virtual version of it. They photograph or scan the original support, analyze the grooves and determine how a needle would pass through them. In this way, the computer calculates how the sound should sound.

In 2003 Berkeley presented the first version of the technique, taking close-up pictures of the grooves with a special camera, as you saw above. The photo must be very accurate, because a groove is at most two millimeters wide and the deviations to the left and right are therefore very small. However, these deviations determine the sound, so if the photo is not correct, the sound will also not be correct.

The researchers tested the system with a recording from around 1950 that could also be played back in the usual way. Below you will find both versions:

Gramophone record with 'Goodnight Irene' played with an old-fashioned record player:>>> Goodnight Irene – record player

Reconstructed version of 'Goodnight Irene':>>> Goodnight Irene – reconstructed

As you can hear, the reconstructed version doesn't differ much from the original. Even less noise can be heard now. This is also apparent from analyzes on the computer (see below).

In 3D

This technique works well for gramophone records, but you cannot reconstruct recordings on a cylinder with it. After all, you don't see depth in a photo. The researchers therefore also developed a 3D variant. They do not work with ordinary close-up photos, but with 3D images made with a confocal microscope (see below).

Below you see a piece of cylinder in 3D, with a view from above at the top right. The graph at the bottom shows how the depth changes when you follow the indicated line. This is in fact just such a sound wave as before with the gramophone record.

The researchers then made the reconstructed sound wave audible again and compared it with a previously played version:

Cylinder with 'Just before the battle, Mother' played with an Archeophone player:>>> Just Before the Battle - cylinder player

Reconstructed version of 'Just before the battle, Mother':>>> Just Before the Battle – reconstructed

Also in this case the reconstruction is as good or even better than the original.

Glass plates

This month another milestone was reached with the optical reconstruction technique (optical because it is based on images). The researchers succeeded in making six recordings from the Volta Laboratory Associates audible (see box).

“The recordings were made on a variety of materials, including rubber, beeswax, glass, tin foil, and copper, as the inventors were trying to find a material that could trap sound,” said Carlene Stephens, curator of the National Museum of American History. “We don't know what's on it; we have only a few cryptic notes on some records and some vague notes in an old Smithsonian catalog.” With the optical reconstruction method we can now hear what the Associates recorded:

(More recordings can be found on YouTube)

The reconstruction of the six recordings is a first step:the aim is to digitize as many historical plates and cylinders as possible. The Smithsonian houses about 400 of the oldest recordings ever made. They can provide insight into everything from the process of invention to speech patterns in the late nineteenth century.

See also:

  • On the website of the IRENE project you will find a lot of information and (sound) examples (in English).
  • More information on the Smithsonian website (in English)
  • More information about the December 2011 results