VisualAudio

Introduction

The project is based on an idea of Mr. Stefano S. Cavaglieri, now chief technology officer at the Swiss National Sound Archives in Lugano. It is based on these two observations:

• On a phonograph record, sound is retrieved by a stylus that follows the lateral and/or vertical movements of the groove. Watching a record through a microscope, the radial displacement of the groove is visible - this means that sound information is visible.
• Many records, in particular the original radio productions (mostly unique) are in a state of deterioration, which excludes any replays by mechanical means. Hence the interest in a non-contact approach.
• The principle of VisualAudio is to take an analog picture of the record, to scan the film, and to process the digitized image using various algorithms (see Fig. 1). The University of Applied Sciences of Fribourg has been in charge of the implementation of this project.

Fig. 1: The VisualAudio concept.

The solution, with its intermediate photographic stage, solves several significant challenges inherently found in archiving systems:

• Speed of the archiving process: the time for the photo-shooting is relatively short.
• Storage of information in an analog film: thus not dependent on a technology becoming obsolete quickly.
• Long film life span of several hundred years: periodic tranfers to new data storage media are thus avoided.

History

The first documented sound recording machine appeared in 1856: the Phonautograph. Edison and Bell introduced the Phonograph in 1877. This recorder captured the sound by the means of a horn ending on a diaphragm, whose vibrations moved a needle, which engraved a soft wax cylinder.

Fig. 2: Antique turntable

Many standard rotation speeds, such as 16, 33, 45, or 78rpm, as well as manufacturers-dependent speeds, ranging from 68 to 84rpm, were introduced. The stylus sizes and shapes were also changing in time.

Before the advent of the vinyl in the 50's, the records were cut or pressed in substances such as wax or shellac. These organic materials degrade over time, and are also prone to attack by certain fungi.

Fig. 3: A few characteristics.

Photography

Photo shooting is the critical part of VisualAudio. It is performed once, at the beginning of the process, in order to archive the record as a film.

It is preferable to properly clean the record before taking the picture, in order to remove dust or foreign particles.

The photographic film has a high resolution of 600 lines per mm, which corresponds to 1200 dots per mm. This resolution is sufficient to follow the groove displacement accuratey.

Scanning process

The record content is stored on a photographic film. The challenge is to be able to recover the sound. A scanner prototype has been built at the University of Applied Sciences of Fribourg.

 

Fig. 4: Latest version of the prototype.

It is made of a glass plate mounted on a rotary motor. The film is placed on the plate, and the digitization of the image is done by a linear CCD camera of 2048 pixels wide, which takes pictures at regular intervals (with frequencies of 25,000 to 200,000 lines per rotation). This camera, combined with the rotation of the film on the plate, delivers a rotary scan and provides a rectangular picture of a ring of the record (Fig. 4). A second motor provides the radial displacement, in order to acquire the next ring.

Fig. 5: Extraction of a ring. The actual width of the ring is 2mm.

Image processing

Once digitized, the ring images are processed and analyzed to determine the positions and displacement of the groove. The first step is to correct the imperfections of the captured images. Indeed, many disturbances can come from various stages of the acquisition process: the disc itself (cracks, scratches, dust), the photography (film grain), or the scanning (dust, optics, CCD sensors).

Then, the groove position is estimated using edge detection algorithms. Once the edges are detected, corrections requiring more complex knowledge about the structure of the image are carried out. Some examples of corrections:

• Interpolation if the groove is interrupted.
• If an edge of the groove is damaged, the information provided by the other side of the groove is useful.

Sound extraction

The groove displacement is converted into an audible signal. This signal is processed by bandpass filters in order to only return the bandwidth of the original recording. Some frequency equalizations (for instance RIAA) are implemented.

The aim of this project is to archive and retrieve the sound as close as possible to the original one. Audio restauration is a feasible step that could be added to enhance the sound quality, but these techniques are contentious. Many papers describe them pretty well and there are commercial applications providing very good results. By default, these techniques are not applied.

Broken records

Many records from the 40's are definitely unplayable. The reason is that the lacquer layer is shrunk. The result is an interesting jigsaw puzzle.

 

Fig. 6: Example of a cracked record and a magnification of a part of it.

As the cracks are due to the shrinkage of the lacquer, we can assume there is no material loss in most of the cases. To solve this problem, in November 2006, we started a project funded by the Gebert Rüf Foundation. The results so far are encouraging. The algorithm basically uses signal features to determine whether two groove parts are contiguous or not.

 

Fig. 7: Two examples. The green and blue pixels respectively are the left and right walls of the groove.

This project is still in a test validation phase, but some sound is already available.

System quality

Reaching the same quality as for a good 33 1/3rpm record replayed on a modern turntable is probably utopic. The signal to noise ratio of our system is situated around 19dB for a good 78rpm record. The ultimate goal of this project is getting sound out of an otherwise forever lost record.

The advantages of such a system are:

• Allows archiving and replaying a record with no surface contact.
• The photographic film is a durable carrier, whose life span is longer than one hundred years.
• The record's state is freezed, and stored in a new format allowing to retrieve the information later using new technologies.
• The system can be used for vertical cut records.

The disadvantages of such a system are:

• Not usable for wax cylinders.
• Some irregularities such as rounded or bulging records, causing no effects when played with a traditional turntable, can influence the sound of VisualAudio.

Conclusions

This project generates a growing interest within both the archivists and the scientists communities. The last results with badly damaged records (i.e. with shrunk laquer) are particularly promising. There is no other way of replaying such records.