Sound Archive
About the Sound Archive
The Sound Archive Project investigates optical methods for scanning archived, historical sound recordings to preserve and reproduce the audio digitally. It focuses on the very earliest recordings, such as:
- Tinfoil and wax recordings created using Thomas Edison’s Phonograph
- Wax cylinders made using Alexander Graham Bell’s Graphophone
- Coarse groove discs such as Emile Berliner’s earliest Gramophone
Many of the artefacts that the project works with are fragile, unstable and at risk of deterioration and as such deemed ‘unplayable’ by conventional playback methods. Dragging a record stylus over them would be unthinkable; the pressure on the surface could cause the recording to shatter, or further damage the groove which holds the vital audio information.
One such recording, made in August 1888 using Bell’s Graphophone, has been historically verified as belonging to Queen Victoria. It is the only known recording of this Royal, and represents a priceless piece of cultural significance.
This important cylinder appears to have been damaged in the past by a conventional stylus and as a result, the Sound Archive Project has worked extensively to preserve this surface and attempt to recover her majesty’s voice from the damaged groove.
The philosophy adopted by the project aims to make a preservation copy of the full recorded surface, not to provide real-time playback. Sound is later reproduced from computer analysis of the digital representation of the whole surface, enabling damaged areas to be recovered using specialist mathematical techniques.
The Sound Archive Project focuses on the following areas of research:
- Optical Surface Measurement – development of metrology systems for mapping the surface topology of cylinders and flat discs
- Audio Signal Recovery – methods of accurate sound reproduction from discrete surface maps of cylinders and discs
Queen Victoria Voice Recording
The Science Museum holds a Graphophone recording which purported to contain the voice of Her Majesty Queen Victoria. Preliminary work had been carried out to recover the audio using conventional stylus in 1992, but results were poor.
Following contact with the Royal Archive, the Sound Archive Project received confirmation that the cylinder was almost certainly Queen Victoria’s voice. Provenance was referenced in the forms of a book entitled “The Lost Voice of Queen Victoria” (written by Paul Tritton and published in 1991 by Academy Books) and a letter held in the Archive from the Queen’s Private Secretary, Sir Henry Ponsonby, to his wife, Mary, dated 29th August 1888.
A man friend of Miss Bauer came here yesterday with a Graphophone. It is different from Edison’s phonograph and has been made by Bell (who we once saw with Johnny), his brother and some others and is very ingenious – no electric or magnetic currents – simply works like a sewing machine with a treadle. Little cylinders revolve which a little machinery marks with your voice – and the contrary returns your voice through a pipe into your ear. Only one can really hear it, but it is very curious. Wernher, one of Queen Victoria’s sons-in-law, spoke in German, Edwards whistled and I laughed – my ‘coachman’s laugh’ – it was most extraordinary the clear way this was reproduced – as often as one liked. The Queen said to me at dinner ‘I heard your hearty laugh this evening’. This was 6 hours after & he says it will keep for years. HM spoke into it – but we told Mr Morse he must not go round the country reproducing the Queen’s words (Sir Henry Ponsonby, Queen’s Private Secretary)
The cylinder was clearly of cultural significance and almost certainly the only recording of Queen Victoria’s voice in existence. To attempt further replay using a contact technique would risk any chance of recovering the audio more clearly. The Sound Archive Project therefore moved their cylinder measuring equipment to the Science Museum and carried out a full surface scan of the artefact.
Subsequent post-processing of the scan data at the University highlighted wear to the grooves of the cylinder, possibly caused by earlier attempts of stylus playback. There was obvious damage to the cylinder recording at the end of the track where a stylus has jumped out of the groove and continued to score the surface. It addition, cross-sections of the groove indicate that the audio track has been cut into again, either with a second recording, or by an incorrectly used stylus.
Work is currently underway by the Sound Archive Project to reconstruct the damaged areas of the recording from untouched areas of the side wall of the groove. This state-of-the-art reconstruction may yet unlock the secrets of Her Majesty’s voice.
Tinfoil Recording
Thomas Edison is attributed as the creator of the first practical Phonograph in 1877. A foil sheet of tin was wrapped around a drum that had a single continuous groove machined around it.
As the drum was hand cranked, the user would speak into the mouth-piece causing the needle to follow the underlying groove while vibrating up and down against the tinfoil in sympathy with the sound. This produced a ‘hill and dale’ impression in the foil, as opposed to the more traditional side-to-side groove associated with modern discs such as Long Play (LP) records.
These tinfoil recordings were very fragile, and even when originally produced could only be played back two or three times before the stylus would perforate the groove.
In summer 2008, the Sound Archive Project received a tinfoil recording belonging to the British Library Sound Archive reputed to contain the voice of Harriet Martineau. The artefact was in a badly damaged condition having been roughly folded into thirds across the sound grooves. One third was detached, while the other two had almost separated. In addition, edges were crumpled, some segments were folded back on themselves and there were tears.
The research at the Sound Archive Project has produced a surface metrology instrument capable of scanning delicate foils using a high precision, optical probe.
The foil, already unwrapped from the recording drum, is laid flat on an air-bearing table and moved back and forth under the probe to measure the height of the foil thousands of times per line scan. By repeatedly measuring lines over the surface, each spaced just microns apart, the whole foil can be mapped.
Once mapped, specially designed software algorithms analysis the groove, following its height changes and converting this information back into an audio track.
In the case of this badly damaged artefact, some sound was extracted from the foil, clearly a female voice. Sadly, without extremely high-end analysis, the audio could not be made more intelligible.
Wax Cylinder Recordings
The Graphophone and latter versions of the Phonograph preferred wax as the recording medium, usually as a coating over a paper or cardboard tube. Wax was more robust and cylinders could be removed and swapped by pushing them onto a slightly tapered spindle.
Techniques were created that allowed the recordings to be duplicated and mass produced for sale. The wax was ultimately replaced with harder compounds such as celluloid (used by the Indestructible Record Company) and a material called Blue Amberol (an early form of plastic akin to Bakelite) by the Edison company.
While the plastic-like materials survive well, many of the most historically significant recordings were made on hardened wax which still has limited playbacks, is fragile and particularly prone to deterioration through fungal mould attacking the surface. As a result of this, the Sound Archive Project undertook a task to create an optical cylinder surface scanner.
Working in a similar manner to the air-bearing system, used for recovering tinfoil recordings, the wax cylinder is loaded onto a rotating spindle and a confocal probe scanned along its length. After each linescan, the cylinder is rotated a fraction of a degree and the process repeated.
Once the entire artefact is scanned, the cylinder’s height data is flattened out and ready for software techniques to find and follow the groove, recreating the audio.
As a comparison between the Sound Archive Project’s metrology solution, and the traditional stylus technique, a 1905 commercial recording called “Beautiful Birds Sing On” (9022: Edison Gold Moulded Record) was measured.
Both versions were provided for comparison and it is clear that the optical technique provides better results, without the need for post-production in the form of equalisation or noise reduction. The most significant benefit of the optical system is that there was no risk of damage to the fragile surface.
Berliner Gramophone Recordings
Emile Berliner was the first person to adopt the now familiar disc recording format, known as the Gramophone in 1888. A continuous spiral groove on flat circular disc, played at a speed of around 70 rpm.
His first commercial recordings, released under the Berliner Gramophone label in Canada during 1892, comprised of a five inch diameter disc pressed on one side only. By 1901, ten inch records were being produced by the now renamed Victor Talking Machine Company; these bear a close resemblance to the now familiar 78’s.
When the Sound Archive Project team were designing their air-bearing metrology system, it was specified to allow the measurement of both flat and rotary mediums. As well as the x, y, z axes necessary to linescan a tinfoil flat sample, the instrument has the additional of a rotary axis, θ.
Partner Organisations
Research is undertaken as a collaboration between the University of Southampton, TaiCaan Technologies Ltd. and the British Library Sound Archive. Additional partners who have supported the project include EMI, Poppy Records and the National Screen and Sound Archive of Wales.
People
- Professor John McBride
- Dr Kevin Cross
- Professor Martin Hill
- Dr Peter Boltryk
- Dr Antony Nascè