- An attempt to remove background noise, clicks, locked grooves and such... -

Version Française

English Version


In April 1877, Charles Cros wrote a paper describing his invention of the paléophone, the first machine capable of recording and playing back sounds, based on the principles of Léon Scott de Martinville, who managed to record sounds without playing them back. In late 1877, Thomas Edison also discovered the possibiliy of recording sounds, and built his first phonograph. With a sheet of tinfoil covering a grooved cylinder, Edison's phonograph did not allow playing back records more than a few times without destroying them. Later in the 1880s, Bell and Tainter improved Edison's phonograph by replacing the fragile sheet of tinfoil with a wax coated cylinder, which could be played many times. They called their machine graphophone. In 1888, Thomas Edison decided to improve his phonograph using Bell and Tainter's wax cylinder, and built his perfected phonograph. The first specimens were offered to important people such as Colonel Gouraud in England, Gustave Eiffel in France, etc. Other people were called by Edison for recording worldwide celebrities such as Brahms, Tchaïkovsky, Aton Rubinstein, etc. Incredibly, most of these recordings survived the years and can now be heard using modern techniques.

This page is about one of these recordings: a white wax cylinder sent to Edison by Colonel Gouraud. This cylinder contains a piano performance called Soft Piano Solo by Mrs Eyre, whose quality is surprising for its time. As a matter of fact, it was difficult to make piano recordings which had a decent sound quality (without excessive distortion). One explanation for the quality of this cylinder is the wax itself: being a derivate of a metal soap, it was much softer than later wax, which could allow a better cut.

Several steps were done to clean this amazing wax cylinder. I say amazing because most of the wax cylinders recorded with piano never gave a good piano sound. This one appears to have been recorded in amazing conditions, probably the very soft white wax gave better results, though very fragile.

Before trying to do any sort of filtering, hiss removal and such, I have removed silent spaces before and after the recording. This reduced the size of the WAV file, and so the time taken by later operations. I have then manually removed the clicks I could hear and see in the wave form. I tried to do some automatic click removal, but my manual declicking happened to be better in most cases. Still, I passed an automatic declick on the manually declicked file to remove whatever the software would find clicking. Finally, I tried to remove the locked groove. This gave a first file, from which I did several attempts of filtering and hiss removal. These attempts are listed in the below below, along with the cleaned files without filtering.

Filtering gave good results. I did three types of low-pass filterings: 2000Hz, 3000Hz and 5000Hz. 2000Hz gave the best results but seemed to be alter the piano sound, though this is not very audible. 5000Hz seemed to leave the piano unchanged, but left some surface noise.

Hiss removal is an amazing helper. It removed a lot of the surface noise, but some of it still remained and played like a robot-like digitized sound. Filtering over a hiss removal gave much better results. Here again, 2000Hz removed most of the surface noise. 3000Hz gave slightly better results but some of the "robot" hiss remained. 5000Hz was better for the piano, but not recommended if you don't like the hiss sound.

I have therefore compiled a list of MP3 files. The first one is the original manually declicked file, without locked groove and silent spaces. The second one had a hiss removal applied. The rest are my attempts of filtering with and without hiss removal.

Cleaned piano (no filtering, no hiss removal) Cleaned piano with hiss removal
Cleaned piano, 2000Hz low-pass filter Cleaned piano with hiss removal and then a 2000Hz low-pass filter
Cleaned piano, 3000Hz low-pass filter Cleaned piano with hiss removal and then a 3000Hz low-pass filter
Cleaned piano, 5000Hz low-pass filter Cleaned piano with hiss removal and then a 5000Hz low-pass filter


These first trials gave quite good results in my opinion. I personally prefer the version with 2000Hz low-pass filter (no hiss removal). However, the piano is better preserved at 5000Hz, but the surface noise is more audible. It is surprising to see how well one could record a piano at that time. Several bass frequencies can be heard in this recording. Of course, there are many programs that can remove clicks, hiss and such. I did not experience them, and it is clear that much better results will come out of a longer study of how to clean the Soft Piano Solo original file. Any suggestion will be very welcome.

David Winter