Version: March 15, 2021

Technical requirements for transfer and processing

by Alfred Meurer, 2021


Playback styli

For playing back 4-min-cylinders and Lioret celluloid cylinders, the original diamond stylus of the Ortofon OM-78 has been used. Width and rounding of its tip are already suitable for these formats.

To be able to play 2-min cylinders, whose groove width is about 0.25 mm, we had to manufacture appropriate stylus tips out of hard borosilicate glass and mount them on the needle carrier of the pickup system. The styli in use have diameters between 0.20 and 0.15 mm and a hemispherical rounding of the tip. These significantly sub-calibre dimensions enable the styli to go down to the very bottom of the grooves even if these are damaged or show casting defects or other deformations. Usually, the acoustic information is better preserved on the bottom of the grooves than on the flanks. In addition, by touching only the bottom, friction is reduced, thus reproduction generates less surface noise. The use of small-diameter styli is also important for playing sections of very high amplitude. Resulting of the round cross-section of the cutting styli used for phonographic recordings, the width of the groove cut into the wax increases with its depth. Therefore, the up and down movement of the groove is always accompanied by symmetrical lateral movement. In overmodulated areas, the lateral movement of a groove intrudes into to the neighbouring grooves, diminishing their width and distorting the acoustic information of the flank section in question.

When playing overmodulated sections, which appear frequently with moulded records, sub-calibre styli deliver the best results, since they always go down to the bottom of the grooves but barely touch the flanks deformed by neighbouring grooves.

A close-up of the surface of an Edison moulded record. The groove section to the left of centre shows an overmodulated area whose strong lateral movement extends significantly into the neighbouring grooves. A broad stylus tip would not go down to the groove’s bottom here.

Stylus pressure

We usually use a stylus pressure between 1.5 g and 3.5 g. In this low-pressure range, abrasion caused even by sub-calibre tips on soft-wax cylinders is negligible. For comparison, the most common original acoustic pickup for 2-min moulded records, Edison's Model C, has a contact weight of about 35 g. And the tiny diamond tip of an Edison Diamond B soundbox, about 0,12 mm in diameter, puts an even much higher pressure on the groove of a Blue Amberol cylinder.

Post-processing the digital signal

When listening to the unprocessed audio signal provided by high-quality electric cartridges, one notices that most cylinders already sound quite natural. However, the electric signal contains much more hiss and rumbling compared to the playback on an original mechanical phonograph. This noise is largely composed of frequencies beyond the sensitivity range of a mechanical phonograph sound box. These frequencies were never recorded, they are a mere result of surface friction, partly caused by impurities and abrasion by former use. Nevertheless, all those unwanted frequencies are precisely transferred by modern phonograph pickups. Such mere noise frequencies can be removed without harm to the wanted acoustic signal. The task is, however, to determine the borderline between noise and recorded signal. Measurements of the frequency response of an ordinary Edison sound box carried out by the German Broadcasting Archives in the 1980s have shown a responsiveness only between 100 and 5000 Hz, with a clear preference for sound frequencies in the region of 500 to 1000 Hz. Beyond these limits, rigorous filtering should be possible without damage. However, listening tests do not confirm the measurement results. In the case of particularly clear-sounding cylinders, there still seems to be a recorded signal beyond 5000 Hz. Low-pass-filtering from 5000 Hz upwards leads to a duller auditory impression than filtering significantly beyond this limit. As a consequence, we refrained from too rigorous trimming of the frequency response and chose a 1600 Hz safety margin to the upper frequency limit determined by the German Broadcasting Archives. We filtered the raw signal with -48 dB beyond 6600 Hz and below 60 Hz. For mechanically copied soft wax cylinders and Lioret cylinders, we lowered the upper limit for the low-pass filter to 6200 Hz without audible negative effects.

But what to do with the remaining acoustic information? We have to decide about what we desire to have as final result. Should it be a frequency response, ...

1. - which imitates the frequency response of a mechanical phonograph and thus essentially amplifies those frequency ranges which were already favoured by the oscillation behaviour of the diaphragm during the original recording?

2. - which attenuates the preference of an acoustic phonograph for frequencies around 1000 Hz and instead raises the flanks of the curve that fall steeply to the left and right of the 1000 Hz peak, thus generating a linear curve between 60 and 6600 Hz?

3. - which amplifies the sound information remaining after the above-described denoising processing evenly without favouring any section of the frequency range?

4. - which strives for a "natural impression" to be determined by listening experience and individual taste?

Good arguments can be found for all alternatives. Filtering according to model 2, however, leads to a rather intolerable result, especially with regard to the upper end of the frequency range. Beyond about 4000 Hz, the sound information recorded on cylinders is only very weak. If we try to amplify this part of the curve to such an extent that its amplitude matches the strongest part of the curve, the significant noise of this now preferred frequency range is also extraordinarily amplified. As a result, we obtain strong hissing, that sounds very unnatural, because the hissing components beyond 6600 Hz are eliminated by the low-pass filter.

Our solution was number 3. Such a treatment neither tries to alter the originally recorded sound characteristics, nor to imitate a historical playback phonograph. Nor does it transform the result to suit anyone’s personal taste. A straight amplification of the recorded signal read out by a modern, high-quality pickup is honest in a sense that it represents the largely neutral modern reproduction of a recording made with all the characteristics of a historical process.

The filtering operations described were performed with the digital version of a graphic equalizer and with low-pass and high-pass filters. In addition, an individually adjusted click filter was used, as well as a carefully applied noise filter to reduce the inherent noise.

The program used for these editing operations is called Audacity – a non-commercial and thus royalty-free audio program – like our website.

Determining the correct speed

In the years before 1902, no general standard for the rotational speed of phonograph cylinders existed. The speed was determined according to the few parameters inherent to the medium: Decreasing speed extended playing time while causing stronger wear and lower sound fidelity; increasing it resulted in slightly better sound quality, volume and durability at the expense of playing time.

Quite at the lower end of the spectrum of cylinder recording speeds are Henri Lioret's celluloid cylinders with mostly 100 to 120 rpm. The fineness of the groove in the robust celluloid material made it possible to achieve good sound quality and volume even with a relatively low surface speed compared to other cylinders.

Soft wax cylinders from the 1890s to around 1902 usually have a speed slightly faster, often between 120 and 140 rpm.

With the introduction of the casting process for mass duplication around 1902/03, the market-leading American producers Edison and Columbia and their European subsidiaries settled on a uniform speed of 160 rpm. The new cylinders thus run faster than the bulk of their softer predecessors, limiting their playing time to a maximum of just below 2 ½ minutes while providing greater volume, improved sound quality and reduced surface wear. Adopting the casting technology, most other cylinder manufacturers followed Edison's and Columbia's lead. However, since even the moulded cylinders were often not recorded at exactly the standardized speed, and many a slower recorded older recording was re-released in the early days of the moulding process, even explicitly stated 160 rpm cannot be relied upon (similar is the case for setting the rotational speed of shellac records at 78 rpm). With a numerically precise definition of this speed, one can be quite precisely wrong when playing a cylinder.

To determine the playback speed, only an accurate ear helps. The correct speed is determined when human voices and instruments sound most natural and familiar. A great relief for the determination of the rotation speed is the very widespread habit of an introductory announcement anteceding the recording. An incorrectly chosen playing speed is quite noticeable right at the beginning. If the score of a piece of music is known, the recording can also be tuned to the correct key. However, a musical piece may have been transposed, so in any case: listen critically!

The revolving speed of all recordings published on my site Wachston (see link below) was determined by ear, nevertheless most moulded cylinders produced by Edison, Columbia and Pathé were in fact played at 160 rpm. The audio files published there are MP3 versions of 32-bit WAV files. As a first step, an unfiltered raw version was created by directly digitizing the output signal of an Ortofon OM-78 magnetic pickup developed for shellac records.


  Write to Alfred Meurer: Visit Alfred Meurer's site Wachston: