Version: June 27, 1999

The evolution of a new cylinder reproducer

by Lloyd Stickells, 1984

Engineer at the British Library National Sound Archive


This picture is of a later version of the player than that described in the text, the main difference being that the basebord here sits on a cabinet which houses the power supply and the circuit boards.


As well as the 2,100 wax cylinders in the Frazer collection (described by Alan Ward in Recorded Sound 85) the British Library National Sound Archive holds several other collections including approximately 500 cylinders recorded by Dr. Arnold Bake in India, Ladakh, Tibet and Ceylon (1931-41) and the 106 cylinders deposited in the Archive by the English Folk Song and Dance Society which include recordings by Lucy Broadwood and Ralph Vaughan Williams. Following the decision by the governors of the then British Institute of Recorded Sound to begin a programme of copying this precious material onto tape, I began work to construct a suitable new machine on which to play cylinders. The resulting reproducer is also suitable for use with commercially published cylinders, of which the Archive holds a small selection of around 2,000, and collections of cylinder field recordings held by other organisations such as the Museum of Mankind and the Universities of Cork and Dublin, which are on loan to the Archive to be copied.

The 'Brown Wax' cylinders used for field recording are vulnerable if mishandled; they are easily scratched or broken. Less traumatically destructive perhaps, but often more insidiously damaging can be the results of poor storage conditions. Wax cylinders are a rich source of food for various forms of mould and fungi, which colonise patches and, in bad cases, the whole surface, giving affected cylinders the appearance of having been rolled in flour. What is beneath the surface, though, is of greater concern than the surface dust, which can be removed. The associated pitting of the grooves is almost impossible to remedy, and the resulting surface noise usually renders any remaining modulation unlistenable.

Because many of the cylinders the Archive holds were rescued from attics, cellars and boiler rooms, and since part of the Archive's function is to make available its holdings for research, study, and to the public at large, it has for some time been a matter of high priority to get the cylinder recordings transferred to a more common and ledd destructible medium, curently open reel tape.

Sound encoded in the cylinder grooves can be transferred to tape either acoustically or electrically. For acoustic transfer, using an unmodified reproducer, only a microphone is needed to pick up the sound emanating from the horn and to treat the cylinder machine as a 'live' performance. This method has a superficial simplicity but also severe shortcomings, not the least of which is the unacceptable wear of the grooves caused by a stylus tracking weight reckoned in more ounces than most modern cartridges track in grams. In addition, the resonances of the reproducing diaphragm and horn may differ from those of the recording horn and diaphragm, reinforcing or cancelling each other, or both. Another drawback is that the machine ideally needs to be in a soundproof and acoustically dead room, or extraneous noises and room ambience could become part of the 'performance'. (If one wishes to reproduce the sound as a listener in a drawing room might have heard it, this is certainly the method to use.) However, to realise the full potential of the recording in the cylinder grooves direct electrical transfer is the only sensible method. There are several ways to do this. One way is to substitute a magnetic pick-up cartridge for the stylus diaphragm and horn on a standard cylinder machine. Alternatively the carriage may be modified to take some form of parallel tracking arm driven by the lead screw mechanism.

Another popular, simple and effective method is to mount the cartridge in a modified i.e. straight, LP pick-up arm, mounted so that the cartridge traces in a gentle arc across the top of the cylinder as it tracks from one end to the other. (Hi-fi buffs might be horrified at the tracking errors this method may produce, but since modulation is vertical, the effect is not as bad as might be imagined, and almost certainly not noticeable above the distortions inherent in the cylinder grooves.) Either approach may benefit from substituting a suitable electric motor for the normally clockwork mechanism of the original.

While both these methods are capable of commendable results, and always give better results than an acoustic transfer, the wider frequency range capability of a magnetic pick-up tends to show up surface noise in the treble region, and the mechanical clatter of the often gear-driven mechanism, combined with the less than precision bearings of most cylinder machines, adds to the wow and flutter inherent in most recordings. After experimenting around the first option I decided to build a device from scratch in order to avoid repeated playings. (It is, or was, possible to buy an off-the-shelf attempt at a 1980s cylinder machine by Art Shifrin, but the performance and facilities offered did not seem worth the $5000 asked!)

The requirements for the necessary device are simple to state: a mandrel on which to mount the cylinder, running in noiseless bearings; a drive system capable of constant, yet controllable speed; and a means of tracking the cartridge along the cylinder, preferably parallel to the centre axis. Various sized cylinders should be accommodated and an indicator should show the playing speed of the cylinders.

Some cylinders (particularly Blue Amberols) have groove pitches of 200 to the inch rather than the standard 100, so a tracking machine would need a two speed capability while locked to the mandrel speed. A pivoted arm is of course self tracking, but apart from the theoretically incorrect path it traces, the necessarily long pivot to stylus length can lead to tracking problems with mis-shapen cylinders ­ and unfortunately many if not most cylinders do warp with the passage of time ­ so this option was discarded.

Some form of parallel tracking device was called for. There are a number of such arms in LP players but, perhaps understandably, most companies seemed reluctant to supply spare units for purposes for which they were not intended. Finally a complete arm and control electronics from a Revox B 795 turntable was obtained. In common with other parallel trackers, the Revox arm is servo-driven along sliding runners. It differs from others in the length of the arm, since in its normal use it sits over the record, not at the back of the turntable. This means it has no arm in the accepted sense; the cartridge is fixed into what almost amounts to a uni-pivot headshell. This results in a very low inertia cartridge mounting, but it does restrict the up and down positioning leeway (which was subsequently found less than that required to cope with excessively fat cylinders).

The tracking length of the carriage was just enough to cover the grooved length of an Edison cylinder, but the internal diameters of cylinders vary enough to cause their longitudinal positioning to vary by 3/4" or more. Fortunately, the arm lends itself to the simplest solution to this problem - lengthening and tracking rods - so the originals were replaced with longer ones.

The whole assembly can be swung horizontally through ninety degrees; this is a necessary manoeuvre in its normal LP playing role to enable the record to be put on the turntable, and useful attribute in getting cylinders on and off the mandrel without risk of damaging the stylus.

Before the arm could be tried on a real cylinder, a mounting for both the cylinder itself and the mandrel, with a motor attachment was needed. For initial trials a simple bracket was made to carry both the arm and a mandrel borrowed from an Edison. I chose a Revox B 77 capstan for the drive motor because of its excellent speed control characteristics and slow running capabilities. With a 2:1 step down drive pulley attached to the shaft and a couple of minor modifications to the control electronics, a mandrel speed variable between about 100 and 250 rpm was achieved, with the added option of halving the speed range in the 'slow' condition, providing a total range between about 50 to 250 rpm ­ more than adequate to cope with most recording speed variations. (In work carried out so far all cylinders have been recorded within the 80-210 rpm range.) The Revox capstan motor has an 'engraved' rotor, which provides the feedback signal for speed stability, and which may also be used as the basis of a speed indicator; the machine now sports a digital speed read-out in rpm.

With this still crude device trials could begin. The arm behaved well; it tracked the cylinders with eccentricities of 2 mm or more. The Revox arm seemed to cope without the excessive thumpings that circumferential ripples give rise to in 'high inertia' arms. Perhaps because of warping with time, some cylinders end up with skewed grooves.

Two things cause the arm problems at present. One is a locked groove. The other is an ungrooved space, when the pick-up sometimes skates across and activates the self-lift mechanism. The restricted vertical movement meant provision had to be made to vary the pick-up to mandrel spindle spacing as required. This did not prove a major engineering problem; the hole for the spindle was made into a slot to allow vertical adjustment.

Having established that the arm seemed capable of all that was hoped for, the next area for improvement was the mandrel bearing, which the relative quietness of the arm had shown to be unsatisfactory. The rather slack brass to steel bearing gave rise to audible rumbling and clanking noise in the output. Rather than adapt the borrowed mandrel, a new one was machined from aluminium bar. The obvious choice of ballrace bearings was rejected in the (possibly erroneous) belief that ballraces are, or soon become, noisy, and so PTFE lined plain bearings were used. Whilst not totally silent, they proved to make far less noise or rumble than the brass bearings on the original mandrel.

The requirement to transfer 5" concert cylinders meant another machining exercise for a mandrel and a redesign of the arm and motor mounting. The arm mounting bracket needed to be taller so that the 4 1/2" diameter mandrel could be mounted, and the variation in spindle height this required meant that the motor had to be mounted at the side of the arm on top of the baseboard, instead of beneath it as before, so that the same drive belt and motor position would do for both mandrel sizes.

Whilst the mechanics are better than those of some other 'home made' machines, two cylinders found in the Fox Strangways collection recorded in northern India in 1911 proved to have such silent and rumble free surfaces that they contained some remarkably realistic bass drum beats, I would have suspected this was due to some electronic fault in the pick-up. They certainly reinforced my belief that most rumble on cylinder reproduction is in the recording. The machine can be improved, though as it stands it will not degrade cylinders played on it. The next important modification is to increase the arm length to enable dictaphone cylinders to be played.

This text (without the image) was originally published in the British Library National Sound Archive's journal Recorded Sound, No. 86 (July, 1984).

Write to Lloyd Stickells:
Main page