The Wommelsdorf Condenser Machine


Heinrich Wommelsdorf was a researcher that developed a series of electrostatic machines, in Germany, in the first decades of the XX century. Between 1902 and 1932 he published a series of papers and patents, mainly describing several versions of what he named the "Condenser Machine". The main version of this machine was characterized by a construction that was similar to a multiplate capacitor, with rotating disks with internal metal sectors forming one set of plates, and fixed plates with also internal inductors forming another set. Electrically, the condenser machine was similar to a fully sectored Voss, or Toepler-Holtz, machine, or a Belli machine with charge collectors separated from the inductors. Wommelsdorf machines were described in most of the books dealing with the subject, as particularly efficient and powerful [1][5][8][26][27]. The final versions (1920) used an open construction [27] with ebonite disks with embedded sectors accessed through slots at their edges, separate inductor plates insulated between ebonite plates, and a neutralizer circuit having a switch at its center. This was practically the last of the classical disk influence machines.

Wommelsdorf described several variations of this machine along the years, including totally enclosed machines, and machines with double rotation, similar to multiple Wimshurst machines, that were the subject of his Dr. Ing. thesis. He also developed Wimshurst triplex machines. Another interesting machine that he developed was known as the Wehrsen machine, after the instrument builder that popularized it. It was a kind of Holtz machine with a disk with embedded sectors accessed through buttons in front of the machine. Wommelsdorf had a firm named "Berliner Elektros Gesellschaft M.B.H.", that produced his machines under the brand name "Wommella".

Three Wommelsdorf machines can be seen at the Museo de Fisica, at the University "La Sapienza", in Rome, Italy. In May 2000 I made a visit to that museum, and could observe the machines in detail. Some items were impossible to observe, as the internal construction of the disks and inductors, but these could be deduced from the old texts. I then decided to build one. This would not be a simple project, as the structure of the machine is rather complicated. But I wanted to see if all that complication can lead to significant gains in performance over, for example, a simpler Voss machine.


I started by October 2000, deciding for a double machine, with two rotating disks and three pairs of inductor plates. The structure would be similar to the structure of the double machine seen in Rome, that also appears in the book by F. Luscia [27], with some small modifications and adaptations. The machine would not be large, with the rotating disks having 28 cm of diameter, but I made it with enough insulation for a maximum spark length of 20 cm (data about old machines suggested that this would be possible), with expected output current of at least100 A. I wanted to see if the doubled inductors and the two sections really increase the output current, and by how much.

I made the disks as a stack of three white acrylic disks with 2.5 mm of thickness, glued with epoxy (Araldite) glue. In the central disk, that have smaller diameter (27.5 cm), I fixed 32 sectors made of adhesive aluminum foil, 16 at each side, anternately. The sectors, 6 cm long, at 0.5 cm from the edge of the central disk, have a thin 0.5 cm tab folding over the edge of the disk, being the only area of the sectors exposed for contact. The sectors cover most of the active area, but the central disk keeps the sectors well insulated one from the other. The glue completes the insulation. The outer disks have an inclined cut at the edge, and so the assembled disks have a groove at the edge, where brushes can make contact with the internal sectors.

I made the inductors plates with paper, with a central aluminum foil strip, sandwiched between pairs of black acrylic plates 2.5 mm thick, glued with epoxy glue. They are accessed only through small holes in the plates. The internal inductors were kept at 8 mm from the edges of the plates, following their external shape, ending at a few cm from the supports of the plates.

This machine has a curious construction, with two lateral structures made of wood, directly connected to the high voltage and insulated from the base by two insulating feet and a flat plate each and one from the other by two long horizontal bars above, that complete the structure. I made the complicated lateral structures with several wood bars and blocks each joined by screws. The insulators were made with 2 cm Nylon rods and black acrylic plates, with the feet fixed to the base with threaded rods and nuts through turned wood bases, and the horizontal bars fixed by screws. The inductor plates are fixed to these lateral structures, to the horizontal bars, and to insulating supports below, by wood blocks, threaded rods and blind nuts.

The axle for the disks is made from a 6 mm steel bar (taken from a discarded printer) supported by two turned wood supports, made in two parts: Conical vertical supports, and horizontal cylinders with ball bearings. The disks are fixed to the axle by a central cylindrical Nylon block where they are screwed by three screws each, and small acrylic disks to distribute the pressure. The block is fixed to the axle by a set screw. The axle is moved by a pulley system, with a small pulley in the axle and a large pulley with a crank at the back of the machine. The speed multiplication in the pulley system is of about 5.6 times (I started with 10 times, but the leather cord was slipping excessively in the small pulley).

The terminals were made by metal spinning. One is a ball, and the other a rounded disk. They are screwed to 1/4" brass bars that slide within tubes mounted in the lateral structures. The tubes end in rounded flanges that support the bars, the outer ones made of brass and the inner ones in Nylon. The tubes are connected by screws to vertical brass bars that connect to the charge collectors below.

The machine has in front of it a complicated network of switches. Two levers at the sides connect the charge collectors to the Leyden jars, and two switches connect the Leyden jars to hooks that serve for external connections. The frontal assemblies are supported by two 2 cm Nylon bars. The switches were made with brass rods, rounded cylinders, and 10 solid brass balls (that I turned in the lathe). The bars, balls, and Nylon handles are interconnected by sections of 3/16" threaded rod. The pivots for the switches are 3/16" brass rods with longitudinal slots, that plug into holes in balls or rounded cylinders.

The fixed Leyden jars were made from plastic beer glasses. I removed their bases (that were separate parts) and made simple wood supports, fixed to the base by threaded rods, washers, and nuts, in a way that allows some freedom for adjustments in the position of the bases. The outer plates of the jars (aluminum foil) touch strips of aluminum foil at the bases, that connect to the rods that fix the bases. Wires below the base connect the Leyden jars to a point spark gap in front of the machine, where connections for pulsed output can be taken. The inner plates of the Leyden jars connect to vertical bars that go to the switch assemblies. The jars can be easily dismounted by unscrewing a ball with a threaded rod in the switch assemblies.

The neutralizer is another complicated assembly of brass tubes, Nylon cylinders, and 8 brass balls, interconnected by internal threaded rods. Brushes made of nickel-chrome wire touch the sectors of the disks, and are put in contact through a switch in the middle of the assembly. An insulated handle allows the rotation of the neutralizer with the machine running and the switch open (there is high voltage in the bars in this condition). The neutralizer is fixed to a short steel bar that inserts in the frontal upright support. A screw with a handle at the side of the support allows the fixation of its position.

All the brushes were made with thin nickel-chrome wires, just 2 to 4 wires per brush. The wires were simply inserted in thin plastic tubes made from wire insulation, folded to be exposed at the outer side of the tubes for electrical contact. The tubes were then inserted in holes in the switch supports, remaining there by pressure. The plastic tubes help in keeping the brushes in the grooves and turn them more resistant to breaking. There are 4 brushes in the neutralizer, 4 in the charge collectors, and two appropriating brushes, that charge the two frontal pairs of inductor plates. These brushes are mounted in short horizontal metal rods, insulated by acrylic blocks fixed to the wood structure, that touch contact buttons with springs at the surfaces of the inductor plates. The back pair of inductors is charged by brushes that take charge from the surface of the back disk, as in a Holtz machine. For symmetry, I added a corresponding pair of these "Holtz brushes" in the central pair of inductors too. The brushes are mounted in metal buttons that make contact with the inductors. I didn't find any document describing this system, but the same system is apparently used in the machines in Rome. The Holtz brushes and the charge collector brushes don't actually touch the disks, but are kept close to them. The others must touch or the machine doesn't start by itself.

Finally, I varnished all the wood parts with several layers of polyurethane varnish and lacquered the metal parts with shellac varnish.


I could test the machine in the first days of 2001. Initially I had to solve problems with startup difficulty (bad contacts in the appropriating brushes) and the driving cord slipping in the small driving pulley (I made a larger one). The first observation was that the spark length was rather low. With the installed terminals no more than 3.5 cm sparks were obtained. I could increase this to 8 cm adding a smaller (9 mm) ball in series with the ball terminal, separated by a short plastic tube. The same trick that I have to use to make my Holtz and Voss machines produce long sparks in humid air. A single smaller ball in the ball terminal was also useful. The disk terminal for the negative side is efficient. Balls there only reduce the spark length. The maximum short-circuit output current was measured as 40 A with 2 turns per second at the crank. This is close to the expected output current of four Wimshurst machines with the same sectors (16 per disk, covering ~1/2 of the active area) and speed. Varying the rotation speed, the current varies in approximately linear proportion, but with significant losses at low speeds. The angle of the neutralizer affects a bit the current, with maximum current at low angle. I was expecting a current close to this value, or somewhat larger. The calculation is as follows:

The total area of the sectors in one disk is calculated as 0.035 m2. When the sectors leave the inductor plates, they are equally charged, and so there is no electric field between the disks. Considering the maximum electric field in air as 30 kV/cm, the maximum charge density at the outer surface is the sectors is calculated as 26.55 C/m2. The disks turn at 11.2 turns/second with 2 turns/second at the crank. Considering that the same occurs with opposite polarity at the other side, and assuming low the current trough the neutralizers with the output in short-circuit, the charge transferred to the charge collectors by the two disks per second, that is, the maximum output current, is obtained as: i = 2 x 2 x 11.2 x 26.55e-6 x 0.035 = 41.6 A. The first doubling is due to the two disks, and the other due to polarity reversal at the collectors.

I made measurements with other configurations too, all at 2 turns/second at the crank, obtaining:
a) 1 disk, 1 pair of inductors: 8 A.
b) 1 disk, 2 pairs of inductors: 17 A
c) 2 disks, 1 pair of inductors between them: 20 A
d) 2 disks, 1 pair of inductors between them and another pair at one side: 29 A.
e) 2 disks, all 6 inductors: 37-40 A.
The cases (a), (b), and {c} appear to indicate that only about 1/2 of the maximum current for those configurations is being generated. A possible reason is the assembly of the disks, where the sectors are intercalated at both sides of a rather thick central disk. The glue, that is not a very good insulator, may be also shielding the sectors from the influence from the most distant inductor. The effect is the reduction of the area used for charge transport by 1/2. Case (d) is consistent with this.
All the measurements were made between one terminal and the neutralizer, kept grounded. Practically the same results can be obtained by measuring the current directly across the output terminals, but the machine becomes more difficult to start in this way.

Similar results, in proportion, were obtained in tests made with another machine of this type, larger (built by Serge Klein, in France, in 2000). I could obtain up to 90 A cranking as fast as possible (4-5 turns/second). This is the same output current of my powerful Triplex machine. The old tables showing the performance of these machines, however, list currents that are significantly larger. Maybe result of very high rotation speed, what also increases the spark length.

The machine shows periodical polarity reversals, with a period that decreases with the output voltage. The neutralizer switch causes an immediate reversal if opened until sparks cease to cross it (more than 1 cm) and closed again after a moment. If the switch is left open, the machine stops as soon as current ceases to flow at the output. These are two useful functions for the switch. With the output at low voltage, or in short-circuit, the machine works with the neutralizer switch open.

The point spark gap produces short sparks when sparks jump in the main terminals. The distance there affects the energy of the sparks, and can be used to control it.

I am not very confident in the insulation properties of the epoxy glue that I used to seal the disks, and some surfaces may have been contaminated by humidity before I glued them, but the insulation of the machine is not specially good, anyway. The Nylon insulators are rather hygroscopic, and the wood structure may emit corona, specially while unvarnished. The insulation between the inductors and the output is also small, and may be the cause of polarity reversals after long sparks, when the sudden voltage drop at the lateral structures can take some charge from the inductors. Low-voltage measurements of resistance shows resistances around 20000 MOhms between adjacent sectors of the disks, and between the terminals of the machine. These are quite low values for an electrostatic machine.

Trying to improve the insulation of the machine, I covered the edges of the inductors with U-shaped plastic strips. This helped, increasing the output current at high voltages. I noticed also that the machine works significantly better with larger Leyden jars. Adding a pair of 200 pF jars to the machine, adding to the normal 50 pF jars that it already had, I could extend the spark length with the normal terminals to 5 cm, and the spark length with an added small ball to 11 cm. These sparks were obtained turning the machine very fast, and with the neutralizer switch widely open, but still sparking (2 cm or more). The machine produces a burst of sparks when the ball terminal is positive, reverts polarity, produces a burst of corona at the ball terminal, and after some time reverts again, repeating the process. If the neutralizer switch is closed, sparks are smaller and the reversals occur with longer period. If it is totally open, the machine reverts polarity at a fast rate, and sparks are also smaller.

In this first version, the machine was prejudiced by the relatively low resistivity of the glue used to seal the disks and inductors. It worked reasonably well when run at high speed, when its high output current compensates for the losses, but startup was problematic. Comparing it with my other machines with disks of this same size, its performance was just regular. Current was as expected and the spark length, although smaller than expected due to losses, was as high as in the other machines. The frequent polarity reversals were a problem, probably caused by insufficient insulation between the inductors and the terminals, but this is almost impossible to avoid in a machine with fixed inductors.

Improvements in the design would start by better insulation, using a different system for sealing the disks and inductors, and by replacing the Nylon insulators. The distances between the disks and the inductors could be smaller, or adjustable, so an optimum value could be located. They are at about 5 mm, what looks high for a machine of this size, but I believe that smaller distance would only result in quicker startup, with larger distances resulting in longer time between polarity reversals. The distances in the original machines were also large. The distances between the disks and the inductor supports could be increased for better insulation. The frontal switches could be longer, to free the connection hooks from the Leyden jars that are partially below them. The central disk of the disk assemblies could be thinner, for more uniform electric field at both sides of the sectors. The Leyden jars could be larger.


I made a new terminal ball, a bit smaller and with a small brass ball (1.2 cm of diameter) directly fixed with a screw. The result was somewhat better than with the glued 9 mm ball, with sparks reaching 12 cm in a dry day. This kind of terminals, ball-plane terminals, require less voltage than regular two-balls terminals for the same spark length.

By May 2002 I made several changes in the materials used in the machine, to improve the insulation. First, I replaced all the critical insulators, that were made with Nylon rods, by Teflon, replaced the leather driving cord by a polyurethane cord, and rebuilt the disks and inductors. I disassembled them, sanded the surfaces clean, and installed new sectors (a bit smaller in width) and inductor plates. The disks I sealed with hot glue, applying beads around the sectors with a glue application tool and remelting the glue in an oven, with the disk assemblies pressed between glass plates. This resulted in some deformation of the acrylic plates, mostly shrinkage, but the disks were still usable, and now with good insulation. I didn't glue the inductor plates, but just covered the paper and metal inductors with two layers of adhesive plastic foil, leaving them between the two plates that form each inductor. The insulation obtained in this way was not enough, and the machine could not generate sparks with more than 4 cm until I reinstalled the covers at the edges of the inductors. With this the machine exceeded its previous performance, reaching 13.5 cm sparks easily, and 50 A of output current at 2 turns per second at the crank. The maximum current, limited by practicable cranking speed, easily exceeds 100 A. Some differences in operation were also observed. The machine works even with the neutralizer switch completely open, and takes longer time to revert polarity, even with its regular Leyden jars. The machine is quite powerful now, although very easy to crank.

Further improvements would be to make new disks, sealed with something that doesn't cause deformation of the acrylic plates, and maybe to make also new inductors, adequately sealed, to eliminate the edge covers. It's interesting to note that the ebonite disks on old machines that I have seen appear to have been made with the groove at the edge machined after the disks were built. The edges of the internal sectors were visible in the groove, being enough for contact. The inductor plates were not glued, but had thick insulating plates folded over the edges.

An old machine

In June 2001 I found an old Wommelsdorf machine in a museum at the Federal University of Juiz de Fora (Juiz de Fora, Minas Gerais, Brazil). It is a small machine with a single 26 cm disk, of the "Wommella" brand, almost identical to the classic machine described by Wommelsdorf in 1920 [p84]. These pictures show a frontal view and a back view of it. The lateral vertical supports and the base are made of wood, with two ebonite insulators at each side and an horizontal ebonite bar above the disk, Two ebonite arms at the sides support the Leyden jars and output switches. No hook is present below the switches. Switches connect the charge collectors to the leyden jars. The output terminals, a ball and a disk, are directly connected to the charge collectors through insulated bars. The disk and the insulators are made of ebonite. The inductors are made from a flexible black material, not very thick, possibly celluloid. They are not sealed, but composed of three pieces, one forming a kind of outer envelope, involving an inner plate, and a section insulating the outer edge from the machine wood structure. In the inner plates there are paper inductors, glued, covering metal strips ending in pins that cross the insulating plates, that are to be inserted in holes in the ebonite structures holding the appropriating brushes. The inductors are held in place by thumb screws, fixed to ebonite blocks mounted in the lateral structures, to the upper horizontal bars, and to a block in the base. The brushes are all made of thick bunches of thin metal blades. The charge collectors brushes are mounted directly in the wood structure. The machine was not in working conditions, but restorable.

My machine: Front view, right side, back view, left side.

Created: 27/01/2001
Last update: 09/01/2003
Developed and maintained by Antonio Carlos M. de Queiroz

Return to Electrostatic Machines