Rotary Spark Gaps

Many high voltage systems have a need for a repetitive switching action at fairly slow rates, i.e. less than 1000 per second. Examples of such systems would be radars (applying power to the magnetron) or tesla coils (impulse exciting the tuned circuit primary). The rotary spark gap is a mechanical means of repetitively shortening and lengthening a spark gap for switching purposes.

Why use a rotary gap?

Sparks are a fairly low loss conductor, particular when short and when carrying high current. Rotary spark gaps are typically constructed so that the gap distance at closest approach is overvoltaged by a factor of 3 to 10. Spark gaps also have the advantage that their "turn on" time is very short, substantially less than a microsecond. Rotary spark gaps are relatively inexpensive and require very little support system electronics, unlike hydrogen thyratrons (which require heater and reservoir power, as well as triggering electronics). Rotary spark gaps do have some jitter in their timing, not only from mechanical sources, but also due to the probabilistic nature of the spark initiation.

The rotary spark gap is best suited to a capacitor discharge or resonant charging type system, because spark gaps are really a "make" switch, and not very good at "breaking" a connection with significant current, because of the arc being drawn. If the load is such that the current oscillates, and the speed of the switch is chosen so that the current goes through a minimum as the gap separates.

Rotary spark gaps are usually operated in air, although sealed rotary gaps have been built for aircraft radar applications, where the change in air pressure would result in large fluctuations in the breakdown voltage of the gaps.

Mechanical configurations

There are a number of mechanical configurations possible for a rotary spark gap. The schematic diagram symbol often used for a rotary gap implies one of the configurations, shown in the following figure. As it happens, this is a poor design for most applications, because as the electrodes inevitably erode, the gap configuration changes, changing the timing characteristics of the gap. Axial endplay in the motor or support bearings also changes the gaps. A gap of this type has been used at very high powers, at 10 megawatts and 1000 Amps, for example, where the gap was set somewhat greater than 0.05 inches. In that application, the electrodes were estimated to need replacement every 1000 hours.

radial pin rotary gap

A better approach, in general, is to have the electrodes move past each other, being parallel and overlapping some distance. There are several approaches to doing this. One, illustrated below, is to have a conductive rim for the rotor and connect to it from a "sparking sector" which is an electrode spaced a small distance from the rim. Brushes could also be used, but erode very rapidly in this sort of application. The sparking sector works well, and with a sufficiently large size spaced very close, the losses are negligible. A rotary gap of this type was used extensively for pulse powers of 500 to 3000 kW. The gap between the sparking sector and the rotor rim was set in the range 0.010 to 0.050 inches. The capacitance between the sparking sector and the rim is fairly large, compared to that between the pin electrodes, so most of the voltage appears across the pin gap.

parallel pin gap with sparking sector

Another approach is to have pairs of fixed electrodes, either arranged so that the moving electrodes pass by the fixed ones, as in the above scheme, or so the electrode ends approach. This latter approach has the same disadvantages as the scheme where the moving electrodes are radial. An advantage of the parallel pin gap is that the motor/bearing axial play doesn't affect the gap distance.


parallel pin rotary gap

This latter approach is also convenient for systems where multiple circuits must be switched simultaneously, as in a Marx generator configuration. In such an application, the mechanical tolerances are substantially tighter, as the pins must be aligned with more than one set of fixed electrodes.

Rotary gaps have also been built using a perforated insulating disk which spins between two fixed electrodes. It is easy to control the interelectrode gap in this sort of system, but erosion of the insulating disk is a real problem. Also, for lower voltages, the disk must be quite thin so that the gap being interrupted is short enough to reliably break down.


Electrode wear

There is some experimental data for rotary spark gaps as used in radar sets from Bell Labs and the MIT Radiation Laboratory. The cathode wear for a gap operating in air at pulse currents of 40 to 170 Amps has been measured at 2-6 mg/amp hour, with the median value being 3-4 mg/amp hour. In a Nitrogen atmosphere, with no oxygen, the wear rate was 10-20 times greater. The anode wear was 1/5 that of the cathode.

Analysis of timing and jitter

Figure xxx shows the two electrodes as they approach and recede. This figure and the following equation assume that the moving electrode is moving along a straight line. The distance between the electrodes is:

Distance = SQRT( DistancefromClosestApproach^2 + (DiamElectrode+MinSpacing)^2) - DiamElectrode

s = omega * r * (t0-t)

(assuming Sin(Theta) = Theta )

The E field strength between the electrodes is then:

E = V / Distance

The time jitter of the gap can be held to less than 10 microseconds with proper gap design.

The impact of thoriated electrodes

If some degree of ionization in a spark gap before breakdown can be assured, then you don't have to depend on electrons emitted from the electrode surface to start the breakdown. The tungsten rod electrodes used in industrial arc welding are typically alloyed with a small amount of Thorium which is radioactive. The TBD particles emitted by Thorium will make the ionization of the gap before discharge more consistent. This will tend to reduce the jitter.

Mechanical stresses in the rotor

The mechanical design of a rotary spark gap should address the sizable stored energy and damage potential in a system with a rotor of significant mass spinning at relatively high speeds. Particular attention should be paid to the method by which the electrodes are attached to the rotor. The "centrifugal" force on these pins is:

Force = mass * omega^2 * radius

For a 10 cm radius (8 inch diameter) disk, spinning at 3600 rpm (omega = 377 radians/sec), the acceleration is 1450 G. A tungsten pin 2.5 mm (0.1 inches) in diameter and 5 cm (1 inches) long weighs about 5 grams. The force on the pin is about 70 Newtons (15.6 lbs). If the disk is made of 1.6 mm (.0625 inch) thick G-10 glass/epoxy, the load on the board is about 2500 psi, well within the allowable design stress for G-10 (around 15,000-20,000 psi). However, if the pins were to be made 1 inch in diameter, they would now weigh about 500 grams, and exert a force of 1600 pounds. The bearing surface is increased to .0625 square inches, making the load some 25,000 psi.

The gap should be surrounded by a "scatter shield" of some material that can contain the parts of the rotor should it come apart during operation. Polycarbonate (Lexan) makes a suitable shield, as would aluminum, steel, or fiberglass. Opaque materials also have the advantage of supressing the significant UV light emitted by the sparks, although, in some applications, the visual appearance of the gap may be desired (e.g. display Tesla coils).

Some rotary gap design data

The following table is taken from Früngel, who in turn took it from Lebacqz.

Electrical Characteristic Resonant Charging DC Charging
Voltage Range (kV)
Peak current (A)
Normal operating voltage (kV)
Normal operating current (A)
Normal average current (mA)
Pulse recurrence frequence (pps)
Pulse duration (µSec)
Pulse power (kW)
# of fixed pins
# of rotary pins
Diam of pins (inch)
Motor speed (rpm)
7-16 5->9 12-22 12-22 5->9
160 >90 220 220 >90
12 8 ~18 ~18 8
120 80 180 180 80
45 100 130 130 70
410 825 800 800 750,375
0.9 1.5 0.9 0.9 1.125,2.25
600 360 1500 1500 300
1 1 2 2 2,1
7 14 10 14 4
0.100 0.188 0.125 0.150 0.125
3500 3535 4800 3400 5600


Früngel, pp 140-149

Glasoe & Lebacqz, Pulse Generators, pp 276-282, pp286-292

Copyright 1997, Jim Lux / 18 Oct 1997 / rotgap.htm / Back to HV Home / Back to home page / Mail to Jim