Research home page

Arc Oscillator

     Oscillations in arcs have been observed since the turn of the 19th century, most notably by the Danish engineer Valdemar Poulsen. Generally such oscillators take advantage of the differential negative resistance characteristic of an arc. That is to say, the arc voltage drop is roughly fixed, so that greater currents passing though the arc produce lower resistance. This instability in resistance makes it possible to transfer energy from the arc to a resonant circuit. Oscillators based on this property tend to be very harmonic rich, and have unstable characteristics. They also tend to be inefficient.

     In the course of studying spark gaps, I observed a few basic properties that lead me to conclude that an oscillator could be designed in a much more direct and efficient manner. Those properties were as follows.

• At a given static voltage, a spark gap breaks over and the resistance drops from a very high value to a low one.
• At a given dynamic voltage, a spark gap will break over, at a value lower than the static breakdown voltage. This is called the restrike voltage.
• A spark gap stops conducting when the current through the gap reaches a very low value.

     C1 and R1 form a power supply for the spark gap. C1 is very large compared to C2, in my case about 10 microfarads. R1 established the effective impedance of the power supply, about 250 ohms here. The spark gap is a pair of thoriated tungsten rods in air, adjustable so that a good breakover voltage can be established within the range of the circuit components. L2 and C2 form the resonant tank circuit of the oscillator, and those values are chosen to given a particular frequency at an impedance very close to the resistance R1. That is to say, sqrt(L2/C2) ~= R1. L2 is a single layer wound on a ferrite core forming a high Q low distributed capacity inductor.

     Here's a picture of the complete apparatus.

     That thing to the right is a home-brew HV DC power supply, using a microwave oven transformer, diodes, and capacitor. Voltage probes are positioned to show the voltage across the spark gap on channel 1, and the voltage at the center of the resonant tank circuit on channel 2. The spark gap is the white tube in the center with the hex nuts.

     The circuit is operated by charging the cap to the breakover voltage of the spark gap. The oscillograph below shows the first few cycles of oscillation that ensue.

     The gap fires at about 620 volts, as can be seen from the initial value on channel 1. As the gap begins to conduct, the tank circuit discharges, producing a counter current in the gap. When the tank current reaches peak value the current exactly matches the DC current from the arc power supply and the gap turns off. The voltage across the gap then rises, limited by the value of inductance in the tank. In this circuit, at this frequency, the restrike voltage is 376 volts. When the restrike voltage is reached the gaps fires again and another cycle of oscillation occurs. This process continues until the voltage on the power supply cap drops below the restrike voltage as shown in this scope shot.

     Note that the roughness of the traces is due to the limited scope sampling rate at this timebase. None of that modulation exists in the real signal as can be seen by using the second timebase to magnify portions of the base signal. Remember, scopes can lie! What is true is the starting and ending voltages, and the time of total discharge, about 3 milliseconds. The oscillating frequency is 500KHz.

     Were this circuit powered by a DC supply capable of supplying enough power at the rated voltage, you'd have a continuous RF oscillator. Such a circuit would need some kind of cooling for the gaps as they would have to dissipate heat that can be safely ignored in this pulse circuit. The circuit can be made to work over a broad range of values. At the low end, frequencies below about 10 KHz have restrike voltages close to the DC value, making it difficult to achieve stable and harmonically pure oscillation. At the high end, the restrike voltage becomes very low and construction of the tank circuit can be challenging. But in principle very high frequencies can be attained.