Investigation Of Spark Gap Electrode Failure

Many facets of spark gap electrode erosion have been studied. In this case, abrupt electrode failure is being investigated. In certain high power spark gaps, a slight change in the operation of the switch has catastrophic results. The electrodes have noticeable pieces of material blown off. It has been suggested that the damage is a result of the material limitation related to the dddt that the spark gap handles [l]. A test circuit has been created to investigate this phenomenon. A Marx bank provides a pulse with a maximum of 120 kV and 12 kA, with a relatively fast rise time. First, any damage to the electrodes will be correlated with the voltage, current, and dddt of the system. Next, a passive magnetic delay technique will be applied to the current pulse to reduce or eliminate the electrode damage. Magnetic delay techniques have been used successfully in a number of situations with high power switches to reduce erosion or to allow a conduction channel to establish before the !Xl circuit current flows [2,3]. In this case, it will be used to reduce the damage to the switch at high dddt. The paper will provide test circuit simulation and operation, along with diagnostic results (with and without magnetic delay) Further in-depth diagnostic evidence will be provided if available. INTRODUCTION Once erected, the Marx bank fires into a load of 8 R in series with the test switch. The test switch gap spacing is set so that the output gap self breaks. The load resistor is actually six water resistors in parallel arranged in a circular manner. Copper sulfate solution is used, with a concentration of 200 grams per liter of deionized water. Water resistors were used for the load as an inexpensive alternative to low inductance, high voltage resistors. Peaking capacitors, (approximately 4 nF), were placed in parallel with the load to provide a boost to the rise time of the output current. The equivalent circuit can be seen in Figure 1. To study the graphite failure mechanism, a system with a high di/dt must be employed. In this case, the target dddt is 350 Nns. The system is designed for a medium current, with a moderately fast rise time, for the following two reasons: (1) to correlate the failure to high dddt and not high current (2) limited availability of very fast diagnostic equipment At maximum charging voltage, the system will deliver a 12 kA pulse with a 30 ns rise time. Current rate of rise is proportional to voltage and inversely proportional to inductance. A Marx bank is used to provide a high enough voltage to the load to overcome the inductance of the system. It also allows a range of output voltage to the load. Although the system geometry was designed for minimal inductance, peaking capacitors were included to insure a fast initial current. The first goal is to establish consistent failure of the graphite electrodes. After that, steps are taken to reduce the liielihood and amount of damage. One simple parameter to change is the size and geometry of the electrodes. Another method is to use a saturable reactor in the output stage to delay the rapid rate of rise in the current. EXPERIMENTAL SETUP The Marx bank has three stages; each stage has a total of .3 uF comprised of two .15 uF, low inductance capacitors in parallel. The bank switches are PI 650 spark gaps, which are pressurized to at least 30 psi to avoid prefire, and triggered with an L-C inversion high voltage generator [4]. The bank is resistively charged through water resistors customized for the geometry of the system. Figure 1. Equivalent Circuit Carefid attention was paid to the inductance of the system, since it is crucial to the dddt. The inductance of the experiment is dominated by the geometry. With this in mind, it was designed to be as compact as possible, keeping voltage holdoff and maintenance in mind. Inductance of the Marx bank and output stage are calculated to be 156 nH and 130 nH, respectively [5]. The experimental setup is shown in Figure 2. Current, current derivative, and voltage measurements were taken over a range of charging voltages. A current transformer is used on one of the ground leads to measure the output current of the system. The sensitivity of the current transformer is 0.1 V/A, and the bandwidth is 35 MHz. A b-dot probe measures the dddt of the system. The probe is a coaxial cable with the inner conductor exposed and curled back to the braid to create a single loop with a quarter inch diameter. The sensitivity of the probe is dI -= 1.5.109Y dt Output voltage is measured using a low impedance resistive divider, using a string of two Watt resistors. The low resistance provides a wide bandwidth; and since the divider is used for pulse measurements only, it is in no danger of large power dissipation. The voltage divider has a ratio of 5000:l with a bandwidth of at least 35 MHz. If the output is connected to 50 R, there is an additional division of a factor of two.