Optimization of switch diagnostics on the MAIZE linear transformer driver.

Abstract

The MAIZE Linear Transformer Driver consists of 40 capacitor-switch-capacitor "bricks" connected in parallel. When these 40 bricks are charged to ±100-kV and then discharged synchronously, the MAIZE facility generates a 1-MA current pulse with a 100-ns rise time into a matched load impedance. Discharging each of the capacitors in a brick is carried out by the breakdown of a spark-gap switch, a process that results in the emission of light. Monitoring this output light with a fiber optic coupled to a photomultiplier tube (PMT) and an oscilloscope channel provides information on switch performance and timing jitter-whether a switch fired early, late, or in phase with the other switches. However, monitoring each switch with a dedicated detector-oscilloscope channel can be problematic for facilities where the number of switches to be monitored (e.g., 40 on MAIZE) greatly exceeds the number of detector-oscilloscope channels available. The technique of using fibers to monitor light emission from switches can be optimized by treating a PMT as a binary digit or bit and using a combinatorial encoding scheme, where each switch is monitored by a unique combination of fiber-PMT-oscilloscope channels simultaneously. By observing the unique combination of fiber-PMT-oscilloscope channels that are turned on, the prefiring or late-firing of a single switch on MAIZE can be identified by as few as six PMT-oscilloscope channels. The number of PMT-oscilloscope channels, N, required to monitor X switches can be calculated by 2N = X + 1, where the number "2" is selected because the PMT-oscilloscope acts as a bit. In this paper, we demonstrate the use of this diagnostic technique on MAIZE. We also present an analysis of how this technique could be scaled to monitor the tens of thousands of switches proposed for various next generation pulsed power facilities.