Replacing a gas switch in a high-voltage pulsed-power system requires more than selecting a semiconductor with a suitable current rating. The switching architecture must also manage voltage sharing, timing, and the pulse characteristics demanded by the load. A study of a MOS-gated thyristor Marx generator examined whether staged semiconductor switching could reproduce some of the operating conditions of an accelerator generator’s thyratron-based switching unit.
The work focused on a unit in the ETIGO-IV pulsed-power generator at the Extreme Energy-Density Research Institute, Nagaoka University of Technology. Its thyratron switching unit operates at a reported peak voltage of 30 kV and peak current of 10 kA, with a pulse width of 2 μs. Those are challenging conditions for an individual semiconductor device, motivating a modular approach in which multiple switches contribute to the output.
Why use a Marx generator for pulsed power?
One straightforward way to raise a semiconductor system’s operating voltage is to connect switches in series. However, a failure or imbalance in a conventional series arrangement can expose an individual device to excessive voltage stress. A Marx generator provides a different circuit approach: its stages are charged and then switched so their voltages combine at the output. In this arrangement, each switch handles a fraction of the overall voltage rather than the full output voltage.
This staged configuration is relevant when engineers need to extend the voltage capability of semiconductor switches. It does not, by itself, guarantee that a system will meet a target voltage or current. The output also depends on circuit parameters, load conditions, switching behavior, and the number of stages.
Designing the MOS-gated thyristor Marx generator
The selected switching device was a MOS-gated thyristor, chosen for its high di/dt current characteristics and relatively fast turn-on time. The design objective was to evaluate the generation of a high-voltage, high-current pulse with a 2 μs width—parameters that are difficult to achieve with one device alone.
Capacitance was estimated using the relationship between pulse duration, inductance, and capacitance: τ = √(LC). With an assumed inductance of 100 nH per stage, the calculation produced a capacitance value close to 5 μF. This estimate provided a starting point for the circuit design; the effective inductance and load still influence the resulting waveform.
Gate isolation and synchronization
Reliable operation depends on the stages switching together closely enough for their voltages to combine as intended. The gate circuits used positive gate-to-source voltage to trigger the MOS-gated thyristors and provide sufficient current to maintain the on-state during discharge.
Because the circuits are not all referenced to ground during discharge, the gate drive required high-voltage isolation. The design used DC/DC converters for galvanic isolation and an optical device to transmit the trigger signal. Measurements of the gate-circuit outputs showed a delay of no more than about 10 ns, indicating closely timed triggering across the circuits.
What the stage tests demonstrated
Tests with different numbers of Marx stages examined whether the output voltage increased as expected. To keep the switch operating conditions comparable between tests, the load resistance was set to the number of stages multiplied by 0.2 Ω. The measured load pulse was approximately 2 μs, and the pulse width changed little as stages were added.
At four stages, the generator produced a maximum output of 1.8 kV and 2 kA. The reported peak load voltage increased approximately in proportion to stage count, while remaining below the ideal scaling: the measured factors were 1.9 for two stages, 2.9 for three, and 3.9 for four. The study also reported a single-unit discharge result of 400 V peak and 2.5 kA peak into a 0.2 Ω load, with a 2 μs pulse at a 1 kV charging voltage.
These are prototype results, not a demonstration of the ETIGO-IV unit’s full 30 kV, 10 kA operating point. The four-stage output remained substantially below that target. The findings instead show that the Marx arrangement could combine stage outputs while maintaining a pulse width close to the intended duration.
Interpreting voltage efficiency and load trade-offs
The study reported voltage efficiency of about 50%, defined in its discussion as the ratio between charging peak voltage and peak load voltage. It also noted that higher load impedance has been associated with improved voltage efficiency in other work. For this design, however, pulse-width and peak-current requirements constrained how high the load impedance could be.
This illustrates a practical design trade-off for pulsed-power engineers: improving one output characteristic may conflict with another. Voltage efficiency cannot be considered independently of the required current, pulse duration, and load conditions. Similarly, adding stages can increase the output voltage, but the resulting system must still be evaluated against its complete operating requirements.
Practical implications for accelerator systems
The MOS-gated thyristor Marx generator offers a way to investigate semiconductor-based switching for applications traditionally served by gas switches. Its modular stages address the challenge of building higher voltage from devices with limited individual voltage capability, while isolated, closely synchronized gate drives support coordinated operation.
The tests established initial evidence of voltage scaling and a pulse width near 2 μs, but they did not reach the accelerator switching unit’s full voltage and current requirements. For system integrators and plant or facility teams assessing similar architectures, the key takeaway is to treat stage count, device stress, gate timing, inductance, capacitance, and load impedance as linked design variables. Prototype results can validate the approach, but achieving a target operating point requires further scaling and evaluation under the intended load conditions.