I have a surge tester setup that uses a high-voltage power supply (up to 1000 V) to charge a non-polarized capacitor bank using SW1 to energize Relay 1 in Image 1. The bank discharges into a waveshape network (inductors + resistors) through a high-power relay, triggered by a push-button. Each activation generates a pulse with a 30–40 us rise time and 100–140 us fall time. See Image 1 for the current relay circuit.
The total series impedance is ~1 ohm, so the open-circuit voltage approximates the short-circuit current when the output is shorted to the bench ground. The supply supports both positive and negative output voltages, and the non-polarized capacitors allow generation of both positive and negative polarity pulses (voltage and current).
I’m looking to replace the mechanical relay (relay 2 in Image 1) with a solid-state switch, as repeated high-voltage/current cycles are causing contact sticking. I’m seeking recommendations on the most suitable switch type (IGBT, MOSFET, etc.), particularly for bidirectional operation.
Image 2 shows an initial concept using a high-power IGBT for positive polarity pulses. However, IGBTs are unidirectional, and I’m unsure how to support negative pulses. Would reversing the IGBT and referencing the gate drive to the high-side emitter work for negative polarity? gate to emitter reference will break with the high voltage at the emitter side. If there's a better topology or practical solution that still supports push-button triggering and an isolated supply, I’d appreciate any guidance.
I’ve also identified a bidirectional IGBT module that appears suitable for my power levels.
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=8ac78c8c869190210186e5fa2b4e6c24

 

I have a revised circuit attached.

 

I suggest NOT using a solid state device unless you can find one rated for that kind of serious abuse. I also suggest adding a magnetic blowout coil to quickly stretch the arc to breaking distance. OR get a relay with contacts that open much farther.

 

I would suggest a solid-state switch like the LCA110. However, the LCA110 can only handle up to 350VDC. Checking in Mouser, something like the S117T might work for you. You want to have the controlling circuit electrically isolated from the load.

 

I suggest NOT using a solid state device unless you can find one rated for that kind of serious abuse. I also suggest adding a magnetic blowout coil to quickly stretch the arc to breaking distance. OR get a relay with contacts that open much farther.

The main reason for the change is to avoid the replacement of relays due to sticking, which in turn causes the low value wave shaping network resistors to fail. I will explore more about the magnetic blowout coil.

 

I would suggest a solid-state switch like the LCA110. However, the LCA110 can only handle up to 350VDC. Checking in Mouser, something like the S117T might work for you. You want to have the controlling circuit electrically isolated from the load.

View attachment 351089

The voltage rating is ok, but based on the current ratings given in the datasheet this part cannot be used to switch high currents.

 

https://menlomicro.com/technology/how-it-works

 

How much current do you need? I have used MOSFETs for high current applications before. Take a look at the MOSFET below.

 

OK, if relay sticking is an issue then another option is two relays in series: One relay doing the contact close portion with the other relay doing the contact open function. So one relay will already be closed and all of the contact bounce completed before the second relay operates to deliver the surge current pulse. The consideration is the current rise time requirement of the test specification.

An all solid state option would be similar in using an SCR to switch on the pulse and then a second transistor in series to open the circuit after the pulse. The second transistor can be arranged to be biased into saturation before the SCR is triggered. IT will need to have the high current rating but it would probably not need the high withstand voltage reading, since it will already be biased into conduction. This might be a simpler solution.