impedance - How can SCRs be effectively paralleled?

I'm designing a precharge circuit which will carry 200A when closed. I'd like to use MCO150 Silicon Controlled Rectifiers, but they won't handle the current. Two in parallel looks like the most cost-effective way to handle the current I need. However, I'm concerned about thermal runaway.

The effective impedance of any device will vary with temperature. If that impedance rises with temperature, two devices in parallel will share reasonably well. The device carrying more current heats up more, and its fraction of the total current drops. But if the impedance of the device drops as temperature rises, the device carrying more current starts carrying even more. Sharing fails, and one device hogs the current until it dies.

The datasheet for this SCR doesn't seem to address the issue of temperature coefficient. Is it just assumed that SCRs have a temperature coefficient in one direction or the other? If the temperature coefficient is negative, is there a way to force the two SCRs to share anyway? Or should I try another approach, like a single device or a large contactor?

Answer

SCRs do not parallel well. Semiconductor junctions, like those in SCRs, diodes, and bipolar transistors, have a forward voltage drop that decreases with temperature. The hotter SCR will therefore draw more of the current making it even hotter, drawing more of the current, etc.

Why do you think you need SCRs? Their main attributes is their latching behavior and the fact that they can be produced with large current capabilities. If you just want the latter, several MOSFETs in parallel would work. These have a positive temperature coefficient so don't exhibit thermal runaway. Still, you need to derate from assuming each of the FETs will share the current equally.

Since FETs look mostly resistive when on, parallelling them not only decreases the dissipation on each, but also the total dissipation. At 200 A, only 5 mΩ will cause a 1 V drop and 200 W dissipation. That won't be trivial to design to no matter what you use as the switch. It will help if this 200 A is only in short pulses with a much lower averge. Take a look at the SCR datasheet and see the forward drop at 200 A. It won't be all that nice either, and you'll have to deal with significant dissipation with the SCR too.

Fortunately 5 mΩ is not that far fetched for a few FETs in parallel. 4 FETs with maximum Rdson of 20 mΩ would do it, and each would dissipate only 50 W if they were to share the current equally. I'd probably derate to 100 W per FET when designing the thermal system.