I have built a valve/tube tester. To provide the various screen and plate voltages I'm doing this:

high voltage regulator

The required voltage is selected by a rotary switch down the stack of zeners. The momentary button is pressed to test the tube. The MOSFET is a 500V device that can handle more than 7 amps. It has protection zeners built in but I've added an external one to the gate anyway. Datasheet here R2 represents the ESR of my power supply. At 150mA its already sagged to 260v but thats fine; the tubes I'm testing draw less than that.

It works as expected. There are two instances, one for screen volts, the other for plate. But every so often I find the MOSFET has failed in one or the other, shorted source to drain.

There is sometimes a 68nF capacitor 'downstream' (depends on configuration), the MOSFETS are mounted on an aluminium heatsink but never get even warm to touch (they only run for a few seconds at a time) and the device itself seems to be well protected and rugged to boot.

What have I missed? Whats the most likely way I'm destroying these MOSFETS and what can I do to mitigate this?

  • \$\begingroup\$ what you've drawn looks pretty good, so maybe it's something you've not drawn. Wire lengths causing spiking through their inductance? I'd be inclined to be paranoid and put a C to ground from the R3/M1 node. M1 SOA is limited to 30 watts for DC operation in the 220 package, how close are you to that? FETs are thermally unstable in linear operation, hence the falling power with pulse length on the SOA. Do you have a 'scope, is it oscillating? Put 100ohms gate-stopper resistor in series, right at the gate, I've had some nasty 'but it's only DC!' follower oscillations when wires are too long. \$\endgroup\$ – Neil_UK Jan 13 at 6:23
  • \$\begingroup\$ I'm very guilty of 'but its only DC'. It should be well within the SOA, typical tube draw s tens of mA. Not oscillating at the moment but that might be a good call. I've had trouble with the tubes under test oscillating. The base connections are selected via patch leads, ie long wires. I'll do as you suggest and add a stopper at the gate. Thanks \$\endgroup\$ – Murray Jan 13 at 7:13
  • \$\begingroup\$ Are you just doing an emissions test for the cathode? Or are you doing a mutual conductance test, varying the grid(s) to check the plate current response? (Mostly, I'm just curious.) Back in the day, a 5U4 diode used as a power supply rectifier would exhibit about \$90\:\Omega\$ plate resistance. Is that what \$R_2\$ is for, here? To replicate the resistance of a pair of these, for example? \$\endgroup\$ – jonk Jan 13 at 19:22
  • \$\begingroup\$ @jonk It tests gm, theres a variable negative source for the grid. Loosely based on this: link . The 730R resistance of the power supply is the small transformer (with poor voltage regulation) and the RC filter. \$\endgroup\$ – Murray Jan 13 at 23:58

The ESR represents the sag at DC, you may get a brief pulse charging that external capacitance that exceeds the (real) Safe Operating Area of the MOSFET.

Assuming there is a largish capacitor on the supply, maybe you can add a series resistor to limit the transient current.

I would not place perfect confidence in the calculated SOA curves on the data sheet- from testing they can be somewhat ... optimistic.

  • 1
    \$\begingroup\$ I wish I'd kept notes about the previous failures. Theres a patch bay and wires before we get to that capacitor so I'm hoping the inrush would be contained but I'll add 100R or so in the B+ line to limit the current. I have two spare MOSFETS left, hope I don't need them... Thanks \$\endgroup\$ – Murray Jan 13 at 7:17

A 25A MOSFET is the wrong way to go when you switch the input with rapid dV/dt with more than 200V and dt ~0, like this configuration and expect to avoid spurious failures. Switching the output is better.

Suggestions; Replace the MOSFET with 10W NPN and delete D1,D2,R1,R4

  • \$\begingroup\$ My thinking was to keep the amount of live high voltage circuit to a minimum but i see your point. I've replaced the MOSFET and added gate and supply resistors as mentioned above. I hadn't considered an ordinary NPN but if it fails again I might just try that. Thanks. \$\endgroup\$ – Murray Jan 14 at 0:06
  • \$\begingroup\$ It's the high capacitance of the ultra low RdsOn switch that makes it rugged yet place a much greater Miller Capacitance Cdg than the Zener has so the Zener cannot conduct to pull down the gate fast enough to pull down Vgs thus exceed its rating for some transient period. \$\endgroup\$ – Sunnyskyguy EE75 Jan 14 at 0:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.