I'm researching a good (and simple) driver for a HV transformer, and found this very simple and elegant circuit on the web:

enter image description here

It uses a feedback winding to turn transistors on in an alternating pattern. I assembled it and it works very well, but later I tried to replace the bipolar transistors with mosfets, to experiment.

The result was awful, the nice sinusoidal output I get with bipolar transistors turns into a very erratic and deformed waveform when using mosfets. Current draw increased a lot, and the secondary voltage was much smaller, the efficiency dropped a lot (by 10x of times).

My question is: is it possible to use mosfets in this configuration? Is the distortion the result of badly chosen mosfets? May I need to change R1, R2 or C1 to different values to make it work with mosfets?

Or are mosfets just incompatible with this self-oscillating configuration?

UPDATE: I forgot to add that the erratic oscillation with mosfets happened only after I changed R1 and R2 placement: I connected them between the gates and GND, instead of gates to Vcc, as in the schematics. Merely replacing the bipolar transitors with mosfets in the circuit above resulted in no oscillation at all, since the gates got tied directly to Vcc, making both mosfets conducting all the time.

So, the description of erratic mosfet oscillation in the original question applies to the circuit with this modification: resistors between gates and GND.

UPDATE 2: Thanks to Andy's reply, I understood what may be obvious for the more experienced: the resistor just bias the base to conduction state, so the feedback winding can make the transistor cut on negative pulses. I have applied the same logic to the mosfet version, biasing the gate to Vgs(th) (Gate Voltage Threshold), and it works.

However, this circuit is not as efficient as one using an active oscillator and a driver, which will ensure rise and fall times are short, and the transistor (BJT or FET) will not be speding time in it's linear region.

  • 1
    \$\begingroup\$ You have a good answer, but I'm curious. Why would you want to replace the beatific BJT with a devil's spawn MOSFET? ;) \$\endgroup\$
    – jonk
    Apr 6, 2018 at 16:31
  • 1
    \$\begingroup\$ It's for learning purposes, and also because I have a reel of 3000 SMD switching mosfets that I would like to use in this project (AO6400), if possible. And a last reason, is to find out if, using mosfets, power consumption would be smaller. :) \$\endgroup\$ Apr 6, 2018 at 16:50

2 Answers 2


Points to note: -

  • R1 and R2 set the base bias current for a BJT. All they do for a MOSFET is set a high gate voltage that causes a lot of current to be taken from the power supply rail.
  • The Base/emiiter region is a forward biased diode hence there is natural limiting of collector current due to BJTs being base fed via resistors
  • The forward volt drop at the base will be about 0.7 volt and any feedback winding voltage may enhance this a little or it may negate that voltage and turn the transistor off (this is what is needed)
  • The feedback level may be nowhere near enough to turn off a MOSFET that is voltage biased on the gate at Vcc.
  • Circuits designed to work for a BJT nearly always don't work for a MOSFET

Try adding two resistors; one on each gate to the common source connection on your circuit. Guesstimate value: 150 ohms. If it stops oscillating then try 220 ohms.

If this doesn't work then try adding small value source resistors such as 10 ohms. If you can't get any of this working satisfactorily then try simulating the circuit and tweaking values in the sim.

  • \$\begingroup\$ Thanks a lot, Andy, I will try your suggestions. I forgot to add in the question that I made a change in the circuitry to make it oscillate with mosfets: instead of connecting the resistors between gates and Vcc, I connected then between gates and GND. I'll update the question with this info, sorry for that. \$\endgroup\$ Apr 6, 2018 at 16:52

Andy aka has correct suggestions, so +1. I have used this ZVS Royer circuit with MOSFETs many times. I changed the biasing setup completely. I used experimental source resistors of 1 ohm to measure the current via a small length of 50 ohm coax going to the scope. When I vary the DC gate bias voltage with a 5K pot, I inspect the source current trace and the drain volts trace for good ZVS. With good wave, TO220 MOSFETs will run cool with no heatsinks when the inverter has no load. If the switching waveform is good, the efficiency will be the same as a driven scheme but it will be cheaper to make and you will not have to worry about load changes or Analog component spreads making the circuit go off-tune and come out of ZVS running hot.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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