I've been trying to build a high voltage generator consisting of a high-voltage transformer and a Cockroft-Walton multiplier.

I have actually tried several different drivers to drive a high-voltage flyback transformer with no luck. My earlier attempts consisted of using an IRF2153 in a push-pull to drive the transformer like an SMPS.

I had minimal success with it but it seemed that I could not get the multiplier to increase the voltage to what I wanted.

I had planned on 15 kilovolts out of the transformer with a 5x multiplier but ended up with only 1.5 kilovolts after the multiplier. This was at about 20 kilohertz or so if I remember correctly. Anyway, I put it away for a while and gave it a rest.

I picked it up again and decided maybe I would go the resonant frequency route and try to drive the transformer with a Mazzilli driver.

I first tested it with a 12-volt 30 amp power supply for the entire source. At first, I had a little better luck with this and got probably a 15 kilovolt arc directly out of the transformer without the multiplier.

But when I connected the multiplier it didn't increase the voltage. I measured the voltage with it connected to the multiplier and apparently, it would draw too much current as the supply voltage would drop off.

So I decided to upgrade my DC supply.

I stripped an old microwave and rebuilt the high-voltage transformer into a step down transformer and rectified the output to 100-volt direct current.

For the gate voltage source, I have the hundred volt split with resistors for a gate-source voltage of 15.5 volts and a maximum output of 1.2 amps.

At first, I was following a schematic that I had found on the internet for the Mazzilli driver. But since I upped the input voltage to 100 volts I made a few changes.

I wanted to aim for a resonant frequency around 10 to 30 kilohertz. So I decreased the series inductance of L1 and the primary coil and also decreased the resonating capacitor down past the recommended .68 microfarad of the schematic to .5 microfarads in hopes that I could get the frequency a bit higher.

I also decreased the pull-down resistors from 10-kilo-ohms to 1 kilo-ohm because I wasn't sure if the voltage would drop off the gate fast enough at 20-30 kilohertz with the given parameters of the gate capacitance, gate resistors, and the 15.5 Supply voltage.

I've been smoking mosfet after mosfet with it in this configuration.

At first, I thought maybe I was overshooting Ids of the IRF460s I had, so I dropped the bucks on a pair of 500 volt 48 amp mosfet ( the numbers are on the schematic I'm going to include).

By my calculations and measurement estimates, I'm pretty sure I wasn't getting more than 48 amps out of regular household 15 amp breaker 120 VAC mains outlet stepped down to the RMS of whatever 100 Volt peak to peak VAC is.

From an earlier revision, I had the mosfet sandwiched between two peltier devices connected to respective fanned heatsinks because I thought the heat might be killing them.

So my question is this I suppose:

Even if my calculations of the resonant LC circuit or something is off and I have just straight current running through one of the MOSFETs in full-on shouldn't it still survive with a max Ids of 48 amps? Or do I have some sort of weird thermal problem in which the peltiers I have on there are actually transferring heat slower than the mosfets being directly connected to the heat sinks or something of that nature? ( Or am I missing something else completely?)

enter image description here

  • 4
    \$\begingroup\$ The most important thing that will destroy your MOSFETs is uncontrolled back emf on the primary. I would recommend that you spend time getting this circuit into a simulator and modelling it properly. I built a 50 kV DC switching power supply with virtually zero mods after first simulating the life out of it. It didn't need more than a couple of amps from 24 volts with no load so you talking about dozens of amps from a 100 volt supply means you are doing something fundamentally wrong. \$\endgroup\$
    – Andy aka
    Sep 16, 2018 at 10:32
  • \$\begingroup\$ @ Andy aka I believe you are right. Checking my memory, I also remember doubling up those CW diodes about the same time as I had it working. If I remember correctly, I looked at it again and had one pair in the wrong place. I think I might have corrected it at the same time I switched up the power supply, trading one problem for another. Old rules of debugging troubleshooting. I appreciate you reminding me to reexamine my process. Thank you. I'll try it again with the original mazzilli design. \$\endgroup\$
    – user198606
    Sep 16, 2018 at 20:15
  • \$\begingroup\$ If you use a MOSFET with a much larger current rating than necessary, then it will have a much larger gate-source capacitance, so you need to lower the impedances of the gate drive circuit.. \$\endgroup\$
    – BobT
    Aug 9, 2020 at 12:07

2 Answers 2


Short version: Build a Mazzilli flyback driver, not the schematic shown above.

I recommend you have a look at Mazzilli flyback drivers online and compare your circuitry to theirs, as your circuits connectivity around the MOSFETs differs to Vladmiro Mazilli's circuit and I think that is the source of your fried silicon.

I would expect to see:

1) a resistor divider to each MOSFET gate from the 100VDC rail (from 100VDC to 100VDC RETURN, with the centre of the divider at the MOSFET gate). This means two resistor dividers.

2) a 12V zener (or thereabouts) between each MOSFETs gate and the 100VDC RETURN (i.e. the bottom of your 890uF cap)

3) a schottky diode (anode) from each MOSFETs gate pin to the other MOSFETs drain pin.

The smaller value resistor (my guess 10k 1W or 5k1 2W) is to turn the gate on quickly but not too quickly. See the MOSFET datasheet. You will have to calculate the value and power dissipation of this resistor. The zener can be circa 12V 1W (in the above examples it dissipates 0.1 or 0.2W) The higher value resistor (my guess 100k) just needs to be enough to discharge the FET when the circuit is no longer powered. Under normal operation the schottky diodes discharge the MOSFET gate when the other MOSFET is switched on.

I hope this helps.


I would also be suspecting the gate drive circuits, MOSFET gates are pretty fragile things. Some other possibilities (in addition to JT's answer)

  1. Spurious oscillation in the MOSFET, usually cured with a 22ohm resistor in series with the gate (or a ferrite bead around the source lead if TO-220) or fix up some sloppy wiring.
  2. High dV/dT , the drains are connected by the resonating capacitor, so when one rapidly turns on , it can put a large negative dV/Dt on the drain of the other, this dV/dT couples to the gate via the miller capacitance, dragging it more than 20v negative , solution schotty clamp diodes on the gate.
  3. High frequency ringing in the transformer, this can couple into the gate as per 2 above, this can be a leakage inductance issue in your wiring or in the transformer. usually needs snubbers somewhere L1 should minimise this possibility, can L1 saturate? This ringing is usually really obvious on a scope. Possibly adding some resistive dividers on gates may help.

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.