Fusor Forums

I know there's a few HV gurus out there, perhaps someone has some
suggestions for replacing standard RC snubbers with something more
elegant.

I have a HV supply (my own design) that uses a pair of MOSFETs.
96V DC is applied to the center tap of a transformer, the MOSFETs are on
the two windings, to ground, switching at around 12 kHz. There's then some
boost windings, and the output goes to another transformer which drives a
standard diode/cap multiplier stack.

As of now, for MOSFET protection, I have diodes across the drain/soure to
protect against negative spikes.

I am also using RC snubbers: a 5.6K and 0.22uF in series, across the two
MOSFET sources, and 11.2K and 0.11uF in parallel with each other, that in
series with a diode, from the MOSFET source to the 96V DC supply
(transformer center tap). All values emperically derived.

The resistors are bukly power resistors, and dissipate a bit of heat. I'm
curious if anyone has tried just using some of the fast protection diodes that
are available today at various voltage ratings, and putting them from the
MOSFET source to ground, with a breakdown voltage somewhat less than
the MOSFET voltage rating?

The only thing you really need to do with snubbing is to suppress the transformer leakage spike to keep the mosfets from breaking down. Most mosfets these days have a repetitive avalanche rating, but it's not really good for them to break them down each cycle. What you want to do is hang a fast recovery diode on each drain with the anode pointed to the drain. Hook the two diode cathodes together. From this junction, hang a parallel resistor -capacitor combination from the diode cathodes to the transformer center tap.The capacitor should be large enough to limit the voltage excursion of the drains to a value comfortably less than the mosfet breakdown voltage. The resistor discharges the capacitor between hits. You want to adjust the RC values such that when the converter is running under load, the capacitor voltage stays above 2 X Vin, but doesn't get too close to the mosfet breakdown voltage. This is to prevent the snubber from diverting a significant potrtion of normal energy transferred through the transformer, which is what will happen if the voltage on the snubber capacitor drops below 2 X Vin. You do essentially the same thing by placing a series combination of diode and transient suppressor zener across each mosfet from drain to source. The diode would be to prevent reverse conduction of the zeners during the switch dead time. The breakdown voltage of the zener would be sized to be higher than 2 X Vin, but less than the mosfet breakdown voltge. The zeners would also have to be sized to dissipate the leakage energy without overheating.

Chris - the snubbers can really make your MOSFETS work hard. The fast Shottky diodes are a good idea. Also, I have found it hard to tell what transients are real and what are artifacts of the scope hook-up.

Dave Cooper

The Motorola HEXFRED diodes are pretty fast and are able to take a beating during reversals. They are very expensive though in the higher voltage and current ranges. Make sure they are located real close to the device they are to protect.

Richard Hull

Snubbers/clippers are a necessary evil. If Chris uses them the way I suggested, they will have a minimal effect on the MOSFETs during steady state operation, as they will be biased as "clippers" rather than your classic load-line shaping snubber. The MOSFETs will take a small hit when initially charging the clipper cap, but they can take this in stride.Ditto for the transient suppressor zeners. Hexfreds might make decent antiparallel commutation diodes for the MOSFETs, but they won't be very useful for snubber duty. What I need from Chris is the breakdown voltage of the MOSFETs he is using, and I could give a rough ballpark guess as to the size network that will be needed. With 90-odd volts on the input supply, 200V MOSFETs would be cutting it too close for any sort of safety margin. This is because with a push-pull topology, the switches see a minimum of 2 X the supply voltage, not including leakage spikes. Even a 250 V device is cutting it close, as it doesn't give much room for the snubber to operate. At least in the case of IR, the next notch up where there are some decent devices to pick from is 400V. If you could find a substantial 300V device, that would also be ok.

Richard,

It's the IRFP460, rated 500V.

I recommend two Hexfreds: one between Source and ground of your mosfet and another between source and drain (in reverse). Mosfets have a built in snubbing diode across the source and drain, but an external diode will help reduce the power dissipation on your mosfet. Some of the books I read mentioned using a CRD network across the source and drain:

|--- S ---+
| C1
| |-------|
| D1 R1
| |-------|
| ---D----+
|
D2
|
GND

The faster the switching response of the snubber diodes the better.

It's important to get your terms straight. The antiparallel diode to which you refer is for commutation of the transformer magnetizing current when both switches of the push pull converter are off - it has nothing to do with snubbing. One can possibly get away with using the intrinsic diode of the MOSFET for commutation if the current is low, but it puts extra stress on the MOSFET to force recovery of the intrinsic diode, as its recovery time is usually relatively slow. I've blown up MOSFETs that way in a half-bridge converter, but it was a fairly large current and old-technology devices. The key to getting the commutation diodes to work is that they should have a forward voltage lower than that of the MOSFET intrinsic diode and a short forward recovery time. Hexfreds could qualify if they are large enough. Steering the current into the commutation diodes is more difficult than it appears, because the intrinsic diode of a large MOSFET like the IRFP460 will have a relatively small forward voltage drop because of its large area. Sometimes designers place a Schottky diode in series with the switch and place the commutation diode across both the Schottky and switch to make sure that the reverse current only flows in the commutation diode.
Attached is a picture of both the commutation schemes I have described, as well as two schemes for drain snubbing/clipping. Returning the clipper capacitor to Vin rather than return allows the use of a lower voltage capacitor. Starting values for clipper capacitor - 4.7-10nF, make the time constant of the clipper capacitor and discharge resistor at least 2-3 X 1/f, where f is the switching frequency. These will be starting values. Start with a small load on the supply, adjust C as needed to keeep the leakage spike voltage < 0.8 Vmax at max load, where Vmax is the max drain voltage rating of the switch. Adjust the clipper resistor so that the voltage on the capacitor doesn't fall below 2 X Vin.
The last scheme shown is a simple zener diode clipper. The series diodes for the zener diodes may not be necessary. For zeners try something like 2 X P6KE150 or P6KE200 in series. You could also return the zeners to Vin, in which case one would use a lower breakdown voltage zener, and the series diodes would be necessary. Be careful of overheating in the zeners, as they will run away thermally if they get much over 100C.

Thanks, I've printed that for future reference.

Mark H

Here is the alternate arrangement for the zener clipper circuit, in case my description was not sufficiently clear.

>The antiparallel diode to which you refer is for commutation of the transformer magnetizing current when both switches of the push pull converter are off - it has nothing to do with snubbing

Yes that is true. In my haste, I didn't fully explain the concept. I was trying to offer a method that I used for my switch mode PS. The idea using the combination of Hexfreds was from a IRF Application note article I read a long time ago. By using a set of Hexfreds one can avoid the need for snubber networks. The Hexfred across the Mosfet bypasses the built in intrinsic as the carrier because the Hexfreds have a faster switching time. By using a pair of Hexfreds reverse current is limited which reduces thermal load.

Unfortunately I can't see your circuit diagram, the image link is currently broken. When I experimented using Snubber networks, I found the performance poor and caused additional thermal stress on the Mosfets. You will also need some hefy resistors and caps if your working with a large currents. The amount of current the snubber is going to handle is limited to amount the RC circuit can absorb it. Large R's & C's going to cost a lot more than a pair of Hexfreds (especialy to deal with heat disapation) In RC Snubbers, when the Mosfet turns back on, there is a current loop which just adds thermal load.

>The series diodes for the zener diodes may not be necessary. For zeners try something like 2 X P6KE150 or P6KE200 in series. You could also return the zeners to Vin

This is fine for low current switching. But if we are talking about a switch mode PS for a fusor, thermal run away is guarenteed with these devices.

Finally I would suggest not using Push-Pull circuits for driving high current applications. The higher the frequency the higher the power losses are introduced. In a Flyback circuit, the losses are acceptable but if your building a high current switch mode PS, those losses become significant. Instead,either put more turns on the secondary, reduce the primary turns (if possible), or use a higher input voltage. Avoid using P-Mosfets (all P devices) are inherently less efficient that N-Mosfets. The P-Mosfets are also more expensive than their N-Mosfets counterparts. JMTC.

Using a Hexfred as an antiparallel diode only gets rid of one problem - the magnetizing current in the MOSFET intrinsic diodes. It will not replace a snubber intended to clamp leakage spikes. The thermal stress caused by turning on the intrinsic diodes has more to do with forcing the recovery of the diodes during the switching time than the actual DC current flow in the diodes.

The snubber schemes shown would be practical for power levels of a hundred watts or so (maybe a couple), provided that the leakage inductance of the transformer are not too large. A properly designed RCD clamp circuit will keep out of the way of the MOSFET drains after the first couple of switching cycles. Adding the diode(s) allows the clamp capacitor to keep some charge between cycles so that the clamp voltage stays above 2X Vin. Simple RC snubbers get in the way and can burn up a lot of power.

A push-pull converter is a much more practical at medium-high power than a flyback, as the switches will run at much lower currents for an equivalent power level. The transformer can also be much smaller, as it is not required to store energy. The best choice for this application would be a resonant half-bridge design, but that would be way, way beyond the scope of this discussion.