How to deal with voltage spikes on secondary synchronous rectifiers?

[KX36]

Hi all,

I have a forward converter with self driven synchronous rectifier FETs as shown (FET datasheet):

**broken link removed**

The waveform I'm getting out of it at full load is as follows (Yellow = Q3 forward FET Vds, Blue = Q4 catch FET Vds):

**broken link removed**
**broken link removed**
**broken link removed**

I don't have much experience with synchronous rectifiers. The waveform is generally as it's supposed to be, with relatively fast transitions roughly mirroring the primary waveform and the slow fall of Q3 Vds (also seen on the primary) is expected the transformer's deadtime when the inductor is still conducting.

The problem is the amplitude of the Vds spikes as each FET turns off. As each Vds is closely related to the opposite FET's Vgs, I'm worried about exceeding the Vgs rating. The Vds spikes were insignificant at light loads, being less than the full step Vds voltage at anything up to 40% load (I skipped straight from 40% to 100%, at 40% the inductor is certainly in continuous conduction). I tried to probe a Vgs but I think I zapped one of the FETs by doing so, so I can't experiment with any snubbers etc until I replace the FETs. In the meantime, I'd welcome any advice about the circuit, especially any relating to this spike.

Note that the body diodes do conduct some current. I did put in a parallel schottky with little effect on the waveform beyond a slightly less negative minimum of Q4 Vds and slightly more ringing on Q3 Vds and vice versa for the schottky in parallel with Q3. I have left footprints for RC snubbers in parallel with each FET but zapped a FET before getting a chance to try any. I knew when I started that I was very likely to run into this issue. Is this just something for RC snubbers to damp, bigger gate resistors etc. or is there another solution?

Cheers,
Matt

 

[treez]

RC snubber for that. It tends to be 220pF and 100R or something similar. Try it on ltspice with the leakage included and it does it well for you.
The more leakage in your transformer, the worse those spikes.
I think parallel UF diodes are needed......schottky only of youre sure you wont ever overvoltage them even transiently.
schotky not like overvolts.

 

[jiripolivka]

Hi all,

I have a forward converter with self driven synchronous rectifier FETs as shown (FET datasheet):

**broken link removed**

The waveform I'm getting out of it at full load is as follows (Yellow = Q3 forward FET Vds, Blue = Q4 catch FET Vds):

**broken link removed**
**broken link removed**
**broken link removed**

I don't have much experience with synchronous rectifiers. The waveform is generally as it's supposed to be, with relatively fast transitions roughly mirroring the primary waveform and the slow fall of Q3 Vds (also seen on the primary) is expected the transformer's deadtime when the inductor is still conducting.

The problem is the amplitude of the Vds spikes as each FET turns off. As each Vds is closely related to the opposite FET's Vgs, I'm worried about exceeding the Vgs rating. The Vds spikes were insignificant at light loads, being less than the full step Vds voltage at anything up to 40% load (I skipped straight from 40% to 100%, at 40% the inductor is certainly in continuous conduction). I tried to probe a Vgs but I think I zapped one of the FETs by doing so, so I can't experiment with any snubbers etc until I replace the FETs. In the meantime, I'd welcome any advice about the circuit, especially any relating to this spike.

Note that the body diodes do conduct some current. I did put in a parallel schottky with little effect on the waveform beyond a slightly less negative minimum of Q4 Vds and slightly more ringing on Q3 Vds and vice versa for the schottky in parallel with Q3. I have left footprints for RC snubbers in parallel with each FET but zapped a FET before getting a chance to try any. I knew when I started that I was very likely to run into this issue. Is this just something for RC snubbers to damp, bigger gate resistors etc. or is there another solution?

Cheers,
Matt

I think the spikes naturally occur due to the inductor in the circuit. Check if their amplitude fits the maximum MOSFET channel voltage, and then all is OK. If the spike amplitude exceeds the voltage rating, then I would connect a RC snubber in series across the MOSFET channel to reduce the spikes, or find a varistor (e.g. by Raychem) for the same purpose.
A ferrite bead on the inductor line may help, too, or use another type of inductor with a ferrite core designed for lower-frequency response.

- - - Updated - - -

As I can see from your schematic, it may be possible to add a capacitor to the gate-source of the MOSFET, to adjust a delay between the alternating MOSFETs, so the overlap would not generate the high spikes.

 

[FvM]

You shouldn't expect perfect switching from a simple passively steered synchronous rectifier. Forward biasing of substrate diodes and respective reverse recovery can be only avoided with an active control circuit. It's also clear that external schottky diodes will hardly carry the complete load current. If you don't want huge reverse recovery spikes, you must go for MOSFETs with fast substrate diode and preferable exact gate control.

On the other hand, if gate voltage rating is your primary concern, add fast Vgs clamping diodes.

 

[KX36]

Thanks for the replies. I do expect the spikes are due to the transformer leakage inductance and not the inductor. The parallel catch schottky was mainly to reduce the loss during transformer deadtime when the FETs are off and the inductor current flows through the body diode which is slower and has a higher forward voltage, rather than to conduct all the current or affect the spike.

I had thought about just clamping the gate with a zener but didn't get the chance before it broke. I have read that it can make ringing much worse with the increased gate capacitance, I will try it out. It may take some research to get the right FET and zener. I will also try out the RC snubber. Not so sure about that gate capacitor but I'll try to sim it.

On the plus side it looks like a gate resistor failed open, so the FET might still be OK. It's been running with these waveforms for some time and hasn't self destructed yet anyway.

Cheers,
matt

 

[treez]

I usually find, if a fet gate resistor fails open, then if the fet gate is left floating (ie no 22k resistor from gate to source), then the ambient noise turns the fet on and it shorts, and then destroys itself

 

[KX36]

It wouldn't surprise me if it has failed or damaged the other one. There is cycle by cycle current limiting and output short circuit protection in the controller IC and it is Q3 not Q4 so I'm hopeful but not expecting much. we'll see. If it has failed I will probably have to wait until I get back from abroad to replace it so that would be a nuisance.

 

[mtwieg]

The spike is only tens of nanoseconds long, that's usually too short for an RC snubber to be effective. Reverse recovery in the diodes is probably the main culprit. It's rare to see passive control of synchronous rectifiers, probably for good reason.

 

[KX36]

The FETs seemed OK. Once I replaced the gate resistor the circuit worked again. Then in the process of substituting snubber capacitors and probing I managed to blow the new gate resistor and I think this time I did damage the FET. It still technically works but measures about 1.4 megohms drain-source and gate-source on the meter where before it measured open. The Q3 Vds waveform is slightly different now too, taking longer to slope up and down and spending less time at the peak. This is very similar to the waveform at a much lighter load. I think I will have to replace that FET.

Anyway, here's the result of me adding RC snubbers. I calculated the snubber for Q4 and then applied the same snubber to Q3. You can see it works a lot better for Q4 but does reduce the amplitude of the spike on Q3. I'm thinking damaging the FET has changed its capacitance among everything else and that's why the waveform's changed and the snubber's less effective, but it might just need a different snubber.

Yellow=Q4 Vds, Blue=Q3 Vds this time, the other way round from last time just to be confusing:
**broken link removed**

I did also manage to measure the amplitude of the spikes on Vgs before adding snubbers, they were much lower than the peak drain voltages. One was 22V compared to 32V peak on the drain, the other was 15V compared to 29V on the drain.

Regards,
Matt