MOSFET ferrite bead placement

[theleakydiode]

Hi, which legs of a MOSFET should I put ferrite beads on? I need to get rid of some very fast spikes as they are coupling into the gate and current sense resistor via miller capacitance.

Should I put one on all three legs or just two? This is for an educational flyback SMPS project (its not mains powered) and the MOSFET is the primary switch. I already have gate resistors but these spikes are very fast.

So which MOSFET legs is it best practice to include them? I'm guessing I'll need them on gate and drain but just wanted some advice as its a bit of a pain to unmount the MOSFET, would like to get it right the first time.

Thanks.

[Zero999]

I don't see how ferrite beads will prevent this. The Miller capacitance is internal to the device and no amount of ferrite beads will mitigate it.

[Leo Bodnar]

Usually ferrite beads go onto the gate terminal but their primary use is it to kill RF oscillations (often caused by sharp turn on.)
Leo

[Leo Bodnar]

...very fast spikes as they are coupling into the gate and current sense resistor via miller capacitance.

Could you re-phrase this? It's not very clear where the spikes are coming from if they are coupling to both gate and wherever current sense shunt is.
Leo

[theleakydiode]

I don't see how ferrite beads will prevent this. The Miller capacitance is internal to the device and no amount of ferrite beads will mitigate it.

Yes however the sharp current spikes caused by this charging up are finding their way onto the gate and through the current sense resistor, if I can squash these spikes a bit then that would be great. The miller capacitance still needs a loop to complete and that would mean a current path involving at-least two MOSFET leads.

Usually ferrite beads go onto the gate terminal but their primary use is it to kill RF oscillations (often caused by sharp turn on.)
Leo

There is a bit of that during turn off well into the Mhz range, but just putting one on the gate would still leave the drain to source capacitance open to sharp pulses which goes through the current sense resistor.

...very fast spikes as they are coupling into the gate and current sense resistor via miller capacitance.

Could you re-phrase this? It's not very clear where the spikes are coming from if they are coupling to both gate and wherever current sense shunt is.
Leo

They come from the primary coil and LCD snubber, there is a small RCD snubber too but the main one is the LCD since several amps flow through the primary at low and its driving a LOPT so that big hv drain pulse is needed. Its well below the MOSFET breakdown voltage but the LCD snubber does make it very ringy as it cycles through its phases.

My aim is to reduce or eliminate miller induced spikes on the gate turning it back on slightly as it reaches just above the turn on threshold, and also to reduce the big spikes on the current sense resistor ramp waveform. Right now they are very fast and cross the threshold where the UC3844 does its thing.

The circuit actually works ok but just this funny business on the gate and current sense resistor is making me want to fix it.

[daedalux]

Have you tried biasing the gate off to ground with a potentiometer to get experimentally a value that will prevent the oscillation?

[theleakydiode]

Have you tried biasing the gate off to ground with a potentiometer to get experimentally a value that will prevent the oscillation?

Yes, I got down to sub 100 ohm and the Miller induced gate spike is still there. If the low impedance path through the chip doesn't keep it off then I'm not sure a pull-down resistor ever will.

Also I traced it down the secondary rectifier diode, problem is the transformer is potted so I can't simply place an RC across it. Any other ways to snub the secondary diode without having access to the anode side? Its also at about 30kV so that adds another spanner to the works.

[Leo Bodnar]

Have you tried removing gate resistor and acceleration diode?  It's not like 3844 has a mighty output driver.  Especially since you are using FB on the gate.

If you are somehow convinced that your problem is miller capacitance, remove miller capacitance (use lower current / Vds MOSFET.)

Leo

[theleakydiode]

I currently have it like this output>12 ohm resistor>reverse diode and 10 ohm resistor in parallel>gate>1k pull down resistor and 18v zener diode.

Putting a small 10n capacitor across the gate resistors or from gate to source reduces the spike a bit so it must be miller induced.

[fourtytwo42]

Could you perhaps post some scope shots of these spikes ?