Hello,

The gate of a mosfet behaves like a capacitor.
You can see the gate of the mosfet as a dead short when the gate pulse is initiated.
The driver you are using has stated that a gate drive current of minimum 6 A is used.
Try to use a stronger external 12 Volts supply and see the difference.

Bertus

But aren't the peak currents handled by the decoupling capacitor? I could try with another 12V/1A power supply.

 

Hello,

Yes, the peaks are handled by the decoupling capacitor, but what if the decoupling capacitor is not recharged quick enough by the power supply?

Bertus

 

The FET has a gate charge of 160 nC at 12 V, which equates to 1.6 mA average current at 10 kHz. A 1 µF capacitor holds a charge of 12 µC at 12 V, which is 75 times the gate charge. The DC-DC converter and capacitor should be more than adequate to meet the requirements.

 

This is my converter datasheet. The only questionable characteristic is that it's unregulated. But if for some reason it could fail, I would like to know more behind that reason.

 

Could it also be that back EMF from the motor is producing more voltage, faster than the flyback diode can clamp?

 

The diode you have chosen is ultrafast recovery. Usually such diodes also have very fast forward recovery (transition time from reverse blocking to forward conduction). You layout is very good for minimizing the inductance that would be detrimental to fast clamping. The FET has reasonably good avalanche capability, so it can probably cope quite well with any very brief overvoltage that might occur due to delay, whatever the cause, in the diode conducting. That said, overvoltage is usually a faster killer of FETs than overcurrent.

The unregulated converter is much more likely to be producing voltage above nominal than below because of the light average load. Though there is more gate charge to contend with if the voltage is high, that really isn't a problem with your circuit. Extra charge is mostly an issue at turn off where it simply adds a bit of delay while the charge is brought down to where the FET actually begins to turn off. Once the FET reaches that point, the fact it took longer to get there is irrelevant. Delay like that can be important in things like an SMPS where it makes very short ON times more difficult to produce. You also have lots of margin between the nominal output of the converter and the voltage it takes to get the FET to very low ON resistance. Usually there is very little improvement in ON resistance after you get to a gate source voltage of about twice the voltage of the plateau in the gate charge curve, which would be roughly 9 V. As Figure 1 of the datasheet shows, there is only very small reduction in Rds for Vgs of greater than 6 V.

 

The diode you have chosen is ultrafast recovery. Usually such diodes also have very fast forward recovery (transition time from reverse blocking to forward conduction). You layout is very good for minimizing the inductance that would be detrimental to fast clamping. The FET has reasonably good avalanche capability, so it can probably cope quite well with any very brief overvoltage that might occur due to delay, whatever the cause, in the diode conducting. That said, overvoltage is usually a faster killer of FETs than overcurrent.

The unregulated converter is much more likely to be producing voltage above nominal than below because of the light average load. Though there is more gate charge to contend with if the voltage is high, that really isn't a problem with your circuit. Extra charge is mostly an issue at turn off where it simply adds a bit of delay while the charge is brought down to where the FET actually begins to turn off. Once the FET reaches that point, the fact it took longer to get there is irrelevant. Delay like that can be important in things like an SMPS where it makes very short ON times more difficult to produce. You also have lots of margin between the nominal output of the converter and the voltage it takes to get the FET to very low ON resistance. Usually there is very little improvement in ON resistance after you get to a gate source voltage of about twice the voltage of the plateau in the gate charge curve, which would be roughly 9 V. As Figure 1 of the datasheet shows, there is only very small reduction in Rds for Vgs of greater than 6 V.

Okay. I have another guess. The source of the FET is connected to the current sensor instead of directly to the reference ground as shown in the gate driver datasheet. Could this be an issue?

I'm trying to analyse all possible causes before burning more FETs.

 

The source connection in terms of the power path (as distinct from the gate drive path) shouldn't be a problem. There is inductance there, as always, but the total path is quite short and tracks are wide. While considering the possibility of small capacitance causing some undesirable noise coupling, I probably would have put the current sensor on the drain side of the FET, but I don't think putting it on the low side is more than very slightly detrimental. A low side current sense resistor is very common in high frequency SMPS circuits and they will typically contribute at least as much inductance as your current sensor.

But I do think the gate driver ground should go as directly as possible to the FET source.

 

Well, I should use a scope and observe what's really happening. So far, I've considered:

1- Stronger 12V supply [no the real problem, I think]
2- RC, RCD snubber design
3- A capacitor across the motor would help?
4- Dealing with voltage spikes somehow

Only the scope would tell me if the problem is at the gate signal, or something related with the VDS.
Fortunately, the driver hasn't suffered. I just replaced the FET, and everything is ok, just that I haven't tried with the load that blew my FET before.

 

Update: I've blown up a second MOSFET at startup of the motor. I've replaced the FET with this IGBT IKW50N60DTP and so far it's working well. Anyways, I will capture some waveforms soon.

 

I've got these waveforms so far, input and output of the gate driver. I haven't been able to capture Vge yet. The new IGBT hasn't failed yet. (That's why I'm now referring to Vge instead of Vgs) Seems like the MOSFET I used before didn't tolerate the kickbacks from the motor. I'll report back when I get Vge measurements.

Input of the driver

Output of the driver: