Low side MOSFET failling at low revs and high torque

[ipadron]

Hi everyone,

I'm working on the design of a 3-phase controller but it fails at full power for a 3 kw motor and low speed (about 100 rpm). At about 60 V and 50 A the controller seems to work well in medium and high speed but at low revs and after a short time (less than 1 minute) any of the low side MOSFET will fail with a short circuit across the 3 terminals. When that happens the controllers short circuit protection seems to work well to avoid a shoot through with the corresponding high side MOSFET.

It has also happened that with the same conditions as before I am able to stop the motor after increasing the load and it seems that everything is fine, but when I check the MOSFETS one of the low side has also died. I understand it must be failed at the same instant the motor stops.

As I said the MOSFET always fail with the three terminals shorted and no signs of burning or smoke. The MOSFET I use are IRFB4410 and the gate is driven through a 10 ohm resistor. The PWM frequency is about 16 kHz and have used slow decay asynchronous and also synchronous trapezoidal control together with Hall sensors.

I would appreciate if anyone have any idea about the type of MOSFET failure and cause. My impression is that it could be an avalanche failure and it may be due to a layout problem.

Any thoughts? Many thanks in advance.

 

[bigmoose]

Lets see what you have in your design and layout for rail caps. There should be low ESR caps directly from the high side drain to the low side source on each leg of the bridge. A good pulse rate 1uF type. These caps will help control the turn off PWM spike that could be driving you into avalanche.

As you said at your power level layout is key. Post up a layout or check Arlo1's posts for his high power layout.

 

[circuit]

50A is battery current, I guess? So at low RPM, when you have 10% duty for example, your lower fet sees constant 500 amps, while upper one sees 10% wide 500A pulses. This can explain lower fet's crystal meltdown. This is the reason why cheap chinese controllers fail at stall and low speed operation.
You can solve this by implementing phase current limiting. Will need current sensors on two phases.

 

[ipadron]

Many thanks for the prompt replay.

I have attached the layout of the main side of the PCB. There are two MOSFET in each high and low sides. I understand that the layout could be improved, especially in terms of the distance between the high and low side MOSFET and I'm actually working in that. The capacitors used in the power line are 470 uF and also 4.7 uF / 100V ceramic capacitors.

50 A mentioned above are actually only 30 A on the battery side. The duty cycle at low rpm (100-200) is about 30%, but its possible that failure may be due to a thermal runaway. The MOSFET are connected to an aluminium dissipation bar and the controller should stop at a maximum bar temperature of 70 C. One solution could be to add a second MOSFET (more costly) or change the PWM control to exchange the PWM between the high and low side so that the MOSFET on both sides have the same power dissipation. This is something to try at least to discard unless someone has a better idea.

View attachment Controladora_V31.pdf

 

[HighHopes]

one of the interesting things that happens at low revs high torque is that there is virtually no Back-EMF yet since the speed is low. so, when a mosfet turns ON it applies practically the full voltage across the motor terminal. at higher RPM the motor has its own voltage (Bemf) which apposes the the DC bus voltage and only the votlage difference between the two (give or take a phase angle) is applied accross the motor terminals. anyway, back to low rpm.. so you put full voltage with nothing countering it and what happens is you generate a low of current since the impedance is low in a motor. the current ramps up real fast, reaching a high peak value, and generates the torque quickly (this is what makes electric motors so great, full torque at zero speed). with high torque generated real fast you have to turn the mosfet off soon or the peak current gets really high.. so this is why the PWM is only 10%.

now you can see, at low RPM there is a very high ripple which is not experienced at higher RPM. with high current ripple comes high peak current (too much?) and also high thermal cycling. the thermal cycling can cause failure just as easy as high peak currents. so .. knowing nothing more about your application then your first post.. there are two things to think about for failures which can occur during low RPM operation. high peak current and/or thermal cycling. to protect against high peak current is to have a current sensor with 100kHz bandwidth and has the ability to shut down the inverter if it exceeds mosfet limit. thermal cycling is a different matter as the bonds flex & bend because different metals inside the mofset has different coefficients of expansion. this problem is enhanced with parallel mosfets because Rds_on changes a lot with temperature and the parallel mosfets might not change at the same rate and so one mosfet will take more current than the other which very often means it takes too much. the solution here is to make sure all mosfets are on the same heatsink with thin even thermal grease and preferably are matched for similar transconducdance performance, or use more rugged mosfet or bigger die.

good luck

 

[peters]

Hello,
as I see the on the layout, the + power supply and the current sensor resistor is placed between the mosfets. I would move them on the left of the bridge (blue layer), then there will be more room and direct wires between high and low side mosfets to add some more low ESR caps.
You have large caps on both side of the current sensor res, but be careful they shunt the resistor, and it reduces the sensor bandwidth. I'd use the 470uF-s only on the power supply before the resistor, and several low ESR ceramic or polypropylene caps on the bridge after the res.

The other thing I think about is the gate driver circuit. Sometimes the low side mosfet opens when the high side is just switching on too fast, because the internal caps of the mosfet form a capacitive divider that raises the gate voltage when the drain voltage raises at high rate (dV/dt problem). You can see a bump on the oscilloscope measuring Vgs of a low side fet after its falling edge. If Vgs fall time is too long compared to the dead time, then this bump happens before the end of the falling edge, and there may be a shoot-through. Make sure this bump is low (max 1-2V) and it is after the falling edge.
You can help this by increasing the dead time or slowing down the high side mosfet turn-on (increase gate resistor). The latter may not fully help, because with inductive load at high current the drain voltage changes very quickly anyway.
Also, if the gate driver circuit is not strong enough, it may not be able to hold the gate at 0V, or the falling edge is too slow.
Which gate driver do you use? The diodes parallel with the gate res are soldered?

 

[ipadron]

Thanks again for all the comments.

It seems the problem I had with the failure of the low side MOSFET is solved. It looks like is related to what circuit and HighHopes had commented. Just had an IRFP4410 on each side and also due to the PWM control used the low side MOSFET get much hotter than the high sides ones and fails. With a maximum current of about 30 A on the battery side and at medium-speed (with a duty cycle of about 50%) had no problems since the temperature sensor will triggers at 65 C and stop the controller before MOSFETs fails due to thermal runaway. But when the load in increased the motor speed is reduced and with a duty cycle of 30 or 20% the current flowing through the MOSFET is too high for the critical temperature of 65 C. Therefore, any MOSFET failed before the dissipation bar reaches that temperature of 65 C. The solution has been to place a second MOSFET on each side that is available in the design of the PCB. I've also changed the PWM control so that both high-side and low-side MOSFET are used for the current decay. Before, the low side MOSFET get much hotter compare to the high side, but also it's true what HighHopes says about different variations in the temperature of each MOSFET due to different Rds behaviour with temperature. Despite using exactly the same MOSFET part number there will be some degrees temperature difference between them.

I will take note about what peters comments on the layout of the capacitors and the shunt current because I have some problems with the current measurement as it varies greatly in the ADC value of the micro despite trying to synchronize the measurement with the PWM signal. I think it may be a matter of improving filtering shunt signal.
Regarding to the drive circuit an IR2110S is used and the diodes in parallel with the resistors are not soldered at the moment. The dead time is 1 us and and adding the diodes effectively would help to avoid the problems you've mentioned.

I have another question. What would be the best way to stop the motor in case of high power and high speed when an event occurs such as motor or controller overtemperature? I understand that if all the MOSFET are turn off at the same time the two phases which are feeding can produce a sudden voltage spike that can damage any of the MOSFET. Perhaps the best thing would be to reduce the duty cycle until the motor stops?