LM5109A: What might be causing my FET driver to blow up?

Hi Cameron,

Sorry to hear of your trouble. We will try to help you get it up and running.

Are all three LM5109s failing, or just one?

I think you are on the right path. Motor drives are notorious for inductive spikes. Have you gone around and measured all the pins with a tightly grounded scope probe to make sure no spikes are present? Ideally, you would use a pigtail ground on your probe.

I don’t think the lack of gate resistors is directly causing the LM5109 to fail. They are mainly used to control dv/dt of the switch node.  

However, insertIng a few ohms of gate drive resistance will slow down your switching and may very well solve your problems.

It could also be due to excessive parasitic inductance in the pcb layout leading to spikes that exceed the voltage rating of the LM5109.

Thank you for your reply.

I have not inspected carefully all of the waveforms for short inductive spikes. I'm thinking it would be easiest to add some clamping diodes around the drivers to just absorb the spikes and hope that fixes it. If it does, it's a pretty good indicator that this was the problem. Would still be handy to confirm that this is the case. Still not exactly sure how the current path makes those spikes show up on the LM5109.

One failure, at least two of the driver ICs seemed to have failed. Other times it was definitely only one driver that failed. It can be difficult to diagnose exactly what failed on each failure. Sometimes it's obvious that a particular IC exploded. Other times we used thermal shipping labels as telltale sign for exactly which ICs are getting hot which usually indicates a failure. Rarely we get lucky and can see a straight short with a multimeter. The rest of the time it's guess and check.

Right now, running the BLDC forward, it spins perfectly with no "hiccups". If I reverse the torque, you can hear definite "hiccups" in the commutation. This is indicating that something is wrong but I can't figure out what it could be. The asymmetry of the failure is particularly confusing since I'm using full sine wave control and am always switching all 6 fets. I'm guessing one part has partially failed and the back EMF in certain phase relationships to the main waveforms causes the glitches. I don't think this is an asymmetry in the motor as it should be wound symmetrically and, before the failure, it ran smoothly in each direction.

Indeed the first version of the new PCB has relatively long traces as it was made in a day in house. So maybe the inductive spikes are mostly from PCB layout issues. We have the next rev of the PCBs arriving on Tuesday and these have much shorter & wider traces and large copper planes everywhere.


On a related note, are there similar gate driver ICs available that are maybe a little more robust?

Hello Cameron,

Thank you for the interest in the LM5109; we will work to help you resolve the concerns in the application.

I do not know if this is the cause of the issues, but I see you mention that the VDD and HB capacitor (Cboot) are 1uF. We typically recommend a larger capacitance for VDD, 10x as a general guideline, so the current sourced from VDD to charge the boostrap capacitor does not result in large VDD ripple.

Please refer to Section 8.2.2.1 Select Boostrap and VDD Capacitor. I am going to make some assumption of large duty cycle of 90% but equation 1 is Qtotal=Qg +Ihbs x DMax/Fsw + Ihb/Fsw. in your application this is 150nC + 10uA x 0.9/31kHz + 0.2mA/31kHz= 150.03nC; so the gate charge is the majority of the energy required. Cboot=Qtotal/deltaVHB, Cboot= 150nC/2.3V= 65nF. We generally recommend margin to allow for pulse skipping and capacitor tolerance, but 1uF is larger than necessary. A Cboot of 220nF to 330nF is likely adequate. Using a guideline of the 10x for the VDD capacitance results in 2.2 to 3.3uF for VDD capacitance. It is good practice to have a lower value ceramic which has lower impedance at high frequency in parallel, so a parallel 100nF is recommended with larger VDD capacitance values.

If there are large voltage spikes from back emf, and significant parasitic inductance in the layout, it is likely there are significant voltage spikes on the driver IC pins. To determine the cause of the failures I first want to suggest limit the possible overshoot and undershoot of the driver output with respect to VSS/VDD, and HB/HS. Add small 1A schottky diodes to clamp the driver outputs. Connect one diode anode to HO and cathode to HB, another diode anode to HS and cathode to HO. Place the diodes as close as possible to the LM5109 IC pins. Place another two diodes clamping LO to VSS and VDD in the same configuration. Add some gate resistance on the driver to MOSFET gate connection to allow the output clamping diodes to have lower current spikes, if there is overshoot or undershoot.

One possible issue with half bridge drivers and parasitic inductance induced spikes is negative voltage on HS, which can be a concern with exceeding the device ratings on HS to VSS. Another possible concern is overcharging the HB to VSS capacitor with the negative HS spike. Can you confirm if the negative voltage spike on the HS turn off edge is lower than the -5V datasheet rating? If you think that is the case, there are other drivers with higher voltage ratings such as the UCC2720xA. It is good practice to have 2.2 to 10 Ohms resistance in series with the boot strap diode to limit the boot capacitor overcharging during negative HS transients. 

See if the addition of clamp diodes on the driver outputs, and adding some gate resistance (2.2 to 3.3 Ohms) helps with the issues. Also confirm if you see excessive voltage spikes that may exceed the device ratings. This will likely be negative spikes or excessive voltage from HB to HS.

Regards

Richard Herring