DRV8323R: GHx and GLx waveforms

Hi James,

I think on both pictures dead time is around 350-400ns and it looks OK. Shape of SHB voltage suggests that current is flowing into SHB half-bridge on both pictures. Ringing on the second picture starts when SHB voltage starts to fall down. I can think of two sources of that ringing, the first is reverse recovery of HS mosfet body diode and the second is capacity coupling between gate and drain of LS mosfet (Cgd). I would try to lower IDRIVEP_LS if it was not already so low (30mA) or improve PCB layout to limit parasitic inductances in LS mosfet gate circuit.

Regards,

Grzegorz

Hi James, 

Thanks for posting your question to the e2e motor drivers forum - 

I've assigned this thread to a team member, and we'll aim to provide a response by end of week on next steps 

Best Regards, 
Andrew 

James Wang said:

I use DRV8323RS to drive PMSM, but the MOSFETs are burned out at high current or VDS_OCP is indicated.

I think it's ringing problem and it seems that the DEAD_TIME does not work during the falling of CHB.

Hi James,

If you think the dead time is not enough and causing the shoot through leading to damaged FETs, did you feel the hi and low FETs were hot before damage occurred? I would carefully touch the FET to sense the temperature. According to the waveform of SHB, it looks normal to me. If a shoot through happened -- both hi and low FETs were ON -- then the signal SHB should have a value in the middle of VM and GND or about 24v during dead-time, but we don't see on the pic, and instead the SHB has a smooth transition from Hi to Lo, or Lo to Hi.

GLB with some ringing, but this should cause no shoot-through as the GHB was clearly turned off 400ns before.

The FET could be damaged not from shoot through current, but from high voltage spikes on VM or GLB and GHB.

Brian

Grzegorz Pelikan said:

I can think of two sources of that ringing, the first is reverse recovery of HS mosfet body diode

There was no current flow through the body diode during SHB was high, and when GHB turned off the low FET body diode conducted, not the top one, so I don't see why you said the HS diode was in reverse recovery. The HS diode only conducts when the LS FET turned off then the HS turned on, but this is not what happened during the LGB ringing. 

Brian

Hi Grzegorz,

Thanks for your reply.

Too low IDRIVEP_LS may affect control performance, and we are checking PCB according to the TI layout guide now.

Other questions about picture 2,

1. The peak of ringing on the second picture is 3.5V which is over MOS Vth. I think there is a risk of short.

2. There is a platform on the GHB falling edge, is this normal?

Hi Brain,

Thanks for your reply.

The peak of ringing is over MOSFET Vth (2.5V) which I think may cause the short. I did not touch the MOSFET, but there were signs of solder seeping out from under the MOSFET. In some tests, the chip also reported MOSFET overcurrent fault.

The motor works fine at low torque, but the MOSFET burns out at high torque, usually at about 10A bus current.

"There was no current flow through the body diode during SHB was high"

On both pictures during dead time SHB voltage rises slightly above VM what is HS mosfet body diode voltage drop and indicates that current is flowing into SHB node.

Grzegorz

Hi James,

"1. The peak of ringing on the second picture is 3.5V which is over MOS Vth. I think there is a risk of short."

When ringing occurs on GLB (the second picture) HS Mosfet is already turned off and that level of ringing may cause SHB voltage to fall more slowly (like around 12V and 4V when SHB/GHB are flattened a bit) . If ringing is higher then it may cause false LS Mosfet turning off  when it should be turned on what would cause increased switching losses and probably more problems. If those possible problems could cause malfunction of HS Mosfet gate driver and lead to a shoot through I do not know. The problem here is not that the GLB ringing is over Vth but when it falls down below Vth.

"2. There is a platform on the GHB falling edge, is this normal?" - please, explain what do you mean by "platform"

"The motor works fine at low torque, but the MOSFET burns out at high torque, usually at about 10A bus current."

The higher currents the higher all voltage spikes and ringings that are expressed by L*di/dt, the higher currents the higher reverse recovery charge of mosfet body diode what also leads to higher di/dt and voltage spikes/ringing. More important than bus current is motor phase current, to measure it you would need current probe. Cheap probe like the one below may be enough (up to 65A)

https://www.amazon.com/Hantek-CC-65-Multimeter-Current-Connector/dp/B06W2KFZLW

To "catch" possible voltage spikes and ringing you need an oscilloscope and probes with at least 60-100Mhz bandwidth. In order to limit noise picked by probes it is better to use ground springs than alligator clips.

To voltage spikes that could kill a mosfet what was mentioned by Brian I would add exceeding mosfet SOA and overtemperature.

It is a very good sign that your board works fine with lower loads, now you need to improve it to work with higher loads, I think good PCB layout with low parasitic inductances and good thermal performance will be essential.

Regards,

Grzegorz

Hi Grzegorz,

Thanks for your suggestion. I will try to measue phase current and voltage at high torque. 

The "platform" I mean is the curve in the red circle and I think I know how it came about.

Hi James,

I think that everything in red circle looks OK. First the GHB is falling down, when it reaches Vth HS Mosfet turns off (SHB rises a bit because of Mosfet body diode voltage drop) then it falls further till it is equal to SHB. After that they both stay at VM+diode voltage drop level till LS Mosfet is turned on what is causing GHB and SHB  to fall to GND level.

Regards,

Grzegorz

Hey James,

It is possible that you are seeing a ringing on the low side gate due to coupling from the switching on the MOSFET.

Lowering the IDRIVE is a method of reducing the effect seen.

This App Note highlighting different ways of tackling these sorts of issues through design.

https://www.ti.com/lit/an/slvaf66/slvaf66.pdf

And here is our board layout guide.

https://www.ti.com/lit/an/slva959b/slva959b.pdf

I would recommend trying them and also measuring VDS and GLx to see if there is any improvement.

Best Regards,

Akshay

James Wang said:

The peak of ringing is over MOSFET Vth (2.5V) which I think may cause the short. I did not touch the MOSFET, but there were signs of solder seeping out from under the MOSFET. In some tests, the chip also reported MOSFET overcurrent fault.

Hi James,

I take a second look at the waveforms and see some odd things:

1. I agree with you that there is shoot-through current (both high and low FETS were on). You said the driver was set to have 400 deadtime, but the FET Gate slew rate is so slow that GLB took more than 400ns to go from 12v to 3v and at that moment the high FET was turned on with GHB also at 3v. I don't see ringing on SHB, so why not increase IDRIVEN_LS to much higher to turn off the low FET quicker? The driver chip can increase IDRIVEN_LS to 2A max. I believe these are big FETs with high gate charge, and so it is better to increase iDRIVE and reduce rise and fall time at the FET Gate to speed up FET turn OFF and ON to avoid shoot through.

2. I don't think the ringing on GLB is the root cause of shoot through current, although I would increase IDRIVEP_LS:  more than 30mA as it took more than 1600ns for GLB to change from 0v to 12V with no benefit. 

3. At high current, 10A you said, slow turn on and turn off FET (more in linear region) generated a lot of heat even without shoot trough.

I bet with high IDrive, the issue will be resolved.

Brian

Grzegorz Pelikan said:

On both pictures during dead time SHB voltage rises slightly above VM what is HS mosfet body diode voltage drop and indicates that current is flowing into SHB node.

Hi,

Let's use the 2nd waveform for this. The inductive current when the HS FET turned off would cause the voltage at SHB slightly higher than VM, as you pointed out, but this was not caused by body diode reverse recovery time.  This might cause the upper FET diode to forward, but not reverse recovery time yet. The lower diode was not conducting because SHB was never going below GND, so no reverse recovery happened to either upper or lower diode.

Brian

Hi Akshay,

Thanks for your reply. We are checking our design and layout according to the guide now.

Hi Brian,

I remeasured the waveforms (pic 1) with scope probes connected to GND of scope when motor spinning today, and it can be seen there is an occasional spike and ringing on SHB. Picture 2 shows the waveform of a single trigger.

I add  RC snubber circuit for each FET, and it seems the spikes disappear, and I will retest the board.

I also tried the high IDRIVE but the ringing was worse on GLB.

Hey James,

Let me know when you are finished testing.

Best Regards,

Akshay

James Wang said:

I remeasured the waveforms (pic 1) with scope probes connected to GND of scope when motor spinning today, and it can be seen there is an occasional spike and ringing on SHB. Picture 2 shows the waveform of a single trigger.

There is obvious ringing on the falling GHB in the 3rd pic that high enough to turn the FET on again; what changed? Why the 2nd pic doesn't have the ringing on GHB?

Brian

The difference is the way the scope probes connect to GND (GND of the board and GND of the scope). It is possible that board ground is also coupled with ringing.

The spike waveform is not affected by the measurement method. The spike is generated when the motor is spining.

James Wang said:

The spike waveform is not affected by the measurement method. The spike is generated when the motor is spining.

If the scope waveform is true, then the issue is why the GHB during turning off, it got down to about 10V and then shot up to 40V again?