This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DRV8323: DRV8323 gate drive fault

Part Number: DRV8323
Other Parts Discussed in Thread: CSD88599Q5DC

Hi all,

After discussing the problem of 'DRV8323 gate drive fault' with you last time,

I modified the drive current to run for a period of time, and the same situation still appeared 

 (IDRIVE:IDRIVEN_HS_120MA,IDRIVEP_HS_60MA  ,IDRIVEN_LS_120MA,IDRIVEP_LS_60MA )

The faulty machine Indicates gate drive fault on the B high-side MOSFET.We measured the resistance values of each pin of MOS, and they were all normal,

Then I configured the  Driver Control Register (DIS_GDF = 1b) to force PWM output,the waveform is as follows

CH2: Continuous high-side source sense pin voltage (SHx) using a regular single ended probe

CH3: High-side gate drive pin voltage with respect to SHx (GHx) using a differential probe

We have no direction on this problem now, do you have any suggestions. 

We want to know more about the internal circuit structure of 8323. Can you provide the information

Please refer to my last question

Attach my email address, you can contact me at any time(zhang_xi@piec.com.cn)

Thanks

ZX

  • I clicked the 'Ask a related question ',but after I posted the question, I didn't see the content of my last question

    Attach the link to my last question again : https://e2e.ti.com/support/motor-drivers-group/motor-drivers/f/motor-drivers-forum/1003469/drv8323-drv8323-gate-drive-fault

  • Hi ZX,

    We closed that thread, but of course we can continue to support you on this new thread while referencing the old one!

    Thanks,

    Matt

  • Hi ZX,

    Sorry to hear that the issue still persists. To understand the issue better, would you be able to answer the following questions:

    1. Is the fault only happening on phase B of the high side FET for this board? How long did the board operate before getting a gate drive fault? 

    2. After the first time you got the fault, would the device then immediately get a fault every time you tried to run it after that?

    3. Did you make adjustments to the layout as suggested in the previous e2e question? 

    4. Even though 60mA/120mA is much better for IDRIVE, with the Qgd of the MOSFET you are using that still may result in a turn on time of 7nC/60mA = 117ns. Most customers consider 100ns-200ns turn on time to be a fast turn on time, and it is possible that since the gate traces are thinner than recommended this could be introducing extra ringing on the gates that could over time result in damage on the GLx pin. Especially if the unit operated for some time before getting a gate drive fault, and then after that there is a gate drive fault immediately upon startup, that would point to the device being damaged possibly due to an abs max violation.

    Regards,

    Anthony Lodi

  • Hi Anthony,

    Thanks for your reply

    1. I did a statistic and the result was that there were seven damaged machines, six of them are phase B of the high side FETand the other one is phase A of the high side FET. The time of the board operate is variable, some are a few days, some are broken when they start up.

    2.Yes,I restarted many times and all reported the same fault

    3.Well, we have changed the layout in the new design, but we already have products in use before this and will continue to use them, so we can only change the IDRIVE in the previous product

    4.I don't quite understand the last sentence. Does It mean that the error reported after operat for some time is most likely because the voltage is out of range?   'an abs max violation' What exactly does this mean? If so, how should we deal with it

    You mentioned possible damage to the GLx pin. Why is it the GLx pin and not the GHx pin because we got ' gate drive fault on the B high-side' ,do I have any way to know for sure, we can only judge by its output waveform now. We tested the resistance of the pin and it is normal

     

    Regards

    ZX

  • Hi ZX,

    Thank you for the additional information!

    It seems like this could be layout related, especially since the majority of the times it is occurring on phase B.

    I apologize, I meant damage on GHx, not GLx. The device has certain voltage limits for the pins (absolute maximum ratings), and if the ratings are exceeded then this can result in damage to the DRV. For the GHx pins, there are certain voltage ratings for the pin to allow for safe operation. The absolute maximum allowable 200ns tranisient GHx voltage is 1V above VCP, and the maximum allowable voltage on the SHx pin for a 200ns transient is 7V above VM and -7V below ground. We are working on a datasheet revision for this device, and the 2 abs max specifications I provided are based on the upcoming datasheet revision. 

    I wonder if voltage transients on the GHx pin or SHx pin resulted in damage to those pins over time. Would you be able to show a waveform of the GHx to ground voltage and the SHx to ground voltage of phase B and phase A on a device before it is damaged? I want to check to see if there is a lot of ringing on those gates. 

    Would you be able to test some of the devices with the IDRIVE settings as follows: IDRIVEN_HS_60MA , IDRIVEP_HS_30MA, IDRIVEN_LS_60MA, IDRIVEP_LS_30MA to see if that resolves the issue? If it does, or if they last longer before damage occurs, then that would point to the issue of excessive ringing on the gates and/or source as the cause of possible damage to the device. Using a lower gate current results in less voltage spiking on the gates and/or source. 

    Regards,

    Anthony Lodi

  • Hi Anthony,

     Thanks for your reply

    Below is the test waveform you need

    CH1:the SHx to ground voltage of phase  B

    CH2:the GHx to ground voltage of phase B

    CH3:the GHx to ground voltage of phase A

    CH4:the SHx to ground voltage of phase A

    1.IDRIVEN_HS_120MA , IDRIVEP_HS_60MA, IDRIVEN_LS_120MA, IDRIVEP_LS_60MA 

    2.IDRIVEN_HS_60MA , IDRIVEP_HS_30MA, IDRIVEN_LS_60MA, IDRIVEP_LS_30MA 

    I found that the waveform of phase B is indeed a bit worse. This should be due to the layout 

    I would like to ask if there is anything else we can do if the layout cannot be changed, do we still need to change the drive current, and if there are any other suggestions besides this

    Regards

    ZX

  • Hi ZX,

    I am a little bit confused about what I am seeing here, the plots you provided don't seem to be during the switching of the MOSFETs. The reason that I say this is because if you align the grounds of the traces on a waveform you will find that the voltage difference between the gate and source votlage would be about 0V (meaning the FET is off). Likely what you measured was during the time when current was switching between flowing through the low side FET on the phase that you were measuring, and then when the low side FET shut off then the current began flowing through the high side FET body diode (which is why you see the phase voltage rise to about VM). For the boards where the layout cannot be changed, I would recommend further reducing the IDRIVE to see if that resolves the issue. The only other thing I could think of would be to solder a thick wire from the GHx pins to the gates of the high side FETs to improve this connection. 

    I would be willing to review your layout once you finalize the revisions and provide feedback. 

    Regards,

    Anthony Lodi

  • Hi Anthony,

    This is a scaled-down version of the waveform sent to you last time

    I can’t put our layout on the public platform, is there any other way to contact you?

    I want to ask another question.We tried for a long time but didn't catch the lethal waveform.

    So in order to enlarge our problem,we tried shorting the SH of phase A and phase B when MOS is switching,

    the PWM of phase A and phase B are synchronized, and the duty cycle is 52.6% and 47.4% respectively.

    The motor is connected outside. As a result, 8323 is damaged and I cannot use spi to read the status register,

    the data read back are all 1. The other damage is that the GH and SH of phase A are short-circuited.

    We set VDS_LVL = 0000b, the mos we use is CSD88599Q5DC,I would like to ask if 8323 will be damaged in this case

    Why can't spi be accessed anymore. Can we not test like this?

    Regards

    ZX

  • Hi ZX,

    By shorting the SH of phase A with the SH of phase B you are creating a short from the power supply to ground. As a result, a lot of current will be flowing through the MOSFETs when both are on (peak current = Vsupply/(2 x rds(on)).

    In the OCP control register, What is your settings for the OCP_MODE and OCP_DEG bits? There is some deglitch time for the overcurrent protection (set in the OCP_DEG bit) that gives between 2us and 8us (depending on the setting) of deglitch time such that the VDS overcurrent monitors don't falsely trip. The driver waits the length of the deglitch time once it detects the VDS trip threshold has been exceeded, and then if the threshold is still exceeded by the end of the deglitch time, then depending on the OCP_MODE setting the driver will trigger a fault and put the outputs in Hi-z mode. The OCP_MODE bit determines whether it is a latched fault (outputs are shut down indefinitely until a reset pulse is provided or the power is cycled), automatic retry fault (The outputs will shut down until the end of the TRETRY time, or if Cycle By Cycle current limiting is enabled it will resume operation at the next PWM edge), report only fault (fault is reported on the nFault pin, but no action is taken), or overcurrent protection is disabled. The rise time is very fast when a short occurs, so current can get quite high by the time the deglitch time has been completed. If you are using the automatic retry feature, then this high current event would be repeatedly occurring after the end of the tretry time, or at the next PWM if it was in cycle by cycle current limiting.. What probably happened is that the short that you induced by shorting the phases resulted in additional ringing due to the really high currents that may have damaged the driver. 

    The issue that you are experiencing seems most likely an issue with the gate switching speed which is causing the MOSFET to turn on very quickly and cause ringing. I wouldn't think that shorting the phases is the best way to reproduce the issue. This is more related to gate drive currents (IDRIVE). 

    I will PM you with my email so you can email me the schematic privately. 

    Regards,

    Anthony Lodi

  • Hi Anthony,

      The OCP Control Register is set as :OCP_DEG = 01b, OCP_MODE = 00b.

      Because we have already mass-produced and they are already in use, it is not a simple matter to modify the software configuration,

      so we want to capture the exact lethal waveform to illustrate the necessity of changing the software

    Regards

    ZX

  • Hi XZ,

    Testing circuit with a shorting the phases or phases to GND may be a good way to find weak points in your design and then by eliminating them there is a good chance you will be able to fix the problem. I would start with lower short circuit currents ie. I would lower voltage supply to let's say 12V (or even lower), set OCP_DEG to 00b and set VDS_LVL around 50% higher than value causing tripping a fault during normal work. If the circuit can withstand short circuits at these conditions it would give time to look for any voltage spikes and ringing on Mosfet gates, switching nodes, LS Mosfet sources, HS Mosfet drains.

    Problem might be related to operating conditions of motor driver like braking motor from high speed and/or with high inertia load (high currents), possible VM overvoltage during braking especially if field weakening is used or load has potential energy (vertical axes).

    Regards,

    Grzegorz

  • Hi ZX,

    I understand the difficulty of modifying the software. Just to clarify, the issue that is being experienced is a gate drive fault, correct? If you were experiencing problems with the VDS monitors then maybe you could consider the test you performed (though it seems like the system wasn't able to handle the test without damage due to high currents. You could try using a much lower voltage as Grzegorz mentioned), but if the issue is with the gate drive fault, this would be a result of the gate of a MOSFET not sufficiently turning on after the end of TDRIVE time. As mentioned before, what might be happening is that ringing on the source and/or gate during switching could be causing the SHx pin to get damaged over time, especially if the ringing reaches the ABS max rating. The damage could be preventing the MOSFET to fully enhance, causing a gate drive fault to occur. I am not sure the best way to show the issue except measure the gate to ground and the source to ground voltage and evaluate the ringing that occurs during switching events to see if it might be coming close to the abs max ratings.

    Did reducing the IDRIVE even further help prevent the fault from occurring?

    Regards,

    Anthony 

  • Hi,

    I forgot to add that probably it is better to make a test short-circuit at the end of 1-2 meter long cable, not directly at driver output connector. Cable inductance should slow down a bit current rise speed and give some time to overcurrent protection.

    Regards,

    Grzegorz

  • Yes, you definitely want to be careful when implementing an external short due to high amounts of current. 

    Regards,

    Anthony 

  • Hi Anthony,

    We have done some experiments these days and feel that we are getting closer to the truth, and there are some questions that need to be confirmed with you.

    The figure below is the first waveform when the damaged IC is turned on.

    IDRIVEN_HS_120MA , IDRIVEP_HS_60MA, IDRIVEN_LS_120MA, IDRIVEP_LS_60MA 

    CH1:the SHx to ground voltage of phase  B

    CH2:the GHx to ground voltage of phase B

    CH3:Current on GHx line of phase B

    CH4:the SHx to ground voltage of phase B

    Change CH1 to measure Vcp voltage

    1.Why does the voltage of SHx rise and fall again after 4 us?

    2.Why does the differential signal have a spike?

    3.Why does the differential signal end up staying at about 1V?

    4.Why does the voltage of Vcp fluctuate?

    We suspect that part of the diagram below is broken

    CH1:Supply voltage

    CH4:the SHx to ground voltage of phase B

    5.We energized between GH and SH without energizing IC as shown in the figure below,

    With the increase of the supply voltage, the differential signal will not continue to rise until it reaches 6V

    We tested the normal IC will continue to rise

    6.We tested the resistance value of GHX to SHx of the good and broken IC is 390K, why is it not the same as the 150K in the specification?

    Please help answer the above questions, if you can please tell us what is broken, so that we can take more effective protective measures.

    Regards

    ZX

     

  • Hi ZX,

    Have you tried switching out the MOSFETs on the damaged board to see if it works again? Or have you tried putting a damaged device on a good board to see if it works? I want to make sure it is the driver that is damaged, not the MOSFET itself. 

    TDRIVE is 4us long (by default, but this can be changed in the TDRIVE bits of the Gate Drive LS Register), and starts when the driver detects a command to turn on (or off) one of the gates. During the duration of TDRIVE, the IDRIVE current is applied to the MOSFET to either turn on the MOSFET or turn it off. After the completion of TDRIVE, the driver will switch to a weaker gate current (called IHOLD) to ensure that the MOSFET remains on. The IHOLD strength that is used to keep a MOSFET on is 10mA. What seems to be happening is that when the driver switches to the IHOLD Current, the current is insufficient to keep the MOSFET turned on which is why you see the drop down to about 1V. This could be a result of internal damage to the device, or damage to the MOSFET itself. The ripple on the charge pump voltage is not unusual, since there is high gate current that is used to turn on a MOSFET when initially switching it on. During this time the charge from the charge pump is rapidly being transferred to the gate of the MOSFET, which can result in a drooping of the charge pump voltage a little bit. 

    You mentioned that channel 4 was the SHx to GND of phase B. Did you mean to say that it is the differential GHx to SHx voltage?

    I will try to test the GHx to SHx resistance in lab tomorrow to see if I get the same results as what you are seeing. I am not sure why you are reading 390k for the resistance there. 

    Regards,

    Anthony Lodi

  • Hi Anthony,

     Thanks for your reply.

    I have tried to put the damaged DRV8323 on a good board for testing, but it doesn’t work. I also tried to replace this damaged DRV8323 with a good one. The original non-working board can work, so I’m sure the MOS is normal.

    I'm sorry I misdescribed it,channel 4 is the differential GHx to SHx voltage.

    Regards

    ZX

  • Hi ZX,

    Thanks for the reply! based on the information it does seem like the damage is internal to the device. The damage seems to be resulting in some leakage current from the high side gate to either the source or ground. Thus it requires more current to counteract the leakage current that is from the damage in order to properly charge up the gate. Once the gate current diminishes to ISTRONG, the leakage current overpowers the gate current and the gate voltage begins to drop. As discussed before, the damage is likely caused from an abs max violation on either the GHx pin or the SHx pin due to transients that occur during switching from high inductance paths on the PCB.

    If you wanted to look into this further I wouldn't mind looking at a plot of the voltage applied vs the differential gate to source of the test that you did where you applied a voltage to the gate through a 10k ohm resistor and grounded the source pin. If you plotted the voltage applied every 0.5V from 0V to 12V then that would allow me to see the trend that it follows. I think that would just further confirm my conclusion.

    Regards,

    Anthony Lodi

  • Hi ZX, 

    Due to my schedule currently it would work better for me to go into lab and run the resistance test on Monday. When you tested the gate to source resistance, did you probe with the multimeter directly on the pins, or did you test near the gate to source by the MOSFET? If you tested near the MOSFET, I wonder if the inductance of the traces could have messed up the accuracy of the reading. Also, did you try switching the position of the probes of the multimeter? I wonder if you may get a different reading if you swapped which probe went on which pin.

    Regards,

    Anthony 

  • Hi Anthony,

      I tested the IC that was detached from the board and directly tested the IC pins.

      I also tested the IC that was not soldered in the warehouse.

      IC alone, nothing can affect the reading

    Regards

    ZX

  • Hi ZX,

    I tested the GHx to SHx resistance of one of the DRV8320S EVMS we had, and I was also getting a resistance higher than expected (about 242k ohms instead of 150k ohms). I will consult internally to figure out why we are seeing the higher resistance. 

    Regards,

    Anthony Lodi

  • Hi ZX,

    I have reached out to a member on my team regarding the resistance reading and he is planning on responding to me next week. 

    Regards,

    Anthony 

  • Hi Anthony,

      Ok, thanks

    Regards

    ZX

  • Thank you for your patience!

    Regards,

    Anthony 

  • Hi XZ,

    I reached out to a member on my team yesterday to see if he has been able to look at the issue with the unexpected resistance measurements and I am awaiting a response. I will let you know once I hear back. Sorry for the delay!

    Regards,

    Anthony 

  • Hi XZ, 

    I had a meeting today with a coworker to discuss the resistance readings on the GHx-SHx pins. There is more information that is needed, and my coworker is going to reach out to the design team to understand better what is going on. Hopefully we can get some more information next week from the design team and work towards a resolution. I will keep you updated.

    Sorry for the delays,

    Anthony Lodi

  • Hi XZ, 

    I have been continuing to communicate with one of the members on the team regarding the resistance readings, and am gathering more information to have more clarification on why the GHx to SHx measurement is the reading that it is. I want to make sure I have the necessary information before providing you with the answer to ensure that the information is clear.

    Regards,

    Anthony 

  • Hi Anthony,

      Thank you for your support

    Regards,

    ZX

  • Hi ZX,

    After looking into this with design and performing some testing, it looks like the GHx to SHx resistor internal to the device is about 480kohms, not 150kohms. We will need to update the datasheet to reflect this. The GLx to SPx resistance is 150kohms. In order to get a good resistance measurement, you could try applying a 100mV load directly between the GHx and SHx pins and measure the amount of current flowing between GHx to SHx. The resistance would be equal to the voltage applied (100mV) divided by the current measured. The measured current should be in the range of a few hundred microamps. I would recommend trying this method on your devices, since I have found that some (not all, but some) of the DMMs have a harder time providing an accurate reading for this particular resistance measurement. 

    Regards,

    Anthony Lodi

  • Hello ZX,

    It's been a while since we last heard from you. Do you still need help on this thread?

    Thanks,

    Matt

  • Hi all,

    We have improved the product by adding diodes, but the rework cycle for production is relatively long. So there are no new questions for the time being

    Thanks for your help

    ZX

  • Hi XZ,

    We're glad you were able to resolve your issue. If you have another question click "Ask a related question" at the top on the screen and we'll be happy to help. I'll close this thread for now.

    Regards,

    Michael