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.

DRV8353R: Negative transients seen at low side mosfet gate when high side mosfet turns ON with 4A of load current

Part Number: DRV8353R
Other Parts Discussed in Thread: DRV8353

Hi,

Am driving a 350W BLDC motor with 37v, 20Khz PWM

Am seeing an undershoot of about -4v at the low side MOSFET gate when the high side MOSFET turns ON. With the no-load current of 200mA everything seems OK but when load increases with the current rising closer to 3A this undershoot starts

Waveform with Phase A LS gate [Cyan] and Phase A HS MOSFET gate [Yellow]

 

IDRIVE setting for all four gates are the lowest 0x00 [50mA]

Observation:

1. At no-load conditions the rise time of the HS gate is more than 150ns so undershoot is not seen

2. But, when the motor load increases the load current also increases up to 5A, at this time the HS gate rise time reduces to <70ns which causes the negative transient 

3. This negative transient increases when the load increases

From the datasheet, I found an Absolute maximum rating for undershoot is -5v max, so as the undershoot is directly proportional to load I am sure this will violate the max transient of -5v soon

Please suggest me a solution to overcome this

FYI, am using a common current sense resistor for all 3 phases

thanks, 

Amutha

  • Hi Amutha,

    Can you send your full schematic as well as the MOSFET part numbers you are using?

    Thanks,

    Matt

  • Hi Matt,

    I used the same parts as in DRV8353RS-EVM board also followed exactly as is from the evaluation board schematic

    please find schematics

    thanks, Amutha

  • Hi Amutha,

    Can you also share your layout either in PDF or Altium format?

    Thanks,

    Matt

  • Matt,

    Please find layout

    thanks, Amutha

  • Hi Amutha,

    Thanks for sharing your design files - We think that the root cause of the negative transient spike at low-side FET could be due to the parasitic inductance of the sense resistor causing ringing at MDRVR1_SP_ABC.

    To confirm this, can you help check and provide the waveforms for GLx (MDRVR1_GLA) on the same time axis as the Rsense voltage (R41) positive terminal (MDRVR1_SP_ABC)? We would like to check to see if the ringing is the same on both nodes.

      • This would be points A and B on the modified diagram below

     Also, we noticed that you have a snubber design included in your circuit (R32 & C32, R36 & C37)

    Thanks and Best Regards,

    Andrew

  • Hi Andrew,

    To confirm this, can you help check and provide the waveforms for GLx (MDRVR1_GLA) on the same time axis as the Rsense voltage (R41) positive terminal (MDRVR1_SP_ABC)? We would like to check to see if the ringing is the same on both nodes.

    ANS: I have already bypassed this resistor by shorting it and adding more ground paths to the FET source pins in the bottom layer to remove this resistor from the current path, but still there is no improvement

    Also, we noticed that you have a snubber design included in your circuit (R32 & C32, R36 & C37)

    ANS: We just copied the DRV8353RS-EVM board's snubber circuit as is 

    I like to explain the issue little more here,

    1. With no load condition and IDRIVE setting to lowest 50mA, I see the rise time at GHx and SHx is >100ns (no transient seen)

    2. But when load current increases, rise time at SHx reduces to <70ns which forces the GHx to rise at the same speed <70ns

    Basically at high load, rise time at SHx and GHx from 0v to VDRAIN is <60ns and after this, at GHx rise time from VDRAIN to (VDRAIN+10v) is very slow (which is acceptable)

    I suspect this transient is due to the fast rise from 0v to VDRAIN at GHx and not because of MOSFET turning ON

    As GHx blindly follows SHx to keep the FET OFF there is no gate capacitance to charge, and it is completely out of control

    Somehow we have to reduce the rise time at SHx, so I increased the CAP in the snubber circuit, but that helped very little (around 10ns increase in rise time)

    As you mentioned sense resistor is inductive, let me capture a waveform by completely removing the sense resistor and see what happens

    thanks, Amutha..

  • Hi Amutha,

    I agree with Andrew. Negative spike looks like the ones caused by di/dt and parasitic inductance between LS mosfet source and GND and its behavior you described is also characteristic for them. It may be caused by something different but I think it is a small chance.

    Usually it happens when HS Mosfet turns on, after LS Mosfet body diode reverse recovery and when HS Mosfet turns off. There a few things that can be done:

    - optimization of  pcb layout to limit parasitic inductances, bypassing ESL of sense resistor usually is not very simple

    - slowing down HS Mosfet, for another DRV I am getting similar time of around 80ns but for Mosfet with Qgd=5.9C and gate current setting 200mA. I would double check gate current settings. Other option may be adding a cap between gate and source of HS mosfet. Something around 1-10nF may work, other option would be HS gate resistor but we would loose strong pull down feature.

    - bypassing some part of parasitic inductances with caps or schottky diodes. If you use DRV8353RS-EVM pcb layout I would try to place MLCC caps lets say around 100nF-1uF with possible small package at the back of pcb between VDRAIN bus and upper teminals of current sense resistors. It might work or not.

    Regards,

    Grzegorz

  • Hi Pelikan,

    Thank you for your comments

    Other option may be adding a cap between gate and source of HS mosfet. Something around 1-10nF may work, other option would be HS gate resistor but we would loose strong pull down feature.

    ANS: Yes, I added 10nF between HS gate and source and that did not help. Reason what I found is GHx rising so fast is due to SHx rising faster, when load current increases, so to keep the HS FET OFF DRV increases the GHx to VDRAIN with the same rate of SHx, only during this time I am seeing the spikes (maybe I am wrong)

    Also, I see around +/-800mV ringing in GLx even at no load condition (no motor connected) when I increase HS IDRIVE to just 100mA, please let me know what exactly causes this ringing even at no load current ?

    thanks,

    Chandra

  • Hi Chandra,

    If cap between gate and source of HS mosfet does not work I would try HS mosfet gate resistor around 22-100Ohm, during such tests damage to DRV8353 and mosfets may happen. To observe results it is better to measure voltage between SHx and GND and watch for voltage slope changes and additionally observe amplitude of the negative spikes.

    You are right about driver following SHx voltage, DRV8353 HS gate driver is a floating type and is referenced to SHx potential. Gate current flows from GHx terminal and returns to SHx terminal of DRV8353. To analyze HS mosfet gate voltage a differential probe is needed or to measure GHx to GND and SHx to GND at the same time and subtract those two voltages.

    Even if the load is disconnected I guess there still will be some current flowing through Mosfet output capacitances, RC snubbers and Mosfet input capacitances from gate circuit.

    If pcb layout is optimal I think there is a good chance that RC snubbers are not needed up to currents around 10-20A.

    During oscilloscope measurements it is good to use ground spring hooks, BW limit turned off and probe set to 1:10.

    Probably I will be available on Monday.

    Regards,

    Grzegorz

  • Pelikan,

    I did capture waveforms from DRV8353RS-EVM board and I do see same spikes even in EVM board.

    Does it mean my probes are inductive or EVM board is not designed well ?

    Also, in my board I captured waveform at very close to the DRV ground and GHx pin there I see lesser transient voltage than in middle of the trace and GND.

    But my question is what design considerations should I make to drive a 1000W motor, as just with 350W motor I am closer to spec violation?

    thanks,

    Chandra

  • Hi Chandra,

    "I did capture waveforms from DRV8353RS-EVM board and I do see same spikes even in EVM board." - That is normal.

    "Does it mean my probes are inductive or EVM board is not designed well ? "- The main problem with oscilloscope probes during measurements on Mosfet based motor drives is picking up noise, by using ground spring hook we can limit that problem but not completely (oscilloscope probe with common pigtail is a quite good near field antenna). If EVM board can work with current that is specified to work with ie. 15 A continuous / 20 A peak at some reasonable rise/fall time lets say around 200ns then I think it is designed well enough. By experience I know that EVM boards are usually designed well enough to make some evaluation tests and they may be a good start point to design your own pcb board but most of them are not ready to use solutions.

    "Also, in my board I captured waveform at very close to the DRV ground and GHx pin there I see lesser transient voltage than in middle of the trace and GND." - Either you moved probe to place where it picks up less noise or transient is limited by pcb trace inductance and DRV8353, these transients usually originate in half-bridge and are the largest in that area.

    Going from 350W to 1000W driver will be a really big step, the main problems that will increase significantly are:

    - limiting overshoots/ undershoots and ringing caused by much higher current and di/dt

    - heat dissipation and its management

    - keeping Mosfets within SOA and DRV8353 within its absolute max specs.

    - choice of Mosfets that will suit your needs ie. voltage, Rdson, package and its parasitic inductance, immunity do dv/dt turn on etc.

    - making a choice between solid and splitted ground design.

    I would spend some time playing with current design and evm board, try to improve it and understand as much as possible how it works. There are quite a lot of materials available online that can help to design motor drives.

    https://www.ti.com/lit/an/slva959a/slva959a.pdf  - that Application Report is a good start. If you like I can find some more materials that I used when I was starting my adventure with design of motor drives.

    Some reference designs like:

    https://www.ti.com/tool/TIDA-010056#technicaldocuments

    https://www.ti.com/tool/TIDA-00281

    https://www.ti.com/tool/TIDA-01485

    https://www.ti.com/tool/TIDA-00436

    may be of some help and as well a good start point.

    Regards,

    Grzegorz

  • Pelikan,

    Thanks a lot for your valuable pointers for the design and support

    Like to know what if I connect a fast switching low leakage diode across LS gate to ground where cathode at gate and anode at ground.

    Will it help in clipping the negative transient at GLx?

    thanks,

    Chandra

  • Chandra,

    No problem.

    Using small high speed Schottky or unipolar TVS diode the way you described should limit those negative spikes. For gate circuit I would use probably some small diode in package like 0805 possible close to DRV (impedance between DRV's GND and diode's anode should be small and connection between diode's cathode and LS gate trace possible short). I used PMEG10020ELRX between LS Mosfet source and ground plane to bypass ESL of current sense resistor, it reduced negative spikes over 2 times, but you will need probably a bit smaller diode. Diode should have low capacitance in order not to limit current that is needed to drive Mosfet.

    After placing diode, please check if it is not causing any new problems like resonance in gate circuit.

    Regards,

    Grzegorz

  • Hi Pelikan,

    Thanks a lot for sharing this knowledge and info! 

    Hopefully these suggested design changes will help in resolving the HS and LS gate transient measurements observed by Amutha.

    Best Regards,

    Andrew 

  • Hi Amutha,

    We will review this thread's info further and align with the team to see if there's any other suggestions that might help.

    In the meantime, please do let us know if Pelikan's suggestions help improve the transients observed

    Best Regards,

    Andrew 

  • Hi Andrew and Pelikan,

    I am working on getting the suggested diode for the experiment, will update you once I get it and complete the experiment

    Once again thanks a lot for the support 

    regards,

    Chandra

  • Hi Amutha,

    Sounds good, let us know once you are done!

    Thanks,

    Matt

  • Hello Amutha,

    I wanted to check in to see if you had any updates here!

    Thanks,

    Matt