BOOSTXL-DRV8323RH: BOOSTXL-DRV8323RH: Custom Board MOSFET damaged

Hi Danny,

Danny Chan said:

My custom board is 99.9% replicate TI reference design BOOSTXL-DRV8323RH + LAUNCHXL-F280025C combo board with minor modifications, few month road test with the combo board without any issues therefore I can concluded custom board damage wasn't caused by 1. software 2. added 100v power capacitor 3. changed to 1mOhm shunt resistor 

What were the minor modifications that were made, is it the 1. software 2. added 100v power capacitor 3. changed to 1mOhm shunt resistor you mentioned or are there other modifications not mentioned?

I'm really not sure if these capacitors would be an issue or not, I do see it in the EVM design so I don't think this is the issue

Heres what the datasheet has 

I think to get to the root of the damage it would be helpful to see the following waveforms and their measurments:

VDRAIN

VCP

SHx to GND

GHx to SHx

Regards,

Yara

Hi Yara,

Thanks for your follow up, I will solder new Custom board and measure those signals again. 

Danny

Hi Danny,

Sounds good!

Regards,

Yara

Hi Yara,

Please find enclosed the waveforms of my Custom Board running under below conditions.

30v powered by Power Supply,  6 inches skateboard motor, motorVars_M1.speedRef_Hz = 150,  Speed Control mode, no-load current = 1A 

VDRAIN 29V

VCP 39.8V

SHx to GND

GHx to SHx

GHx to GND and GLx to GND

Danny

Hi Danny,

Nothing looks too concerning in these waveforms and it seems like you are well within the specs of the device, are you still experiencing issues like you did on your last board?

Regards,

Yara

Hi Yara,

Thanks for your quick reply.

Yara Al Shami said:

are you still experiencing issues like you did on your last board?

Not yet, my last 3 boards were damaged with 10s Li-po battery + e-Bike motor, your help in first step is confirm the waveform are within specs. so I will start test with 10s li-po + e-bike motor on tomorrow.

I will post my result here again ! 

Danny

Sounds good!

Hi Yara,

Bad news again, but little progression was made !

The new custom board can survive under no-load test because I have modified the TI universal_motorcontrol_lab example especially the updateExtCmdPotFreq subroutine, inside this function with below codes were removed, the flagEnableRunAndIdentify turned on all the time outside this subroutine.

if((objMtr->cmdPot.flagEnableCmd == true) && (objMtr->faultMtrUse.all == 0))
    {
        objMtr->flagEnableRunAndIdentify = objMtr->cmdPot.flagCmdRun;
        objMtr->speedRef_Hz = objMtr->cmdPot.speedSet_Hz;
    }

yesterday I started the e-bike full load test with SPEED CONTROL e.g. motorVars_M1.flagEnableSpeedCtrl = 1 without any issues, then I changed to TORQUE CONTROL e.g. motorVars_M1.flagEnableSpeedCtrl = 0, and both motorVars_M1.fluxCurrent_A = and motorVars_M1.startCurrent_A = 15A, CUTOFF CURRENT to ~ 30A ... unfortunately after 1 minutes my bike stop working and my BMS detected CUSTOM BOARD Input shorted (VBUS and GND) together. 

step 01 : to office immediately examine my CUSTOM BOARD confirmed both VBUS and GND shorted together.

step 02 : then removed all 6 MOSFET out of the board and measure it individually, confirmed PHASE A's Upper and Lower MOSFETs damaged with both DRAIN and SOURCE shorted together.

step 03 : without MOSFETs on board, I can measure the DRV chip's input to check if it's damaged, the result was many pins shorted together.

Yara, any idea or suggestion ? yesterday the full load test was just 10 to 15 A maximum 40V li-po battery, so I don't think it's MOSFETs over current or any other issues, the IDRIVE wasn't aggressive and the VDS cut off is high but within MOSFET spec. 

Danny

Hi Danny,

I am really unsure about your capacitors C56, C58, C60. Is there a a purpose for these capacitors? What IDRIVE setting were you using?

Regards,

Yara

Hi Yara,

Thanks for your quick response, the C56, 58 and 60 were follows TI's BOOSTXL-DRV8323Rx reference design which is connected to SHUNT SPx instead of GND 

my IDRIVE setting with 18K resistor => Trise 266ns and Tfall 133ns 

Danny

Hi Danny,

Is Trise 266ns and Tfall 133ns  what you measured Trise and Tfall to be? with the MOSFET you're using my calculations show that your rise and fall time are more around 400ns and 200ns?

Are you able to recreate the last failure? Can you recreate it while capturing waveforms? GHX and GLx? Are you ensuring that your inputs are timed correctly?

Regards,

Yara

Hi Yara,

Yara Al Shami said:

Is Trise 266ns and Tfall 133ns  what you measured Trise and Tfall to be?

do you mean SHx to GND and GHx to SHx ? I have attached above for your review 16 days ago, can you help to check it again ?

Yara Al Shami said:

with the MOSFET you're using my calculations show that your rise and fall time are more around 400ns and 200ns?

sorry I don't get it, I use 18K resistor to GND for IDRIVE setting of course I based on 8mOhm as MOSFET Qdg, anything wrong ? 

If this's too fast or incorrect, why the no-load testing without issues ? as I said it's look like when I switched to CURRENT MODE with fluxCurrent_A and startCurrent ~20A, CUTOFF CURRENT to ~ 30A  it's damaged immediately.

so my question is on SIGNAL perspective what's is different between FULL-LOAD and NO-LOAD ? on NO-LOAD testing the speed can reach to it's limited e.g. 250Hz but FULL-LOAD almost immediately damage, if it's damaged by OVER-CURRENT ? how's the VDS or software OVER CURRENT cut-off can help ? my VDS setting ~75A cut-off is VDS reliable for protection ? 

Yara Al Shami said:

Are you able to recreate the last failure? Can you recreate it while capturing waveforms? GHX and GLx? Are you ensuring that your inputs are timed correctly?

don't think I can capture it while my bike is running, all my previous testing / signal capture were at office with NO-LOA conditions.

Yara Al Shami said:

GHX and GLx? Are you ensuring that your inputs are timed correctly?

yes I think so.

Danny

Hi Danny,

I guess I'm asking how did you come to the conclusion that your trise and tfall were 266ns and 133ns? Through calculation using the formulas in this FAQ(https://e2e.ti.com/support/motor-drivers-group/motor-drivers/f/motor-drivers-forum/796378/faq-selecting-the-best-idrive-setting-and-why-this-is-essential) I calculated it to be 400ns and 200ns.

I think what might be happening is an abs max violation during normal operation of the ebike, often times I see customers typically use DRV835x for ebike and scooter applications as it can better account for conditions during operation that would cause VDRAIN and VCP to spike.

Can you estimate what the values of VDRAIN and VCP are throughout the operation of your bike?

Regards,

Yara

Hi Yara,

Yara Al Shami said:

I guess I'm asking how did you come to the conclusion that your trise and tfall were 266ns and 133ns? Through calculation using the formulas in this FAQ(https://e2e.ti.com/support/motor-drivers-group/motor-drivers/f/motor-drivers-forum/796378/faq-selecting-the-best-idrive-setting-and-why-this-is-essential) I calculated it to be 400ns and 200ns.

yes I read that articles before, based on the formula I use 18K to AGND for IDRIVE setting.

since my MOSFET Qdg ~ 8nC, IDRIVE setting = 18K to AGND which is 30 / 60mA

the result Trise should be 8nC / 30mA = 266 ns and Tfall should be 8nC / 60mA = 133 ns

if I was wrong, can you tell me what's your formula for the result 400ns and 200ns ?

as we discussed before my original BOOSTXL-DRV8323RH board come with R1 and R5 unpopulated mean IDRIVE Hi-Z 120/240mA which is much much faster that what I currently use, but BOOSTXL-DRV8323RH can survive, why ? 

Yara Al Shami said:

I think what might be happening is an abs max violation during normal operation of the ebike, often times I see customers typically use DRV835x for ebike and scooter applications as it can better account for conditions during operation that would cause VDRAIN and VCP to spike.

That's why I would like to consult you, I am not question about DRV chip because I have tested few months on BOOSTXL-DRV8323RH + LAUNCHXL-F280025C combo

Yara Al Shami said:

Can you estimate what the values of VDRAIN and VCP are throughout the operation of your bike?

Can I measure or records VBUS (VDRAIN) during FULL-LOAD test e.g. capture the max. VDRAIN value but not sure can capture the spike !

Do you think the DRV8323 damaged pattern was look liked VDRAIN and VCP were violated by spike ? 

OR do you mean once my eBike on FULL-LOAD the VDRAIN and VCP can be damaged by spike ? if yes how to prevent ? I use one BIG 100V 2200uF line capacitor, this big capacitor was added to my BOOSTXL-DRV8323RH + LAUNCHXL-F280025C combo to simulate my custom board without issues ...

Hi Yara,

Last time you ask if I have Altium Designer version, now I can convert from KICAD and upload to altium 365, if you willing to hep, I can send you my username and password for your review my design !!!

Danny

Yara Al Shami said:

I do believe so yes. Unfortunately though I cannot log into your account to review the layout.

it's because of TI's policy ? it's the temporary account without any confidential information, can you help ? 

Danny

Hi Danny,

The only way I could provide a layout review is if you send me the Altium file via email.

Regards,

Yara

Hi Yara,

I will try to convert and send you via email before that I would like to ask you one more question :  

In boostxl-drv8323rh board TI use 2 x 330uF 63v.

I use very big 100v 2200uF capacitor, do you think it's one of the reason cause spike ? if not bigger is better then what's optimal value of this capacitor ?

Danny

Hi Danny,

I'm actually not sure about this but I did find some content that might give you a bit more information:

https://www.ti.com/lit/an/slvaft0/slvaft0.pdf?ts=1726057141483

https://www.ti.com/video/6265440667001

Regards,

Yara