DRV8701: PWM version - Intermittent predrive or OCP fault

Colin,

Can you please reattach your schematic using the "paperclip" icon above?  It is not properly displaying.

Colin,

I am also suspecting a pre-driver Fault (PDF) as described in section 7.3.13.4 of the datasheet.  Can you try as an experiment to remove the additional FETs by just pulling the gate resistors?

Also, we recommend isolating the gates of parallel connected FETs with a 10 ohm resistor.  For example, R61 can be zero ohms, but R65 should be 10 ohms.  I am not sure if that is causing your problem, but another experiment.  

From your first post:

When the fault state is active, the fault line transitions low when sending higher speed PWM commands and the motor will not move, but sending slower PWM commands still moves the motor with no fault.

Can you elaborate more on this?  Higher speed is "higher duty cycle"?  Are you slowly ramping from lower duty cycles to higher duty cycles or quickly jumping between speeds?  

Hi Ryan, thanks for the quick response.

I actually thought of removing 4 redundant FETs yesterday (didn't think to remove the gate resistors which would have been easier). Everything seemed to be working fine and we could not reproduce the fault condition. However, a catastrophic failure occurred about an hour into testing and some traces on the PCB burned up...so it seems we need either the extra traces and/or FETs to deal with the current.

I can try using 10 ohm resistors for the parallel FETs on a new board tomorrow.

To answer your question regarding PWM, the controller keeps both IN1 and IN2 pins at logic high at all times when not in use (100% duty cycle on both pins, which is equivalent to brake mode) until we want to move the motor, at which point we begin to decrease the duty cycle of one of the input pins for forward or reverse. This is done at 20kHz and seems to work well when it's working.

When the fault state is active, this means smaller duty cycles (higher speed ~= 25-60% duty cycle) for one pin are causing the hbridge chip to pull it's fault line low, and larger duty cycles (lower speed commands ~= 60-100%) still work. I have not measured the exact duty cycle threshold yet where the fault occurs, but it seems consistent and does not matter if we change rapidly to a higher speed or with a gradual ramp up - the result will be the same for roughly the same range of duty cycles.

Hi Ryan, from my original post: "MOSFETs being used: NVMTS0D4N04CLTXG".

RDS(on) = [0.3, 0.4] mOhm, where VGS = 10 V, ID = 50 A. Maximum motor current is about 30A for this application from past measurements.

I can see that the total rated nC for these MOSFETs is slightly high, if I'm calculating correctly for the driver chip circuit: Qg < I_VCP / f_PWM = 12mA / 20000Hz = 600nC. Total gate charge for these FETs is 341nC, but doubled due to the extra parallel FET (682nC)? Do you think this could cause the fault?

It looks like I won't have time to swap the jumpers to 10 ohm resistors until next week or early January now, but I will let you know the results when I do. Please let me know if you can recommend a replacement FET and I will try that as well.

Thanks for your help!

Colin,

Definitely could be at the limit for this device with those FETs and removing one FET seems to solve the problem.  

I am not sure why 1 FET did not work as 30A should be easily achieved with the 0.4mohm Rdson of a single FET.  Power dissipation is <1W with a single FET considering I^2R losses.  900*0.0004 = 0.36W.  I wonder if there was a over-voltage condition that could have damaged it.  

This FET is also 40V and Rdson is 0.8mohm at 10V, so right at 2x the ON part, but still should be sufficient for 30A.  The package is considerably smaller at 5x6mm.  And total gate charge is about 1/3 of ON part.  Unfortunately it is not pin-to-pin due to the size difference.

Hi Ryan,

I was able to install a new driver PCB today, but before I did I swapped some gate resistors out. I did not have 10 Ohm on hand, so used 20 Ohm for every parallel FET (for example, R52 = 0 Ohm, R57 = 20 Ohm, etc).

I used a thermal camera to watch the driver as I started the motor and the motor doesn't turn at all now, but the driver makes a high pitched whining noise and the FETs with the 0 Ohm jumpers start to heat up very quickly (at rest, the case temperature of a FET is about 38C, but will heat up to ~55C within a second or two when trying to drive the motor). I haven't left it for longer than a few seconds, worrying I might damage the board. The hbridge IC also starts to heat up at the same time as the FETs.

I will solder in all 0 Ohm jumpers again to make sure the board itself isn't faulty, and then try an alternative FET when I can.

Colin

Colin,

Thanks for the update.  There is no good reason for the driver to heat up like that.  Do you have another PCB to try?

Tried again today with the 0 and 20 ohm resistors still in place, thinking the controller firmware may have been out of date when I last tested, so I reflashed the controller.

The motor was still having a difficult time moving, but would start up slower than normal, and 1 pair of FETs would heat up within a few seconds from ~25C to about 70C. Resetting the speed command back to 0% would cause the FET to cool down again to 30C or so, and I could repeat this, but didn't keep this state for more than 5 or so seconds at a time. So it's possible the FET temperature could have increased above 70C if  allowed to keep conducting.

I replaced all gate resistors back to 0 ohm jumpers and now the motor is moving again at normal acceleration and speed. With all 0 ohm gate resistors, some FETs will still get warm when starting the motor (the specific group of FETs depends on direction), but the temperature is very reasonable with a maximum of 40-50C and only when the motor ramps up to speed.

I noticed that the alternative MOSFET you suggested is not only the wrong size, but also has a different pin out, and would require soldering jumpers from pins to pads (the gate is on pin 4 as opposed to pin 1 of original parts). I know they aren't TI parts, but I found these potential alternatives and wondering if you could comment on using them with this driver IC:

Alternative#1 - NVMTS0D7N04CTXG

Alternative #2 - TK099V65Z,LQ

Thanks,

Colin

Colin,

The Toshiba part is 100mohm and 650V part.  That is moving in the wrong direction on Rdson and overkill on voltage.

The ON part seems reasonable and a good alternative.  

Hi Colin,

I don't have experience with DRV8701 but I have some with DRV8711 and stepper motors.

Probably you are right that gate charge of two NVMTS0D4N04CLTXG is too much for DRV8701.

I would try to minimize Mosfet losses by balancing switching losses and conduction losses.

Probably even with single NVMTS0D4N04CLTXG switching losses are much higher than conduction ones.

I would try solution with single NVMTS0D4N04CLTXG (no gate resistors if there are not other problems like ringing and EMI)

then I would try single NVMTS0D7N04C and compare temperatures. This way I would know which direction I should go, whether

lower RdsOn and conduction losses but higher switching losses or opposite. 

You can also measure with oscilloscope Mosfet switching times and estimate switching losses, if they are much higher than

conductive losses then I would go for Mosfets with higher RdsOn but lower gate charge.

One disadvantage of faster Mosfet switching are possible problems with ringing and EMI but those

can be minimized by proper PCB design and other solutions.

Best Regards,

Grzegorz

Colin,

It just crossed my mind that bulk capacitor of 470uF (if it is electrolytic type) may be a bit

to low value for 30A ripple current.

Regards,

Grzegorz

Grzegorz,

Thank you...all good suggestions!

Hi Ryan,

I ordered the same package FET with lower gate charge a few days ago and they arrived this morning (NVMTS0D7N04CTXG). I had time today to swap out all 8 transistors and everything was looking good on power up. I am still using all 0 ohm jumpers on all 8 gates. I did not see the recommendation here to use single transistors until after swapping the parts unfortunately

About 30 minutes into testing, the breaker supplying the motor current tripped, and I found that a few of the FETs were severely damaged with shorts between gate, source and drain, and shorts between several pins on the driver chip, including GH2 to ground. I have replaced the driver IC and the damaged FETs, and the motor is turning again now, but I will be monitoring current draw very carefully when I continue testing tomorrow. Not sure what could have caused this fault, possibly too much heat while swapping parts, or possibly a voltage spike above the max Vdss of the transistors? (Vdss max = 40V) Temperature is very reasonable, nothing more than 50C maximum that I can see anywhere on the circuit board, even while ramping up the motor.

Thanks for the suggestion Grzegorz, I will order some 1000uF caps to try out in place of the 470uF's, and connect a differential scope across Vds of the high side FET to observe switching time.

I think switching losses can be approximated using this formula? Power = 1/2 * Vin * Io * (T_rise + T_fall) *f_switching [Hz]

I may try removing the redundant 4 transistors if I can get this new part to solve the original fault condition, without causing damage. It would be nice to simplify the circuit and reduce part count.

Thank you for your help!

Hi Ryan,

In the last few weeks I have only had a handful of days to test the driver with the lower gate charge replacement FETs. I have seen the same behaviour 3 times now on two separate PCBs (both of same design), where 1 or more of the FETs surrounding M3B (Q9, Q10, Q12, Q13) have blow through damage and shorts between drain source and gate. I will almost always see a short from gate-source, source-drain internally through the DRV8701 chip as well, and so need to replace the driver chip every time a failure occurs. This is with all 0 ohm jumpers and the 1000uF bulk capacitor, in place of the 2 ohm gate resistors and 470uF cap respectively. The FETS on the M3A output seem not to be affected by this failure, but could be that the motor is typically being run in one direction when it is stopped that this is typical.

There is a large relay in line with the motor voltage controlled by a latching estop button, and the failure seems to happen when this estop is either pressed or released. I assume that the failure is blow through and caused by a massive voltage spike at the drain of the high side fets, and is either coupled into the gates to drive them and accelerate the burnout (gate voltage maybe exceeding 20V, over max Vgs as well), or there is reverse voltage from motor inductance/inrush burning out the internal FET body diodes. I have measured 24V line from drain of the high side fets to ground using a scope and I can see occasional ringing, but not more than a few volts (I think 5Vpk-pk max, but forgot to get a screen capture). Most of the time, hitting the estop produces no ringing at all though. I have also measured Vds with differential probes of each of the M3B pairs of fets and I can see that the PWM signal being sent looks normal at 20kHz, although I have not calculated losses yet, and also more concerned about what is happening during the estop press or release now, not so much what is happening during normal operation. 

To try to work around this estop failure, I added 15V zeners to the gate-source of each fet pair (max driving voltage should be ~10V according to the datasheet at 24V VM, so I think precise 15V zeners should help protect against any transients), and I added external schottky diodes across DS to supplement the weak internal fet body diodes. I also tried adding two snubbers consisting of a 100ohm resistor in series with a 10nF ceramic cap, both of which were placed across DS of each pair. I have seen the PCB fail with this snubber and without it.

I thought that the large estop relay, whose coil shares the driver PCB ground, might be causing damage with negative spike on the ground plane when switched, so I completely isolated the coil and powered it straight from the battery for testing. I have also seen that there was a relatively high resistance ground path from input power to this H bridge and the MOSFETs, so I soldered an 18AWG wire straight from the negative battery input on the PCB to the sense resistor ground connection in case there was increased inductance due to the layout.

None of these fixes has solved the issue and the same FETs are still dying (or the driver chip is still dying which kills the FETs in turn). The estop relay does not cut 24V power to the DRV chip VM input, only to the MOSFETs, but I suppose a voltage spike could travel back through the relay contacts on closing from the motors to the driver chips...

So the original issue looks like it can possibly be solved with the new fets (the tdrive timeout fault), but it comes with the new issue where the FETs will become randomly damaged with an estop press. 

Not sure what else to try at this point without completely redesigning the PCB to accept alternative FET models, but if you have any ideas, please let me know. I can attach scope captures if there is anything specific you would like to see, or motor data.

Just tried installing the TK099V65Z,LQ alternative MOSFETs I had ordered last month, as a final effort to get this driver working today. Although the Rdson is much higher, I think they should still be able to handle the ~15A motor current (worst case loaded motor uses about 15A, measured with a clamp ammeter, although I am trying to design for 30A total).

I also figured that the higher voltage ratings might avoid blow through destruction from whatever is causing a fault. When I put these in for all 8 mosfets of the h bridge, the driver made a high pitched whining noise and the motor would not turn. I was able to capture some waveforms of the A and B motor points to ground, as well as the fault line from the driver. I was about to measure the gate to ground, but I believe the driver failed before I could get a waveform on the screen. Now the fault line is always low.

Fault signal when trying to send forward full speed:

Motor point B measured with respect to ground while sending full forward speed PWM:

Zoomed in B to ground signal (same signal as above). Seems to be close to 2.5us t_drive limit at around 2.0us, but still under. Although the board failed before I could measure the gate voltage, this tells me that the fault should not be related to t_drive, and the gate voltage would have to be well above 1V, well within the 2.5us by the time it is fully driven

And measured A with respect to ground as well:

As I stated in my last post, there is some ringing on the 24V power rail, approximately 5V as can be seen in these images, but I don't think that should be enough to damage the IC, which is rated for 45V, and not sure why it can't turn the motor at all... 

Although the data sheet says I can share the power for both motor and IC, should I be isolating the two, or adding a better filter to the VM pin?? It says 10uF minimum, but maybe in this case I need something like 100uF???

Colin,

You definitely do not want to isolate the power connection between the DRV8701 and the external FETs.  This will impact the high-side VDS monitor that compares the voltage drop across the drain of the external high-side and source.  Since we don't have a separate pin for monitoring the drain voltage, the device internally uses VM as the reference so it should be at the same potential as your external FETs.  

Is it possible to share the schematic showing this external relay that you use to cut power to the driver and FETs?  If the motor is driving when doing this, it will become a generator when removing power and current will flow back through the body diodes of the FET and pump the supply voltage up.  A large bulk capacitor can help to absorb some of this energy, but the voltage still may be increasing beyond the rated voltage of the FETs and the driver.  I suspect this is why the driver is being damaged.  Have you tried to capture the voltage at the DRV8701 VM pins when hitting this ESTOP?

You might find this blog interesting.

https://e2e.ti.com/blogs_/b/industrial_strength/archive/2013/12/13/art-of-stopping-the-motor-vm-pumping

I have measured VM while hitting estop, but not really while moving - I have tried placing the oscilloscope on a cart and moving it with the machine, but it's difficult to measure while moving, without shorting something or ripping the probes away from the board. I have placed the machine on blocks as well, but then the motor is not loaded realistically. In all cases that I have measured VM, I can only see a +5V ringing at worst when I hit the estop, and at ~26V battery supply, the maximum is just over 30V.

Here is the power input section with reverse polarity mosfets:

  And the relay that is activated by estop:

Something I am confused about, is that if the motor is acting like a generator and spiking the supply voltage above max VM, wouldn't the mosfets have to be in coast mode, with at least one high side driven? We are sending a brake command to the motors (IN1 = 1, IN2 = 1) when we hit the estop or just when we set the speed back to neutral (0% speed). If the motor is set to low side slow decay and only the low side mosfets are on, could that cause a large voltage spike on the ground? 

The AVDD pin seems to be shorted to ground internally through the drv chip after testing with the TK099V65Z,LQ alternative transistors. None of the transistors failed this time like they typically do using the ON parts (I think due to the very large voltage rating of this part). I assume this means that there was a large voltage spike on the VM pin, which propagated through to the 4.8V LDO, burning a short through it..

Would it be advisable to add a small value resistor between 24V and VM (maybe a few ohms to limit current?), or a few uH inductor, as well as a larger capacitor in place of the 10uF on the VM pin? I might also add a TVS diode from VM to ground.

Hi Colin,

Rds of TK099V65Z is to high for you application even if two of them a connected in parallel.

TK099V65Z Rds for 80 deg Ta is 0,12 Ohm so for two in parallel is 0,06 Ohm.

DRV8701 OVC protection level is 1V so OVC trip current for two TK099V65Z in parallel is around 16A.

At the moment probably you are experiencing OVC fault.

On the last oscillogram I can see quite substantial ringing of switch node voltage, around 4V below GND

and around 4V above VM. That is probably for current around 16A and for higher currents ringing should

be higher and would violate DRV8701 Max node pin voltage (GND-2V and VM+2V).

The project that you have undertaken is quite tricky and I would make it in steps.

1/ I would make the board to work without any faults at normal conditions for some time (with no switching of VM Motor)

If it does not work I would decrease voltage/current and look for any problems that I mentioned before.

2/ Once the board works at normal conditions I would check ringing and voltages if they don't

violate DRV8711 and Mosfet max ratings.

3/ If they are within safe margins I would try to increase VM and current by 10 to 50% and see

if board still works with no problems.

4/ Then I would try using VM relay and solve any problems related to that relay.

Regards,

Grzegorz

Thanks Grzegorz,

Just so I understand you correctly, for step 1 above, do you mean not to disconnect power to VM, as in do not turn the power on and off? If so, I never disconnect VM right now, only the mosfet supply through the estop relay (VM is supplied to the DRV chip upstream of the relay), but I am not switching the relay right now, only focusing on signals to and from the DRV chip. I cannot easily change the voltage of the system as it runs from 2x 12V batteries. I would have to only connect 1 to decrease to 12V, but would require some rewiring, and is not typical of the system voltage.

I've started by reconnecting a new board to power but no motors connected to look at the signals without noise from the motors. All the signals without motor connected look like clean square waves with very little ringing or overshoot (I measured about 600mV undershoot on one of the motor outputs. Signals still look relatively clean when I connect a motor, but I have the machine up on blocks to run without full load.

Vgs of a low side pair looks to have some ringing to about -1.2V:

Gate drive time seems to be less than 500ns, till FET is driven on full:

Not sure how to interpret Vgs of a high side gate (ground on M3B, probe on gate, scope ground isolated from driver ground), as the peak to peak looks very close to the max Vgs of the FET I'm using now, and I think the source is being pulled down as the motor gets up to speed:

Voltage across the motor looks fairly clean, but also slight negative voltage spikes to ~1V:

This is with 1000uF bulk cap at motor output, but also added 15uF electrolytic at VM pin of drv chip in addition to a 10uF ceramic, and a 40V TVS zener to avoid damaging any more driver ICs. I am again using the original MOSFETS (NVMTS0D4N04CLTXG), which only issue was occasional fault signal with unresponsive motor (not the drv chip failing with the alternatives).

I'm thinking the original issue was not actually tdrive, as these FETs seem to drive well within a microsecond and within the 2.5us limit, so possibly the negative spikes or overvoltage was causing the chip to fault (mabye Vcp was dipping too low - although I have checked Vcp with a voltmeter during a fault, I have not measured with a scope yet). I think the next step would be to test with a loaded motor off blocks and see if the noise or ringing gets worse. If the fault occurs again, I will recheck tdrive as well as voltage spikes on the VM pin and on the motor nodes.

Hi Colin,

I meant not to disconnect 24V_ToMotors from VM pin of DRV8701 (of course please make sure the operation of cart/machine is safe).

Operating at lower power supply voltages and lower currents gives time to perform measurements if failure occurs too quickly after start.

This way you can avoid problems with frequent components replacement.

Ringing in Mosfet circuits is usually caused by parasitic inductances and current changes ie. di/dt, higher currents give usually

higher di/dt and higher ringing.

If the motor is not connected there should be no ringing.

It's quite essential to understand nature of mosfet ringing and pcb design to reduce that ringing. It is possible to find some good explanations on internet and youtube.

You can search for "mosfet ringing" "mosfet body diode" "mosfet parasitic inductances" etc.

There are also some good  materials about pcb design like  https://www.ti.com/lit/an/slva959a/slva959a.pdf.

That lecture shows the complexity of motor driver design process  www.youtube.com/watch

First oscillogram. It looks like LS mosfet gate is not fully charged to around 9.3V (9.3V - according to DRV8701 datasheet), probably because of too high mosfet gate charge.

It's charged initially just to around 5V but it should be charged to around 9.3V, there should be no that second slope, that part should be flat at around 9.3V.

Third oscillogram. It would be good to confirm that high Vgs voltage by measurement with two channels. One channel between GND and switch node, the second

between GND and HS mosfet gate. If the difference is 20V then there is a problem.

Best Regards,

Grzegorz

Colin,

To monitor 24V_ToMotors for peak values you can use some good (not very slow) DMM with Peak MAX function.

Best Regards,

Grzegorz

Colin,

Just another good video

https://www.youtube.com/watch?v=OhIChZWMVUQ&list=PL9B4edd-p2ah3Uj_HhARAW6jOe5IbvvVN&index=14

BR,

Grzegorz

Thanks for the information! I have tried to follow the layout recommendations in the datasheet as closely as possible. Yes, I see now that the gate voltage appears to be charging rapidly and then slows down, which is likely the weaker holding current from the driver as it transitions from tdrive.

Here is a differential signal on the high side, probe 1 on gate to ground, and probe 2 on source to ground:

This also seems to be around, or just above 20V...again, not sure if there is some abstract reason why the voltage might appear higher on the scope than it really is in circuit..but that seems like a hazard for damaging mosfets, many of which typically have +/-20V max for Vgs.

Colin,

Can you measure voltages on VM and VCP pins on DRV8701?

If VCP voltage is 20V higher than VM then DRV8701 may be damaged, then I would replace DRV8701

and measure the voltages again with motor not connected.

Regards,

Grzegorz 

This is odd: I'm certain that I'm measuring the same points with the same channels and probe settings (same attenuation) as yesterday, but today I am seeing 5 to 6V on gate to source for the same high side mosfet. I'm wondering now if this is a scope issue or if the chip or mosfet is behaving unpredictably. Either way, I can still see the gate charging too slowly, and the peak voltage is lower than the expected 9.5V, possibly not enough time to reach 9.5V during each PWM cycle.

The VM voltage is very stable with very little ripple and no noticeable spikes (when starting/stopping the motor, or running continuously) at 25.2V, and VCP  has about 1Vp-p ripple with mean voltage of 35.2V and also looks very stable. If I see 20V on the high side GS again, I will try to remeasure VM and VCP and post images here.

It looks like adding the additional 15uF capacitor on the VM pin has improved operation after testing today.

Colin,

Thanks for the update.  Is the motor moving as well as expected?