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.

DRV8873SEVM: Works only once, then destroyed

Part Number: DRV8873SEVM
Other Parts Discussed in Thread: DRV8873, L293, L293D

We bought the evaluation module. Our software engineer used it in my absence and said it worked only once.

I bought another one and together we wired it up carefully. We are driving a linear motor (24 V, 20 W). We used the TI software to control it.

The linear motor moved in one direction. It got to the end point of its travel and stopped by a switch internal to the linear motor that disconnects the power. At that point the fault light on the evaluation board went red and the SPI indicates all ones. The chip is now dead.

This is what happened to the first evaluation board. Now I am convinced our software engineer did nothing wrong and we are two out of two failures with this chip.

Why?

(And can I get a refund?) 

  • Hello Iain,

    Thank you for posting to the forum.

    Can you provide more information about the motor in your application? What is it's nominal stall current at 24V? In certain cases, if the load stall current is very high and last for less than the over-current protection deglitch time (3µs), the device may not protect itself and damage can occur to the driver. This is my first theory.

    Did you try driving another load?

    Was there any burnt or visible damage on the driver IC? It may be the case that another component on the EVM was damaged and not the driver. A good test to veify if the chip is functional is to disconnect the load and measure the output voltage. What is the output voltage when load is disconnected?

    Regarding the refund, let me talk with out team internally and see what we can do about that. I will let you know as soon as I have a response.

  • Hi Iain,

    For the refund, go to this link to reach out to customer service. I can help you with any technical questions but all of the refund request must go through our customer service.

  • Driving linear motor

    Pablo,

    Thanks for the quick reply. I have attached a picture of the linear motor. This is close to the type of product we use:

    https://louieactuator.en.alibaba.com/product/60331531201-806680499/Xtl_Linear_Actuator_24V_for_Industrial_Use.html

    As the arm moves (left to right in the photo), a microswitch in the mechanism disconnects power when the arm has reached the end of its travel. The movement of the arm ceases by the removal of the applied dc from the motor and not because the motor is stalled.

    We don’t see any magic smoke from the DRV8873, but the fault LED is always on, the IC never recovers after a power cycle, reports all 1s on the SPI, and both output voltages are at ~100 mV (with or without a load).

    We did not try to use a resistive load (and cannot try that now the IC is dead). But I am unconvinced that would have made a difference. I think the issue occurs because the motor is removed from the IC whilst the IC is driving the motor. (It is removed by the microswitch.)

    We have previously used the old L293 through hole part on the previous design of this board:

    https://www.ti.com/product/L293

    It has no problem driving this motor reliably, but it isn’t available in SMT.

    It seems that disconnecting the motor whilst the driver IC is driving the motor should be a reasonable thing that the IC must handle. I suspect it cannot, because the motor worked well as it travelled in one direction (which took 10 s) and then the IC died once the motor arm got to the end.

    And it died on two separate evaluation modules!

  • Hi Iain,

    Disconnecting motor from driver by microswitch is probably causing high di/dt (current is disappearing really fast).

    High di/dt is causing high voltage spikes on parasitic inductances in board and probably kills the DRV8873.

    I would buy 10-20 of DRV8873s for replacement and make a few trials with evm boards.

    1. If you don't use sense resistor I would try to connect SRC pins directly to GND under the drv8873.

    The point is to reduce inductance between SRC pins and GND.

    2. I would try RC snubbers between motor outputs and/or between motor outputs and GND/VM.

    3. Another solution would be building your own board with places for optional components like

    schottky diodes / capacitors/ RC snubbers between outputs and GND and VM, between SRC pins and GND.

    The key is to keep possible low inductances of traces and components (short and wide traces), small footprint

    components.

    Regards,

    Grzegorz

  • Hi,

    It looks like R27 is some kind of option resistor, DRV8873 probably does not use LS current sense

    resistor at all. I would definitely tie SRC pins to GND under the DRV8873.

    Description of SRC pins in datasheet is "Power FET source.Tie this pin to GND through a low-impedance path"

    Regards,

    Grzegorz

  • Grzegorz,

    Your point is well taken. The equivalent circuit looks like this:

    LM and LG are inductances on the PCB external to the IC. LW1 and LW2 are the inductances of the wires going to the load. When the current is broken by S1 as you say, di/dt is huge across LM and LG. Vh is easily driven below ground and Vl is easily driven above Vm. This results in destruction of the IC.

    The interruption of the current in LW1 and LW2 should have no effect unless the voltages are so great that there is an arc across S1.

    I still believe that the sudden disconnection of the load is something that a designer should accommodate. I naively assumed the evaluation board from TI would be designed to be bullet proof and that one can take the schematic for it and if necessary remove what might not be required. It seems I was wrong!

    As stated above, we have used the L293 on previous boards. This has worked well for years. The data for that IC (link is above) recommends external protection diodes in §8.1 and §8.2 (and we put them on our board):

    There is no mention of such a requirement in the data sheet for the DRV8873. Figure 36 of the data sheet does not show protection diodes. Perhaps the designers of this IC have forgotten what was done on the old parts developed years before. Whilst there are large number of fault conditions incorporated into the DRV8873, the most basic — that the load my suddenly become disconnected — is unaccounted for.

    The item you reference on page 4 might not entirely solve the problem, even if the inductance is 1 nH. However, even that suggestion was certainly not heeded by the designers of the evaluation board for the DRV8873!

  • Hi Iain,

    Thank you for the detailed information.

    The IC has internal body diodes which are used to allow current to flow when the FETs are disabled.

    I think the best way to protect against the voltage spikes is to add clamping diodes, like TVS diodes, on the output terminals to keep the VM voltage below the absolute max ratings. This FAQ, has more details about reducing voltage spikes and can be helpful.

    The DRV8873 does not have open-load detection. We do have other devices with open-load detection but in those devices, the outputs will be disabled in the case the load is disconnected which is not what your applications needs.

    The DRV8873EVM was design for most common uses. Your application is unique, disconnecting the load while there is current through the output terminals is something that we recommend, so it was not taken into account when designing the board.

    I think the next step will be to replace the damaged ICs on the EVMs with new ICs and add the safety circuitry to limit the output voltage. This should prevent damage to the IC and you can continue to evaluate the DRV8873. I'll be happy to help out if you need more help.

  • Iain,

    First of all, I am sorry you are having issues with our device.  

    The DRV8873 has the diodes integrated...body diodes across the FETs.  You can see this in multiple figures throughout the datasheet.  The L293D (optional part number) also has the diodes integrated.  I think you are using the L293, so you added them externally.  

    Are you using the same power supply between EVM testing and your actual L293 board?  Some supplies can't sink current and the supply voltage will get pumped up quickly.  Adding more bulk capacitance or a clamp on the supply side would help to avoid this.

    If you are able to order more devices and replace them on the EVM, you could put a scope on the supply and see if this is in fact the issue.

    Regards,

    Ryan

  • Ryan,

    I am aware that MOSFETs have a built in body diode. However, I think you misunderstand the purpose of the diodes I have suggested adding and why the built in body diodes cannot achieve that purpose.

    I am sure the power supply is not the issue. The motor is just 20 W and in each case we have the same 100 W PSU driving our old board and the EVM.

    We plan to replace the damaged IC with new ones, add the required protection diodes, and resume testing.

    I still maintain that a designer should assume that the most obvious failure is that a connector comes off and the load is removed. The EVM should have been designed with this in mind!

    I will report later whether the updates to the board resolve the issue.

  • Iain,

    Thank you for confirming the supply is not the issue.  Look forward to your results.  Will keep this post open until we hear back.

    Regards,

    Ryan