Hi team,
a customer has came across to motor driver failures. Please see the description.
DESCRIPTION
We are seeing a low rate of unusual component failures in product that has been delivered and is in use by customers (hospital nurses). When the motor driver fails, it fails to drive the motor in one direction, but continues driving the motor in the other direction. There are no external indications of damage to the silicon such the component becoming warm, changes in power consumption, or changes to pin voltage levels, such the charge pump output voltage, and no signal from the status flag. Is this type of failure mode familiar to your design/support staff?
MY ANSWER DESCRIPTION
please provide schematic (your MCU and the MC33926) with voltage levels and part values.
CUSTOMERS REPLY DESCRIPTION
I've attached a simplified schematic showing the relevant circuits. . The operating voltage is 9.5V-16V (lithium battery), and the motor stall current is 2A. All signals to the motor driver have been verified to have correct levels and are identical for both good and failed assemblies. The charge pump voltage comes up to a level well within the specification range when the bridge activates, even in the direction the motor won't run.
This is the third generation of product that uses the same motor drive circuit, and the boards now failing have been used in production for 2-3 years without seeing enough failures of this type to garner attention, as it has now. We've already made a very considerable effort to diagnose the cause of these failures, so I have a lot of information I can pass along.
- A failed assembly can drive the motor in one direction, but not the other. The failure always seems to be the same rotational direction.
- The failures we see occur with printed circuit assemblies that have passed functional testing at multiple production levels. The failures are sometimes seen in delivered product after a considerable period of normal use by a customer.
- In most cases, the failure can be reproduced on the bench reliably after removal from the product. Bench testing uses a 12V power supply rather than lithium battery, wire instead of cables, and a different motor.
- Almost all of the failed assemblies being operating correctly at some point, even after being reproduced on the bench reliably many times. For example, one assembly I tested extensively on Tuesday worked on Wednesday after sitting on the bench overnight.
- We've never seen an failed assembly that began working once again begin to fail. One previously failed assembly was tested for many hundreds of hours without failing again.
- Seeking to understand the problem, I've tested the half bridges separately exercising each bridged transistor alone (a diagram is attached). In healthy assemblies, all four FETs drive the motor. On a failed assembly tested Wednesday, the current through the high-side FET on one side did not turn on fully (current was half) and neither low-side FET could be turned on. The status flag was asserted when I attempted to turn on the low-side FETS.
I thought a continuous 16V supply voltage might be the reason. Most probably I was wrong. Please see below.
MY ANSWER DESCRIPTION
you have done a great job with testing the failed MC33926. I was thinking to test the logic, whether the INV input changes the logic state of the IN1 and IN2 and thus change the direction of the motor. But I see you have done that too.
We have an evidence of destroyed MC33926 when it was continuously supplied by 28V. Although it is stated in the datasheet, that it should withstand the 28V in Steady-state, the MC33926 has been primarily designed for 12V systems and is optimized for such systems. It is tested at 28V considering a double battery condition while jump starting a 12V system.
According to your description, the supply voltage reaches up to 16V. It is an only thing, that comes to my mind, that the continuous 16V supply voltage triggers some protection and after bench test, when its supplied by 12V it returns to normal.
Would it be possible for you to add an LDO, between the battery and Vpwr pin? I know it will increase the price of the device and it should be stated in the datasheet.
I have found one for a price of 1$. Please see this link and this link.
Another way to deal with the failed MC33926 would be to send them to the NXP Quality team for a failure analysis. To do this, please contact your device NXP Authorized Distributor directly and ask them to initialize a CQC (failure analysis), they should be able to start this process.
It is also strange, that both low side drivers cannot be turned on, but the motor still can turn one way. According to the datasheet, to turn the motor one way or other, one high side and one low side transistor must be turned on and the other two turned off. Please see a picture attached.
CUSTOMERS REPLY DESCRIPTION
The 16V I mentioned is the value we use when designing new circuits and includes a considerable margin above the actual value. In practice, our 12V Li polymer battery should never see even the 15V sometimes used to charge 12V lead acid batteries. Additionally, the assemblies often continue to exhibit the problem when run from 12V on the bench. It was during the second day of testing that one of the three failed assemblies in this last batch began working perfectly the next day. That the failure sometimes goes away for no apparent reason and never reappears is a distinct and well documented characteristic of this problem.
When the low-side drivers didn't turn on with the motor connected to 12V I may not have been clear in stating that the status flag was asserted when this occured. The IC clearly detected something wrong. The same assembly drove the motor at half current on one high-side FET and ran perfectly using the other high-side FET.
I wanted to get your assessment without steering you toward a particular root cause, so I didn't tell you two key detail in my last message. Two popcorn part values was changed inadvertently by our contract manufacturer when they copied the design used in our previous generation products.
1) The status flag pull-up value they used is 4.7 ohms, which means the current drawn when the status flag is low will just barely exceed the specified maximum. For several reasons, it seems very unlikely that this is the cause of the problem.
2) The charge pump capacitor used on the CCP pin was increased by a factor of ten to 33uF, putting it 3.3 times the .01uf maximum specified in the datasheet. Taking this to be root cause doesn't seem like a good fit because changing the capacitor to one with a proper value on a failing board doesn't correct the problem. Additionally, the charge pump is cleanly producing the right voltages and there is no noticeable difference between the voltage in the working and failing directions. Also, there are .01uF X5R capacitors with values that are -20%/+100%, and I assume the chip designers took this and a sizeable design margin into consideration when specifying the CCP cap value.
If you have access to the chip designers can you ask them what would happen if we used the wrong CCP value, and if there is mechanism that would explain the fact that the failures continue occuring if the CCP cap is replaced with the correct value?
I will take you up on your offer to perform a failure analysis and contact our distributor. Barring feedback from the chip designers saying the CCP cap could be the cause, this is a worthwhile next step.
With Best Regards,
Jozef
