Hi!
We designed a valve controller with the DRV8312. It was produced and assembled into customer devices since June 2015. 797 out of 800 pieces work reliably. The latest batch shows failure rates of 100%. Another batch from a different supplier shows the same behavior.
The valves work as expected first. After 40 up to 80 shifts they fail (see tests below).
All failing controllers show the same symptom. One of the three phases does not work anymore. The high side FET does not switch. Apart from that, the system is fine. Moving the valve to a position where the motor first run on the remaining two good phases leads to a slow but complete movement. Shifts starting at the broken phase do not turn on the motor in any way. You can just see the other phases pulling the damaged phase to ground through the motor coils.
The motor shifts the valve during just 200ms. Then in real application it can cool down for at least 1 minute. In tests there is a cooling period of 1 second. The shift rate was proven to have no impact on lifetime.
I did a bunch of tests and checks:
1. supply voltages in idle mode:
- 24V ripple 14mV; ok
- 12V ripple 10mV; ok
- 3.3V ripple 10mV; ok
- Vreg (5V) 13mV; ok
2. supply voltages while moving
- 24V peak drop 1.74V; ok; peak overshoot 80mV; ok
- 12V peak drop 500mV; ???; peak overshoot 40mV; ok
- 3.3V peak drop 63mV; ok; peak overshoot 40mV; ok
3. observed movement and failing
- valve moves as expected for 40 up to 80 shifts, finally during speeding up one phase stops working
- in rare cases the valve starts working for a few movements but then finally fails
4. current limit
- 6.3A; ok
5. over-/undershoot of PWM signals
- 0V
6. maximum current:
- the calculated maximum current through motor coils is 7A. Measured values are at about 4A
- recommended operating conditions according to the datasheet are 15A(peak) and 16A(absolute maximum rating; "Transient peak output current (per pin), pulse width limited by internal overcurrent protection circuit")
7. buffer capacitance PVDD (24V)
- reference design: 1000-2000µF
- in system: 1880µF
- tested with extra 3300µF, directly soldered to controller connection: reproducible longer lifetime (about 190 shifts) but failing as well
8. buffer capacitance VDD + GVDD (12V)
- reference design: 381µF
- in system: 271µF
9. soldering of thermal pad
- visible solder junction
10. design notes according to DS chapter 9.2.1.2
- current requirements: 5%-10% of load current; ok
- voltage decoupling: ok
- overcurrent threshold: resistor ok, test for higher limits done
- sense resistors: not needed
- output inductor selection:
- recommended: >>10µH
- in system: 1.87mH; ok
- note, that high current slew rate like from shortcuts might damage the driver!: does not apply
11. massively coupling power and signal grounds
- increased lifetime (about 360 shifts), but finally failing as well
12. temperature checks
- 30-40°C at case
- thermographic check shows short temperature shifts of about 2°C after each shift.
13. logistic checks
- both suppliers buy from well known and reliable distributors
- both tested batches have different datecodes
For further design information, please find attached a PDF with schematics, layout and real view.
These are my findings so far. None of the values seem critical to me. However, the design that worked so reliably suddenly fails.
Anyone out there, who has an idea, what's going wrong here? Every hint is much appreciated!
Thanks to everyone, having a look on this.
Best Regards,
Ralf
