4

Is it possible to electrically damage a microcontroller (or any chip) in such a way that it becomes very sensitive to mechanical stress?

Some background:

We are developing a board with 7 identical microcontrollers, each running the same motor control firmware. After running without incident for dozens (sometimes hundreds) of hours, the 7th microcontroller starts behaving oddly. Specifically, the output of its internal core voltage regulator begins drifting away from its 1.8v target. Even more strangely, this output voltage now seems sensitive to mechanical stress. Giving the board a slight twist reliably causes this voltage to fluctuate (usually upwards), and become quite choppy.

Extensive probing and measurement has failed to detect any fault in the PCB itself. The decoupling capacitors were removed, measured (with no problems found) and replaced. As far as we can tell, the microcontroller is receiving correct and stable voltages to its power pins, which do not fluctuate when the board is stressed. Fluxing and reflowing the faulty chip does not fix the problem, but replacing the chip does fix it until, after some hours of use, the problems appears in the new chip.

This behaviour has been seen on twice on two boards, always on that 7th same microcontroller.

It looks like there is some kind of problem inside the chip. One hypothesis we are pursuing is a high voltage spike from the motor driver is somehow making its way to that chip and causing subtle damage.

Is it possible to electrically damage a microcontroller (or any chip) in such a way that it becomes very sensitive to mechanical stress?

JRE
  • 67,678
  • 8
  • 104
  • 179
Rocketmagnet
  • 26,933
  • 17
  • 92
  • 177
  • Other than the unique microcontrollers, what is the same and different about the various positions? Is the seventh position closer to an electrical input, for example? Are higher voltages or currents closer to it? Does that position on the PCB experience more mechanical stress? (Also, what is an electrical _insult_?) – JYelton Jun 24 '22 at 20:07
  • Also, do you have any images (that you can share) of the board? One thing I've learned as well, is that during manufacturing, some processes would impart stresses on the board and result in (for us) cracked capacitors on a small percentage of boards, but always on the same component. – JYelton Jun 24 '22 at 20:09
  • 1
    What kind of package are the microcontrollers in? It's not uncommon for mechanical stress to move a bandgap for example, but why it would happen only after some electrical overstress is an interesting problem. – John D Jun 24 '22 at 20:09
  • @JohnD - The chips are QFN68. We don't know that there was any electrical stress. It's just a hypothesis we're pursuing. – Rocketmagnet Jun 24 '22 at 20:15
  • @JYelton - The board isn't very large, but there are a couple of notable differences about the 7th MCU. Firstly, it seems to run very slightly hotter than the others (55ºC rather than º50C), probably because it's near the LDOs, and also near some Ethernet chips which run warm. Also, looking at the PCB layout, I can see that it receives its 5v supply at the end of a long track involving several vias, and having already delivered power to the other MCUs. Also, I should note that the board is 0.8mm 6-layer flexi-rigid. – Rocketmagnet Jun 24 '22 at 20:17
  • 2
    _"So, Is it possible to electrically damage a microcontroller (or any chip) in such a way that it becomes very sensitive to mechanical stress?"_ - Yes, it's possible. – Bruce Abbott Jun 24 '22 at 20:52
  • 2
    @Rocketmagnet I've had CPUs (from ADSP-21xx from Analog) which worked well for a year or two, then suddenly we were getting 50% failures to software testing performed at every boot-up (darned glad that code had been added.) We reported it to Analog and they found the problem. They'd moved the IC wafer manufacturing to a different FAB and their own testing had verified the chips. They asked me for the source we used to detect the problems, verified things on their end, modified the FABing, and our parts started working again. Might consider that there is an unknown (now) latent issue. – jonk Jun 24 '22 at 21:37
  • @jonk - That's amazing. For now I'm going to assume the problem is our fault. – Rocketmagnet Jun 24 '22 at 23:06
  • 1
    Perhaps a silly question, but did you test the grounds? Also does the chip have wettable flanks, and an exposed pad underneath? Finally - you could try swapping chips from position 1 and position 7 and see if the fault follows the chip. – Bryan Jun 24 '22 at 23:15
  • 1
    @Bryan, swapping chips to see if the problem follows is one of the best tips offered. With so many rabbit holes (PCB defect, external parts defect, design defect, etc.) this one helps exhaust spending too much time on the wrong path. – icodeplenty Jun 25 '22 at 15:05
  • @Bryan - The chips have wettable flanks, which all look good. No visible gaps, dry joints, cracks, voids or whiskers. We have not tried swapping the a pair of chips on the board. I might try that this week. – Rocketmagnet Jun 26 '22 at 20:31
  • 3
    @Rocketmagnet - Hi, Now that you know the cause of this problem (as you kindly explained in your [next question](https://electronics.stackexchange.com/q/625339)) please would you write a [self-answer](/help/self-answer) to *this* one and accept it (which you can do immediately, as it's more than 48 hrs after you asked the question). That will effectively close this question and make a record of the cause. Otherwise people might spend more time on this, when the cause has now been found. Thanks. – SamGibson Jun 28 '22 at 11:36

0 Answers0