STM32G0B0RET6 failure rate

[Zdiode]

We are on the second generation of prototype here.

On the first generation I had an MCU fail for reasons I couldn't trace.  This was from a set of six.

I have now powered seven of the second generation two have had MCU failures.  The second failure was on the prototype's initial turn on.  No code had been loaded into it.  A hole blew out of the part when power was applied.

All failed parts then shunted the VCC to ground with a low impedance, less than 100 ohms.

I have included part of our schematic.  Do you see anything that could explain this failure?

Best regards,

Bill Grimm - Zdiode

[TDK]

> The second failure was on the prototype's initial turn on.  No code had been loaded into it.  A hole blew out of the part when power was applied.

Definitely suggests overvoltage on a pin.

Schematic looks okay.

I would guess one of the signals coming in exceeds the allowable voltage, possibly due to a spike on startup.

If parts are still responsive, you may be able to determine which pin or port is blown by set each pin high (set pin as output and set GPIO->ODR) and see which one still reads low (read GPIO->IDR).

Probably this spike is happening on the other parts too but it just hasn't caused failure (yet). Putting a scope on each incoming pin and monitoring during startup may be informative.

1/6 and 2/7 are way above the expected failure rate of the chip.

[Andrew Neil]

---

@TDK wrote:

Definitely suggests overvoltage on a pin.

---

+1

---

@TDK wrote:

I would guess one of the signals coming in exceeds the allowable voltage, possibly due to a spike on startup..

---

+1

The schematic shows that there is (at least) 24V elsewhere in the system; also lots of external connections with "motor" in the name...

And no sign of any protection or suppression on any of the external connections.

PS:

Another possible cause is applying signals to the MCU's pins when the MCU is not (properly) powered and/or grounded ...

[Uwe Bonnes]

Can you explain what you intend with the combination of Q1/D1 and Q2/D2?

Otherwise your error for me also looks like overvoltage.

[Zdiode]

Q1/D1 and Q2/D2 are reverse voltage protection.  I have wondered if one or the other is sufficient.

[Zdiode]

If I can do it I will do it.  At least this is a good technique to remember for the future.  Thank you!

None of the damaged parts so far have been in any way usable.

[Zdiode]

On the first one that failed in this set, I removed the part after it fried.  I will test the prototype with its pins open and see if I see any overvoltage.  Perhaps with the processor removed, I  will see the high voltage.

I am not in the office today due to a big snow storm.

Best regards,

Bill

[Uwe Bonnes]

Either Q or D is enough. And D is the more common and probably cheaper solution.

[Zdiode]

I agree.  Q1 and Q2 need to go.

I just took the prototype that has the MCU removed and probed every pad on the processor footprint.  I got no voltage higher than 3.298 volts.  I probed every pin again and it indeed verified that no voltage at the MCU is higher than 3.298 volts.

This verifies my reading of the schematic.

I suspected workmanship problems.  I have three prototypes here that have not seen power.  I ohmed out the 3.3 volt line and they all checked the same.  However they did read different from a known working prototype.

The second one I pulled out verified my workmanship concerns.  The photos are attached.  C11 had come off entirely.  C10 had broken loose from its positive terminal and pulled up the trace on the negative terminal.  It seems like cold solder joints were to blame for this.

Unfortunately I can't trace this gross problem to the failed MCUs.  C10 and C11 are fine on the prototypes that failed.  I have examined the prototypes for solder splashes, but not found any.

Many thank yous for the suggestions.  I especially like the idea of reading the bits to find the one where the voltage came in.  Unfortunately my MCUs are all latched down.

Any other ideas are welcome.

[Andrew Neil]

---

@Zdiode wrote:

I probed every pin again and it indeed verified that no voltage at the MCU is higher than 3.298 volts..

---

But that's just steady-state.

You said failures occur at startup - so you need to be looking for transients ...