STM32F105RBT6 Processor overheating

[klevente]

Dear Support Team,

I have problem with STM32F105RBT6 overheating,

After several months of operation, the processor will break down, overheat and burn out completely.

Power consumption is constantly rising, but it works flawlessly even with higher power consumption, and then reaches a temperature where the processor shuts down and burns out completely.

In general, 2-3 of 100 pieces of electronics have this phenomenon, this product has been in production for 5 years, the defects have appeared in the last two years.

The fault cannot be eliminated during the testing process, it usually occurs after 6-8 months of continuous operation.

Can such a fault be caused if one of the GND or VCC pins is not soldered properly?

Where can I try to find the error?

Attached to the wiring diagram, the electronics are powered by a MeanWell 230VAC / 24VDC power supply

Best regards,

Levente

[Danish1]

In my experience, stm32 are pretty reliable and don't burn much power under normal use. On a failed unit, can you identify which component(s) have got hot?

What do you mean by "The fault cannot be eliminated during the testing process"?

Do you mean that even during an extended soak test, you don't see the power-consumption rising on boards that will (eventually) fail?

But it's possible to abuse them in various ways, which might lead to excessive power consumption and eventual failure.

My first suspicion would be your power-supply and/or other voltages arriving at the stm32. Consider the LM1117 linear regulator that takes the +5V down to +3.3V. If the GND on that were to disconnect, then the full +5V might be presented to the stm32 - well outside absolute-maximum ratings and likely to lead to burn-out.

Also, if any of the signal lines were to rise to +5V, unless they are on 5-Volt-tolerant pins, then the +5V will get in that way. I don't know stm32f105, but many stm32 5-Volt-tolerant pins are only 5-volt-tolerant when Vdd is powered; can that ever happen?

Do you drive high currents on any of the stm32 pins (say above 4 mA)? That will increase heat-generation but I/O is pretty robust.

Do you switch any inductive loads e.g. relays? That can be disastrous unless you contain the back-emf of the coil during turn-off.

This is all pretty vague, but hope it helps,

Danish

[klevente]

Dear Danish,

Thanks for the comments,

Here are the answers:

On a failed unit, can you identify which component(s) have got hot?

Yes, the R15 resistor in front of the LM1117, which is 2 Ohm, breaks from the high current that flows through it, when the electronics fail then the current consumption of the processor goes through the R15 (2 Ohm) resistor.

Traces of overheating around the LM1117 are clearly visible on the printed circuit, but no other part in the power supply section is destroyed other than the mentioned resistance…

What do you mean by "The fault cannot be eliminated during the testing process"?

Do you mean that even during an extended soak test, you don't see the power-consumption rising on boards that will (eventually) fail?

The electronics are tested for a total of 2 hours, during which no fault occurs.

Also, if any of the signal lines were to rise to +5V, unless they are on 5-Volt-tolerant pins, then the +5V will get in that way. I don't know stm32f105, but many stm32 5-Volt-tolerant pins are only 5-volt-tolerant when Vdd is powered; can that ever happen?

There is no + 5V signal on the STM32, there is a max + 3.3V signal on each pin.

Do you drive high currents on any of the stm32 pins (say above 4 mA)? That will increase heat-generation but I/O is pretty robust.  

The maximum load is 3.7mA / pin.

Do you switch any inductive loads e.g. relays? That can be disastrous unless you contain the back-emf of the coil during turn-off.

NO.

I have two questions,

1. Can such a fault be caused if one of the GND or VCC pins of STM32 is not soldered properly?

1. It is possible that an LM2576 DC/DC IC sends a voltage much greater than +5 V to the LM1117, which it cannot regulate to + 3.3 V, and for a moment a higher voltage than allowed is applied to the STM32, which destroys that?

I'm waiting for other ideas.

Best regards,

Levente

[Danish1]

I would not expect open-circuit on any of the Gnd or Vdd lines to cause high current latchup/failure.

However, it might cause misbehaviour of the stm32, because it would not have all the decoupling-capacitors; maybe also drawing high currents from the GPIO pins could cause entire ports to fail through internal power-rails on the cpu chip failing. (There is a process called electromigration, where excessive currents eventually cause tracks to fail).

(The fact that you ask this suggests that you have cause to suspect this. Maybe your pcb layout guy took some "liberties" and did not wire up all the Vdd and/or Gnd pins. I do not think that fixing the layout would be enough to solve your problem).

If you get a voltage higher than 4.0 on Vdd even for a short time, (the absolute max rating), there can be permanent damage to the stm32 which might include destructive latchup causing a short-circuit between Vdd and Vss. The absolute maximum voltage on LD1117 is 15 V (20 V on LM1117), so momentary failure of your switching-regulator could cause LD1117 to misbehave.

[klevente]

Thanks for the quick reply.

I will redesign the power supply with more over voltage protection...

Best regards,

Levente