STM32L4A6VG 3.3V breakdown

[SL61]

Hi everyone,

I designed a custom board and now I found two MCUs (STM32L4A6VG) are breakdown. The MCU is running at 3.3V and all VDD, VDDA, VREF+, VDDUSB and VBAT is connected to the same 3.3V source. Here's what I found.

For the initial trial, I did 2 boards and one became very hot at MCU after running for several days. The resistance between 3.3V and GND is around 10 ohms. The MCU can still run even it is very hot.

For the second trial, I change the TVS at 12V input and the clamp voltage is 14V. This time I did 7 boards and one failed to run just after powering on. The resistance between 3.3V and GND is around 80k ohms. I removed the LDO and the resistance dropped to around 30k ohms. I applied 3.3V using an independent power source and the current was over 600mA (voltage is 1.8V so there's not enough power. Using higher current may destroy the MCU completely). Now the resistance between 3.3V and GND is around 400 ohms.

Any suggestion is welcomed.

Best,

SL61

[Danish1]

The buck converter will generate spikes of its own. What power-supply-decoupling do you have very close (physically and electrically) to the stm32? Quantity, value and dielectric?

You show the TVS clamp, buck converter and stm32 micro all sharing the same ground. How have you ensured that spikes that should have been clamped by your TVS don't cause voltage-drops along this ground and thereby reach the stm32?

It is possible for spikes to come in on other pins that will trigger "destructive latchup" - essentially turn on the thyristor that is present between Vdd and Vss on all CMOS devices. Modern CMOS is much better than what was around in the 1970s but it can still happen.

Does your board connect to anything inductive (motors, relays, loudspeakers)?

Do you have any wires coming from off the board, that might pick up electrostatic discharge?

Hope this helps,

Danish

[SL61]

Thanks for your reply. ​The buck converter (TPS565208) and the MCU are on the same board. Power pins (VDD, VDDA, VREF+, VDDUSB) are using 1uF/10uF + 100nF caps for decoupling.

For TVS, could you explain more on that? My understanding is that voltage higher than 14V will be clamped in nano seconds but voltage less than 14V can still pass through the buck converter and the LDO to the MCU and damage the MCU?

For the load, we don't use any inductive load and wires coming off the board are protected by TVS diodes except SWD pins and debugging UART pins. Perhaps I should add a series 33-ohm resistor on UART lines?

For the buck converter, I didn't see voltage spikes during startup and stable operation. When will it generate spikes and what is the cause?

Thanks for your time.

Best,

SL61

[emil28]

Hello,

Simple question: did you checked the good pinout in you schematic ? May be two pins are inverted.

600 mA on your uC is not normal, it looks like there is a short-circuit somewhere.

Hope this help.

[SL61]

Hello everyone,

Any suggestion about this problem?

Best,

SL61​

[emil28]

Hello,

It's hard to find the solution with the information we have. The only suggestion I have is a short-circuit on one or several GPIO. Did you check these configurations ?

Regards.

[SL61]

I agree that it is hard to find the solution, just want to know more about STM32. Configuration should be fine since other boards with the same configuration run well. Besides sudden ESD strike when debugging, like touch the MCU directly, there is a chance during producing the board (these boards are prototypes thus components are placed by hand). I know using X-ray may help to diagnose the problem but not practical at this moment. Perhaps I should protect those debug pins as well.

Best,

SL61

[TDK]

> Configuration should be fine since other boards with the same configuration run well.

This is faulty logic and the source of much frustration if you follow it. Things can, in fact, run just fine even though there are errors. Something running fine doesn't imply there are no issues.

It is unlikely that an ST chip, which is running fine off of a 3V3 power supply, would suddenly and spontaneously produce a short circuit. I expect there is something else you either do not know or are not telling us. Perhaps an input to the chip is out of tolerance.

Using an LDO after your buck-converter provides a reasonable amount of safety. You could put a large Schottky diode from GND -> 3V3 to protect against reverse polarity spikes, but I doubt this is the issue.

[SL61]

Thanks for your reply. Actually agree with you and seems I need do more tests and find out ​the reason. I cannot upload the full schematic since it is still under development so that makes diagnostic harder.