STM32H7 failing (dying) if there is oscillation on the power supply.

[JMill.1]

We are using an STM32H743ZI in our application and just got the first prototypes. I was using a current limited power supply for initial test to protect the circuitry, however, something has caused two micros to fail so far after only a few power on cycles. I have subsequently discovered that limiting the power supply current was actually causing the 3V3 regulator to continually drop out and restart. See below for a plot - yellow is the 3V3 rail.

However, I'm not sure how this could kill the processor (both units have failed short to GND and are unrecoverable). There is nothing in the Errata to suggest sensitivity. Has anyone else experienced this phenomenon? What can be done to protect the device if the rail suddenly drops out? I would expect the Brown out detect to protect the device?

[Uwe Bonnes]

Blue voltage goes far beyond 4.0 Volt, the absolute maximum for 3.3 Volt capable supply pins. If that blue voltage is connected direct to the STM32, permanent damage will happen!

[JMill.1]

Hi Uwe - the blue voltage is the input to the 3V3 regulator. The IC never sees this voltage, only the yellow trace is relevant.

[Uwe Bonnes]

From my obeservations, a short between power rails inside the chip only happened when I had overvoltage on the part. Carefully check for that condition again.

[Danish1]

You don't disclose other bits of your prototype circuit.

What could be happening is that while Vdd is good, other parts of your circuit are charging up to reasonable voltages (<= Vdd).

Then Vdd collapses rapidly.

The other parts of the circuit might stay up - due to capacitance (intended or otherwise) and a significant discharge path might be through the stm32 itself. This can cause destructive-latchup.

I see from an stm32h7 data-sheet "Power Supply Management" that while the main Vdd <= 1.0 V, all other Vdd lines (VddA, VddxxUSB) must remain below Vdd + 0.3 V. They say "During the power-down phase, VDD can temporarily become lower than other supplies only if the energy provided to the microcontroller remains below 1 mJ. This allows external decoupling capacitors to be discharged with different time constants during the power-down transient phase."

Hope this helps,

Danish

[JMill.1]

Thanks Danish. This is an interesting possibility. Power supplies for the chip itself are all on the same rail (see below). However, there are other circuits on 3V3 that may be staying alive longer - I'll look into that. What do you think would happen if those ICs were trying to drive signals into the STM as the rail is collapsing? It should be noted, perhaps, that on the second board that failed I didn't ever program the chip, so all pins would have been in the default input state.

Thanks,

Jamie

[TDK]

If other ICs are driving a pin past VDD + 0.3V, it can cause issues including killing the chip. Depends on how much current you're injecting.

I don't think the VCAP pins should be connected to each other. They aren't connected on ST board. Doubt that's the underlying problem.

NRST has an internal pullup. External pullup doesn't hurt, just FYI in case you want to save a BOM component.

[Tesla DeLorean]

Agree on the VCAP

Really need to walk the netlist, make sure you don't have any disconnected islands or combined nets.

Do a complete pin review.

Remove the broken part, and check voltages and pins against part diagrams.

Work with your local FAE, and have them review both schematic and layout.

[JMill.1]

Thanks all, I will contact an FAE for review. However, on the VCAP question - I'm looking at the datasheet and Table 25 has a note that states "1. When bypassing the voltage regulator, the two 2.2µF VCAP capacitors are not required and should be replaced by two 100nF decoupling capacitors." Does anyone know what is meant by this? They state a nominal operating condition at 2.2uF per pin in the table, then recommend an order of magnitude smaller...?

[TDK]

If you want to supply your own ~1.2V supply for VCAP for whatever reason, then you can do so if you disable the voltage regulator. In this case, the bypass caps don't need to be as large since your supply should be stable on its own.