New STM32L433 pins pull high??? Damaging opamps

Question

I have an STM32L433, fresh from RS components. I soldered it to my board. PA6 is connected to the output of an opamp - INA181A1.

There are another 2 INA opamps on the board, connected to other pins.

I have now had 2 INA181s die (actually I am not sure they died, since I unsoldered the first and replaced it assunming it was broken... the second I unsoldered just the output leg and may have damaged it mechanically...). The board is not even programmed yet. There is actually not yet even a programming header...

The net between the opamp and the MCU is pulled high. If I hold the MCU in reset state (connect the debug header nRES to ground) the opamp-MCU net floats. As soon as I release the reset pin, it pulls high.

The other 2 channels to not do this! The other 2 channels behave exactly as expected!

There is no board error. There is >500kOhms between the net and ground/3v3 when not powered - this pull IS coming through the MCU. I observe there is ~50mA extra drawn from my power supply between when the MCU is held in reset and when reset is released.

Are my new chips pre programmed???am I missing some line in the datasheet/reference that says "PA6 will pull high when new"?

The boards were supplied without MCU and opamps and gate drivers. I have populated them myself. I have built 2 boards with F303CB chips, which work fine, using internal opamps, and 1 with L433CC. The pinout appears to be the same in terms of power, boot0 etc.

The schematics can be seen in https://github.com/davidmolony/MESC_FOC_ESC/tree/Rearrange_Vin - this is an open source EBike controller I work on in the evenings. I have been riding around with the F303CB version, pushing ~80phase amps and getting up to ~40km/h, so the board DOES work. Edit: Schematic Offending opamp is op2, U9

Answer 1

If the MCU is unprogrammed, it will automatically go to factory bootloader.

Bootloader uses PA6 as SPI1_MISO, which will be push-pull output.

It might go high-impedance if PA4 used as SPI1_NSS is pulled high, so that is something you might try, but not much else.

Answer 2

An aside: I made that sort of a mistake at one point and the lesson is that you must avoid connecting an opamp output to anything without a series resistance. That resistance may be already present for functional reasons and then you wouldn’t add it, of course. And sometimes the function of the circuit may preclude it. But generally it’s a bad idea to just hard-connect an opamp output to any input pin. It gets very hard to debug opamp circuits without such resistors, and sometimes the lack of resistance may cause latch-up-like behavior - not an actual SCR latch-up, but the circuit may start in a “nonsensical” operating point and stay there because with no series resistors large currents may flow from the output, that would never be present in normal operation, and an undesirable steady state is entered which will persist until power-down.

I had this problem when somehow the baord vendor put a bunch of dodgy chips on the board - they didn’t look like fakes, but more like packaged dies that passed die test and “died” in packaging. Those were double op-amps where only one op-amp would work, but with the circuits being connected hard, you couldn’t tell which ones were the source of the problem and which ones were misbehaving just because of incorrect inputs.

Many op-amps will misbehave when input currents are too high, eg. because there’s a diode clamp between the inputs. MCUs can also misbehave then, especially if you manage to overheat the pin driver/mux and eg. its performance degrades. Say that the opamp output goes to an ADC input. I have experienced several mainstream MCUs where a short circuit current from an op-amp (or any other source of similar strength) will degrade the ADC mux at high ambient temperature and the ADC performance will be poor or even none (stuck at 0 or FS). This is usually precipitated by a pin conflict just like you experienced.