Nucleo board health test ?

[acapola]

I have 19 nucleo-u5a5zj-q, my program successfully run on 18 of then but fails for 1 (I did not modify the board nor connect anything else than the STLINK USB cable). Failure here means it detects a hardfault.

Is there anything that I can use to verify that the failing board is fully functional ? (or at least the MCU on the board).

Notes:

• In this ticket I do NOT ask about fixing my program (I created another ticket for that, see https://community.st.com/t5/stm32-mcus-products/what-is-the-cause-of-this-hardfault/td-p/702478 ).
• Another board exhibit a similar failure but with another program. 

[Pavel A.]

Without more details on what your program does and what exactly fails - can only give the following advice: get 6+ more factory-new boards and test on them. If no new board fails, assume these 2 boards are defective. Else you'll need to get into debugging and details.

[acapola]

Hi Pavel, I do not think that testing on more boards is the way to decide if the board is defective or not. I have seen incorrect software working on several hundreds of devices and fail on few. The few failing parts were not defective, they just happened to be "just within the specs" while others could tolerate the incorrect setting because they had more margin.

Anyway, I had to get 7 more boards, so I did the same test. I updated the original description has a result. It turns out that actually a single board was failing with this exact program, the second one failed with an earlier version but does not fail with that one.

[Pavel A.]

Testing on more boards by itself of course is not the way to decide if a board is defective. But this can help to understand the margins of the boards to be considered good and decide what to do: adjust the specs or invest in the software.