STM32L4 - HSI16 accuracy / over time stability

[aritec]

On our project 4 (out of 1200) devices startet to have problems after a few weeks of working. It turned out they had a HSI16 of 15MHz @ the default calibration settings. A recalibrating brought it back to 16MHz, but questions arises:

1. How can the HSI16 come out of frequency in a very stable environment ?
2. Do we need to check and recalibrate the HSI16 (MSI) in a timed base manner?

MCU: STM32L496QGI6P

Thanks for any comments

Armin

[Tesla DeLorean]

Honestly something you should perhaps discuss with your FAE. Not sure the general population has done this kind of long term testing or evaluation.

Temperature and voltage could be most direct causes.

Long term might look for other components absorbing moisture from the environment, or drying in low humidity ones.

Perhap board level stresses.

I don't think there are any aging factors like there would be for crystals.

Might consider replacing the MCU and see if the issue follows the board or the IC

[waclawek.jan]

> HSI16 of 15MHz @ the default calibration settings

Sounds like the possible calibration extreme.

Wasn't the calibration register overwritten by 0 (possibly due to software bug)?

JW

[aritec]

It was reproducable with every restart.

Here you see the relation HSI16 frequency to calibration value (0..127).

The measurement was done with the LSE 32768Hz 100ppm mean of 10x LSE trigger divided by 8.

  

[waclawek.jan]

Isn't this that problem where Cube overwrites the default trim value with an incorrect one?

https://community.st.com/s/question/0D53W00000XFdiNSAT/what-is-the-correct-default-value-for-rcchsicalibrationdefault-for-stm32l431it-is-going-to-be-corrected-in-the-hal-libraries

JW

[aritec]

Thank's for the reply, but this has been fixed a while ago.