Strange behaviour when writing to RTC calibration register

[jvog.sen]

Hi everyone,

I am trying to trim the internal RTC of an STM32MP157 to run slightly faster. According to the reference manual, you need to set bits CALP and CALM[8:0] in RTC_CALR to do so.

Here is what I did (using devmem2:(

# Enable write access to RTC_CALR
./devmem2 0x5C004024 h 0xCA
./devmem2 0x5C004024 h 0x53
# Write calibration factor (CALP=1, CALM=327) to RTC_CALR
./devmem2 0x5C004028 h 0x8147

Reading back the contents of register RTC_CALR returns 0x0147 instead of 0x8147, which corresponds to CALP=0.

Measuring the RTC deviation confirms that CALP does not get set to 1.

I am aware that using hwclock with adjtime can easily work around this issue. Yet not knowing what is going wrong here keeps bugging me...

[PatrickF · Accepted solution]

Hi,

this behavior is explained because our SW does not use default values and sets PREDIV_A = 0 and PREDIV_S = 32767 (to allow better accuracy on Linux time restore on low power modes exit as subsecond granularity is linked to PREDIV_S).

See relevant RTC section to calibrate in that case:

Regards,

In order to give better visibility on the answered topics, please click on 'Select as Best' on the reply which solved your issue or answered your question. See also 'Best Answers'

[jvog.sen]

Thanks, now it makes sense!