RTC Time drift on NRF9151 chip set.

[Junyoung Jeon]

Hello,

My name is Jeon. I am using the RTC function on a custom board that utilizes the same chipset as the nRF9151. I have configured the RTC to trigger an alarm every 20 minutes in order to send MQTT messages to my database. Under normal conditions, the system functions reliably.

However, after approximately 12 RTC alarms, a time drift of about 3 minutes occurs, as shown in the image below. Following this drift, the interval returns to the expected 20 minutes. I have searched through various discussions on this platform, and some users have reported that similar issues may occur when using multiple threads.

I also attached my code below. If you know how to fix or need more information, just let me know.

DRI_Sensor (3).zip

[Susheel Nuguru]

Hi,

I went through your attached project for second pair of eyes and I am confused the way you are using different clock features.

When you print timestamps using the library functions like date_time_now() or time(), you are actually showing wall-clock time, not RTC time. Wall clock is different than the RTC counter time. So if the modem receives a more accurate time fix (for example, after being offline or in PSM), your log shows a sudden shift. When logging, it is best to print both values (RTC ticks or k_uptime_get() and date_time_now()) to clearly see whether this is the confusion where the alarms are drifting or not. Even after printing these two separately and if you still see drift in both date time and RTC time, then the source is that RTC is drifting. So it is the clock that is sourced to RTC that should be drifting then.

Are you sure that you have configured RTC to use XTAL and not the internal RC? Assuming that you have XTAL LFCLK on your board. Else you probably can use SYNTH derived from HFCLK to source it. Check your hardware to see if it has XTAL and check the device tree node &lfclk to see if it is using the XTAL if you have it on the hardware. An example of how these configurations are set can be seen in this devzone thread.