Variations in the Main Cycle Using k_timer

[Fumiya]

Hello.

I used `k_timer` to create the main cycle (5 ms), but there is a deviation of ±1 ms.

I’ve determined that the cause is the UART interrupt callback, but I don’t know how to minimize this deviation.
Note that the interrupt handling took approximately 25 μs.

Based on the above, I have two questions:
① Is a deviation of ±0.5 ms unavoidable in the Zephyr OS thread-based implementation?
② What is the recommended approach to minimize this variation on the nRF54L05?

My development environment is as follows:
　SoC used: nRF54L05
　SDK version: nRF Connect SDK v3.1.1
　Zephyr OS version: v4.1.99
　Development tool: nRF Connect for VS Code

[Vidar Berg]

Hello,

Could you provide a minimal sample project that reproduces this issue, so I can test it here? I don't see how the UART callback could explain a 1 ms deviation when according to your measurements it only blocks for about 25 us. Could you also say a bit more about what you’re trying to achieve with this periodic timer interrupt? I may be able to suggest another approach.

Best regards,

Vidar

[Fumiya]

Dear Vidar,

Good evening.

This issue has been successfully resolved.
The cause of the main cycle offset was that a while loop was executing a 1ms wait within the k_timer handler (ISR).

Therefore, we resolved the issue by using a semaphore within the k_timer handler to simply send notifications,
and by moving all main processing to the main() function.

I apologize for the inconvenience.

Best regards,
Fumiya

[Vidar Berg]

I am glad to hear that you were able to identify the issue, thank you for the update. I will mark this as resolved.

Best regards,

Vidar