Title: nRF54L15: Stopping WDT (WDT31) in PM state entry does not decrease sleep current (~2 µA)
Device: nRF54L15
NCS / Zephyr Version: ncs 3.4.0 and zephyr 4.4.0
Description:
Hello,
I am trying to minimize System ON sleep current on the nRF54L15 by stopping the watchdog (WDT31) prior to entering sleep, and re-starting it upon waking up.
According to the electrical specification, active WDT consumes approximately 2.2 µA (LFXO) / 2.8 µA. However, when successfully executing the stop sequence in Zephyr's PM notifier (pm_state_entry), I do not observe any measurable reduction in sleep current.
My Stop Procedure:
- Configured
WDT31->CONFIGwithSTOPEN = 1during initialization. - Inside
pm_state_entry:- Set
WDT31->TSEN = 0x6E524635 - Triggered
WDT31->TASKS_STOP = 1 - Polled
WDT31->RUNSTATUSuntilRUNSTATUSWDT == 0(confirmed it completes). - Cleared
WDT31->TSEN = 0.
- Set
- In
pm_state_exit: Re-configured and re-started the watchdog.
Questions:
- Does executing
TASKS_STOPon the WDT peripheral actually power down its internal logic to save the ~2–2.8 µA listed underIWDT0/IWDT1, or does the base consumption remain because the Low Power domain / LFCLK (PCLK32KI) is still requested by the Zephyr kernel/system timer? - Is the ~2.2 µA – 2.8 µA figure in the datasheet primarily the overhead of the LF clock source (LFRC/LFXO) rather than the digital logic of the WDT itself?
- What is the expected net current difference between an actively running WDT vs. a stopped WDT when the device is in System ON sleep and LFCLK is already kept active by other peripherals?