RTC hangs after a while, Watchdog does not reset

[kammutierspule]

I'm using RTC to wakeup every 500ms,
I have Watchdog enabled but after a while (about half.. or an hour...) it will not wakeup anymore and watchdog will not reset.

I measured the current on that hang state and it consumes about 2.90uA (this matches the expected 1.5uA RTC + 1.5uA Watchdog consumes )

I was expecting, whatever happen, that Watchdog will reset the MCU.

Some more details:

- I'm not using any SoftApp
- I'm using internal LFRC
- It is feeding the Watchdog before going to sleep
- I tested with an hardware fault, the Watchdog works

Is there something that could make the Watchdog not work ? and RTC stop to work?

I found some similar topics:

https://devzone.nordicsemi.com/f/nordic-q-a/43961/nrf52832-in-sleep-after-a-period-of-operation-the-watchdog-of-the-equipment-did-not-work/172299

https://devzone.nordicsemi.com/f/nordic-q-a/14321/wdt-halts-rtc-after-nvic_systemreset/54658

https://devzone.nordicsemi.com/f/nordic-q-a/64765/rtc-fails-to-wake-chip-from-systemon-sleep-nrf52840

https://devzone.nordicsemi.com/f/nordic-q-a/68666/app-timer-stops-working---need-a-power-cycle-to-fix-issue/281629

https://devzone.nordicsemi.com/f/nordic-q-a/51559/nrf52-wdt-lfclk-sourcing

Some more details:

Every time I go to sleep I set the LFCLK: nrf_clock_task_trigger( NRF_CLOCK_TASK_LFCLKSTART );
I init the RTC before go to sleep and uninit it after return.
I'm using nrf_pwr_mgmt_run(); to enter in sleep.

[run_ar]

Hi,

Once set up and started, the watchdog will run until it is reset. There are two things I can think about that might cause this:

1. is if you configure the Watchdog to halt while in sleep. See Temporarily pausing the watchdog

2. is if you are using system off: WDT: WDT configuration is cleared when entering system OFF, but is seems to me you are not using system off?

[kammutierspule]

Hi,
I believe I tested 1. and it works ok and 2. is not the case, as it is waking from RTC.

I found that on my configuration sdk defines, it was wrongly setting the LFCLKSRC to use external XTAL and I'm using internal RC.

But why does it work with the wrong settings on the first place?

I tested the following (pseudo) code, with the wrong setting (LFCLKSRC configured to use XTAL but not populated) it will crash in about 5 minutes:

int main( void )
{
	HAL_OutInit( 1 );

	nrfx_wdt_config_t config = NRFX_WDT_DEAFULT_CONFIG;

	nrfx_wdt_init( &config, NULL );
	nrfx_wdt_channel_alloc( &s_wtd_channel_id );
	nrfx_wdt_enable();

	NRF_CLOCK->LFCLKSRC = 1; // XTAL, is not populated, will crash in about 5m
	//NRF_CLOCK->LFCLKSRC = 0; // RC, works

	HAL_OutSet( 1, true );

	HAL_mcu_sleep(500);

	while( 1 )
	{
		HAL_OutSet( 1, false );
		HAL_mcu_sleep(25);
		HAL_OutSet( 1, true );
		nrfx_wdt_channel_feed( s_wtd_channel_id );
		HAL_mcu_sleep(15);
	}
}

Does this different settings explains the behaviour or could be something else?

I was expecting that with the wrong setting it shouldn't work anytime.

[run_ar]

Hi,

From WDT documentation: When started, the watchdog will automatically force the 32.768 kHz RC oscillator on as long as no other 32.768 kHz clock source is running and generating the 32.768 kHz system clock

From Clock documentation: If the LFXO is selected as the clock source, the LFCLK will initially start running from the 32.768 kHz LFRC while the LFXO is starting up and automatically switch to using the LFXO once this oscillator is running. The LFCLKSTARTED event will be generated when the LFXO has been started.

Unfortunately I am not sure why it crashes after 5 minutes. Do you have a reset reason or similar?

[kammutierspule]

run_ar said:

while the LFXO is starting up and automatically switch to using the LFXO once this oscillator is running. The LFCLKSTARTED event will be generated when the LFXO has been started.

So since I start the LFCLK to user in the RTC (to wake up) but maybe I never check for the LFCLKSTARTED? it should be running from LFRC?

But so, at some moment  it will "automatically switch" to the LFXO (not present) and will never recover from that?

[run_ar]

I think that could be the issue, yes. But I have not verified it. Normally you should have a loop to wait for LFCLKSTARTED. i.e. nrfx_clock_lfclk_is_running

[kammutierspule]

ok I was just wondering if there was a known (defined?) behavior of this condition