Accurate ADC measurement without the Soft Device

RE: Accurate ADC measurement without the Soft Device

Eliot Stock — Wed, 21 May 2014 14:03:59 GMT

Understood, thanks.

RE: Accurate ADC measurement without the Soft Device

Håkon Alseth — Wed, 21 May 2014 12:46:51 GMT

normally you want your program flow to look like this: int main() { init(); while(1) { /* do something*/ } // return will never execute return 0; }

What you want to do in your while-loop is to execute the functions that you need to run in a loop, then go to sleep (WFI for instance). Since you're using the app_timer, most of your functionality will be executed via interrupts, so you can just put a while(1){__WFI();}

Prior to the while-loop you must have all the init-functions. Like creating and starting the app_timer instance and other init functions.

RE: Accurate ADC measurement without the Soft Device

Eliot Stock — Wed, 21 May 2014 11:09:49 GMT

Thanks. No, what's shown above is the entire main() function in main.c, there's none of the usual while loop stuff at the moment. Should there be? It seems that anything after:

NRF_TIMER0->TASKS_SHUTDOWN = 1;

won't execute anyway? Or will it?

RE: Accurate ADC measurement without the Soft Device

Håkon Alseth — Wed, 21 May 2014 11:01:16 GMT

Hi Eliot,

Using the RTC (app_timer) to create this 30 sec tick is the way to do it, especially when thinking of current consumption.

Returns from main aren't recommended in embedded applications. I do not know on which level you're calling this return, but if this is set in the main-while loop, then you'll return to an unknown state. I assume this is some sort of function call, since there's no sleep-call at the end of it?

RE: Accurate ADC measurement without the Soft Device

Eliot Stock — Tue, 20 May 2014 11:40:26 GMT

Hi Håkon,

I think what you've addressed above is helping reduce power consumption while reading the ADC. But that only takes a fraction of a second. What about reducing consumption in between samples? (I sample once every 30s for now.)

I've tried using NRF_TIMER0->TASKS_SHUTDOWN for this and it seems to work but I want to check my approach. My main() looks like this (note the return from main, which seems a bit weird).

	uint32_t err_code;
	app_timer_id_t timer_id;

	timers_init();
	battery_level_init();

	start_clock();

	err_code = app_timer_create(&timer_id, APP_TIMER_MODE_REPEATED, generator_test_timeout_handler);
	APP_ERROR_CHECK(err_code);

	err_code = app_timer_start(timer_id, APP_TIMER_TICKS(30000, 0), NULL);
	APP_ERROR_CHECK(err_code);

	// Shutdown and wait for timer to timeout.
	NRF_TIMER0->TASKS_SHUTDOWN = 1;
	return 0;

battery_level_init() sets up the ADC. start_clock() starts the LF clock explicitly, since we're not using the Soft Device. The timeout handler is simply:

void generator_test_timeout_handler(void * p_context)
{
    UNUSED_PARAMETER(p_context);
    battery_level_read();
}

in battery_level_read() I use your "safest approach" above.

Seems to work, even with the debugger, but I haven't yet tested the power consumption. Does all this seem OK to you?

RE: Accurate ADC measurement without the Soft Device

Eliot Stock — Thu, 08 May 2014 11:19:37 GMT

No problem, this looks like the lowest power consumption approach to my problem too. Thanks.

RE: Accurate ADC measurement without the Soft Device

Håkon Alseth — Thu, 08 May 2014 11:11:44 GMT

Eliot, I must apologize for going off-topic. This is a continuation of the WFE-talk in the comments area.

As events and interrupts are closely coupled in the MCU core, you'll need to enable this in the INTENSET register prior to using SEV/WFE.

you enable SEV instruction in SCB->SRC.
You enable the event/interrupt for HFCLK-started (INTENSET)
Start the task enabled in 2)
The first WFE will cause the MCU to go to sleep and wake up on EVENTS_HFCLKSTARTED.
SEV will set the Event Register in the core register.
Second WFE will see that the event register is set, and clear this. No sleep will be executed at this call.
EVENTS_HFCLKSTARTED and IRQn is cleared.

Safest approach: I had a chat with our nrf_soc developer, and was told a better implementation. The above usage of WFE/SEV/WFE can cause you to wake up immediately if the internal event register is already set.

To be certain that you wake up on the correct event, you should do like this:


NRF_CLOCK->INTENSET = CLOCK_INTENSET_HFCLKSTARTED_Msk;
// Enable wake-up on event
SCB->SCR |= SCB_SCR_SEVONPEND_Msk; 
NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
NRF_CLOCK->TASKS_HFCLKSTART = 1; 
while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0) 
{
  __WFE();
}
NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
NVIC_ClearPendingIRQ(POWER_CLOCK_IRQn);

In general, I recommend reading up on WFI/WFE/SEV instructions: infocenter.arm.com/.../index.jsp infocenter.arm.com/.../index.jsp

BR Håkon

RE: Accurate ADC measurement without the Soft Device

FormerMember — Thu, 08 May 2014 09:45:56 GMT

Håkon, can you please explain the code snippet that you've posted. What is the advantage of using the method that you have posted to enabling the interrupt for POWER_CLOCK and using WFI? And why SEV between WFE?

RE: Accurate ADC measurement without the Soft Device

Håkon Alseth — Wed, 07 May 2014 14:51:46 GMT

You can save a bit of power by using SEV/WFE instead of polling (been meaning to learn how this WFE works): NRF_CLOCK->INTENSET = CLOCK_INTENSET_HFCLKSTARTED_Msk; // Enable wake-up on event SCB->SCR |= SCB_SCR_SEVONPEND_Msk;
//NRF_CLOCK->EVENTS_HFCLKSTARTED = 0; NRF_CLOCK->TASKS_HFCLKSTART = 1;
__WFE(); __SEV();
__WFE(); NVIC_ClearPendingIRQ(POWER_CLOCK_IRQn); NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;

Hope this helps on power consumption.

RE: Accurate ADC measurement without the Soft Device

Eliot Stock — Wed, 07 May 2014 14:21:13 GMT

Looks good, thanks very much.

RE: Accurate ADC measurement without the Soft Device

Håkon Alseth — Wed, 07 May 2014 13:49:27 GMT

Hi Eliot,

You'll have to manually start the 16M XTAL in a sequence like this:

    
if (!(NRF_CLOCK->HFCLKSTAT & (BIT_0 | BIT_16)))
{
    NRF_CLOCK->TASKS_HFCLKSTART = 1;
    while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0);
    NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
}
    
/*
Do something, then stop the HFCLK.
*/
NRF_CLOCK->TASKS_HFCLKSTOP = 1;

Best regards Håkon