Interrupt energy intensive

[boris bergman1]

When I program an interrupt, the NRF52840 requires 250uA extra. I read something about the sense interrupt, but I cannot find any working example. How should I implement an interrupt when my power is on a budget?

static const struct gpio_dt_spec signal = GPIO_DT_SPEC_GET(ZEPHYR_USER_NODE, signal_gpios);
static struct gpio_callback _interrupt;

void _isr(const struct device *dev, struct gpio_callback *cb, uint32_t pins) {  
  LOG_INF("isr given");
  k_work_schedule(&_proces, K_NO_WAIT); 
}


void attachinterrupt(void) {

	__ASSERT(device_is_ready(signal.port), "custom device not ready");
	if(gpio_pin_configure_dt(&signal, GPIO_INPUT)) {
    LOG_ERR("error configure signalgpio");
  }

	if(gpio_pin_interrupt_configure_dt(&signal, GPIO_INT_EDGE_TO_ACTIVE)) {
    LOG_ERR("error configure interrupt signalgpio");
  };
	gpio_init_callback(&_interrupt, _isr, BIT(signal.pin));
	if(gpio_add_callback(signal.port, &_interrupt)) {
    LOG_ERR("callback for interrupt not set");
 };
}

I tried to replace gpio_pin_interrupt_configure_dt with a sense configuration, but I don't really get how this is suppose to work:

#include <hal/nrf_gpio.h>

nrf_gpio_cfg_input(signal.pin, NRF_GPIO_PIN_PULLDOWN);
nrf_gpio_cfg_sense_set(signal.pin, NRF_GPIO_PIN_SENSE_HIGH);

Also, I read a pin interrupt consumes more than a port interrupt. How would I implement a port interrupt (my system needs only 1 interrupt)?

[Vidar Berg]

You may also consider enabling level triggering instead of edge triggering by replacing the GPIO_INT_EDGE_TO_ACTIVE flag with GPIO_INT_LEVEL_ACTIVE. Edge triggering uses GPIOTE IN instead of GPIOTE PORT events, which increases the idle current to ~17 uA. PORT events are suitable for detecting lower frequency inputs.

[boris bergman1]

Thanks, but that'll only work when I configure my hardware to have a latched interrupt. This will cause an infinite loop, which I don't know how to resolve. I can immidiately disable the interrupt from the ISR routine. This won't ensure tough things are not executed twice. What triggers also belong to GPIOTE PORT event?

It does save ~12uA, thanks!

[Vidar Berg]

Are you saying that you need to clear the PORT event within your interrupt handler to prevent the interrupt from firing repeatedly? I'm also not sure I understand which hardware changes you're referring to. Could you please post a snippet of your current code?

boris bergman1 said:

What triggers also belong to GPIOTE PORT event

Any GPIO input enabled with sense can trigger the PORT event.

[boris bergman1]

Vidar Berg said:

'm also not sure I understand which hardware changes you're referring to.

I meant I configured my peripheral to signal 1 until _data_handler is executed.

Vidar Berg said:

Could you please post a snippet of your current code?

static void lsm6dsl_data_handler(struct k_work *work);
K_WORK_DELAYABLE_DEFINE(_proces, _data_handler);

void _isr(const struct device *dev, struct gpio_callback *cb, uint32_t pins) {  
  gpio_pin_interrupt_configure_dt(&signal, GPIO_INT_DISABLE);
  LOG_INF("isr given");

  k_work_schedule(&_proces, K_NO_WAIT); 
}

void attachinterrupt(void) {

	__ASSERT(device_is_ready(signal.port), "custom device not ready");
	if(gpio_pin_configure_dt(&signal, GPIO_INPUT)) {
    LOG_ERR("error configure signalgpio");
  }

	if(gpio_pin_interrupt_configure_dt(&signal, GPIO_INT_LEVEL_HIGH)) {  //GPIO_INT_EDGE_TO_ACTIVE
    LOG_ERR("error configure interrupt signalgpio");
  };
	gpio_init_callback(&_interrupt, _isr, BIT(signal.pin));
	if(gpio_add_callback(signal.port, &_interrupt)) {
    LOG_ERR("callback for interrupt not set");
 };
}

Vidar Berg said:

Any GPIO input enabled with sense can trigger the PORT event.

I don't really get how this translates to the events in Zephyr.

[Vidar Berg]

PORT events are used when you use the *_LEVEL_* flags with the gpio_pin_interrupt_configure_dt() function, and IN events are used when you use the *_EDGE_* flags. I don't know what your input signal looks like, but are you not having this problem when using GPIO_INT_EDGE_TO_ACTIVE?

boris bergman1 said:

I meant I configured my peripheral to signal 1 until _data_handler is executed.

Sorry, but I'm not sure what you mean by this.

[boris bergman1]

Vidar Berg said:

Sorry, but I'm not sure what you mean by this.

I mean my trigger signal is HIGH(1) until my function (_data_handler) reads the register of my peripheral, then it turns LOW(0).

An GPIO_INT_EDGE_TO_ACTIVE trigger will only trigger once, when the signal goes from LOW to HIGH. a level trigger will trigger indefinite as long as this signal is HIGH.

That's the reason I turn of the trigger in my ISR (see my previous code snippet).

[Vidar Berg]

I now understand the problem. I had only tested the GPIO_INT_LEVEL_ACTIVE with button presses and didn't realize that it would keep firing interrupts as long as the input signal is active. The solution is to revert back to using the GPIO_INT_EDGE_TO_ACTIVE flag and set the 'sense-edge-mask' property in your gpio node in the devicetree to enable the sense mechanism to be used for edge detection.

[boris bergman1]

Thanks, this works and uses exactly the same current as GPIO_INT_LEVEL_ACTIVE.

&gpio0 {
    sense-edge-mask = <0xFFFFFFFF>; //<0x400>;
};

Only, if I use 0x400 it consumes more. (pin 11)