Why is current consumption unstable in k_cpu_idle() after GPIO interrupt?

[Yusuke]

Current consumption during  k_cpu_idle() after GPIO interrupt does not take the same value. Why does this happen? Any idea? Thank you for reading this post in advance.

<Environment>
- DK v0.8.5(SB43 cut)
- modem firmware v1.0.0
- nrf; v1.0.0
- current measurement tool: multimeter

I configure GPIO11 as interrupt pin. I connect it to GND and disconnect again and again by jumper wire. After I connect it back to GND, the DK sometimes continuously consumes 25 uA and in another time 150 uA. 

CONFIG_TRUSTED_EXECUTION_NONSECURE=y

CONFIG_GPIO=y
CONFIG_SERIAL=n
# CONFIG_SERIAL=n in spm program
CONFIG_LOG=n

CONFIG_AT_HOST_LIBRARY=n

CONFIG_REBOOT=y

#include <zephyr.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <gpio.h>
#include <device.h>

#define GPIO_IN_INT GPIO_IN_PIN11_Pos

struct device *gpio_controller;
struct gpio_callback gpio_cb;
bool is_interrupted;

void gpio_test_interrupt(struct device *gpio, struct gpio_callback *cb, u32_t pins) {
    is_interrupted = true;
}

u8_t init_gpio_interrupt() {
    int8_t ret;
    
    ret = gpio_pin_configure(gpio_controller, GPIO_IN_INT, (GPIO_DIR_IN | GPIO_INT | GPIO_INT_EDGE | GPIO_INT_ACTIVE_HIGH));
    if (ret) {
        return -1;
    }
    
    gpio_init_callback(&gpio_cb, gpio_test_interrupt, BIT(GPIO_IN_INT));

    ret = gpio_add_callback(gpio_controller, &gpio_cb);
    if (ret){
        return -1;
    }

    ret =  gpio_pin_enable_callback(gpio_controller, GPIO_IN_INT);
    if (ret){
        return -1;
    }

    return 0;
}

void main(void)
{
    gpio_controller = device_get_binding("GPIO_0");
    if (!gpio_controller) {
        sys_reboot();
    }

    if (init_gpio_interrupt() != 0) {
        sys_reboot();
    }

    while (1) {
        k_cpu_idle();
    }            
}

[Yusuke]

I added and measured the current consumption, but the same symptom happens.

Can you measure the current with nRF9160 DK on your side? Do you see the same symptom?

    ret = gpio_pin_configure(gpio_controller, GPIO_IN_ACCEL_INT,
			     (GPIO_DIR_IN | GPIO_INT | GPIO_INT_EDGE |
			      GPIO_INT_ACTIVE_HIGH | GPIO_INT_DEBOUNCE | GPIO_PUD_PULL_DOWN));

[Hakon]

Why do you connect it to GND when you have the GPIO_INT_ACTIVE_HIGH flag set? Shouldn't it be GPIO_INT_ACTIVE_LOW for it to trigger when touching ground?

[Yusuke]

Hi, Hakon. sorry for a late reply.

This seems to work out. The current consumption is not stuck at 150uA.

I don't really understand why this works though.

    ret = gpio_pin_configure(gpio_controller, GPIO_IN_ACCEL_INT,
			     (GPIO_DIR_IN | GPIO_INT | GPIO_INT_EDGE |
			      GPIO_INT_ACTIVE_LOW | GPIO_INT_DEBOUNCE));

[Hakon]

You were probably not triggering the interrupt handler properly. Connecting the jumper wire to ground makes the GPIO pin go low, so the pin needs to be active low in order to trigger an interrupt. Does that make sense?

[Yusuke]

>the pin needs to be active low in order to trigger an interrupt. 

sensor A port ----interrupt signal line ----> nRF9160 port

sensor A releases a pulse signal when it detects motion. This line is active high. 

So with the above setting (GPIO_INT_ACTIVE_LOW), nRF9160 detects the falling edge of the pulse.

GPIO_INT_ACTIVE_HIGH should have the same result detecting the rising edge and actually it does.

However, the current consumption is different between the two experiments.