Zephyr/NCS - gpio_pin_interrupt_configure causes ~15ua increase in consumption

[Christian Holland]

As instructed to post here by carlescufi  andrzej

We're using Zephyr on at least 4 of our products and have thousands of boards with NRF chips in the pipeline; we're trying to get the consumption down as much as we can on our project, but we noticed that the code for enabling an interrupt on a button is bringing our consumption up.

After adapting the latest NCS Zephyr sample of the "basic/button", I can see that the use of gpio_pin_interrupt_configure causes a 15ua increase in consumption. I have attached the project here so you should be able to build it and reproduce it yourself. I have added a few things to get the consumption down as much as I can to the sample so that there is no noise when focusing on the consumption changes after the function call is made.
You will notice that when this line is commented out (you may have to comment the rest out before the while loop too as it's dependent), the consumption goes down to ~3ua, but when it's left in, the consumption sits at around ~15ua. I have tried this on Zephyr 2.3.0 and 2.3.99 (latest) and seen the same results on both

button_sample.tar.gz

[Sigurd]

Hi,

I see that this got resolved in the Zephyr slack channel.

In short, the EDGE interrupt in Zephyr uses GPIOTE IN event, and is therefore affected by this errata: https://infocenter.nordicsemi.com/index.jsp?topic=%2Ferrata_nRF52832_Rev2%2FERR%2FnRF52832%2FRev2%2Flatest%2Fanomaly_832_97.html&anchor=anomaly_832_97

While waiting for an improvement in Zephyr for this, a workaround is then to use the GPIO_INT_LEVEL_ACTIVE option instead, which uses the GPIOTE port event, and then alternate GPIO_INT_LEVEL_INACTIVE/ACTIVE in the interrupt handler.

[Andrew Porter]

Yes - thank you - we have a working solution toggling the interrupt between INACTIVE/ACTIVE.

We have more power consumption issues to solve - as we re-enable more parts of our original code - should we continue to ask in slack or carry on it this ticket ?

[Sigurd]

This case here was about the gpio_pin_interrupt_configure() function. If you have new power consumption issues, you can open a new case for that.

[Andrew Porter]

It looks like we nailed it - nothing outside of the expected power profile now in all use conditions  \o/

[xuan0802]

Dear all,

I have the same power consumption issue when configuring edge interrupt, 

err = gpio_pin_interrupt_configure(button_dev, FUNC_BUTTON_PIN, GPIO_INT_LEVEL_ACTIVE); --> I changed to level interrupt but when I press the button, it could not run to call back function.

I am using NCS 1.4, 

Thank you,