Long Delay between triggering GPIOTE event and his execution

RE: Long Delay between triggering GPIOTE event and his execution

Markus Aerni — Wed, 01 Aug 2018 13:45:15 GMT

OK with

nrf_gpio_pin_dir_set(ONEWIRE_DATA_PIN, GPIO_PIN_CNF_DIR_Output); // One Wire Direction output
nrf_delay_us(60);
nrf_gpio_cfg_sense_input(ONEWIRE_DATA_PIN, NRF_GPIO_PIN_NOPULL, NRF_GPIO_PIN_SENSE_LOW); // OneWire data Pin Input

all works !

Thanks

RE: Long Delay between triggering GPIOTE event and his execution

Markus Aerni — Wed, 01 Aug 2018 12:07:08 GMT

hmm OK

in my interrupt i do

LED_ON;

nrf_gpio_pin_dir_set(ONEWIRE_DATA_PIN, GPIO_PIN_CNF_DIR_Output); // One Wire Direction output
nrf_delay_us(60);
nrf_gpio_pin_dir_set(ONEWIRE_DATA_PIN, GPIO_PIN_CNF_DIR_Input); // One Wire Direction Input

LED_OFF;

but after that the interrupt wont work again.

RE: Long Delay between triggering GPIOTE event and his execution

Håkon Alseth — Wed, 01 Aug 2018 11:44:18 GMT

Yes, this will change the GPIO from input to an output. Do note that GPIOTE PORT (or IN[n] for that matter) only works if it's an input.

RE: Long Delay between triggering GPIOTE event and his execution

Markus Aerni — Wed, 01 Aug 2018 11:42:14 GMT

Hi,

Thanks this works perfect !

Much faster.

Just another question.

is it possible to change direction with this pin configuration ?

nrf_gpio_cfg_sense_input(ONEWIRE_DATA_PIN, NRF_GPIO_PIN_NOPULL, NRF_GPIO_PIN_SENSE_LOW); // OneWire data Pin

as i did before with ?

nrf_gpio_pin_dir_set(ONEWIRE_DATA_PIN, GPIO_PIN_CNF_DIR_Output); // One Wire Direction output

Best regards,

Markus

RE: Long Delay between triggering GPIOTE event and his execution

Håkon Alseth — Wed, 01 Aug 2018 08:59:15 GMT

Hi,

The interrupt latency for the Cortex M0 is 16 cycles == 1 us @ 16M.

The remainder of the time is processing in the nrf_drv_gpiote library (interrupt handler), as this library handles both PORT and IN[n] events. The PORT event uses the SENSE functionality, which is a level trigged signal, not a edge-trigged signal. In order to detect toggle operations, you emulate edge-triggering by inverting the SENSE (going from HIGH to LOW polarity), which may account for the processing time.

You can switch to use the IN[n] events by setting "high accuracy" when configuring your pin, which shall decrease processing time needed: GPIOTE_CONFIG_IN_SENSE_TOGGLE(true)

GPIOTE IN[n] usage does use more current than the PORT event.

If this is too much of a jitter, I'd recommend that you implement GPIOTE PORT directly, like done here: https://github.com/NordicPlayground/nrf51-powerdown-examples/blob/master/system_on_wakeup_on_gpio/main.c

Best regards,

Håkon