Why does toggling a GPIO pin happen in constant time, but not set and clear for GPIO?

[Goose]

In testing the timers on the dongle, I have made a GPIO pin be controlled by the timer triggering.

So the following has a constant time between edges:

    case NRF_TIMER_EVENT_COMPARE1:

        nrf_gpio_pin_toggle(DBG_TIMING_PIN_1); // This happens with a constant spacing between signal edges

        break;

While this does not:

    case NRF_TIMER_EVENT_COMPARE1:

        // This creates a small impulse, but the timing between these impulses vary

        nrf_gpio_pin_set(DBG_TIMING_PIN_1);

        nrf_gpio_pin_clear(DBG_TIMING_PIN_1);

        break;

Am I understanding the GPIO library incorrectly? The first piece of code generates a 50ms pulse between edges, while the second one produces a varying width pulse between each impulse.

[Håkon Alseth]

Hi,

So the following has a constant time between edges:

The first piece of code will invert (toggle) the GPIO on every time the NRF_TIMER_EVENT_COMPARE1 occurs. 

While this does not:

 This second piece of code will set and clear within the NRF_TIMER_EVENT_COMPARE1 interrupt, and you'll see a short pulse every 50 ms (that seems to be your CC value)

 

It'll look like this if you scope and compare the two:

 

Kind regards,

Håkon

[Goose]

Hi, perhaps my question would be more clear if I asked to compare it set then clear with toggling twice:

double_toggle:

    case NRF_TIMER_EVENT_COMPARE1:

        nrf_gpio_pin_toggle(DBG_TIMING_PIN_1);

        nrf_gpio_pin_toggle(DBG_TIMING_PIN_1);

which produces the following waveform that I expect

compared to set_then_clear:

Goose said:

    case NRF_TIMER_EVENT_COMPARE1:

        // This creates a small impulse, but the timing between these impulses vary

        nrf_gpio_pin_set(DBG_TIMING_PIN_1);

        nrf_gpio_pin_clear(DBG_TIMING_PIN_1);

        break;

which produces pulses at random times and sometimes according to the timer.

I am looking at the time between of the pulses and not the width of the pulse high-level.

Both tests were done on the same piece of code without changing any other lines. Is this a bug in the SDK?

[Håkon Alseth]

Your logic analyzer isn't catching the smaller spikes.

Try setting the level (1.8V), and the sampling rate to >50 MSPS:

Kind regards,

Håkon

[Goose]

Thank you Håkon,

It appears that I was indeed witnessing aliasing.

I haven't tested for these specific examples, but I did witness an improvement when increasing the sample rate for a different piece of code which is also toggling the pins similarly.