Delay from GPIOTE Event until a Task is started over PPi on nRF52

[Pascal Künzi]

Hi all,

I measured the delays between a GPIOTE event and a Task (in this case toggling with another GPIOTE) through PPI.

The results are showed below:

For the nRF52 it's arround 380ns.

The same measurements were done with the nRF51.

Here it's arround 250ns.

Originally i am coming from counting the pulses of a high frequency signal ~3.5MHz which works fine on the nRF51 but doesn't work on the nRF52.

I couldn't find any information about this topic. So i was wondering:

1. where that difference is coming from
2. if these delays are accurate and valid for other tasks too
3. what happens if too many pulses appear?

On the nRF52 the output pulses look as followed:

It looks like the chip can only handle some but not all of them which leads to weird on-off times.

With the nRF51 it looks like that:

All of the input pulses are processed and generate the expected output.

I hope you can help me and that there is a possibility to get the same results on the nRF52 without adding a flip flop to divide the HF-signal by two.

[Ole Bauck]

Do you see any difference in the delay if you enable constant latency mode (NRF_POWER->TASKS_CONSTLAT = 1)?

[Pascal Künzi]

Hi Ole,

Thanks for the input! Yes I do see a difference. The delay is reduced to the ~230ns which I also measured with the nRF51. But the nRF52 stil can't handle all the rising edges. It seems to me, that the cpu is not able to clear the event register before the next rising edge appears.

[MartinBL]

Hi Pascal. What kind of signal generator do you use, what is the signal peak voltage, and the nRF52's supply voltage?

[Pascal Künzi]

Hi Martin. The nRF is supplied with 3.3V and the signal generator is an Agilent which does generate a sharp and proper 3.3V square signal.

[MartinBL]

We have managed to reproduce the issue, but have not found the root cause. One workaround that seems to work (with a frequency generator) is to lower the signal voltage to ~0.2V below VDD. Can you please try this and tell me if it works?

PS: if you use a nRF52 DK kit there are some protection diodes in front of the chip reducing the input voltage a bit. So make sure to measure VDD_nRF and do not assume that your supply voltage is the same as the voltage supplied to the nRF52.