nRF52 GPIOTE interrupt latency

RE: nRF52 GPIOTE interrupt latency

Sonya — Wed, 24 Oct 2018 07:35:40 GMT

I use your example. The measurement result is similar with you. But if I test the LPCOMP interrupt latency, the latency is longer than the GPIOTE

RE: nRF52 GPIOTE interrupt latency

Kasper Leonhardt — Wed, 24 Oct 2018 06:02:47 GMT

Do you use the softdevice. Have you measured the latency? The SDK is OK if you do as in my example.

RE: nRF52 GPIOTE interrupt latency

Sonya — Wed, 24 Oct 2018 02:38:12 GMT

Hello, I also have a long GPIOTE interrupt latency. I use the the original interrupt service routine from the SDK. I do not how to skip the original interrupt service routine from the SDK. I am a novice for nrf52832. Can you give me some advice. Thanks~

RE: nRF52 GPIOTE interrupt latency

Kasper Leonhardt — Wed, 11 Apr 2018 10:22:10 GMT

I found this table in the documentation of the SoftDevice:

RE: nRF52 GPIOTE interrupt latency

Kasper Leonhardt — Wed, 11 Apr 2018 10:16:53 GMT

This could explain it. But I am suprised that it takes the softdevice 1.25µs to make a relative jump. I think that Nordics developpers should improve this. I do not think that the SoftDevice needs any GPIOTE events, so every event should be routed to the user application.

Thank you for your help.

RE: nRF52 GPIOTE interrupt latency

Kenneth — Wed, 11 Apr 2018 09:24:48 GMT

There is some delay due to forwarding in the softdevice, can you try without the softdevice alltogether for comparison?

RE: nRF52 GPIOTE interrupt latency

Kasper Leonhardt — Wed, 11 Apr 2018 06:34:56 GMT

Nope. CPU was in most active mode. No sleeping or stand-bying.

RE: nRF52 GPIOTE interrupt latency

Kenneth — Tue, 10 Apr 2018 14:53:00 GMT

What state is the CPU in here? Maybe you have called __WFE(), then there is some additional wakeup time.