SPIS Interrupt Latency. Delay due to Zephyr Tasks

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Sigurd Hellesvik — Tue, 16 Apr 2024 11:51:39 GMT

As far as I can tell, disabling idle is not really a use-case so we have no direct support for just turning it off.

When you call k_sleep(), the app will check scheduling and if there is nothing to do, I think it will go to idle.
Maybe try to do a while(1){} loop to never enter idle mode?

Alternatively, maybe you can use a hook to check if the CPU is about to enter Idle, and then trigger a task?

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Chris R — Mon, 15 Apr 2024 22:30:57 GMT

Well, I guess because it's an interrupt driven event I can't really determine exactly when it will happen in order to pre-emptively wake up to shorten the delay.
If you had a way to permanently disable Idle I could at least figure out the tradeoff / increase in power.

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Sigurd Hellesvik — Mon, 15 Apr 2024 07:41:48 GMT

First of all, removing idle will increase power consumption, so be aware of that.

[quote user="Chris R"] How would I go about keeping the MCU out of the IDLE state? Is there a way to force Zephyr to not go into this IDLE state?[/quote]

Do you mean permanently, or disable idle temporarily then enable it again when you are done with the tight timings?

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Chris R — Mon, 15 Apr 2024 01:54:37 GMT

Hi Sigurd,

Thanks!
That actually makes sense...

One follow up question - How would I go about keeping the MCU out of the IDLE state? Is there a way to force Zephyr to not go into this IDLE state? Does idle here mean the IDLE zephyr OS task?

Best regards,

Chris

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Sigurd Hellesvik — Fri, 05 Apr 2024 11:49:25 GMT

I asked a collegua about this and they say the following:

"
https://infocenter.nordicsemi.com/topic/ps_nrf5340/chapters/reset/doc/reset.html?cp=4_0_0_3_9_11#unique_287630547

tIDLE2CPU
tIDLE2CPU,CONSTLAT

3 us means that the system isn't fully gone to systemonidle, while >10 us means that the clock tree is running.

Does this make sense to you?

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Sigurd Hellesvik — Tue, 02 Apr 2024 13:52:43 GMT

We have been away on Easter vacation, but we are back now.
I will continue looking into this now

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

elarson — Tue, 26 Mar 2024 14:23:05 GMT

Hi Sigurd,

I'm working with Chris on this issue. The 12.5us shown here is the most common value, with very few that have no delay. We have also seen that running with and without the debugger can affect the delay we observe, having the debugger attached somehow reduces the delay.

RE: SPIS Interrupt Latency. Delay due to Zephyr Tasks

Sigurd Hellesvik — Tue, 26 Mar 2024 11:58:31 GMT

Hi,

[quote user=""]However, we are encountering a ~12.5us increase in latency when a background task is running. [/quote]

Is the latency the same every time, or does it vary a bit?

Regards,
Sigurd Hellesvik