Softdevice uses CPU resources for hardware based events (PPI w/out IRQs...) - why?

[donbosley]

NRF52840 - SDK 15.0 + Some register based custom functions.  S140, 6.0.0.

In order to work around some of the hardware limitations with SPIM in the context of a project, we've built a solution that uses PPI to drive a chain of peripherals. 
In a non-softdevice application it works flawlessly - only using CPU after a number of data transfers have been executed. 

In the Image below we have 8 groups of 16 individual chip selects, all driven via PPI - at the end of the 8th transfer an actual IRQ is issued, a software flag is set and we block process all of the data. The top line is the SPI CS toggling, and the bottom is CPU active low  - it's pretty clear the CPU is only active during the IRQ and block processing.

Once SoftDevice is enabled, we appear to see "responses" to every single event generated, even though none of the peripherals used in the PPI chain have hardware IRQs enabled - is this expected / normal? If so, why?

[Einar Thorsrud]

Hi,

Just having the SoftDevice present should not cause any additional CPU activity. And even if it is enabled, but not advertising, scanning or in a connection, you should not see any frequent regular CPU usage (you could see LFRC calibration every few seconds depending on configuration, but that's all). Can you describe all changes you have done between the two tests? What is the difference in your code?

[donbosley]

I've isolated all of the pieces and the only difference now is the SDK config - I turned SD off with the alternate SDK config and it's generating the interrupt every time. So SD is not the culprit, but given that the CC on the counter isn't being matched, I'm hard pressed to understand what is causing the CPU to turn on.

[donbosley]

So in the end it was a constant defined improperly in our code... sorry for the wasted time, and thank you for the help.

[Einar Thorsrud]

No worries - that happens to all of us sometimes. Thank you for letting us know