Power consumption higher in nrf_pwr_mgmt_run() with soft device disabled?

[JamesB7]

This surprised me.

I disable advertising, stop my app timers, and then loop "while (gpio_sample()) { nrf_pwr_mgmt_run(); }"

I have five low-accuracy nrf_gpiote buttons running, and am waiting for a button press after an event. It should sleep into the low micro-amps while waiting for the event.

If, after disabling advertising, I DO disable the soft device nrf_sdh_disable_request(), it pulls 0.5 mA

If I do NOT disable the soft device, it pulls 0.036 mA

Why would this be? I'd expected it to be lower, but more importantly, I'd expected it to be the other way around.

[JamesB7]

I also notice that sd_power_system_off() pulls 0.029 mA. Shouldn't it pull 0.8 microAmp? Low-accuracy nrf_gpiote is PORT, yes? There's nothing else on the PCB. Is there something in particular I might need to turn off?

[Susheel Nuguru]

That seems a bit wierd, Can you help me reproduce this at my desk? if softdevice is disabled, then the :__WFE sleep should be same as sd_app_evt_wait from the softdevice. Unless we forgot to patch an errata workaround around the sleep which most likely is done implicitly by the softdevice API.

[JamesB7]

I can send you the code. How would you like me to send it to you?

It's a very simple adaptation of a sample.  (The nRF SDK is patched as linked in that one support page, to support nRF 52805, by the way.)

[JamesB7]

Worst case, can the Soft Device be left on, but get down to the under-1uA power level? Or does it always keep timers on that make that impossible?

[Susheel Nuguru]

you can attach the code here in the "insert" option in your reply. If this is simple an no confidential information is there in the code, then we can keep this thread public so as to help others. 

[Susheel Nuguru]

Softdevice uses RTC and keeps lfclk on for housekeeping for the softdevice scheduler. So the currents for RTC when enabling softdevice and not having any BLE traffic should be around 2uA. If you disable the softdevice completely (in addition of stopping advertising/scanning/connections) you will be able to save this addition 2uA.

[JamesB7]

It turns out the problem was 100 kOhm on the debugger, even with power and ground disconnected. I'm down to 1 microAmp now. Thanks!