nRF52 LFCLK ~40 uA consumption with SRC=Xtal

[damaki]

Hello,

I'm seeing higher than expected power consumption when running an RTC peripheral using LFXO as the LFCLK source compared to using the internal RC. Specifically, I'm observing a constant ~40 uA current consumption using the external 32 kHz Xtal as the LFCLK source, compared to ~3.3 uA using the internal 32 kHz RC as the clock source. I was expecting to see only 2-3 uA current consumption using an external 32 kHz LFCLK source.

My hardware setup is a nRF52840 Preview Development Kit, measuring the current consumption with the Power Profiler Kit.

My software is enabling the following peripherals only:

• LFCLK (SRC=Xtal or Rc, BYPASS=0, EXTERNAL=0)
• RTC0 (running at 32 kHz tick rate)

RTC0 is only configured to generate OVRFLW and COMPARE0 interrupts, and COMPARE0 is used to wake up the CPU every 1 second, and then the CPU goes back to sleep with WFI. The average current measurements I stated above were captured while the CPU is asleep, but RTC0 is running using LFCLK.

I am not using any SDK, as I am not using C/C++.

The *only* thing I am changing is the SRC for LFCLK to either Rc or Xtal. I am not explicitly enabling any other peripherals or the HFXO, so I don't understand the large difference in power consumption between the two LFCLK sources.

What could be causing such a high current consumption using the external Xtal compared to the internal Rc oscillator?

edit: corrected hardware setup (I have a Preview DK, not one of the newer DK).

[damaki]

Some additional information:

The chip marking is: N52840 Q1AAAA 1716AK

I perform the same errata workarounds in my startup code from system_nrf52840.c in the nRF5 SDK. I.e the workarounds for erratas 36, 66, 98, 103, 115, 120, and 136.

[Stian Røed Hafskjold]

Did you add any checks to determine that the LF crystal is actually starting up? It needs to start up before you can enable the RTC. Something like this:

NRF_CLOCK->TASKS_LFCLKSTART = 1;
while(NRF_CLOCK->EVENTS_LFCLKSTARTED == 0){}
NRF_CLOCK->EVENTS_LFCLKSTARTED = 0;

If this is not working, can you please post the hex file here, both the LFXO and the LFRC one. Then I can try to measure here on a preview DK.

[damaki]

Hi Stian,

Thanks for the reply and suggestion.

I am already checking the that the LFCLK is starting before using the RTC with the same code you posted (start the LFCLK, wait for it to start, then clear the event). I have also confirmed that the LFCLK is starting correctly and with the correct source by inspecting the CLOCK registers via a debugger once the code has started.

I've attached the two hex files using the LFRC or LFXO clock source. I built these by only changing the line of code which sets the LFCLKSRC register. The attached programs should both blink LED2 with a period of 500 ms on, 500 ms off.

blinky_lfrc.hexblinky_lfxo.hex

[Stian Røed Hafskjold]

I tried measuring both of the hex files on a Preview DK with the eng. A chip revision as you are referring to. With a Power analyzer I got 3-4 µA on both the hex files you sent. And with the PPK I got 44 uA on both hex files.

I recommend you to get a newer version of the DK. The DK version corresponding to QIAAAA (eng A) was the first version released and it was not very stable. Not really sure where the 40 µA comes from but this is not related to LFXO vs LFRC.

[damaki]

Hi Stian,

Thanks for testing these on your end. It's good to know that with a power analyzer the current consumption meets my original expectation.

I'll get a newer version of the DK for performing current measurements.