Optimisation level in SES - nrf52832 + ICM20948

RE: Optimisation level in SES - nrf52832 + ICM20948

haakonsh — Wed, 26 Aug 2020 08:36:52 GMT

You need the HFCLK for ESB operation, I suggest you rather disable ESB than the HFCLK as that should also clean up any resources like HFCLK.

RE: Optimisation level in SES - nrf52832 + ICM20948

SL06 — Tue, 25 Aug 2020 15:49:17 GMT

Hi, I have found that the main culprit is the HFCLK used for the ESB. I have to do more measurement and testing today.

nrf_drv_clock_hfclk_release(); // does nothing

NRF_CLOCK->TASKS_HFCLKSTOP = 1;  //  power consumption drop from 200µA to almost nothing, less than 40µA.  Its time to get my scope to get more precise data.


Now the problem is that I don't know how to restart the clock.

Calling the clocks_start routine does not do the trick.

void clocks_start( void )
{
    NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
    NRF_CLOCK->TASKS_HFCLKSTART = 1;
    while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0);

}

Any suggestion ?

RE: Optimisation level in SES - nrf52832 + ICM20948

haakonsh — Tue, 25 Aug 2020 08:33:51 GMT

I might be able to see what it is that is using the excess power from the waveform.

Are you using a custom board or a DK?

RE: Optimisation level in SES - nrf52832 + ICM20948

SL06 — Mon, 24 Aug 2020 15:21:25 GMT

Yes I do have a scope, but my voltmeter gave me the same value (an average of the voltage). I am measuring voltage across a 10 ohm resistor (ground connection) on my external board. I have open tread about that, but found no obvious solution. The next step will be to strip the program to the basic and had section by section and see were it goes wrong. I will get there...

RE: Optimisation level in SES - nrf52832 + ICM20948

haakonsh — Mon, 24 Aug 2020 14:40:19 GMT

Yeah, that's not right. Do you have a proper scope of the current consumption?

RE: Optimisation level in SES - nrf52832 + ICM20948

SL06 — Mon, 24 Aug 2020 14:23:00 GMT

Problem solved.

I have forgot to set the printf option to float in the new build configuration.

The other problem was likely with the application size. Removing an unused large matrix variable solved the issue and now the compiled program run normally, using optimisation level 1 or 2.

I still have to work on the power consumption, 200µA in resting mode. I should be able to do better.

RE: Optimisation level in SES - nrf52832 + ICM20948

haakonsh — Mon, 24 Aug 2020 14:13:31 GMT

Can you share the relevant code?

Best guess is a loop where the conditional check uses a variable that is not declared as volatile.