power consumption is too high on my nrf52 SoC

[Juliusc]

Hi, 

I am currently trying to decrease my power output, for this purpose I used a power profiler and found this pattern:

I am using a test firmware that just disables all unused gpio ports, start ble and then goes into while(1){ __WFE();}

power consumption is never below 400uA, the spikes seem fine (BLE advertising spikes I presume).

s there anything you can tell me about why this would happen?

Even if I turn of BLE I still got 200uA on average, so some of the problem is probably in my configuration (but. already increased the adv intervals) and part of it is the chip by itself. I am using a custom PCB but am pulling down all GPIO ports

Best, Julius

[Vidar Berg]

Hi Julius,

Unused pins can be left floating without affecting the sleep current as GPIO pins are configured as inputs with the input buffer disconnect after reset. 

Are you developing with the nRF Connect SDK or the nRF5 SDK? 

Best regards,

Vidar

[Juliusc]

nRF5 SDK

I see I noticed that the current is a bit higher when I dont set certain gpio pins to clear specifically. 

[Vidar Berg]

There wasn't much to see in that function. Do you have a DK you can try this on? Alternatively, could you share your whole project here or in a private ticket so I can try it here?

[Juliusc]

this is what I get running the exact same code on a PCA10056 DK

[Vidar Berg]

You still have a high idle current, although it's much less noisy. Either way, I need to know what your code is doing to understand what's causing this.

[Juliusc]

I could try to trim it down, its basically only the ble file then anyways, I could send you the params Im using without the services

how do I know if WFE() actually works? I got the feeling it doesnt really shut down the uC.

[Vidar Berg]

The idle current would not have dropped below the CPU run current if the system did not enter sleep. The CPU run current ranges from approximately 2.8 mA to 6.3 mA depending on the supply and regulator configuration. There might be something that is either requesting system resources, such as the 16 MHz clock in sleep mode, or causing leakage from the IOs.

Do you measure a different sleep current if you comment out everything except the main loop in your main() function while running on the DK?

[Juliusc]

if i run it on the DK with only the WFE main loop it takes 20uA

[Juliusc]

I actually noticed something while investigating, I used the following lines in my ble setup: 

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

and when commenting them , the average went down significantly (at least on the DK, I will test this on my custom PCB later). Though I think I included this because there was a problem with the SD but I cannot pinpoint what exactly it was. What are your thoughts on this?

[Juliusc]

still it looks like something is requesting air time every ms, the advertising event only occurs every 500ms as defined, thats whats so weird

[Vidar Berg]

Writing '1' to NRF_CLOCK->TASKS_HFCLKSTART will start the 32 MHz crystal oscillator, and the crystal oscillator is not automatically released in sleep. The run current with the oscillator active is about 200 uA. 

[Juliusc]

ok thanks so I will leave this out and I also found out that my saadc sampling is responsible for most of the drainage besides that.