Current consumption difference - nrf5340 running at 128MHz vs 64MHz

[sshenoy105]

Hello,

While working on optimizing battery life, i noticed something strange.. This particular code actually does almost nothing other than setting low power mode (instead of low latency) and mostly system is sleeping (don;t worry about the fact that current consumption is not in single uAs).

nrF Connect SDK 1.9.1
Custom board with nrf5340-CLAA-R7
Used PPK-II to get the screenshots pasted above.
Secure build.. BLE image with both cores flashed (But BLE is disabled.. I do not call bt_enable)

1. First image (where you see current spikes up to 4mA, the CPU runs at 128MHz.. Second image is the case where we run with default 64MHz clock. The only difference bwteen the two cases is the following line of code.

  //make sure application CPU core is running at 128MHz
nrfx_clock_divider_set(NRF_CLOCK_DOMAIN_HFCLK, NRF_CLOCK_HFCLK_DIV_1);

and I comment it out when i like to run at default 64MHz clock..

Questions:

1. Is the API used to configure the system to run at 128MHz correct ?

2. I read one line somewhere that said not only will CPU consume more power while running at 128MHz, sleep current will be higher too.. is the difference in consumption shown above in line with what you think it should be  ? Baseline current difference is almost 4x 

3. There is a spike every 170ms or so.. Like I mentioned, my system is doing absolutely nothing other than zephyr booting, setting low power, shut down some of the peripherals (that were enabled using prj.conf .. like I2c, spi1 etc.).. The spike is up to 4mA while running with 128MHz but only about 180uA while not running at 128MHz..  Any idea what this could be  and why every 170ms  and why the difference in current consumption to that extent ?

[Stian Røed Hafskjold]

Hi, I'm able to reproduce what you are seeing here. It looks to me like the clock tree is always on when the clock source is configured to be 128MHz. I will have to run this through the HW designers, and see if this is intentional or not, and if there is a workaround. Will keep you updated.

Best regards,
Stian

[Stian Røed Hafskjold]

I was just looking through the electrical specs and found that this is in fact a known issue:

ION_IDLE3	System ON, wake on any event, power-fail comparator enabled					1.3		uA
ION_IDLE3,128MHz	System ON, wake on any event, power-fail comparator enabled, clock=HFINT128M					785		uA

https://infocenter.nordicsemi.com/topic/ps_nrf5340/chapters/current_consumption/doc/current_consumption.html?cp=3_0_0_3_3_0_0#unique_2095571962

I'm also getting around 700 uA when enabling 128MHz, so I would say that your 400 uA base current is actually lower than expected. It could be due to chip to chip variation, or different voltage or regulator settings, etc.

[sshenoy105]

thanks for the response.. what are our options given this is going to affect our battery life significantly ? and running at 64MHz is not really the option we want to use

[Stian Røed Hafskjold]

Hi, there's some more information here: https://infocenter.nordicsemi.com/topic/ps_nrf5340/chapters/pmu/doc/pmu.html?cp=3_0_0_3_0#topic

So basically what you need to do is to configure the clock to 64MHz before the CPU goes to sleep. The easiest option, if you want to run the CPU at 128MHz during CPU intensive tasks, algorithms, etc, is to toggle the frequency before and after these specific algorithms. If you always want 128MHz when the CPU is running, the best option is probably to modify the idle thread to toggle back to 64MHz before WFE.