Zigbee low power end device

[rozpruwacz]

Hi,

What is the lowest possible current consumption for nrf52840 dongle dev board with zigbee ?

I'm using this for development and i'm getting 20-30uA in idle. Is it possible to go down with current consumption to 2-3uA ? I know that it is possible without zigbee because I prepared a minimal test app that just calls k_sleep in main and then the current consumption is about 2-3uA. But when I created minimal test app with zigbee I cannot get lower than 20uA.

Below is my minimal zigbee app:

#include <zephyr/kernel.h>
#include <ram_pwrdn.h>

#include <zboss_api.h>
#include <zigbee/zigbee_app_utils.h>
#include <zb_nrf_platform.h>

void zboss_signal_handler(zb_bufid_t bufid)
{
    zigbee_default_signal_handler(bufid);
    if (bufid) {
        zb_buf_free(bufid);
    }
}

int main(void)
{
    zb_set_rx_on_when_idle(ZB_FALSE);
    power_down_unused_ram();

    zigbee_enable();

    k_sleep(K_FOREVER);
    return 0;
}

CONFIG_SERIAL=n
CONFIG_GPIO=n
CONFIG_CONSOLE=n
CONFIG_BOARD_SERIAL_BACKEND_CDC_ACM=n

CONFIG_ZIGBEE_ADD_ON=y
CONFIG_ZIGBEE_APP_UTILS=y
CONFIG_ZIGBEE_ROLE_END_DEVICE=y

CONFIG_RAM_POWER_DOWN_LIBRARY=y

CONFIG_NET_IPV6=n
CONFIG_NET_IP_ADDR_CHECK=n
CONFIG_NET_UDP=n

[Maria Gilje]

Hello,

I see that you have made use of the low-power recommendations here. That's good.

Are you able to share your measurements you are taking of the idle mode with us? The value is close to the RAM on, GPIOTE input (event mode) scenario for the sleep current consumption on nRF52840. This mode has full RAM retention though, and with the power_down_unused_ram() function the actual consumption is more expected to be somewhere between a "no RAM retention" and a "full RAM retention" scenario. Do you see any difference when enabling and disabling the function?

To reduce power consumption in general, you can increase the sleep time by increasing the long poll interval and keep alive timeout.

Best regards,

Maria

[rozpruwacz]

More info.

I power the board with 3.3V into VDD pin, VDD shorted with VDDH. The DCDC buck converter is enabled by default board dts. I can see that in most basic test case (jus k_sleep in main) it makes difference whether VDD and VDDH is shorted or not, when not shorted then the current is around 13uA, with shorted the current is around 3uA.

I also made a full test case, scene controller + power management clusters, commands sent to coordinator on button press, battery voltage reporting using ADC. And this test case uses exactly the same amount of current in idle as in minimal zigbee test case attached in the first post.

So if the minimal zigbee and full zigbee test cases consumes the same amount of current in idle, and taking into account that the full test case must use some kind of button triggered event to wake up it looks like both cases use it.

[Maria Gilje]

Hello,

Thank you for your patience.

rozpruwacz said:

What You mean by that ? I do measurements with multimeter. Today I will try w oscilloscope.

My apologies. I wrongly assumed that you were using the PPK2 to measure the current on your board and had a file with the measurements.

rozpruwacz said:

I also made a full test case, scene controller + power management clusters, commands sent to coordinator on button press, battery voltage reporting using ADC. And this test case uses exactly the same amount of current in idle as in minimal zigbee test case attached in the first post.

So if the minimal zigbee and full zigbee test cases consumes the same amount of current in idle, and taking into account that the full test case must use some kind of button triggered event to wake up it looks like both cases use it.

I will try to reproduce your measurements and forward the findings internally. I may have missed a configuration to reduce the current consumption in idle mode.

Best regards,

Maria

[Maria Gilje]

Hello,

I did some preliminary measurements today on an nRF52840 DK and I measured current in the 1.5-2mA range for just the MCU, but for the whole DK the measurements were higher (33-40mA). But this still shows that the current consumption for Zigbee SED should get to around 2mA for the MCU when the device is idle.

I used the Zigbee Light Switch from the Zigbee R23 Add-on to test.

Here are my measurements in the ppk2 and csv file formats.

ppk-measurements-1405.zip

Best regards,

Maria

[rozpruwacz]

That is great news ! So it is possible to lower the current consumption and there must be something wrong with my setup.

Do You ave any samples for sleeping device on nrf52840 dongle ? Because I still get 20uA with light_switch example but I had to adapt it to nrf52840 dongle.

One thing that I just noticed is that the CPU is waking up every 1 second when in idle. Is this correct behaviour or result of misconfiguration ?

[Maria Gilje]

Hello,

It looks like I messed up my units. My measurements were for the mA order, but you are of course talking about µA. My apologies for the mistake.

So my test did not in fact show lower draws than yours.

rozpruwacz said:

Do You ave any samples for sleeping device on nrf52840 dongle ? Because I still get 20uA with light_switch example but I had to adapt it to nrf52840 dongle.

We don't have any SED samples which support the nRF52840 Dongle out of the box.

How did you modify the sample? Could you share the files which you changed and added to the sample?

One source for higher power draw can be that some peripherals are initialized. They can be disabled in a devicetree overlay and some instructions are found here: https://docs.nordicsemi.com/bundle/ncs-latest/page/nrf/test_and_optimize/optimizing/power_general.html#disable_unused_pins_and_peripherals

I'll also do some more work where I make sure to disable unused peripherals, but let me know if you have any progress in the mean time. I'll get back to you early next week.

Best regards,

Maria

[rozpruwacz]

Hi,

I have finally found the culprit. It was the ncs,zigbee-timer chosen node. I used the dt overlay from light_switch sample:

/ {
	chosen {
		ncs,zigbee-timer = &timer2;
	};
};

&timer2 {
	status = "okay";
};

I disabled timer2 and removed ncs,zigbee-timer node and current consumption went down to 3-4uA in idle which is a very good result.

From the ncs-zigbee code I can see that the same can be achived by selecting CONFIG_ZIGBEE_TIME_KTIMER=y which is described in the docs as:

"Configures the ZBOSS OSIF layer to use Zephyr's system time as the Zigbee time source"

But this is not clear to me what are the consequences of that setting. Is it ok to use CONFIG_ZIGBEE_TIME_KTIMER=y ? I can see that this setting is selected only in ncp sample. Why is that ?

PS.

Another tip to keep current consumption low when using gpio buttons with native zephyr input subsystem (CONFIG_INPUT=y) is to set sense-edge-mask parameter in DT:

&gpio0 {
    sense-edge-mask = <0xffffffff>;
};

&gpio1 {
    sense-edge-mask = <0xffffffff>;
};