mesh_stack_init and Low Power

[Jerome ML]

Hello,

I'm working with the nRF5 SDK for Mesh v4.2.0.

During the init phase, I call the function mesh_stack_init()

After doing so, the consumption is on average 2.8 mA, even when doing :

    for (;;)
    {
        nrf_pwr_mgmt_run();
    }

And the MCU is never in idle state.

How to desactivate all the functions initialized by mesh_stack_init() in order to put the MCU in low power state ?

Thank you.

[Hung Bui]

Hi Jerome, 
When you enable mesh stack the radio will be active and stay in RX mode all the time. 
To achieve low power you need to disable the stack. 
Please have a look here: https://devzone.nordicsemi.com/f/nordic-q-a/57450/reliably-disabling-mesh-stack-quickly-for-low-power

[Jerome ML]

Thank you for your answer.

I tried this, and also to disable everything that I could, but it doesn't work.

PS : I made a lot of research and tests before posting my question, and I wouldn't have posted it if there was a publicly available and functionnal answer.

[Hung Bui]

Ok if you have done that and the current consumption still high, I suspect the high current was from something else. 

Have you made sure you power reset the chip before you do the measurement ? After flashing the chip will remain in debug mode hence the CPU will stay awake regardless what you do. A power reset needed. 

[Jerome ML]

The chip is not in debug mode. If I don't call the function mesh_stack_init(), the average power consumption is 2.3 µA.

I did what you suggested :

uint32_t retval = proxy_stop();

if (retval == NRF_SUCCESS)

{

retval = nrf_mesh_disable();

}

And I still have an average power consumption of 2.8 mA.

The INTENSET registers give me :

<t:      14010>, main.c,  410, Peripherals enabled:
<t:      14013>, main.c,  436,   RTC1:   	0x10002
<t:      14016>, main.c,  460,   RNG:    	0x1
<t:      14019>, main.c,  475,   MWU:    	0x1000001

So, it seems that the stack is not disabled.

I didn't find any example that disables the mesh stack. If you have one, I'm interested.

[Hung Bui]

Hi,

I added this:

 case 2:
        {
          uint32_t retval = proxy_stop();

          if (retval == NRF_SUCCESS)

          {

              retval = nrf_mesh_disable();
              if (retval != NRF_SUCCESS) 
              {
                  __LOG(LOG_SRC_APP, LOG_LEVEL_INFO, "error = %d \n",retval);
              }


          }
        break;
        }
        case 3:
        {
          retval = proxy_enable();
          if (retval == NRF_SUCCESS)
          {
              retval = proxy_start();
          }
          if (retval == NRF_SUCCESS)
          {
              retval = nrf_mesh_enable();
          }
           if (retval != NRF_SUCCESS)
           {
            __LOG(LOG_SRC_APP, LOG_LEVEL_INFO, "error = %d \n",retval);
            }
          break;

        }

Into button_event_handler() in light_switch server in SDK v4.2 and I can see when I press button 2 on nRF52832 DK the current dropped down to a few uA. And pressing button 3 will wake it up again (up to 12mA). 
(There is an issue when enabling it when it not provisioned that I'm looking into, but I guess it's not relevant to this thread)

Please make sure you don't use RTT log when you measure the current, it would keep the CPU awake. 
I attached the hex file here in case you need for testing. 

5001.light_switch_server_nrf52832_xxAA_s132_7.0.1.hex

[Jerome ML]

I did what you said, once more, and it still doesn't work.

I also added in the callback of the NRF_MESH_EVT_DISABLED event  :

mesh_adv_stop();
scanner_disable();

Some timer is still running and wakes up the MCU.

Try to do something like this in your own code, and tell me if the LED blinks :

for (;;)
{
    nrf_pwr_mgmt_run();
    nrf_gpio_pin_toggle(LED_1);
    nrf_delay_ms(100);
}

How to be sure that the RTT log is not working ?

[Jerome ML]

I managed to get to 33µA.

It's still more than 10x what it's supposed to be to get back to the original condition :

"If I don't call the function mesh_stack_init(), the average power consumption is 2.3 µA"

[Hung Bui]

Hi Jerome, 

Could you let me know what made the current dropped down to 33uA ? 

I did what you suggested and it's true that the CPU was woken up occasionally. I suspect it's the RTC still running and still trigger some interrupt. I need to check with the Mesh team to see if we can turn that off as well. 

However, one thing you can do to reduce the CPU waking frequency is to remove: 
rtt_input_enable(app_rtt_input_handler, RTT_INPUT_POLL_PERIOD_MS); inside start() function.

I found that it's the main reason the CPU stay awake. The default period is 100ms. 

In addition, if the device is not provisioned, you can call mesh_provisionee_prov_listen_stop() before you disable mesh. But I would suggest to test after the device provisioned. 

With those modification I'm seeing around 5uA. (CPU woken up once a second(estimated), I will need to look into what causes this).

I attached my binary here. You press button 2 (P0.14) on the DK to disable mesh and press button 3 (P0.15) to enable mesh. 

0488.light_switch_server_nrf52832_xxAA_s132_7.0.1.hex