nrf52832 power consumption when disabling Twi driver (i2c)

[Luca Nisti]

Good morning,

It has been implemented a sleep mode to save battery when the board is not used.

To do that we follow the documentation and implemented the SystemOn mode and we've seen huge improvement in the power consumption (both in sleep and running mode)

When sleep mode is engaged all the peripheral are turned off.

To save power due to the twi driver we use the lines of code below.

Although we see good power save we think it is not enough.

The board when in sleep mode consumer around 550uA, but if the twi driver is not initialized at all, the consumption is around 280uA. So I guess there someting we can improve in the twi driver shut down sequence.

We are using nrf5 SDK 17.0.2

        /* Uninit all the TWI instances, this will put the hardware in a known state and stop any ongoing transfer, standard shutdown */
        for (uint8_t i = 0; i < I2C_BUS_N; i++)
        {
            nrf_drv_twi_disable(&busToTwiInstance[i]->twi);
            nrf_twi_mngr_uninit(busToTwiInstance[i]);
        }

[Håkon Alseth]

Hi,

 

I suspect that this erratum is the root cause of the higher current consumption while in sleep:

https://docs.nordicsemi.com/bundle/errata_nRF52832_Rev3/page/ERR/nRF52832/Rev3/latest/anomaly_832_89.html#anomaly_832_89

 

Could you try to implement this and see if you are able to reach the expected current consumption?

 

Kind regards,

Håkon

[Luca Nisti]

Good afternoon, I tried the following code with no success

    NRF_TWI_MNGR_DEF(nrfTwiMngr0, MAX_PENDING_TRANSACTIONS, TWI_INSTANCE_ID0);
    NRF_TWI_MNGR_DEF(nrfTwiMngr1, MAX_PENDING_TRANSACTIONS, TWI_INSTANCE_ID1);
    static const nrf_twi_mngr_t* const busToTwiInstance[I2C_BUS_N] =
    {
        &nrfTwiMngr0,
        &nrfTwiMngr1
    };
    
    

    nrf_drv_twi_disable(&busToTwiInstance[0]->twi);
    nrf_twi_mngr_uninit(busToTwiInstance[0]);
    nrf_drv_twi_disable(&busToTwiInstance[1]->twi);
    nrf_twi_mngr_uninit(busToTwiInstance[1]);

    *(volatile uint32_t *)0x40003FFC = 0;
    *(volatile uint32_t *)0x40003FFC;
    *(volatile uint32_t *)0x40003FFC = 1;
    *(volatile uint32_t *)0x40004FFC = 0;
    *(volatile uint32_t *)0x40004FFC;
    *(volatile uint32_t *)0x40004FFC = 1;

[Luca Nisti]

Håkon Alseth  I am happy to run a debug session and report o you register status if you need to understand what is the issue.

Thank you for support

[Håkon Alseth]

Hi,

 

If custom design; Is this repeatable on all your boards, or just one?

  

Luca Nisti said:

The board without mcu consume around 200uA.

Does this mean that you physically remove or power off the nRF?

Luca Nisti said:

Using the same firmware without initializing the I2C (twi0 an twi1 using easy dma) the consumption is around 250uA as expected and how we should want to have with the i2c disabled.

Could you verify that the SDA and SCL is set high, either by pull-up (via NRF_GPIO->PIN_CNF register), or externally?

If these pins are floating, you will see excessive current leaked.

 

Can you please share your sdk_config.h?

I want to explicitly see if you are using NRF_TWI or NRF_TWIM, by checking the defines "TWI0_USE_EASY_DMA" and "TWI1_USE_EASY_DMA".

If one or both are set to '0', try setting them to '1' and see if this has any impact.

 

Kind regards,

Håkon

[Luca Nisti]

Good morning Håkon Alseth 

1. We measured the board consumption removing the mcu from the pcb, plus it is consistent with calculation done by the hw engineer.

2. I checked the SDA and SCL pin status during the sleep mode and they are set correctly to save power

3. Attached the sdk config file. I can confirm we are using the NRF_TWIM and TWI0_USE_EASY_DMA and TWI1_USE_EASY_DMA

[Luca Nisti]

The results are consistent with different boards.

[Håkon Alseth]

Hi,

Luca Nisti said:

1. We measured the board consumption removing the mcu from the pcb, plus it is consistent with calculation done by the hw engineer.

Luca Nisti said:

The results are consistent with different boards.

Thank you for confirming.

Luca Nisti said:

2. I checked the SDA and SCL pin status during the sleep mode and they are set correctly to save power

Good to hear.

Luca Nisti said:

3. Attached the sdk config file. I can confirm we are using the NRF_TWIM and TWI0_USE_EASY_DMA and TWI1_USE_EASY_DMA

It looks like you have forgotten to attach the .h file?

 

Is the log module disabled? Should be sdk_config.h::#define NRF_LOG_ENABLED 0 to disable logging, as uart will cause the peripheral clock tree to run in sleep.

 

Kind regards,

Håkon

[Luca Nisti]

Here the file.

The NRF_LOG_ENABLED is set to 1 but module is not initialized at boot.

How does it explain the consumption when the i2c is initialized or not initialized?

4274.sdk_config.h

[Luca Nisti]

Hi Håkon Alseth .

I can confirm the high frequency clock is still running when the i2c is configured and disabled when firmware goes to sleep mode.

If the i2c is not initialized at all the high speed clock is off when in sleep mode.

Register checked is CLOCK->HFCLKRUN

[Håkon Alseth]

Hi,

 

Luca Nisti said:

Register checked is CLOCK->HFCLKRUN

This shows only if the task has been triggered:

https://docs.nordicsemi.com/bundle/ps_nrf52832/page/clock.html#d910e3076

 

Which it will, as the SoftDevice will use the external HFXO (start and stop when finished), as the radio is unable to run on the HF_INT (which has % tolerance)

 

Can you please test with setting NRF_LOG_ENABLED to '0' and see if this has any impact?

 

Kind regards,

Håkon

[Luca Nisti]

I can confirm there is no changes in power consumption, compared to the 2 cases above.

[Luca Nisti]

Håkon I think I've found the reason of the power consumption.

The hfclock need to be released. I though once no peripheral was using it, it was suspended but I was wrong.

Calling the nrf_drv_clock_hfclk_release when going to sleep fix the issue.

Thank you for your support.