Periodic current spikes while sleeping when twis is enabled

[bearsh]

I need the nrf52 as a I2C slave, so I set up a very minimal example:

void i2c_event_handler(nrfx_twis_evt_t const *p_event)
{
    switch (p_event->type) {
    case NRFX_TWIS_EVT_READ_REQ:
        nrfx_twis_tx_prepare( ... );
        break;
    case NRFX_TWIS_EVT_READ_DONE:
        break;
    case NRFX_TWIS_EVT_READ_ERROR:
        break;
    case NRFX_TWIS_EVT_WRITE_REQ:
        nrfx_twis_rx_prepare( ... );
        break;
    case NRFX_TWIS_EVT_WRITE_DONE:
        break;
    case NRFX_TWIS_EVT_WRITE_ERROR:
        break;
    case NRFX_TWIS_EVT_GENERAL_ERROR:
        break;
    }
}

static void i2c_init(void)
{
    nrfx_twis_config_t config = {
        .addr = { 3, 0 },
        .scl  = PIN_I2C_SCL,
        .scl_pull = (nrf_gpio_pin_pull_t)NRFX_TWIS_DEFAULT_CONFIG_SCL_PULL,
        .sda = PIN_I2C_SDA,
        .sda_pull = (nrf_gpio_pin_pull_t)NRFX_TWIS_DEFAULT_CONFIG_SDA_PULL,
        .interrupt_priority = NRFX_TWIS_DEFAULT_CONFIG_IRQ_PRIORITY,
    };
    nrfx_twis_init(&i2c_dev, &config, i2c_event_handler);
    nrfx_twis_enable(&i2c_dev);
}

void clock_event_handler(nrfx_clock_evt_type_t event) {}

static void clock_init(void)
{
    nrfx_err_t err_code;
    err_code = nrfx_clock_init(clock_event_handler);
    APP_ERROR_CHECK(err_code);
    nrfx_clock_enable();
    //nrfx_clock_lfclk_start();
}

static void power_management_init(void)
{
    ret_code_t err_code;
    err_code = nrf_pwr_mgmt_init();
    APP_ERROR_CHECK(err_code);
}

static void idle_state_handle(void)
{
    ret_code_t err_code;

    if (NRF_LOG_PROCESS() == false) {
        nrf_pwr_mgmt_run();
    }
}

int main(void)
{
    clock_init();
    power_management_init();
    i2c_init();

    for (;;) {
        idle_state_handle();
    }
}

which seems to work so far. the chip wakes up, receives the i2c msg and is able to send back data.

but if I look at the current consumption, I see a periodic peek (700Hz) and I have no idea by what it may be caused:

if I don't initialize the i2c, it looks like this:

any idea?

[Stian Røed Hafskjold]

This is probably the chip operating in refresh mode. This is the lowest power consumption mode, and it works by regularly charging up a capacitor to provide power to the chip. When the chip consumes more current the spikes will occur more often because the capacitor is emptied faster.

[bearsh]

so this means I have somewhere a small leak? in the first picture the average current is around 30uA. so it would also mean that the pins, once configured as twi, leak most of the 30uA because if disabled, the refresh interval is much higher?

[Stian Røed Hafskjold]

You should get about 10uA when TWIS is enabled and waiting for an interrupt. Testing I've done here seems to indicate that this corresponds to a refresh spike frequency of 350Hz. But there are chip to chip variations, so this might be higher as well. To see if this is chip to chip variation or the code you are running can you try to flash the attached hex file, and see what average current you get?

nrf52840_xxaa.hex

Thanks

[bearsh]

Thanks for looking into this. The 10uA doesn't sound nice but at the end it explains the behavior.

I assume your hex file is for a nRF52840 DK? unfortunately I don't have that board, I only have a nRF52 DK. For the nRF52840 I have a DK and module from ublox.

[Stian Røed Hafskjold]

Yes, that was for the nRF52840. Here is one for nRF52832:

nrf52832_xxaa.hex