nRF52840 Sleep, System-on, wake on LPCOMP - Unexpectedly High Power Consumption

[jumpingfool]

We have recently ported a design from the nRF52832 to the nRF52840 using SDK 15.2.

During power profiling we have found the new design with the 840 chip is pulling more power than expected during power-on sleep (approx 25uA vs an expected 4uA). This problem is not evident on the design with the 832 and there are no significant differences in either hardware or firmware, other than to support the change to the 840 SoC.

We have traced the additional power draw to use of the Low Power Comparator:

nrf_drv_lpcomp_config_t config = NRF_DRV_LPCOMP_DEFAULT_CONFIG;

config.hal.detection = NRF_LPCOMP_DETECT_DOWN;
config.hal.hyst = NRF_LPCOMP_HYST_50mV;
config.input = NRF_LPCOMP_INPUT_0;

err_code = nrf_drv_lpcomp_init(&config, lpcomp_event_handler);
APP_ERROR_CHECK(err_code);
nrf_drv_lpcomp_enable();

Removing the last line in the code snippet above to enable the LPCOMP reduces the power-on sleep current to the expected ~4uA.

The same issue was reported here some months ago by Darko but there was no response:

https://devzone.nordicsemi.com/f/nordic-q-a/37439/nrf52840-lpcomp-power-consumption/166996#166996

The nRF52840 product specification doesn't explicitly list the "sleep, system-on, wake on LPCOMP scenario" in the electrical specifications but we were expecting performance would be similar to the 832. Is this not the case?

Any suggested workaround to reduce the current draw when using the LPCOMP to wake the system from sleep on the 52840?

[Jørgen Holmefjord]

Hi,

The typical run current with LPCOMP enabled in nRF52840 is 20 µA. The difference from nRF52832 come from higher leakage current in nRF52840 when this peripheral is enabled. The current will be greatly reduced by putting the chip in System OFF sleep mode (~1 µA), if this could be an option for you.

What are you using the LPCOMP for in your design?

Best regards,
Jørgen

[Jørgen Holmefjord]

In fact, the LPCOMP is the same, and its isolated run current is about the same in nRF52840 as in nRF52832. It is the power domain that is enabled with the LPCOMP in system on mode that is much bigger in nRF52840. Unfortunately, there is no workaround for this.

How long is the pulse you are detecting? If you do not need hig sample rate, you could look into using the SAADC in low power mode, and use limits to only trigger interrupts when you go below the threshold.

[jumpingfool]

Pulse width is < 1ms so it is not viable for us to use continuous sampling via SAADC - we need the comparator.

This seems very odd, what's the point of having an LPCOMP on the chip if it's going to use almost as much power as the standard comparator? Having the LPCOMP peripheral onboard was one of the reasons we targeted the Nordic platform.

Any chance this may be improved in future revisions of the nRF52840?