nRF51 TEMP sd_temp_get() missing codes, noise

[Mark Rages]

Hi,

I'm running a nRF51422QFAC and using sd_temp_get() to read the chip temperature twice / second.

I put it in a temperature chamber and swept the temperature from 0F to 100 F (18 - 38 C) slowly, over an hour. I saw some results that I can't explain.

Missing Codes

Here is a histogram of the temperature values from sd_temp_get():

It looks like some codes are missing entirely from the output, even though the temperature was swept slowly over the range.

Noisy Readings

Here is the sequence of the temperature readings from sd_temp_get():

There is some kind of weird noise in the readings.

Zooming in:

It looks like the count dips down occasionally. These dips are always 8 or 9 counts, never less.

Is this a known behaviour?

---

EDIT: Added call to sd_power_mode_set((uint8_t) NRF_POWER_MODE_CONSTLAT)

Here is the histogram. The same codes are missing:

Here is the timeseries. The same noise is present:

And zoomed-in, the noise looked the same:

[Susheel Nuguru]

Hi,

This could be because you are using XTAL cryxtal clock for 16MHz and the softdevice is sometimes using this crystal to calibrate its internal clocks and sometimes using internal RC HFCLK. The TEMP module is depending on which HFCLK it is using it.

Can you take your samples with setting constant latency on which keeps the HFCLK on always.

sd_power_mode_set((uint8_t) NRF_POWER_MODE_CONSTLAT);

This wil lincrease your power consumption by about 500uA I think but that is different thing.

[Mark Rages]

I tried that (see edited question) but the readings look the same to me.