nRF52832 Temperature Sensor

[SteveHx]

We need to sense temperatures in the upper end of the operating range, specifically in the 40C to 70C realm. We have three different units, and we are finding that the temp sensor specs appear to be violated. Using a calibrated thermocouple monitor, with a thermocouple place right on the chip, one unit reports spot on, a second unit reports a bit over 5C error, and a third unit shows over 10C error. The ±5C spec is pretty sloppy to begin with, but 10C is unacceptable. The piecewise linear calibration registers are poorly documented. We are attempting to deploy a beta test of 50 units in the field. Are we expected to have to do an individual calibration on each unit at 50C in order to get ±5C or better accuracy?

[Turbo J]

Careful: Its pretty easy for an MCU chip to heat itself 10°C over ambient just with the CPU (or the radio). If you wanted reliable temperature readings, use an external sensor.

Recalibrating is a terrible idea, the temp values are used to correct temp dependent chip behaviour especially for the radio stuff.

[Priyanka]

Hi,

The ±5C spec is pretty sloppy to begin with

This mentioned here is the worst case scenario and mostly you should be able to get something less than this.

For better and proper measurements, I would suggest you to try using an external temperature sensor and yes, it's better to have batch wise (if not piece wise) calibration of the units.

Regards,

Priyanka