mpc5674 temperature

[damiencarroll]

The temperature reading for the MPC5674 seems to have changed for one with a batch code 2016 versus ones we have with a date batchcode 2014 or earlier.

When powered up the processor internal temperature is measuring 70 degree C with 2016 batch code whereas the earlier ones 2014 measure 40 degree C which agrees with our thermal camera.

It seems the coefficients in the register are less for these newer 2016 processors which causes the temperature to be calculated higher as the ADC feedback is the same for all batch code MPC5674's we have.

Has anything changed in the thermal calculation formula or in the process for putting these coefficients into the register that is causing the difference?

[tsairo]

We have just recently found that some individual MPC5674F processors measure out-of-spec temperature sensor values. Preliminary investigation of one unit shows that the saved calibration values are quite noticeably different from a normally behaving unit, even though the raw ADC values from the sensor and the bandgap reference are relatively close between the units in room temperature (in 14-bit resolution about 5900 for the sensor channel 128, and 3990 for the bandgap channel 45). The difference in calibration values leads to 'normal' units showing a calibrated temperature of around 30 degrees Celsius which is rather close to external package temperature, and the misbehaving one showing roughly 5 degrees Celsius.

Comparison of the factory calibration data at 0xFFFEC000:

Failing unit: 9F69 1883 CFCC 

Normal unit: 9E84 1689 CFA6

So, for both units the the calibration points are indicated to be Thigh=145 and Tlow=25, but the calibration values are higher for the failing unit - especially the value at Tlow is significantly higher.

For the record, the inspected failing unit has following marking on package:

SPC5674FMVY3
3M17W
QHT1326G

Are there any previously known  problems with the temperature sensor calibration data?