LP-MSP430FR2476: MSP430FR2476 Temperature Sensor Calibration Coefficients are incorrect.

Hi Hunter,

1) It appears that this bug was reported previously since the MSP430FR2476 can support up to 105C internally and the code was just transposed across multiple devices with different maximum temperatures. I will check with the team to see the status of updating the code. https://e2e.ti.com/support/microcontrollers/msp430/f/166/t/931441?MSP430FR2476-MSP430FR267x-Demo-ADC-Sample-A12-Temp-and-Convert-to-oC-and-oF

2) For the ADC calibration coefficients stored in the Memory Browser, the data is stored in little endian (LSB first), so if you look at the data in 8-bit Hex - TI Style, it should make more sense. 

In terms of your last question, could you clarify with an example by what you mean "the equation uses these calibration values is as y coordinates on the slope of the transfer function for the temperature sensor" and "the slope of the function is negative"? Although the code is using an 85C calibration value, the equation implemented in line 142 of the code example matches the equation in section 1.13.3.3 of the User's Guide.

I'm basically rewording the section 1.13.3.3 as I understand it. If i'm reading it correctly, the TLV values are stored in mV and are stored as the measured voltage at two temperatures, those two points are 30C and 105C. 

When I read the memory for 1.5V Reference i get the data below. If we use all decimal in the give equation, do you agree that 2108 and 1350 are the correct values based on the endianness and memory locations?

This indicates that the calibration value was lower (less mV) at 105C than it was at 30C. This is what doesn't make since. When I heat up the part, the ADCMEM0 value goes UP. So one would expect the 105C calibration value to be bigger than the 30C calibration value.

The same document indicates in section 21.2.7.8 that the expected transfer function has a positive slope. Where the 105C reading is higher than the 30C reading. Unless I'm interpreting the calibration data wrong, the 105C value is less than the 30C value, which is wrong.

With the example code the equation returns approximately 30C no matter whether the part is heated up or cooled down.

If you plot the equation using CALADC_15V_30C = 2108 and CALADC_15V_105C = 1350. You get a negative slope.

Is this macro somehow not reading the memory as expected? Seems that it matched the Memory Browser and the Expression Window.

#define CALADC_15V_30C *((unsigned int *)0x1A22) // Temperature Sensor Calibration-30 C

Hi Hunter,

Apologize for the slight delay, I was consulting with the team in regards to how the internal temperature sensor works and what the expected behavior for the behavior should be in regards to internal die temperature and ADC memory value. 

It is expected for the ADC value to be related to the change in temperature, therefore, you will see the ADC value increase if the temperature increases. 

According to the User's Guide, the equation in Section 1.13.3.3:

V_sense = TC_sensor * Temp + V_sensor is what calculates V_sense, the sense voltage that gets converted into an ADC value based upon the 1.5-V reference voltage. The slope of the equation, TC_sensor, is positive, which explains the positive slope between ADC value and temperature. 

In regards to the temperature equations we have been referencing, the ADC calibration values at 30C and 105C are just correction offsets, meaning that the offset values are linearly interpolated because offset values inside of the MCU ADCs varies greatly with temperature. These equations calculate what the internal temperature should be based on the offsets of the room and high temperature. 

In regards to seeing temperatures at around 30 degrees no matter how much the part is cooled down or heated up, could you provide some test points / screenshots of the Expressions window at a high temp and cold temp? For instance, I see temp = 2092.0 and IntDegC = 103.41 in the screenshot, what temperature was this test with?

Hi Aaron,

Thanks for the help. These are the captured values. Can you demonstrate the calculation?

Calibration Values: 30C = 2108, 105C = 1350. All values in decimal

I don't have this part but I checked on a FR5969 Launchpad and found calibration values of: $1a1a - 2721 and $1a1c - 3184. This is consistent with the ADC count increasing with increasing temperature.

From the guide the correction would be: (ADC(raw) - 2721) * ((85-30)/3184-2721)) + 30

There appears to be a problem with your correction data. Have you checked the CRC for the TLV data to verify it is correct?

David,

Agreed. Your calculation and example confirms how I expect the calibration values to be interpreted and the equation to work. So it does seems like my calibration data is wrong on this unit.

I believe however my data set passes CRC check.

It's not unusual for (early) Launchpads to be populated with pre-production "X" devices ("XMS", or "XFR" in this case), and for "X" devices to have a few quirks around the edges.

Hi Hunter, 

What is the lot code number of your device? Please see section 5 of the Erratasheet to find the lot number on the MSP430FR2476 device.