using TS_CAL1 value on STM32C011 seems to make calculated temp less accurate

[danmcb]

I'm using the internal temp sensor on a C011 chip *actually the STM32C0116 dev board) and trying to  get an approximate reading in C, using __LL_ADC_CALC_TEMPERATURE_TYP_PARAMS macro.

I understand that this won't be very accurate, but it seems that when I use the T_CAL1 value stored in ROM the result actually gets significantly worse. Here is my code:

#define VDDA_APPLI                     ((uint32_t) 3300)
#define TEMPSENSOR_TYP_CAL1_V          (( int32_t)  760)
#define TEMPSENSOR_TYP_AVGSLOPE        (( int32_t) 2530)
#define TEMPSENSOR_CAL_VREF            ((uint32_t) 3000)


int32_t get_last_temp_reading(void)
{
	uint16_t t_cal1 = *TEMPSENSOR_CAL1_ADDR;
 return (int32_t)__LL_ADC_CALC_TEMPERATURE_TYP_PARAMS(TEMPSENSOR_TYP_AVGSLOPE,
         //TEMPSENSOR_TYP_CAL1_V,
		 t_cal1,
         TEMPSENSOR_CAL1_TEMP,
         VDDA_APPLI,
		 raw_value,
         LL_ADC_RESOLUTION_12B);
}

My raw_value is 787. When I use the typical datasheet value of 760, I get a result of -20. If I use t_cal1 (which has a value of 1048) I get -134

What is going on? OK, I don't expect much accuracy, but using the internal cal value should at least make things better not worse?

[RobK1]

What temperature are you expecting to see?

The t_cal1 value is an ADC reading, not the value in mV. Multiply by 3000, the divide by 4096 to get the voltage.

[danmcb]

I expect to see somewhere near ambient - 25 to 30.

I am just plugging numbers into the macro. In all honesty I spent a good hour trying to relate the macro to the datasheet without success. it is not even clear to me what reference is used internally for the temp sensor.

[danmcb]

using the datasheet and trying to do it manually (screenshot below):

Avg Slope Code = 2.53 x 4096/3000 = 3.45

sense data = 787 x 3.3 / 3 = 866

TS_CAL_1 = 1048

Temp = ((866 - 1048) / 3.45) + 30 = -23

I am just putting the numbers into the equation, exactly as described, so something is incomplete or unclear here.

[RobK1]

Well, the 787 you quote as raw value does indeed correspond to -20°C, according to the calculations given in the manual. Using the t_cal1 value it's -22.8

I would expect to see an ADC count of around 1030 at 25°C

• Are you meeting the required minimum sampling time of 5µs?
• Are you meeting the required start-up times given in the datasheet?

[TDK]

In addition to these:

• Are you calibrating the ADC on startup, before it is used?

[danmcb · Accepted solution]

indeed, I think sampling time was the issue. I went back into CubdeMX, slowed down the ARC clock and the sampling setting to about 10us, and now I get results of about 70 degrees, which are believable.

I still cannot make sense of the supplied macro in the library, but if I do it myself, I get these numbers now.

Thanks very much for your help.