STM32L433 strange temperature sensor readings

cschumann · ‎2017-05-29

Posted on May 30, 2017 at 08:50

Actually I am working with a STM32L433CBY controller. When using the internal temperature sensor I got some inconsistencies and strange readings. Maybe someone can clarify the issues.

1. Reference temperature TS_CAL2

The

http://www.st.com/content/ccc/resource/technical/document/datasheet/f7/a0/fc/27/24/4e/4f/3f/DM00257192.pdf/files/DM00257192.pdf/jcr:content/translations/en.DM00257192.pdf

explains on page 40 both calibration points. TS_CAL2 is defined to be 130�C.

The

http://www.st.com/content/ccc/resource/technical/document/reference_manual/group0/b0/ac/3e/8f/6d/21/47/af/DM00151940/files/DM00151940.pdf/jcr:content/translations/en.DM00151940.pdf

introduces on page 423 (center) the formula for using the calibration points. Here TS_CAL2 is defined to be 110�C.

Any Idea which value is the correct one?

2. Oversampling

I get different values when enabling 16x oversampling for the ADC. At room temperature (~25�C, VCC=3.3V) I get an ADC reading (after a division by 16) of 0x2FE. When I deactivate oversampling I get 0x343 for the same temperature on the same hardware. Is there any limitation regarding oversampling of the internal temperature sensor? Did I miss something in the errata sheet? (PS: Tried it also on different controllers, same result.)

3. Unrealistic readings of the temperature sensor

When I fill in the values from (2) into the formula, I get quite strange results. First I converted the readings into millivolts to compare them to page 155 of the datasheet. The datasheet states that I should get about 0.76V at 30�C.

0x02FE ==> 620mV (much too low, even for 25�C)

0x0343 ==> 670mV (better but still to low!)

When I use these values for the formula (RM0394, pp423) with the calibration values

TS_CAL1 = 1031

TS_CAL2 = 1371

I first have to normalize the readings to 3.0V (since the calibration was done at that voltage).

TS_DATA = 0x343(=835) * 3,3V / 3,0V ==> 0x396 (=918)

T = (110�C-30�C) / (1371 - 1031) * (TS_DATA - 1031) + 30�C

= 80�C / 340 * (TS_DATA - 1031) + 30�C

= 0,235�C * (TS_DATA - 1031) + 30�C

= 0,235�C * -113 + 30�C

= 3,44�C (*brrr, coooold*)

When replacing TS_CAL2=110�C with TS_CAL2=130�C I get

T = 0,294�C * (TS_DATA - 1031) + 30�C

T = 0,294�C * -113 + 30�C

T = -3,2�C (*coooolder*)

Any ideas what's going wrong?

Carsten

#temperature-sensor #adc-stm32l4 #stm32l4

AvaTar · ‎2017-05-30

Posted on May 30, 2017 at 09:14

Can you tell what hardware (board) you are working on ?

I don't have any L4xx board, but I vividly remember some peculiarities of other discovery boards.

These boards could be powered by either USB (via the ST-Link part), external 5V, external 3.3V, and sometimes even via target USB. And all these supply sources were decoupled via Shottky diodes, which made for a 'floating' supply voltage, below the specified 3.3V.

If one used Vcc as reference for the ADC on those boards, similar problems arose. Perhaps applies to your case.

cschumann · ‎2017-05-30

Posted on May 30, 2017 at 09:25

The board is a custom PCB. Voltage is derived from a 5V Supply (bench PSU) using a linear regulator on the board. The 3V3 is quite stable.

Analog parts are decoupled from digital parts using a ferrite. Blocking capacitors are present near the uC according to the hardware guide.

When I did the tests, the only enabled peripherals were RCC (for PLL), ADC and DMA.