Inexplicable offset on internal temperature sensor of STM32F446

[Andreas Fischer]

Posted on January 04, 2017 at 13:19

I have a Nucleo-F446RE Eval Board with a STM32F446 RET6U. The internal temperature sensor shows valid values of about 34°C with constant cpu load.

With the same program code on a STM32F446 RET6 soldered on the target hardware the internal temperature sensor shows a value of about -200 °C.

Both µCs have a factory-calibrated temperature value in the corresponding register, which is used for the temperature calculation. At the Eval board and the target hardware all supply voltages are basically the same. Why is there an difference? At several identical hardware versions, the temperature sensor shows always this behavior.

[waclawek.jan]

Posted on January 04, 2017 at 13:23

What are the raw readings and the calib values?

[Andreas Fischer]

Posted on January 04, 2017 at 13:42

Nucleo-F446RE Eval Board STM32F446 RET6U:

 TS_CAL1: TS ADC raw data acquired at temperature of 30 °C, VDDA= 3.3 V Reg: 0x1FFF7A2C Val: 0x04B003B6 Reg: 0x1FFF7A2D Val: 0x0104B003  TS_CAL2: TS ADC raw data acquired at temperature of 110 °C, VDDA= 3.3 V Reg: 0x1FFF7A2E Val: 0x400104B0 Reg: 0x1FFF7A2F Val: 0x66400104   STM32F446 RET6 soldered on the target hardware:  TS_CAL1: TS ADC raw data acquired at temperature of 30 °C, VDDA= 3.3 V Reg: 0x1FFF7A2C Val: 0x04AD03AF Reg: 0x1FFF7A2D Val: 0x0204AD03  TS_CAL2: TS ADC raw data acquired at temperature of 110 °C, VDDA= 3.3 V Reg: 0x1FFF7A2E Val: 0x400204AD Reg: 0x1FFF7A2F Val: 0x66400204

[Andreas Fischer]

Posted on January 04, 2017 at 13:58

ADC Data Register

0x40000000 + 0x00010000 + 0x2000 = 0x40012000  Offset: 0x40012000 + 0x4C = 0x4001204C

RAW Data of ADC data register of STM32F446 RET6 on target hardware (several read outs):

0x000000F60x000000F70x000000F10x000000EF0x000000F40x000000F30x000000ED   RAW Data of ADC data register of Nucleo-F446RE Eval Board with a STM32F446 RET6U (several read outs): 0x000003CB 0x000003CF 0x000003D1 0x000003CF 0x000003D0 0x000003CF 0x000003D1

[waclawek.jan]

Posted on January 04, 2017 at 14:07

The raw data are way off. They should be around 0x3Bx.

Don't you use the ADC in non-12-bit mode?

[Seb]

Posted on January 04, 2017 at 17:34

If the software is the same, other possible root cause could be :

- If the package has Vref supply pin, it might be different voltage than Nucleo (or not connected)

- The supply voltage of the MCU is unstable or has big ripples

- The pinout of the 2 boards are different and the ADC channel is wrong.

-  Quick oscilloscope probing of all the pins on Nucleo (including supplies and oscillators) vs Target board might show clues

Good luck!

[Mert Karadeniz]

Posted on December 11, 2017 at 13:53

Hi,

I have the same problem and I tried the everything that is mentioned above. Is there anybody know what the actual problem is?

Kind regards,

Mert.

[waclawek.jan]

Posted on December 11, 2017 at 15:02

What are the raw readings and the calib values?

[Mert Karadeniz]

Posted on December 11, 2017 at 16:46

Hi,

Calibration values

TEMP30_CAL_ADDR = 0x3b1

TEMP110_CAL_ADDR = 0x4ad

Raw values

0x492

0x48e

0x491

0x492

0x493

0x48e

[waclawek.jan]

Posted on December 11, 2017 at 16:53

That's credible. What's the supply voltage or VREF?

JW