On board HTS221 temperature sensor read values higher than expected

[Whoing]

Environment

- Board: B-U585I-IOT02A (STM32U585AIIXQ)
- STM32CubeIDE, FreeRTOS (CubeMX-generated), CMSIS-RTOS v2
- Driver: ST vendored HTS221 component driver (`hts221.c` / `hts221_reg.c`),
  untouched, accessed through the public `stmdev_ctx_t` of the object layer
- I2C2 @ standard mode, internal pull-ups, 1 Hz sampling in one-shot mode

Hi, application on the **B-U585I-IOT02A** discovery kit (STM32U585, FreeRTOS, STM32CubeIDE). The onboard HTS221 humidity/temperature sensor is driven by ST's vendored driver from BSP.

Symptoms that i am experiencing: Right after reset (cold boot), the HTS221 reports **~20 °C, then over the following minutes of operation the reading rises steadily and settles at **~30 °C**, while the actual room temperature stays ~20 °C.
The drift is smooth and monotonic (thermal-looking), not noisy or jumpy.

I have already verified that BSP driver conversion formula is correct. Also HTS221 internal heater is disabled.

My sensor configuration:

HTS221_Set_One_Shot(&hts_obj);
hts221_humidity_avg_set(&hts_obj.Ctx, HTS221_H_AVG_512);
hts221_temperature_avg_set(&hts_obj.Ctx, HTS221_T_AVG_256);
/* heater bit cleared in CTRL_REG2 */
HTS221_HUM_Enable(&hts_obj);
HTS221_TEMP_Enable(&hts_obj);

My current hypotheses is that  on board component heating up and cause sensor reading to be distorted. 
Has anyone encountered it and whether it is known bug?

[KDJEM.1]

Hello ​@Whoing ;

 

I recommend you to look at STM32StepByStep:Step4 Sensors usage - stm32mcu wiki and precisely section “2.2. Example: Getting temperature values using the HTS221 sensor and displaying them on a terminal” and follow the steps.

The purpose of this section is to provide a step by step description on how to interface with the HTS221 sensor in B-U585I-IOT02A discovery board to get temperature values and to display them on a terminal.

 

Thank you.

Kaouthar

[Whoing23]

Hello ​@KDJEM.1 ;
Thanks for the answer. I have used the BSP component and compared conversion formulas against the datasheet of the sensor itself. they are correct. So there is no problem on software side for sure. I would like to know the details of how board was traces. I want able to find information about whether it has a cooper layer or not. I suspect that heat from other components is transferred to the sensor, which cause wrong reading.

 

Thank you.
 

[TDK]

If board temperature increases, it is expected that the sensor will report this. The sensor measures the temperature at the sensor.

[Whoing23]

Common sense. I only wanted to determine whether it is a known problem with this  board, or this problem in my particular board

[Ozone]

>  Also HTS221 internal heater is disabled.

Are you sure ?

> The drift is smooth and monotonic (thermal-looking), not noisy or jumpy.

With a nearby souce heating the sensor up, the rise would be exponential (not linear), with a constant decrease of delta T until it reaches equilibrum.

Have you tried to deliberately warm/heat it, e.g. with your finger ?
Or cool it, e.g. with coolant spray ?

[Andrew Neil]

> The drift is smooth and monotonic (thermal-looking), not noisy or jumpy.

With a nearby souce heating the sensor up, the rise would be exponential (not linear),

But an exponential rise would still be monotonic: ie, it only goes up, not down.

[Whoing23]

But an exponential rise would still be monotonic: ie, it only goes up, not down.

Actually it stabilizes around 30 degrees and then hover around 30 - 32 degrees.

[Andrew Neil]

​@Whoing As ​@TDK said, an onboard sensor is going to detect the temperature around it - so is the board (particularly in the vicinity of this sensor) actually warming up ?

 

Do any of the other sensors also read temperature? If so, do they give consistent results?

 

Have you tried a separate sensor, placed in contact with the HTS221 ?

[Whoing23]

I have measured the board temperature with a thermocouple. Yes, it really is 30 degrees. The sensor and the board really heat up. I think that the board has a copper layer, so when other components heat up, the heat is transferred to the sensor, even though it is far away from the component. 

[Ozone]

That makes sense, copper has at least an order of magnitude smaller thermal resistance.

However, this is why those ST products are called “evaluation boards” - they are not made to fit a specific use case, but to demonstrate general capabilites and allow for prototyping.
For your application, you probably want a thermally detached sensor.