STM32H725 temperature sensors

[mihail_dobrev]

Hello,

I am trying to measure Junction temperature of the STM32H725IG device.

ADC3 is configured as an errata say to have only temperature channel 17 and continuous measurement. 

Temperature I see seems to be between 60 and 80 degree Celsius.

I tried also to enable Digital temperature sensor, expecting that both sensors will measure Junction temperature, but difference between readings is too big - around 12 to 15 degree Celsius less from DTS than from ADC3.

Is this normal behavior and which sensor I shall use in order to obtain  the real junction temperature, as it is needed to calculate the device lifetime? 

Regards,

[STTwo-32]

Hello @mihail_dobrev and welcome to the ST Community :smiling_face_with_smiling_eyes:.

Before using the temperature channel, did you calibrated the ADC3 to eliminate the offset error. If no, please refer to the page 1111 of the RM0468 and ensure to follow the calibration steps before any mesurement. For the DTS, calibration should be performed automatically before any mesurement except for the case of quick mesurement mode.

Best Regards.

STTwo-32 

[AScha.3]

Hi,

if looking at ds and basic values :

DTS

analog:

So a "basic" error at about (-7 and + 5 ) = 12° difference is "ok" . According to your tests.

Because the analog sensor needs the ADC to get a value (and added possible error here  (need run calibrate on ADC first and set recommended sampling time -- i think you did)) i would prefer the DTS at temp more around ambient.

So just do a check : set the cpu to a very slow clock, to have very low heat /dissipation , and look, what temp its giving then; should be only some degrees above ambient (on the board).

But from ds big tolerance if temp is high , compared to ambient :

The coefficient might be 30% off...so without a calibration this could be more "guessing" at 100° or more.

----

If clock at 25M , only 4...7mA , at 3v3 about 20mW or less dissipation:

So you could prove , the readed value is close to the real/true temp at ambient temp.

But on the analog sensor you have the stored calibration value for 110° :

So if you want to know the chip temp at high clock + dissipation, this should be the better (really tested on this chip) value, to check chip temperature at around this 110° .

[Amel NASRI]

Hi @mihail_dobrev ,

The article How to measure the junction temperature of an STM32H7 recently published should answer your request.

-Amel

[mihail_dobrev]

Hi,

I already used this article, but thanks anyway.

I tried also to enable calibration for ADC3 measurements.

Now ADC3 measures temperature that is lower, but  again I have a difference between Analog and digital sensors about 10-15 degrees. 

After calibration of ADC3 - now DTS temperature is higher than analog.

Both sensors I think should show close temperatures, depending where on the die they are placed - like closer to the M7 core or not. 

-Mihail 

[mihail_dobrev]

Thanks,

Will try , but as this is a whole system with a lot of software it is not easy to measure temperature on low cpu frequency.

-Mihail