Calculating Vybrid CPU core temperature with the internal temperature sensor

[bhuvanchandradv]

Hi!

At the moment the internal temperature sensor readings are calculated using the formula[ Temp = 25 - ( (Vtemp -Vtemp25) / m) ]

with optimal slope-coefficient constant at all temperatures provided in Vybrid datasheet, we observe a difference of ~20 'C from

the internal temperature reading and the external temperature on SoC package. We believe there could not be much temperature

differnce. When using the cold/hot slope-coefficients constants provided in AN3031[1] there is ~4'C difference between the internal

cpu temperature and the external temperature on SoC package. Can some one please confirm the temperature difference between

the internal CPU core and the external SoC temperature on package.

[1] https://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CBwQFjAAahUKEw...

Best regards,

Bhuvan

[jiri-b36968]

Hi Bhuvan,

1. Temperature sensor is specified at 3.0V VDDAD. Do you use 3.0V or 3.3V?

Data from 2 samples shows this voltage at 3.3V is slightly different.

Sample 6      0.69851V

Sample 1      0.70012V

2. Second think is ADC precision. For 10 bits 1LSB is 3.2mV. Which are 2-3 °C. But precision of ADC is less. For example for -+3LSB it will be +-5°C due to ADC. Presume that you use hardware filtering on ADC.

Internal temp sensor is dedicated for approximate internal temperature measurement - not for precise measurement. Anyway by calibration you can achieve precision in °C.

/Jiri

[bhuvanchandradv]

Hi Jiri,

> 1. Temperature sensor is specified at 3.0V VDDAD. Do you use 3.0V or 3.3V?

We use 3.3V as reference.

Seems the slope coefficients provided in AN3031 doc is also considering VDD as 3.0V.

How can we calculate the slope coefficients when VDD of 3.3V is considered.

There is no data available in An3031 doc for calculating slope coefficient with VDD as 3.3V.

> Data from 2 samples shows this voltage at 3.3V is slightly different.

>

> Sample 6      0.69851V

>

> Sample 1      0.70012V

I'm sorry, i didn't understand this.

> 2. Second think is ADC precision. For 10 bits 1LSB is 3.2mV. Which are 2-3 °C. But precision

> of ADC is less. For example for -+3LSB it will be +-5°C due to ADC. Presume that you use

> hardware filtering on ADC.

>

> Internal temp sensor is dedicated for approximate internal temperature measurement - not for

> precise measurement. Anyway by calibration you can achieve precision in °C.

Agreed.

Best regards,

Bhuvan

[jiri-b36968 · Accepted solution]

Hi Bhuvan,

Slope coefficient remains. Only offset at 25°C is little bit higher 696  -> 699 mV when 3.3V is used.

/Jiri

[bhuvanchandradv]

Hi Jiri,

> Slope coefficient remains. Only offset at 25°C is little bit higher 696  -> 699 mV when 3.3V is used.

Ok, thanks for the confirmation.

Best regards,

Bhuvan

[jiri-b36968]

Hi Bhuvan,

You are welcome - please close the thread by clicking on correct answer button.

/Jiri