STM32F373 SDADC Offset Error

[MaxMueth86]

We want to use the SDADC to measure the temperature of a PT100 sensor. The measurement should have a very high accuracy of +-0.5Kelvin. Therefore, it is important to have a low offset error. The datasheet lists very large values ​​for single-ended mode, e.g., 2000uV for VREFSD+=3.3. 

Differential mode cannot be used because the number of channels would then be insufficient (We need 18 channels, and the STM32F3 has only 11 differential Channels).

There is a product lineup where the offset error is specified as < 1 LSB.

Why is the offset error specified in the datasheet so high in single-ended mode ?
Can it be reduced by using a buffer op-amp at the input ( The buffer op-amp would mean a very low source impedance and would reduce e.g. bias errors ) ?

I have also read the AN4207 section offset calibration, but there are no details about single ended or differential ended offset errors after calibration

[KDJEM.1]

Hello @MaxMueth86 ;

Are you using Single-ended mode (zero reference) for gains 16x or 32x?

I recommend you to look at SDADC_TempMeasurement and check your software.

This example aims to show how to use the 16-bit resolution Sigma-Delta Analog-to-Digital converter to perform accurate temperature measurement.  The SDADC is configured in single ended offset.

In this example the temperature measurement is performed in two steps:

1. Temperature sensor calibration

2. The temperature measurement

Thank you.

Kaouthar

[LCE]

What's the targeted measurement range?

What kind of signal conditioning do you have in mind?

"High accuracy" usually uses some kind of calibration, the higher the accuracy the more on the user side, meaning like calibrating each sensor including the complete measurement chain, saving correction factors and offsets in EEPROM, ...

[MaxMueth86]

Hello Kauthar,

We want to use single-ended mode because the hardware should support six PT sensors with 3-wire measurement, and there aren't enough pins available for differential-ended measurement.

We don't want to calibrate the ADC inputs at 2-3 points, as we do with the evaluation board and evaluation software (that would require additional effort during the end-of-line test, and we would prefer to use an external ADC).

When calibrating with an external reference, the input offset of a maximum of 1800uV at VREFSD+=3.3V can of course be detected and eliminated.

If this input offset cannot be calibrated, 1800uV can already mean a deviation of several degrees Kelvin,

depending on the current flowing through the PT sensor.