STM32 ADC Internal Calibration - how reliable is it?

[unscripted]

The STM32L4 ADC includes a calibration step, with HAL it is performed with function HAL_ADCEx_Calibration_Start. When I skip this step I get better accuracy than when including it.. Why could this be?

[AndyC_772]

IIRC in some devices there's a required time delay between enabling the voltage regulator which powers the ADC, and the ADC actually working correctly. If your code doesn't include a pause, it can end up running the calibration before the power supply to the ADC is stable.

This is, not surprisingly, worse than not calibrating at all.

[unscripted]

IIRC in some devices there's a required time delay between enabling the voltage regulator which powers the ADC, and the ADC actually working correctly. If your code doesn't include a pause, it can end up running the calibration before the power supply to the ADC is stable.

This is, not surprisingly, worse than not calibrating at all.

I took that time delay into consideration. My micro has an external +VREF pin so I'm using a precise voltage reference. I wonder if even using this external ref I still need to perform the micro internal self-calibration.

[uer166]

I used HAL for the self-cal routines and always found the results to be better after the calibration, than before. The ADCs aren't terrible, but not great either, so a little extra performance is welcome with the cal. I have no idea what it actually does internally, but it seems to work..

[raw1]

Dear Unscripted,
 I am running into the same issue now on a STM32L471RGT6. did you ever resolve your issue and if so can you share any details?

Thanks a lot

RAW1

[wek]

If ADC conversions are better without calibration than with calibration, then you either did something incorrectly during calibration, or, more probably, do something incorrectly during the conversions.

VREF+ instability is one of the candidates, as said previously; too short sampling times are one of the other possibilities.

JW