We are using an MKE18F512VLH16 MCU in our application. We have noticed variation in the ADC calibration results which results in measurement variation in the voltage of sensor inputs. We performed a test where we performed around 3500 calibrations on 4 different MCUs without powering down or resetting the MCUs to determine the range of values we get for calibration registers ADCx_CLPS, ADCx_CLP3, CLP2, ADCx_CLP1, ADCx_CLP0, ADCx_CLPX, and ADCx_CLP9. The data is contained in 4 tabs in the attached Excel file.
1) The software checks each of the calibration register vales according the Sec. 4.3 of AN5314 ADC Calibration on Kinetis E+ Microcontrollers so all of them are in range. We see what appear to be "outlier" calibration register values. Is the amount of variation we are seeing normal?
2) We have tried to implement the median calibration method described in the last paragraph of 16-bit SAR ADC calibration. We have several questions about this?
a) Is this method intended for the Kinetis E Series 12-bit ADC calibration?
b) The method appears to describe determining the median value of each register independently, so the resulting calibration would not necessarily be a coherant set of calibration values. Is this what is intended?
c) The median calibration method did not achieve much improvement in the ADC calibration variation. It only reduced the probability of observing the "outlier" calibrations but did not eliminate them. Is there any further guidance the NXP could provide.