TMS320F28035: ADC Calibration and Total Unadjusted Error

HS WG-FAE said:

why is the ch-to-ch error not relevant?

Each channel will fall within the TUE of the ADC.  The ch-to-ch error does not contribute to TUE; it only describes the amount of variation that should be expected between channels while within the TUE.

HS WG-FAE said:

Calibrating gain error: Why is the NL +/-4 LSBs for the source error added? The 2.1LSB of the REF5030 should be enough.

The +/-4 LSB INL comes from mismatches in component values within the ADC itself so it would be independent of the external REF5030.

HS WG-FAE said:

Calibrating gain error: Why is the resistor divider needed? Can the REF voltage be directly used on the ADC channel?

Sampling REF (when driving VREFHI) directly will not allow detection of positive gain errors (when ADC conversion is higher than ideal) because the conversion will saturate at full-scale 4095.

Hello,
some further questions:
- the 2nd table in the last thread showed the calibrated values of the F28034. The column of the F2807x and F28004x is still shown uncalibrated. Then eg the Referece GE should also be +/- 0 LSB.
- F28034: A production calibration will not make sense when periodically calibration of offset and gain is implemented. Right?
- F28004x: A production calibration would eliminate resistor and reference errors. Then customer would disclaim the periodically calibration. Does this make sense? (Of course the temperature will not be eliminated).

Regards, Holger

HS WG-FAE said:

the 2nd table in the last thread showed the calibrated values of the F28034. The column of the F2807x and F28004x is still shown uncalibrated. Then eg the Referece GE should also be +/- 0 LSB.

Yes.  The old thread is focused on F2803x so it is assumed that only the F2803x is being trimmed.  There is also a secondary message from the comparison table, which is that the gross TUE improvement from in-system calibration of F2803x is much greater than the improvement that might be gained from in-system calibration of F2807x or F28004x.

HS WG-FAE said:

F28034: A production calibration will not make sense when periodically calibration of offset and gain is implemented. Right?

Correct, the in-system calibration will supercede the production calibration, although a production calibration could be used as a reference for gross-fault detection.  For example if the in-system calibration is more than X% different from the production calibration, then something may have gone wrong.

HS WG-FAE said:

F28004x: A production calibration would eliminate resistor and reference errors. Then customer would disclaim the periodically calibration. Does this make sense? (Of course the temperature will not be eliminated).

I'm not sure that I follow your question.  Do you mean that the production calibration would be more accurate because the resistor and reference values can be measured directly by a calibrated bench instrument?

I think that it would be beneficial to collect accurate measurements in production and store the values to flash so that the in-system calibration will have the benefit of knowing the real values of the resistor and reference.

in the 3rd question regarding the F28004x customer want to do only production calibration and abstain the periodically calibration. Through production calibration they could detect resistor and voltage reference tolerance.