2
\$\begingroup\$

I’ve some different ADC channels, extended/standard channels and Precision channels.

So far I know should the TUE (total unadjusted error) be calculated like this: \$ TUE = \sqrt{e_{offset}^2 + e_{gain}^2 + e_{DNL}^2 + e_{INL}^2} \$

For the 12bit ADC datasheet, that gives:

\$ TUE = \sqrt{(6^2+4^2+1^2+3^2} = 7.87 \$ (If looking at the datasheet it must correspond to the extended channel without current injection)

and for the 10bit:

\$ TUE = \sqrt{0.5^2+0.6^2+1^2+1.5^2} = 1.96\$ (If looking at the datasheet it must correspond to the precise channel without current injection)

This is confusing me, but my guess is however that the stated Offset error, Gain error, INL and DNL corresponds to different channels in each of these datasheets. Datasheets ADC 12bit and 10bit The 10bit Datasheet (see page 80-87)

Now if I’ve thought right above : I do wonder how I can calculate the compensated maximum error after adjustments/calibrations for the Offset error and Gain error?

My guess is that I could just simply take away the offset error and gain error for the TUE equation for the 12 bit “extended channel without current injection” and the 10 bit “precise channel without current injection”.

But how do I do for the other channels which don’t have the Offset error, Gain error, INL and DNL stated? How do I calculate the maximum error after calibration/adjustment for them?

I hope someone can be kind and help me.

Best regards

\$\endgroup\$
6
  • \$\begingroup\$ To me, your calculations are far too scientific. In the real life, the worst case is always considered (and far too often it is what happens). So do not add squares, just sum errors. Then, what is "input current injection"? I have never seen anything of this kind on ADC datasheets. You need to explain it or give a link to datasheet. \$\endgroup\$
    – Master
    Commented May 31, 2016 at 9:58
  • \$\begingroup\$ And one more comment: your ADC specs are really bad. You can easily find 10 bits / 12 bits ADCs with less than 1 LSB INL and less than 0.25 LSB DNL on the market, they are about 2 USD. If you really care about precision - just use them and focus on drifts. \$\endgroup\$
    – Master
    Commented May 31, 2016 at 10:01
  • \$\begingroup\$ But if I just sum the error sources the total error will be higher than for the TUE equation? “input current injection” is the leakage current that can occur between channels if the input channel voltage is higher than the ADC reference voltage. This do not affect the current measured ADC channel but it will affect the other channels that will get an induced voltage on its channel. This may give those ADC a higher ADC reading than it actually should be. ----------------- to be continued (next comment)---------------- \$\endgroup\$
    – user103776
    Commented Jun 1, 2016 at 10:57
  • \$\begingroup\$ -----------Continued-------- I don’t remember where I found that explanation but if I’ve understood it right that would be the explanation. The datasheet link has been added just under the snap of the datasheet. “The 10bit Datasheet (see page 80-87)” Yeah I agree a better ADC would be the easy way, but I can’t motivate an extra component if I can’t prove that I’ve done research that proves it won’t work with a calibrated ADC. \$\endgroup\$
    – user103776
    Commented Jun 1, 2016 at 10:57
  • \$\begingroup\$ You should not add DNL it is included in INL. And you forgot quantization. \$\endgroup\$
    – David
    Commented May 27, 2020 at 13:31

0

Reset to default

Browse other questions tagged or ask your own question.