The specified A/D datasheet performance shows a typical SNR of 69dB. Measurements on the EVM give values of 81.8 dB. Does this family of parts usually run that much better than datasheet values?

[LCE]

Maybe you have done the measurements with oversampling ?

Or some other calculation error?

Or you just checked SFDR?

The ~ 12dB difference look suspicious... ;)

[WBori.1]

Oversampling is not involved (that also was done as an independent measurement with the expected results of slightly less that 4 bits improvement).

We used ENOB = (20log(FSR/StDev)-1.76)/6.02 where in this case FSR = 2^12 and StDev = 0.332 from 1024 samples collected using an LM4140 to generate a 1.25V input on the 0 channel. (The mean sample value was 1542.03 **** as expected)

[LCE]

I'd rather use RMS values, like full scale RMS input voltage compared to RMS noise over the used bandwidth.