Hi Johnny
In the meantime i did some more measurements.
First I reduced the MCLK from 1MHz to 500kHz. The linearity error and also the steps are only half.
Looks like the change in the input resistance get smaller.
Furthermore I measured the linearity by using your EVAL-AD7767 input stage with the recommended 15Ohms serie and 2,2n to GND at each ADC input.
Linearity looks better but still away from 3ppm and steps are recognisable too.
Values are averaged over 128 samples. Due to the high Filter cut off frequency the signal is a little bit noisy.
Is there any chance to get the 3ppm or am I beating a dead horse?
Stefan