I want to achieve the best linearity possible from the AD5791B at low frequencies, below 1k Hz, but I'm concerned about the buffer amp performance.
The AD8675 is recommended for the output buffer but it's minimum CMRR spec is only 114dB which equates to 2ppm. Would this not significantly degrade the AD5791's linearity of .25ppm typical, 1ppm max? The AD8675 datasheet doesn't have anything to say about the linearity characteristics when operating as a unity gain buffer. It's possible that the CMRR error could be linear thus manifesting itself only as a gain error, but some op-amps have very large linearity errors with changes in common mode voltages - e.g. AD8618.
It's not feasible for me to bench test this as the chances of me obtaining an 8675 with CMRR close to the 114dB minimum specification are low.
The ADA4077-1 looked a promising alternative with a minimum DC CMRR of 132dB, 150dB typical, although a bit slower and noisier. But fig 38 shows the typical CMRR as only 105dB from 100Hz to approx 3kHz! Surely the scale is in error? The ADA4522-1 datasheet is similar - 160dB typical DC CMRR but fig 30 shows typical CMRR to be only 110dB from 10Hz to 100Hz. It can't have lost 50dB in the first 10Hz and then lost 0dB to 100Hz.
On the other hand, the AD8675 typical CMRR spec is 132dB yet fig 20 shows the typical CMRR to be approx 10dB better at 142dB from 10Hz to 1kHz. Which figures are we to believe?