Unfortunately I'm not measuring ideal resistors but rather coils so the impedance changes pretty drastically from 1KHz to say 20KHz. I would like to be able to accurately run a sweep like this and get accurate results.

For this I can understand that two hardware components needs to be altered a few times, these are the feedback resistor and the calibration resistor.

This brings me to my two questions: The feedback resistor is calculated according to Vpp*Rfeedback/Zunknown*PGA gain and this should fall within the ADC working range. Any other consideration here?

The calibration resistor is said needed to be close to the Zunknown. How much? What error will be introduced as the impedance is set further away?

No - we have not modelled this error. Too many variables.

Suggest multi-point gain factor calculations for each of the frequency increments as in eval software.

Also the main error is introduce by distance from calbration impedance - so if acuracy starts droppoing significantly another calibration step needs to be made.

A little bit of benchwork required to find your system ranges to fit accuracy requirements.

Liam