Although the required input source impedance can be estimated assuming a 0.5-V initial error and exponential recovery during sampling (acquisition) time, this estimation over-simplifies the complex interaction between the converter and source, thus yielding inaccurate estimates. Thus, this design uses an iterative approach with the converter itself to provide reliable impedance values.
To determine the actual maximum source impedance for a particular resolution and sampling rate, two subsequent channels are set at least 95% of the full-scale range apart. With a 1xVREF range and 2.5 VREF, the channel difference is at least 2.375 V. With 2xVREF and 2.5 VREF, the difference is at least 4.75 V. With a source impedance from 100 Ω to 10,000 Ω, the conversion runs at a constant rate and a channel update is issued that captures the first couple samples after the update. This process is repeated at least 100 times to remove any noise and to show a clear settling error. The first sample after the channel update is then compared against the second one. If the first and second samples are more than 1 LSB apart, throughput rate is reduced until the settling error becomes 1 LSB, which then sets the maximum throughput for the selected impedance. The whole process is repeated for nine different impedances from 100 Ω to 10000 Ω.