The determined data nicely follow the predictive model and thus easily fall within maximum possible error, which is unlikely to measure.

One notable difference between measurements and predictions occur at -40°C as seen with gain errors in Table 9-12 and Table 9-14. This can mostly be because the model's parameters assumed linear temperature coefficients; however, R_EXT can easily have non-linear, even parabolic, temperature coefficients that changed polarity going cold. Best efforts was made to get lowest prediction errors and future efforts can employ non-linear temperature coefficients. Tradeoffs were made so that prediction model at least matched the 125°C range.

The other high prediction error occurred with the V_{OS, EXT} at V_CM=0-V in Table 9-17. This can be a result of the change in I_{B, CM} slope when V_CM < V_S, which can be seen in Figure 2-2. If accurate leakage input bias current is required at this level please post an inquiry on our public forum for more information.