What am I missing here? Thermistors are inherently nonlinear. Using a Wheatstone bridge or a current mirror won't change that. The reason Wheatstone bridges were so popular is that they made differential measurements easier in an age where dynamic range was limited and subtraction had to be done by analog means.
Configuring your thermistor with a precision resistor in a voltage divider, driven from the same reference voltage as the Arduino (or whatever) will give you equally good performance. If you want linearity, apply the Steinhart-Hart equation.
I had some issues correctly biasing (selecting the divider) for the voltage divider circuit on a previous project, so this was just an experiment with a different technique. But that project had other issues (resistive load sensors that were out of their load range), and reading into it now I agree that the voltage divider should be just as fine. I glanced at the Steinhart-Hart equation, and i'll give a shot at implementing it when I try to calibrate the sensors, thanks! --Andrew (Terg.is)
Configuring your thermistor with a precision resistor in a voltage divider, driven from the same reference voltage as the Arduino (or whatever) will give you equally good performance. If you want linearity, apply the Steinhart-Hart equation.