The primary objective of this whole exercise is to get the best possible resolution out of the temperature measurement.

The procedure so far has helped establish the optimum values of Vfeed and Rfeed, in order to get the best resolution. There is one more avenue for possible improvement, and that is adding a shunt resistor across the thermistor. Normally this will simply reduce the overall performance. If you experiment with a few values (try 10000 to 100000) you will find it tends to “roll” the resolution curve to the right. It frequently also lifts the whole curve.

Nevertheless, it is still worth experimenting with, particularly with the AJ18 and the OEM36 using the main power supply (eg 24V) as Vfeed.