OK, I have got some real world data.
This circuit is designed to provide a 0 - 2.5V output from a PT100 temperature sensor with 100Ω = 0V and 138Ω = 2.5V (it is nearer 2.6V but the gain can easily be adjusted).
The data and plot below shows what happens when I implement this circuit in anger. I have the reference input (TVREF) and the PT100 input (TVPT100). PTDIFF(2) is the output of the amplifier stage and Theoretical PTDIFF is the calculated value for what PTDIFF(2) should be with the component values selected. Unfortunately this plot is not large enough to show how non-linear the response of the circuit clearly.
This is really giving me a headache.
Thanks.