Distrortion issues as a result of the device physics?

I strongly suggest that you don't go there.

To a first order, most of your linearity issues should be seen using linearity tests (two tone IIP3 or similar) as a function of the circuit architecture, transistor model and bias point characteristics.  

A lot of what is seen there can be directly observed based upon the bias point, a  DC bias curve set, and the input signal amplitude. Dealing with all of the second order effects are generally not too necessary.

Will the BSIM4 (4.5 ??) reflect this stuff accurately? The model is very capable of it. ***However*** you got to remember that a lot of the model parameters are not properly set. The model definition is ok, but the numbers put into the model are not accurate. Need to discern between those two. They are very different things. A foundry can use the BSIM99.9 for 1picometer CMOS (just joking) but if they dont set the model values correctly, all the defaults get you back to a much simpler model.

Oh, and one other comment - Just because the system is differential does not mean that second order harmonics go away. Greatly reduced, yes. Eliminated no.


Jerry