Hi,
Consider the following two circuits, being an integrator with identical component values (R, C, OpAmp impedances, gain, ...) in its single-ended and fully differential version (the amplifier is a behavioral model basically just a VCVS):
Ignoring the nonlinear resistance in the circuit and I confirm with transient simulations that both give the same identical numerical results up to numerical precision.
However, then I add the simple nonlinear resistor (just a pvccs) described by
i = g1*v + g3*v^3
Note that there is no even order term! Now the fully differential circuit gives an error (compared to the linear version) that is exactly (up to numerical precision) a factor of 4 smaller than the single ended version.
This can't be coincidence. Why?