ATF36077 S-parameters and non-linear model differences

Hi,

I'm trying to desing an oscillator using ATF36077. I noticed that when I use a non-linear transistor model (ph_hp_ATF36077_19940627 component or Statz model with parameters downloaded from Avago website) I get different results than when using S-parameters file (I have the same biasing - Vds 1.5V, Id 10mA).

When I simulate S-parameters for wide frequency range (1-18 GHz) the curves are quite similar - but there are slight differences. At the frequency of my interest (9.5 GHz) the differences are important - the same feedback circuit gives negative resistance in one case and positive in the other case.

I understand that the non-linear model is an approximation of the real thing, but in this case the differences are so big that whole design is completely unreliable. Am I missing something?

The schematic consists just of the transistor, input port connected to the gate, output port connected to the drain and source connected to GND. There is also a biasing circuit in case of non-linear model. Biasing voltages are applied through 1 uH inductors, so they shouldn't influence the results. And the outputs are connected through 10 pF capacitors.

Perhaps someone has an example schematic that compares measured S-parameters and S-parameters simulated using transistor model?

Best regards

Piotr