We are observing unexpected results for transient simulations with S-parameters. The models used for simulation are 'Amplifier2' and 'Mixer' from 'ads_behavioral' library. If the S-parameters entered into the models are real then the results are as expected, but if the S-parameters are complex, then the results are not as expected. We are suspecting error with simulations not considering the imaginary part of S-parameters. Is there something we are missing that is used to simulate complex values of S-parameters.
The test bench used for simulation is as given below:
Since S21 = polar(10,30) and other S-parameters are zero, we expected a gain of 10, with a phase shift of 30 degrees. But it can be seen in the below graph that phase shift is zero and gain is 10*cos(30) (only real part of S21).