PLECS is running into numerical truncation problems due to your choice of parameters:
- large DC offset, i.e. your filter input (1330)
- large time constant (2000)
- very small "Rel. perturbation for Jacobian" in the "Steady-State Options" of the AC Sweep (1e-6)
As a consequence, when PLECS tries to linearize the model, it gets a plain 0 for the Jacobian matrix.
If you increase the "Rel. perturbation for Jacobian" to 1e-4 (which is the default), you will get the expected result. But if you then also increase the filter input or the time constant by a factor of 100, you will run into the same problem again.