I am simulating a direct conversion mixer with LO and an RF signal that are very close in frequency. As a result, the simulator has a small beat frequency. PSS analysis simply fails to converge as long as the beat frequency is low. It converges just fine if the RF signal is taken away and only the LO signal is present.  

Can anyone suggest changes in my technique or other things I can do to successfully run the simulation with both LO and RF signals present. My end goal is to estimate IIP3 which requires running PSS simulation with LO and first RF.

Raz

Frank is right. QPSS is the way to go for the base analysis, with QPAC, QPXF, and QPnoise being used for the various small signal analyses.

You will also want to take a look at Accurate and rapid measurement of IP2 and IP3 at http://www.designers-guide.com/Analysis/. It describes a way of computing IP₃ that is much faster than the traditional approach. That paper describes the technique in terms of an LNA and PSS/PAC, so you will have to adjust the procedure to fit a mixer and QPSS/QPAC.

-Ken