ACWWong
|
I think your issue is that your input tones are too small, so the intermodulation products on the output are somewhere near the simulator noise floor. Try increasing the input signal levels (upto 10dB backed off from Gain compression point) to say 10mV and try again. As you have now increased the tones by 20dB, the third order distortion product will now increase by 60dB and should now be big enough to give you a stable reading given an adequate tstab (check tstab on a transient run). Anyway you would probably want to do a couple of signal levels to confirm a consistant IIP3, say for a 100mV gain compression, i would be trying say input tones of 5mV, 10mV and 20mV to confirm the third order linearity. If you needed to lower the pss numerical noise floor use "high order" option at the cost of simulation time. Transient and dft should give the same result, and with good use of time step control and dft point you can get very low numerical noise floor. for the differences between pss and qpss check out the cadence documentation int he instatllation path, or spectre -h pss or qpss for qucik summary.
cheers aw
|