Differential mixer simulation problems with 4.4.6

I am trying to simulate transient and conversion gain for a
differential Gilbert cell with Spectre.

In my testbench I have used a "psin" ports in RF, LO and IF terminals
(pos. node connected to an ideal balun and neg. node to gnd).

I have simulated the conversion gain both with PSS&PAC and PSS&PXF
analysis.

PSS&PAC provides about zero difference in the outputs (i.e. output
signal level at the IF psin port is somewhat -250dB), when the output
is taken as differential, before the balun, or single-ended after the
balun.

When taking the output as single-ended before the balun at the IF
output, the result is somewhat reasonable.

With PXF the result at the same point seems a bit too optimistic, the
difference is more than 10dB.

What to do and which simulation method to trust?

-Jack

 

I don't see any particularly good reason why PAC should give you different
answers from PXF, if you've set up the simulation the same way. If you have a
single ended input - with a PAC magnitude of 1V, and plot the output either
differentially, or single ended - then you should get the same results as PXF
where you specify the output as differential, or the same single ended output,
and plot the PXF results from the same input results.

If you're getting different answers, I suspect you've not set things up the same
way for the two different simulations.

Regards,

Andrew.