This is really in the wrong board (it should be in the RF Simulators board) - but I expect Ken will take care of moving it.
I'm not familiar with the paper you mention, but some general observations:
- You don't need to use a port to use the RF analyses (just in case you thought that was needed)
- I don't believe (if my memory is correct) that when using timedomain (strobed) noise it produces input referred noise. However, you should be able to use Results->Print->Noise Summary to do that (I think)
- I suspect your refsideband should be 0. Think of it this way, it is used to describe the frequency that your input signal should be relative to the output frequency. So fin=k*PSSfund+fout. Even so, this only affects computation of input referred noise (or noise figure, if that was what you wanted in other circumstances); the output noise is unaffected.
- Setting the pac magnitude on the source isn't used here, because you're not doing a pac analysis.
You could also use pxf in sampled mode to compute the gain, potentially.
The clock frequency could affect the noise, depending on the shape of the noise sources or if there is any bandwidth limitations in the circuit, but small changes are unlikely to have a major impact.
Sorry for not directly answering your questions - I've not really simulated this type of circuit with SpectreRF or thought about it enough, so I'm just answering some of the obvious things that come to mind as I read your post.
Regards,
Andrew.