How to check loop gain poles for HBSTB simulation

Hi,

I'm working a 60GHz multi-stage PA project and have some problem of checking its stability.

From my point of view, the K-factor method (Rollet-Criterion) doesn't work for multi-stage design.

For transient analysis, it has problem dealing with the snp files that respresents the transmission-lines and transformers from my EMX simulation. I have read many suggestions (e.g. 7 habits of successful S-parameters) of how to set nport. However, even I set my EMX from 0 to 300GHz (to cover 5th harmonic of 60GHz) with 100MHz step, the transient simulation cannot converge and exhibits oscillation over 100 times of my VDD, which to me is impossible to be true. I can try to further extend the EMX frequency but I don't think it is accurate above 300GHz.

For the pz analysis, I believe it cannot work with frequency-dependent component like nport (snp).

The HBSTB anlysis seems to be the most proper solution. I can successfully run it and get some useful results. For the simulated gain/phase margin, I'm not sure whether it can tell the stability as now the loop gain is a BPF function instead of low pass in the op-amp case. I followed the video in the cadence forum on how to create a Nyquist plot from the stb analysis. However, I disagree with the conclusion in the video claiming that stability can be told only by the encirclements of (-1,0). For Nyquist stability checking, we also need the information about poles of the loop gain. Therefore, I wonder whether HBSTB analysis can provide this loop gain poles information or is there any other (better) way to determine the stability in my case.

Besides, if there is anything wrong about my understanding to all these different analysis, please feel free to point it out.

Thanks.