I am trying to simulate a frequency multiplier. The oscillators are simple ring oscillators oscillating at 2GHz made form inverters.What the multiplier circuit is that it cancels the fundamental tone and retains/preservs the 3rd harmonic(3,6,9...) and cancels the other harmoics like 1,2,4,5.......
I am interested in looking at the Phase noise at the combiner output oscillating at 6GHz. The output signal has a voltage swing of around 900mV. The setup does not have any ideal blocks in it. In the PSS options form I donot check the oscillator option (I read a post suggesting this in case of a driven circuit, which I assume this multiplier to be)
When I run pss using shooting method it does not converge. I have tried changing the tstab (increasing it), changing the method from trap to gear2 etc. But I can't get the pss to converge.
When I run PSS using the HB method the pss does converge but the Phase noise at the comes out to +150dbc/Hz which is a unrealistically large value (opposed to the theroratical results).
Running a pss+pnoise for a single ring oscillator works fine. I would be grateful if someone could guide me what could be the problem or if this is the right way of simulating this kind of circuit,
I cannot post the image so here is a link to the circuit diagram
A couple of thoughts came to mind - perhaps you have already explored these.
1. Are all three ring oscillators operating at exactly 2 GHz? The concern is that if they are not, one of the intermodulation products may have a very low relative frequency (i.e., for example 2 GHz/1.999 GHz => 0.001 GHz) which could make pss think the circuit has not settled.
2. Are you using an errpreset of "conservative" or setting the maximum integration timestep to a low value consistent with the expected maximum frequencies of interest? Since your circuit appears to be very sensitive to the creation of intermodulation products - whose frequencies will be higher and less than your fundamentalof 2 GHz - I was thinking you may need to force a maximum integration timestep to get the proper waveform fidelity.
3. Have you examined the converging HB waveforms and do they look at all realistic?
Thank you for your detailed update!
 - Great - this a good thing!
 - I've never needed to use anything less than a maxstep of 1/100 of the period of interest for stiff oscillators in pss in the MHz range. Hence, I suspect with your 2 GHz design a maxstep value of 1/100 is more than sufficient. Perhaps you might look at the step size being used in the pss transient analysis to determine how close to the minimum steps are to 1/100 of the the 2 GHz period.
 - Sounds good - this might provide some insight (hopefully!)