Ken,

You can try to read this:
"Computation of large-signal S-parameters by harmonic-balance techniques
Rizzoli, V.; Lipparini, A.; Mastri, F.
Electronics Letters , Volume: 24 Issue: 6 , 17 March 1988
Page(s): 329 -330"

As for me, I like the following definition: "a brutal attempt to force nonlinear circuits to obey linear circuit theory"  ;)

Hello Ken,
>
> I have difficulties to reply from the website you attached below, so I
> do it here.
>
> LSSP is originally for matching purpose, because s11 and s22 is major
> parameters determines operation point of PA. Also the harmonics
> distortion, and most important, the output power.
>
> Since you can measure S-para of PA at high power level directly, then
> we use an indirect format to measure S11 and S22, in this way, s21 and
> s12 is not measured. Actually, S21 is wrong anyway (or not valid) in
> testbench measurement.
>
> The indirect measurement is connect PA input and output with two
> slotted line stub, (it behaves like our port-Adaptor), then we tuned
> two slotted line to different power level, and then disconnect the
> slotted line, we measure the slotted line instead. On end of the
> slotted connected to source is terminated to 50Ohm, the other end of
> slotted line which is connected to the PA is actually s11*. We do the
> same thing for output.
>
> Since PA Z0 is usually working below 10 Ohm, so that the Smith chart
> is actually normalized to 10 Ohm or PA Zo instead of 50Ohm.
>
> The beauty of smith chart is only A Z-plot, but also a Reflection
> coefficient plot, and VSWR plot.
>
> The input and output impedance of PA is determined in this way. It
> cannot use to do stability or other small signal S-para  features. Of
> course there is way to extract the linear S-para from Non-linear one,
> but no body do that.
>
> The modern application of LSSP is mainly for modelling of PA (input
> and output Impedance modulated the behavior of PA) and give a clue of
> ACPR and IM3, because you can find a sweet spot of input and output
> matching condition that product maximum power, highest PAE and minimal
> distortion (ACPR), etc.
>
> To understand more, you can search MTT-S proceeding, conference, there
> are lot of stuff on this topic. Following I attach one for you.
> (loadpull a.k.a. LSSP)

i use LSSP in ADS very often (last job), and the measured result is quite good. it is for finding matching indeed. i may need to run PSS/PSP for this purpose in spectre.

There's now an LSSP wizard in ADE to set up the PSS simulations and post-processing
to output large signal S parameters. It's under Tools->RF->Wizards->LSSP; however, it
currently requires a "magic" setting in your .cdsinit file since it is in engineering release.

It's covered in the PA tutorial in the MMSIM60 USR2 installation -
<mmsimInstDir>/tools/spectre/examples/SpectreRF_workshop
The example there has the setting in the .cdsinit file needed.

I don't recall exactly when it was introduced, but it's there in IC5141 USR3/MMSIM60 USR2.

Regards,

Andrew.

Only large signal S11 generated by either ADS or SpectreRF LSSP simulation is useful. Large signal S22 would be useful too if it was calculated under input-driven condition (of course, in that case, S22 would be a function of input power, but so is S11), but both ADS and SpectreRF LSSP would set the input to be passive (meaning just 50 ohm termination) when calculate large signal S22. I don't know how other people make use of that S22.  :-?