?Extracting [S]/[T] matrix of partial section from inside of simulation setup in HFSS

I have: Wave port 1(whole setup airbox boundary 1)-> [~50Ω matching microstrip]([S]o) -> [strip structure under investigation]([S]f) -> [~50Ω matching microstrip]([S]o) -> Wave port 2(whole setup airbox boundary 2)

Problem: I need to get ε and γ(or λ) of the section inside the setup. I am trying to get [S]f (or corresponding [ABCD]f (or call it [T]f) first.

Supposing, HFSS 13 gives me [S]s matrix of the whole 2-port setup:
I can't get correct values via [T]o^-1*[T]s*[T]o^-1

i.e. I am getting values of ε like 1E-12 instead of about 10 for example.
Excluding this To*Tf*To model HFSS also gives strange ε, λ results, but I was ignoring this due to stacked nature of setup.
I tried to simulate this problem analytically by cascading build blocks with Matlab and it proved OK.
But seems like HFSS's "results" matrix is not just a simple thing one could suppose...

Please, can somebody give me a hint, is there something special on HFSS's output [S]matrix or how do I need to change my way of solving the problem?

It would help if you post the HFSS project file here. I never used S-parameter models before but I will try to help.

Did you check that the S-parameter file was in the correct format? db vs linear? polar vs rectangular?

So, here's the most sane(most simple) sample
This project

On 10 GHz it gives matrix:
S = [ -9.862774E-003 + 6.824515E-003i, -8.291410E-001 + 5.356211E-001i; -8.291156E-001 + 5.356098E-001i, -1.001636E-002 + 6.588526E-003i]

I convert it with //D.A.Frickey IEEE trans. on MW V.42,N.2,Feb.1994//

and then there are two ways, use reciprocal:
β=±(j/len)ACOSH((A+D)/2), Zo=√(B/C)
or more detailed:
e^jβlen=(A+D)/2±√((A+D)²-4Δ)/2 ...
ε=(βc/ω)²
So, I get
β=285.3299662 + j1.451008798 m^-1 (actually, it must be Re(β)>0 & Im(β)<0, or vice versa... this is suspicious)
ε=1.853383138 + j0.01885077005
λ=2.202 cm
Zo=52 Ω


But HFSS gives me
β=316
ε=2.27 (more realistic)
and λ=1.99 cm

The difference is too large for the same matrix.
What am I doing wrong

//Actually this is just a first part of extraction, cause I need to simulate input, output strips and substract them from the whole simulation. i.e. [ABCD]in^-1*[ABCD]setup*[ABCD]out^-1.
//But numbers go mad this way.

//Interesting that rough ACOS(S21)/len gives correct signs of β(γ), but signs change when taking into account all S(or ABCD) elements.
//So I don"t have doubt in calculations, but have a suspicion that somethin wrong with data matrices, or maybe the way I use them.

^up

...please, tell me if you have an idea

:( nobody can't tell anything about those S-matrices...

....up....

up........

..........