I don't know what simulator you use.

But "length" parameter exists in W-element of HSPICE which is absolutely industrial standard transmission line model.

Generally Nport is not converted to RLGC model.
See http://www.designers-guide.org/Forum/YaBB.pl?num=1237187544

Length information is used for generating RLGC model efficiently from S-parameter in W-element of HSPICE.

[S Model form of W-element]
Wxxx in1 <in2 <...inx>> refin out1 <out2 <...outx>>
+ refout <Smodel=modelname> N=val L=val

[Parameter Description]
 Wxxx : Lossy (W-element) transmission line element name.
 
 inx : Signal input node for xth transmission line (in1 is required).

 refin : Ground reference for input signal
 
 outx : Signal output node for the xth transmission line
          (each input port must have a corresponding output port).

 refout : Ground reference for output signal.

 Smodel : S Model name reference, which contains the S-parameters of the transmission lines.

 N : Number of conductors (excluding the reference conductor).

 L : Physical length of the transmission line, in units of meters.

sheldon wrote on Dec 24^th, 2009, 3:28am:

hi Sheldon ：
       Thank you very much! the tool I used is spectre. Now I think Nport is much better. In my s4p file , the minimum frequency point is 30Mhz,  is it too big to caculate the DC response?
                                                           Best Regards
                                                                   kenny

Kenny,

  A couple of comments:

1) Before using a component you might want to review the manual.
   It appears that the version of MMSIM you are using supports
   dc extrapolation, it is listed in the instance parameters. You
   might want to try the dc extrapolation --> unwrap.

2) If you concerned about dc extrapolation, then you might want to
   try MMSIM72. An additional algorithm for dc extrapolation has
   been provided along with some other new features.

                                                                   Best Regards,

                                                                       Sheldon

sheldon wrote on Dec 29^th, 2009, 8:16am:

Hi !
    Thank you so much!
    I have read the manual and try unwrap ,but it makes no sense. When I use ADS to simulate the transmission line with my s4p file ,the result is right , but in spectre not . The difference between in spectre and  ADS is the common voltage of output, it is lower in spectre .Is it problem of spectre?

Kenny,

  As previously mentioned, you should review the manual,
in particular, with regard to passivity checking (see item #1
below).

1) The check passivity function includes an option to enforce
   passivity. I believe that passivity enforcement has always
   been supported by the simulator.
2) As previously mentioned, there is an additional option
   for dc extrapolation. This option has worked for me
   when I have had common-mode level issues.
3) Please note that casuality enforcement is supported
   for the nport from MMSIM72

                                                 Best Regards,

                                                    Sheldon

Kenny,

  You need to do a quick check, spectre -h nport | grep unwrap. For
your case, you probably need to use the dcextrap=hpunwrap option.
Just checked my version IC61 and it is not a supported instance
parameter so you will need to manually edit the netlist and run the
simulation from the command line. This should address your dc level
issue. If it does not you should inform your Cadence support person.

                                                                   Best Regards,

                                                                      Sheldon

I used NPORTs in a commercial design. There was a balun that I didn't have time to optimize an accurate broadband lumped element model from the Momentum results. The Momentum dataset seemed to match the Cadence NPORT results, and the large signal SpectreRF results matched the AC/s-parameter results.

I have also used NPORTS to incorporate vendor s-parameter models of SMT inductors for matching networks.

The biggest problem area seemed to be multiport models of packages. The multiport models are only needed if coupling is a big issue (and it can be for some specs). So, I would only try to use these for isolation simulations and extracting lumped element equivalents for simplier models which don't capture the coupling effects.