The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Transmission Lines and Other Distributed Devices >> conversion of .s8p into SpectreRF netlist
https://designers-guide.org/forum/YaBB.pl?num=1237187544

Message started by anirban on Mar 16th, 2009, 12:12am

Title: conversion of .s8p into SpectreRF netlist
Post by anirban on Mar 16th, 2009, 12:12am

I have a S parameter file in touchstone format (.s8p extension). I want to convert this into SpectreRF netlist such that I can use that for frequency as well as transient simulation. Can anyone help me regarding  conversion of .s8p into SpectreRF netlist. Thanks in advance.  :)

Title: Re: conversion of .s8p into SpectreRF netlist
Post by pancho_hideboo on Mar 16th, 2009, 6:22am


anirban wrote on Mar 16th, 2009, 12:12am:
I have a S parameter file in touchstone format (.s8p extension).
Nport of Cadence Spectre can read S-Parameters in Touchstone Format directly.
See "spectre -h nport".

Title: Re: conversion of .s8p into SpectreRF netlist
Post by anirban on Mar 16th, 2009, 9:05pm

I have already checked but I have not found how to generate equivalent model so that it can be used for transient simulation.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by pancho_hideboo on Mar 17th, 2009, 6:14am


anirban wrote on Mar 16th, 2009, 9:05pm:
I have already checked but I have not found how to generate equivalent model so that it can be used for transient simulation.
What on earth did you check ?
Did you read "spectre -h nport" surely ?
If so, it seems that you couldn't understand functions of nport at all.

See the followings.
http://www.designers-guide.org/Forum/YaBB.pl?num=1185454065/3#3
http://www.designers-guide.org/Forum/YaBB.pl?num=1227194040/8#8
http://www.designers-guide.org/Forum/YaBB.pl?num=1231076499/8#8
http://www.designers-guide.org/Forum/YaBB.pl?num=1174585355/1#1

After studying convolution, rational linear macro model fitting and identification as linear state space model, etc.
again read "spectre -h nport" surely.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by Andrew Beckett on Mar 20th, 2009, 7:58am

There's also an application note on sourcelink.cadence.com - search for "nport" and restrict your search to application notes. You should find          "Using the nport Component from the analogLib Library". It's a good overview of using nport.

Regards,

Andrew.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by pancho_hideboo on Mar 20th, 2009, 8:50am


Andrew Beckett wrote on Mar 20th, 2009, 7:58am:
"Using the nport Component from the analogLib Library".

Cadence has insisted on rational fitting(linear macro model fitting) about s-parameter in transient analysis and denied convolution over long long time.
But Cadence is denying rational fitting now.
Cadence recommend user to use convolution and don't recommend rational fitting now.

This situation is also very true for HB Analysis.
Cadence has strongly denied HB Analysis over long long time.



Title: Re: conversion of .s8p into SpectreRF netlist
Post by Andrew Beckett on Mar 20th, 2009, 10:15am

This isn't true. Cadence haven't "denied" these things.

Convolution was supported in transient for years - way before rational in fact. The problem was supporting convolution in shooting in PSS; it's not that trivial, and the rational approach was created precisely to solve that problem. In practice, the rational method was not fully developed, and in the meantime we had done further work to come up with a way of supporting convolution in shooting PSS. That happened in around 2004.

Shooting methods are (and still are) excellent for solving strongly non-linear periodic steady-state. For many years harmonic balance methods in other simulators could not handle anything other than very small circuits and very linear circuits - they failed to converge, or ran out of memory. That changed as some of the numerical techniques first used in SpectreRF were also applied to harmonic balance in other simulators - so other simulators became able to handle larger circuit problems very effectively and robustly.

It became apparent that SpectreRF could also implement a harmonic balance engine using the same numerical techniques, building upon Cadence's experience of iterative Newton solvers, thus giving users the choice of algorithms - to use either shooting or harmonic balance as appropriate.

All tools take time to develop and evolve and take advantage of new techniques. It's not a matter of "denying" anything - it's just a matter of prioritizing resources and doing development work. Sure mistakes get made along the way as with any business, but hopefully in the long run the right thing gets done!

This app note is from 2005, and harmonic balance was added to SpectreRF nearly 4 years ago now and has had considerable development since then. So even if Cadence had "denied" anything, that's not exactly current information.

With harmonic balance, using s-parameter files is simpler, because you're simulating a frequency domain description and directly simulating in the frequency domain.

(yes, I work for Cadence, so this is probably a biased view).

Regards,

Andrew.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by RFICDUDE on Mar 28th, 2009, 6:37pm

Ah, the endless battles between time and frequency with nonlinear simulation.

I recently was forced to waste company time on the nport model. The default convolutional interpolated model was interpolating a negative real resistance at very low frequencies. This caused the nport model to oscillate severly during transient simulation.
The solution was to switch over to the "rational" model; however, the settings had to be played with a bit to get a reasonable fit and model size.

Hopefully Cadence will invest more time in to refining the nport model options and fitting routines.

I am happy this method for handling distributed components is available, but robustness and accuracy are also important factors in using and trusting a model.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by pancho_hideboo on Apr 10th, 2009, 8:57pm


RFICDUDE wrote on Mar 28th, 2009, 6:37pm:
The solution was to switch over to the "rational" model; however, the settings had to be played with a bit to get a reasonable fit and model size.
It is very rare that "rational" gives good results compared to "convolution".
However we have no choice except for "rational" in Envelope Analysis in Cadence Spectre.

I don't use Cadence SpectreRF mainly for RF circuit.
But due to lack of Agilent licenses, I often have to use Cadence Spectre reluctantly.

When I run Envelope Analysis in Cadence Spectre using following S-parameter Library,
I often encounter simulation problems such as blow up or fake oscillation.
http://www.murata.co.jp/cgi-spara/search.pl?primary=lqg15hn&lang=en
http://www.murata.co.jp/cgi-spara/search.pl?primary=grm15&lang=en

When I can get Agilent licenses, I resimulate same circuit using Agilent Envelope Simulator such as GoldenGate.
Of course I can get reasonable results very quickly.

I'm always very honest because I don't have to defend or promote any specific vendor's tool.

Title: Re: conversion of .s8p into SpectreRF netlist
Post by RFICDUDE on Apr 17th, 2009, 11:38am

Thanks, I plan on checking out Golden Gate very soon.

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.