How to compare what Spectre interprets from the raw s-parameter file?

Do you have access to Cadence Online Support?  https://support.cadence.com  

There is an appNote entitled  7 Habits of Highly Successful S-Parameters which I strongly recommend that you read.

7 Habits of Highly Successful S-Parameters 

How the S-parameters are handled by Spectre depends on the interpolation method chosen

• bbspice, a rational approach.   
• linear/spline, a convolution approach

For both approaches, an impulse response (linear/spline interpolation) file or rational fit (bbspice interpolation) file is cached which speeds up subsequent simulations. By default, nportirreuse is turned on (caching and re-using IR file). The nport impulse response or rational fit file has version control, so if the nport parameters change or the software version changes, a new impulse response or rational fit file is automatically generated.

Before I go into more detail than you'd ever want to hear <grin>, please download the appNote, review it, and let me know if you have questions.

best regards,

Tawna

Tawna,

Thank you for the appNote. I downloaded it.  Well written and informative. You wrote it.  Someone in my group told me bbSpice has issues with multiple ports (>10) and trouble with high frequency accuracy matching and to use linear for best correlation. What are the trade offs between rational and convolution interpolation? How do these work and what is your experience with both. I see there are 4 choices. What is the other rational method? 

My original question was is there a processed file that is viewable that I could  overlay and compare with Spectre fitting to the raw touchstone file.
Another question can you recommend a good EM extraction tool for easier fit? Are there differences?

I suspect that you the info about bbspice and a large number of ports is outdated - maybe that was the case in the early days of bbspice, but I've used it successfully with a very large number of ports. Similarly the accuracy is fine - provided that you have a reasonable range of frequencies in your s-parameter file.

bbspice is only suitable for passive networks; in general it's the best choice for such networks. Linear is suitable for other types of network. You should not use the older "rational" choice (it's pretty old and has not been touched in many years; in fact we are planning to remove the choice in a future release).

You can directly open Touchstone s-parameters in ViVA so you can compare them - and  you can also compare against the fitted parameters that bbspice produces. There is also a Tools->S-Parameter comparison in ADE which allows you to compare them rather than just comparing them visually.

As for an EM tool - that's probably not something we'd recommend. We have a number of connections partners who provide such tools...

Regards,

Andrew.

Hi Bobtest,

Thanks for the kind words.   

I agree with Andrew's comments.  I believe he's answered most (if not all) of your questions.  

I've successfully simulated with both bbspice and linear/spline up to a 405 port nport.  (That just happens to be the largest nport I recall dealing with.)  In general, bbspice handles large numbers of ports better than linear/spline.   The most important thing is to use the proper interpolation method for the DUT.  As far as tradeoffs, that's described in the appNote on pages 19-20.   

There is one situation where bbspice (any rational modeling approach) has dcOpt accuracy issues.  However, this is an intrinsic limitation of the rational modeling approach in general (not a problem specific to bbspice).  S-parameters are bad at modeling impedances far from the reference (Zo) and small relative errors can be amplified.   This is discussed in Article 20432379.  

Should you have more questions, feel free to contact Customer Support.  You are welcome to put my name in the Case cc list.

best regards,

Tawna