Re: [SI-LIST] : Via model

About this list Date view Thread view Subject view Author view

From: Todd Westerhoff ([email protected])
Date: Fri Aug 04 2000 - 09:01:51 PDT


You've got a good question, and a valid concern, but the answer is long and involved. Let me see if I can give the short answer coherently.

The return path - the thing you're ultimately concerned about - is a function of a lot more than just the via model. As a matter of fact. although the via model certainly has something to do with it, the real issue is the presence or absence of high-frequency decoupling between the signal's reference planes in the vicinity of the via. The via model itself could not predict this, and thus, wouldn't give you what you're after.

Analyzing return paths in the time domain requires a tool that properly models the power supply, decoupling capacitors and power planes, in addition to the trace being studied. That requires that the planes be modeled as a mesh with their associated non-ideal characteristics, such that the localized effects of power demand (i.e. switching current), decoupling and the transition of a signal between reference planes can be analytically predicted.

That also means that the modeling / analysis time would jump from seconds/minutes to hours/days, depending on the complexity of the model. As far as I know, there are no current commercial analytical solutions to the problem you propose - and that's not for lack of desire. It's because of the complexity of the problem.

The biggest question - and the biggest challenge - with simulation tools is understanding the limits of their accuracy, and in that light, the correct way to apply them to a problem, and the conclusions that can/cannot be drawn from the results. Most commercial SI tools make the assumption that reference planes behave ideally, and that trace behavior can be predicted from the 2-D cross-sectional characteristics of the trace and its distance from its respective reference planes. That assumption, and a few others, makes it possible to simplify the modeling problem to the point where circuit extraction and simulation can take place as an interactive process - point to a trace, and get a result.

If you want to get a more accurate answer, you can, but the cost is a substantial increase in the complexity of the circuit extraction, modeling and analysis processes. Ask anyone who's using a 3-D field solver in conjunction with SPICE what their process is, and how long it takes to generate a model and run analyses. It ain't seconds, and it ain't minutes.

So - the real question is - which approach do you use, and when? I would argue that you're best off using the interactive tools to study the problem and hone-in on a potential solution to a problem, keeping in mind the limitations of the tools, their applicability to the problem at hand, and the likely amount of error you expect in the analysis. Once you're close, you can use a more detailed, and lengthy analysis process to close in on the ultimate answer, should you need to do so.

In the particular case of return path analysis, I'm not sure that you'll find any commercial that satisfy your need. The alternative is to use an expert-based rules system to examine the design, checking against a known set of design rules. This isn't an analytical approach, but it serves to automatically and highlight areas that deserve thought (and possibly) detailed analysis using some other method. A common example is the case where a trace runs across a split plane with no localized decoupling (between the planes). The additional loop inductance caused by the discontinuity in the return path degrades the signal's edges and frequently causes the circuit to fail. The case you propose is similar - the transition between layers changes the signal's reference plane, and will have the same effect - increasing the loop inductance. There are commercial tools available that deliver the expert-based rules checking capability, which (at least) provide a fast way to highlight problem areas, even if they don't analytically predict the effect on the circuit.

  Anyway, that's my $0.02. Hope you found it useful.


At 10:02 AM 8/3/2000 -0700, Cosmin Iorga wrote:
>Does anybody know a field solver that can extract and simulate
>a via model from a board database? The via model should include
>both signal and return path if the transmission line changes the
>reference plane. The following picture tries to illustrate it:
> ______________________________
>|______________________________| <-- trace 1 (top layer)
> ||
> ___________________________ || __________________________________
>|___________________________ || ________________ground plane 1____|
> || ||
> signal via --> || || <-- ground via (return path)
> || ||
> ___________________________ || ___||_____________________________
>|___________________________ || ________________ground plane 2____|
> ||
> _||________________________________
> trace 2 (bottom layer)--> |___________________________________|
> The signal propagates through trace 1, which is a microstrip
> line referenced to ground plane 1, then through the via and
> trace 2, which is a microstrip line referenced to ground plane 2.
> The return path for the current consists of ground plane 2,
> ground via, and ground plane 1. The transition from the top
> to the bottom layer will look more inductive if the distance
> between the signal and ground vias increases. So far I've tried
> XTK and SPECCTRAQuest but none of them could extract correctly
> this via model.
> Best Regards,
> Cosmin Iorga
> (818) 874-7149
>**** To unsubscribe from si-list or si-list-digest: send e-mail to
>[email protected] In the BODY of message put: UNSUBSCRIBE
>si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
>si-list archives are accessible at

    Todd Westerhoff
    Product Marketing Director | High Speed Systems Design | Performance Engineering
    Cadence Design Systems | 270 Billerica Road | Chelmsford, MA 01824
    ph: (978) 262-6327
    fx: (978) 446-6798
    email: [email protected]
    internal information website:

**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected] In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at

About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Nov 22 2000 - 10:50:57 PST