Re: [SI-LIST] : Tools for high-speed interconnect design

About this list Date view Thread view Subject view Author view

From: Ray Anderson ([email protected])
Date: Tue Aug 08 2000 - 21:09:42 PDT


As you pointed out, digital designers like to see time domain waveforms
as the result of their simulations. A time domain simulator (a spice
variant, XTK, HP ADS or other time domain solve engine) is commonly
utilized. Inserting accurate models of discontinuities into such simulations
can sometimes be a challenging problem.

There are a number of ways such models can be developed:

Frequency Domain Methods:

From field solvers or from frequency domain lab measurements s-parameters can
be determined. As you mentioned, by means of optimization, a lumped (or
distributed) topology can be fit to the synthesized or measured data to create
a discontinuity model that can be quite accurate over some defined bandwidth.

Alternatively, the same s-parameter data can be mathematically manipulated
(extract poles and zeroes) to form a rational function (i.e. a ratio of
polynomials). Pade' functions are one common example. The derived rational
function can be implemented in a spice simulation as polynomial controlled G

A third method is to directly mathematically extract circuit element
values from measured or simulated s-parameter data. The derived circuit
elements (usually R's, L's, and C's) can then be used in a spice simulation.

One "gotcha" of the frequency domain methods utilizing measured
characterization data is that some sort of de-embedding must be performed to
try and separate the desired discontinuity data from the undesired fixturing
data. Sometimes a non-trivial exercise.

Time Domain Methods:

Methods exist to utilize time-domain (TDR and TDT) data that is measured in
the lab to create models that are suitable for time domain simulations.

A nice thing about the time-domain methods is that you can take advantage
of time windowing to extract the data relevant to the discontinuity of
interest and to ignore responses caused by fixturing.

TDR and TDT measurements allow one to extract parasitics, equivalent
circuits, impedance and delay, frequency dependent Zo, and s-parameters.

Methods that may be employed include: Model optimization, Peeling algorithm,
time domain network analysis, and rational functions.

The model optimization method is somewhat similar to the freq. domain
optimization method except one optimizes element values in a time domain
simulation instead of a freq. domain simulation.

The peeling algorithm is implemented in the Tektronix IPA-310 and 510
systems, the tool offered by TDA systems as well as Agilent (HP).

Time domain network analysis is a mathematical method currently
under development.

The rational function method is a time-domain adaption of the frequency domain
method of the same name.

I won't get into any of the details here. (it can get quite involved), however,
my colleague Madhavan Swaminathan of the Packaging Research Center at Georgia
Tech and his students have done extensive work on both time domain and
frequency methods of modeling and have published much of the work and have
several new upcoming pubs in queue on the subject.

There are many variants on the methods I've outlined, and I'm sure many
other methods I haven't mentioned here. The point is, once you have either
measured or synthesized characterization data for a discontinuity, there
exist numerous methods for introducing the data into a time domain
spice (or spice-like) simulations. Each method has it's pros and cons.
Depending on exactly what you are doing some methods may be preferred
over others.

-Ray Anderson
Sun Microsystems

> Hello, everyone:
> This topic has been brought up several times but I still want to have
> some idea from you as to what tools you use and when you use them.
> For digital designers, one of the major concern is the waveform they get
> at the devices pins. Such job is most efficiently done using traditional
> tools like HSPICE, XTK, etc., which can handle the active device models
> with non-linear characteristics.
> However, when the speed of signal get faster and faster, the
> discontinueties such as vias has to modelled properly. To get this
> done, full wave field solver such as Ansoft HFSS, which was
> traditionally aimed for RF/microwave/antenna design may have to be used.
> But the problem with field solver is that they usually does not support
> active device models (correct me if I'm wrong). My question is, how do
> you guys utilize results from field solver in your circuit simulation ?
> As far as I know, one approach is to build an equivalent circuit for the
> interconnect under study, optimize the RLC values so that the S
> parameters from the equivalent circuit match that of the HFSS
> simulation. Is this approach the common one used by SI engineers ? Any
> other approachs that you use and the tools you used ? Please share with
> me your experience.
> Thanks in advance.
> Perry

**** 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 - 10:51:00 PST