Introduction
The Samtec Golden Standard is a reference structure for validating and debugging
electrical simulation software and high speed/high frequency test methods.
A simple mechanical design makes it quick and easy to import into analysis software,
yet the structure provides a rich set of easily observed electrical behaviors. Several
electrical characteristics can be defined by analytical solutions, so precise performance
benchmarks can be established.
The Golden Standard was designed with a special focus on validating coupled and
lossy models. It can be used to determine model accuracy for single-ended and
differential impedance, delay, S-parameters, and single-ended near-end and far-end
crosstalk.
The Golden Standard structure has been implemented into a physical PCB assembly
for use in validating laboratory measurements. The PCB assembly includes SMA
connectors for attachment to industry standard instrumentation. The Golden Standard
board is useful in evaluating new test procedures and measurement post-processing
techniques. The SMA and transition area of the assembly also lend themselves well to
analyzing de-embedding methods.
1.1 Reference Structure Overview
Learning to use new simulation or modeling software is often a daunting proposition.
Problem set up is usually unique from tool to tool, and terminology often varies as well.
The user is often faced with multiple options for parameters such as launch type,
boundary conditions, and convergence criteria. In some cases, solution run time versus
accuracy trade-offs must be considered.
After the user becomes familiar with the software and generates initial output, the
solution must be verified. Simulations are prone to a wide variety of potential errors at
many steps in the process. Some error sources are under the user's control, and those
are best understood by experimentation. Other sources of error are inherent in
numerical processing and underlying assumptions of the software and are thus beyond
the users control. But an understanding of these limitations, and the type and
magnitude of errors to which they are prone to introduce, is critical to efficient
application of various software packages.
Verifying simulation results against a well characterized and understood standard allows
the user to become familiar with the magnitude and sources of error in various
simulation methods. When the sources and level of errors are understood, the user can
exercise engineering judgment to determine when a particular simulation process is
“good enough” for the problem at hand.
A good reference structure should have parameters of interest which can be
characterized with analytical, closed form equations. For this reason, coaxial structures
or simple conductors over ground planes are often used as simulation standards. Such
standards can work well for accuracy determination in many common situations.
Coaxial standards can provide absolute numbers for certain parameters, and their
geometries are easy to construct in most modeling software. They can also be
manufactured with great precision. Such physical standards can interface directly to
calibrated instrumentation for correlation to lab measurements.
However, coaxial standards are not well suited for studying or validating simulations
involving coupling. And since the energy is almost entirely contained within the
structure, coaxial references do not lend themselves well to studying effects of varying
boundary conditions in field solvers.
To study and quantify these effects, a structure consisting of multiple conductors over a
ground plane is often used as a reference. Absolute accuracy of the analytical solution
might not be as great as in the coaxial case, but it is usually sufficient for the task at
hand. But such structures do not lend themselves well to physical construction and are
often difficult to connect to laboratory instruments for model or test procedure validation.
Well-characterized physical references are useful for more than just simulation
validation. Standards are also important for quantifying the accuracy of test procedures.
Testable standards are also necessary for “closing the loop” on complex model
validation.
For most complex structures such as high speed connectors, analytical solutions do not
exist. A typical method for validating a model for such a complex structure would be to
compare simulated data with test data obtained from a physical sample. However,
testing such complex structures often presents many potential sources of error. Thus,
an engineer can be caught in an endless loop trying to determine whether the
simulation or the measurement is “more correct”.
A good example of a potential source of errors in testing is the transition fixture which
converts from the coaxial test instrument interface to the connector under test. The
effects of such fixtures can in some cases be subtracted from the measurement using
mathematical normalization routines. If great care is taken in fixture design, structures
can be designed which can be accurately measured or modeled. If fixture
characterization is complete, the effects of such structures can be removed from the
measurement using de-embedding techniques. However, de-embedding methods are a
potential source of error as well.
A standard reference structure which can be easily used in both simulation validation
and in laboratory testing can be used as a “stake in the ground” upon which the
accuracy of both simulations and measurements can be referenced. Correlation
between a simulation process and a measurement procedure is first established with
such a “Golden Standard”. When acceptable agreement between the two is quantified,
an unknown structure of similar behavior can be expected to show similar agreement.
Samtec set out to develop a reference standard that addresses all of these needs.
Specifically, the goal was to develop a reference structure with the following attributes:
• Simple geometry for ease of model creation
• Accurate analytical solutions for several key parameters
• Sufficient losses for validating lossy models
• Multiple conductors for validating:
o Differential and single-ended impedance
o Differential and single-ended loss
o Differential and single-ended propagation delay
o Single-ended crosstalk
• Easily fabricated
• Easily tested in the lab |
