The result of this effort is the Samtec Golden Standard. The Samtec Golden Standard
has been used for several years internally at Samtec. It has proven to be an invaluable
training and investigative aid for evaluating simulation and modeling packages and
approaches and for evaluating test procedures and data post processing.
1.2 The Ideal Golden Standard Design and Construction
The Golden Standard is based on a two conductor coupled microstrip design. The
geometries were developed with consideration for future manufacturability. The design
is illustrated in isometric and cross section views below.
The structure can be described briefly as two lines of 0.075 inch width separated by a
space of 0.025 inch. Each trace is separated from the top ground plane by 0.025 inch
gaps. Conductors are considered ideal lossless copper, 0.0014 inch thick. The upper
and lower conductors are separated by a dielectric 0.059 inch thick. The total width is
1.00 inch. The total coupled length is 5.15 inches.
The ideal solution requires an infinite ground plane below and on both sides of the
microstrip traces, but very "wide" ground planes are sufficient. Three to four times the
width of the traces is found to be acceptable. To reduce the problem space for more
universal applications and to match with the physical design of the Golden Standard
board, the total width of the reference structure is limited to one inch.
A similar situation occurs with the upper and lower ground planes. In the ideal solution,
the upper and lower ground planes are equipotential. Again, in many field solvers, this
can easily be constrained.
In the Samtec analysis, material properties were chosen to closely approximate
standard FR4 PCB laminate. The values used actually aren’t much of a concern in the
virtual model, as long as identical values are used in both the analytical calculations and
the model.
1.3 Derivation of Golden Standard Analytical Solutions
In this paper, we focus solely on frequency domain characterization of the Golden
Standard. However, Samtec has performed many time domain simulations and
correlations using several commercially available software packages.
A theoretical discussion of transmission line theory is included in Part II of this
document. An in-depth discussion of the calculation of the analytical solutions can be
found there as well. An analytical solution can be obtained for the magnitude and
frequency of maximum coupling for near-end crosstalk (S31). See Part II, Reference (2)
for details.
For the Golden Standard, a maximum coupling of 13.3 dB at a frequency of 324 MHz is
calculated. The coupling maximum should repeat at multiples of 324 MHz. Phase
change will also occur in S31 at 324 MHz and related multiples. This provides several
easily discernable markers at low to mid frequency for comparing against simulated
data. It is interesting to note that the S31 marker is a function of the coupled line length,
and the added length of the SMA connectors and “Y” break out region have very little
impact on data from the Golden Standard physical boards.
An analytical solution also exists for the maximum coupling frequency for far-end
crosstalk or S41. Details are provided in Part II, Reference (3). The S41 marker occurs
at a higher frequency (8-13.6 GHz) in the Golden Standard structure, so it serves as a
good high frequency validation point.
An important feature of the S41 resonance is that its frequency is highly dependent on
the lossy characteristics of the dielectric material. For example, for a lossless dielectric,
the resonance should occur at 13.6 GHz. Losses tend to lower this frequency. For
example, using our assumptions for the FR4 material properties, we calculated the
resonance would occur at 8.3 GHz. Thus, the S41 marker is an excellent test for
validating material property assumptions.
The S41 marker can also be used as a validation of S21 data. Often, S21 measurements
of coupled line structures exhibit an unexpected, deep null. This is often described as
“anomalous” and sometimes erroneously attributed to impedance mismatch or possibly
radiation. Rather, this resonance is caused by coupling to the adjacent conductor
where the energy appears as far-end crosstalk. The two transmission lines are in
essence functioning as a broad band coupler. If the model is working properly, the S21
null will occur at roughly the same frequency as the S41 maximum. The slight
discrepancy between the maxima and minima is due to the even and odd mode
properties and losses in the substrate. See the following plot.
1.4 The Physically Realizable Golden Standard Design and Construction
Moving the Golden Standard from the ideal, virtual world to the real world presents a
more complex set of problems. Any manufacturing process introduces dimensional
tolerances which must be considered and quantified. The standard device must be
constructed from real world materials as opposed to ideal mathematical
representations. This introduces the significant challenge of characterizing the electrical
properties of the materials and establishing tolerances for those properties.
For example, with the Samtec Golden Standard structure, some critical physical
dimensions are conductor thickness, dielectric thickness uniformity, trace widths, trace
pitch, ground plane spacings, overall flatness, and length.
A similar challenge is faced in accurately characterizing the electrical properties of the
dielectric material such as loss tangent and dielectric constant. Nominal values are
usually attainable, but batch to batch variations can be much more difficult to quantify.
To establish correlation at frequencies above the 10 GHz range, Golden Standard
boards have been precision milled from tight tolerance substrate materials. Golden
Standard boards which are included in the free kit are manufactured using high end
commercial etch and laminating processes and tight tolerance materials. While
tolerances are not as tight as those obtained with a milling process, they have been
found to be acceptable for many correlation studies up to several GHz.
Golden Standard boards are constructed from FR4 laminate dielectric. FR4 is used in a
wide variety of applications and is readily available. Variability in the dielectric
properties of FR4 is not as well controlled as in some more expensive substrate
materials. Variation in dielectric properties can become an important factor in
differences between board samples at higher frequencies. Samtec Golden Standard
boards are constructed from substrate materials which are carefully selected to ensure
acceptable physical and electrical consistency from batch to batch.
One of the greatest challenges in constructing a physical board is in devising a means
to connect it to test instrumentation. For the Golden Standard boards, SMA connectors
are used as they are nearly universal. The transition between the circular coaxial
structure of the connector to the planar PCB microstrip structure can introduce effects
not predicted by the analytical solutions in some parameters at higher frequencies. |
