Simulation Guidelines for Wideband Ground Backed Coplanar Waveguide Transmission Lines

Full text

Simulation Guidelines for Wideband Ground
Backed Coplanar Waveguide Transmission Lines
Jay W. McDaniel
Advanced Radar Research Center, School of Electrical and Computer Engineering
University of Oklahoma, Norman, OK 73019, USA
Email: jmcdaniel@ou.edu
Abstract—This paper presents a series of guidelines to transi-
tion ground backed coplanar waveguide (GBCPWG) from the
design phase to finite-element-method (FEM) electromagnetic
simulations. A two step process is discussed including a design to
circuit simulation and circuit to electromagnetic simulation. The
majority of the paper focuses on the differences between circuit
simulations and FEM simulations including port definition,
side ground placement, and excitation of higher order modes.
Careful consideration of these topics are crucial to achieve ideal
performance over wide bandwidths, but also develop accurate
simulation models. A DC-20 GHz CPWG transmission line is
used as a design example throughout this paper.
Keywords—Keysight ADS, Coplanar Waveguide (CPWG), AN-
SYS HFSS, transmission line, wideband.
I. INTRODUCTION
The recent trend in modern wireless and radio frequency
(RF) system design is to integrate several subsystems onto a
single printed circuit board (PCB) through the use of multi-
chip module (MCM) techniques [1]. This is accomplished by
placing several bare die onto an underlying substrate and wire-
bonding to a transmission line that interconnects them. As next
generation systems continue to demand larger bandwidths and
greater sensitivity requirements, the need for wideband and
low loss transmission lines will continue to grow. Furthermore,
with the desire to implement systems on unmanned aerial
vehicles [2], unprecedented reductions in size, weight, power,
and cost (SWaP-C) will require the designs to also achieve
superior electromagnetic shielding effectiveness.
Ground backed coplanar waveguide (GBCPWG) has been
used extensively in circuits as an alternative to microstrip
(MS) transmission lines. This is primarily due to GBCPWG
being less radiative than MS offering lower loss and greater
isolation. The latter being critical to reduce electromagnetic
interference (EMI) within the system. In addition, frequency
dispersion is typically small resulting in better impedance
control and wideband performance. While several papers de-
tail the design equations of GBCPWG [3], impact of via-
stitching placement [4], and parallel-plate mode suppression
[5], very little literature is available in regards to the process
of transferring designs from circuit-based simulations to finite-
element-method (FEM) electromagnetic simulations. Improper
simulation setups and can lead to undesirable and/or incorrect
Fig. 1. Ground Backed Coplanar Waveguide geometry.
simulations results, which might not be caught until first test
articles are measured.
In this paper, differences between circuit and FEM simula-
tions, and their affect on simulations results, is discussed. An
emphasis is placed on port definition, side ground placement,
and excitations of higher order modes. These are three critical
differences between the two simulation types and often cause
disagreements when the simulated results are compared. The
two software packages used for this analysis are Keysight’s
Advanced Design System (ADS) and ANSYS’ High Fre-
quency Structure Simulator (HFSS). A DC-20 GHz GBCPWG
transmission line is used as a design example and a series of
guidelines are provided throughout the paper.
II. CPWG DESIGN
A 3D model of the GBCPWG is shown in Fig. 1 to
illustrate the transmission line’s geometry. The characteristic
impedance of the GBCPWG trace is determined by the ratio
of the center strip width (W) to the gap width (S) given
a specific substrate thickness (h). Using equations from [3],
a DC-20 GHz GBCPWG transmission line is designed. The
substrate material chosen for this example is Rogers 4350b
(r= 3.66 and tanδ = 0.004) with 1/2 oz. copper and a
thickness of h=30 mil. A gap width of S=15 mil is chosen
and a trace width of W=53 mil is solved for to achieve
a characteristic impedance of 50 ohms. The length of the
transmission line is set to 360 mils, which corresponds to a
half-wavelength at 10.25 GHz. The CPWG trace is modeled in
ADS and placed between two 50 ohms terminations. Simulated
results are shown in Fig. 2.

Fig. 2. ADS simulated S-parameters of the DC-20 GHz GBCPWG.
III. ADS VS. HFSS SIMULATION COMPARISONS
For the remainder of this paper, simulation differences
between ADS and HFSS are compared and guidelines to
make an effective transition are discussed. After the ADS
simulations is complete, the most common error is to instantly
jump to a full-scale HFSS model without verifying the HFSS
setup. There is a significant amount of additional effort that
is needed to setup a HFSS simulation compared to an ADS
simulation. If the HFSS setup is not made incrementally more
complex and verified at each step, it is very difficult to down
select what issue is causing a simulation mismatch between
ADS and HFSS. This can lead to an inaccurate tunning and
optimization scheme that might not be noticed until first
articles are fabricated and measured resulting in wasted time
and effort. A major contributor to simulation mismatch is the
wave port in HFSS and is discussed in detail next.
A. HFSS Wave Port Definition
In HFSS, wave ports are used as an excitation source to
propagate a wave along the GBCPWG transmission line over
a defined frequency range. While wave ports are required
to reside on the radiating surface boundary, they have the
flexibility to renormalize or not renormalize the port to 50
ohms. The ability to not renormalize is useful when running
port only simulations to extract the characteristic impedance
at the port. This in turn provides knowledge on how to tune
the GBCPWG geometry to achieve the desired characteristic
impedance. However, HFSS simulation accuracy is highly
dependent on the geometry of the wave port that is defined in
the 3D model.
In general, there is very little information in regards to wave
port geometry for GBCPWG. This is primarily due to the
wave port geometry being very dependent on the GBCPWG
geometry, relative permittivity of the substrate material, etc.
that makes it challenging to derive a one solution fits all
approach. The wave port should be made large enough to
sufficiently capture the field distribution, but small enough to
Fig. 3. GBCPWG field distribution in HFSS port field display.
not excite higher order modes. As a first order approximation,
coplanar waveguide (CPW) wave port dimension guidelines
can be used. The height of the wave port, starting at the lower
ground plane, should be approximately (or greater than) four
times the thickness of the substrate or gap width, whichever
is larger. The width of the wave port should be approximately
(or greater than) ten times the center trace width or gap width,
whichever is larger. It is noted that if a high dielectric constant
material is used, the 10x guideline might be too conservative
and result in higher order mode excitation. Regardless, it
is critical that the width of wave port intersect the side
grounds. If the width is too narrow or too large (in the case
of finite side ground planes) and does not touch the side
grounds, the wave port sides will be left floating and act as
signal conductors, which will provide an incorrect simulation
response. Fortunately, because HFSS has a port field display,
the field distribution can be analyzed after initial simulation to
verify that the correct mode is being excited and a sufficient
amount of the field falls within the wave port geometry.
An example of a GBCPWG HFSS wave port field distribu-
tion is shown in Fig. 3. The wave port has a horizontal length
(L) of 165 mil and a vertical height (H) of 150 mil. The
width is eleven times the gap width instead of the trace width
even though the trace width is larger. However, notice that the
majority of the field is contained between the center trace and
the lower ground plane as well as from the center trace to the
side grounds, as is to be expected, and a sufficient amount
of the field distribution is captured within this smaller wave
port. A port only simulation is ran and the port characteristic
impedance is shown to average 49.99 ohms over the DC-
20 GHz bandwidth. Therefore, given an accurate port field
display and desired characteristic impedance, a final check is
to simulate the HFSS design and compare the S-parameters
to ADS. However, before that, it is important to discuss side
ground placement differences between ADS and HFSS.
B. ADS and HFSS Side Ground Placement
In order to accurately compare an HFSS simulation response
to the ADS simulated response in Fig. 2, an understanding of
the side ground placement between the two software packages

Fig. 4. HFSS GBCPWG 3D model with ADS equivalent side grounds.
Fig. 5. Simulated S-parameter comparison between the ADS and ADS-
equivalent HFSS.
must be understood. In ADS, the side ground placement is
effectively modeled to be perfect right at the edge of the
gap. In other words, the current return path is from the center
trace, across the gap, straight down to the lower ground plane
right at the gap edge, and then back towards the source.
The side ground plane is assumed to be infinitely wide away
from the gap edge, so there is no finite side ground effects
included in the simulation. Furthermore, the direct path from
the side ground to the lower ground plane is assumed to
be perfect along the entire length of the transmission line.
These assumptions are what allow ADS to simulate GBCPWG
structures in the matter of seconds.
The same boundary conditions described above can be
enforced in HFSS by appropriately modeling the GBCPWG
structure. This is accomplished by modeling the GBCPWG
transmission line as usual using desing parameters from sec-
tion II, but placing the design in between two solid PEC
rectangles as shown in Fig. 4. It is noted that the HFSS model
Fig. 6. HFSS GBCPWG 3D model with stitching vias.
in Fig. 4 was used to create the port field display in Fig. 3;
therefore, the port definition accuracy has been confirmed. The
HFSS model is simulated and the S-parameters are shown in
Fig. 5 along with the ADS simulated results. The S-parameters
show excellent agreement between the two software packages
indicating that a representative ADS-equivalent HFSS model
is achieved.
There are two main differences between the simulations,
which are the slight shift in the reflection zero around 10 GHz
and a slight decay in S11 performance at higher frequencies.
The reflection zero shift is due to the effective relative permit-
tivity in the HFSS model being slightly smaller than the ADS
model. This is attributed to HFSS more accurately capturing
the field distribution above the GBCPWG trace. The effective
relative permittivity of the HFSS model is 2.45, which is close
to the 2.56 value extracted from ADS. The slight degradation
in S11 is common in wideband simulations and is due to the
increased inductance of the copper material as a function of
frequency. Overall, an accurate HFSS model is confirmed and
can be used as the control for the remainder of the paper.
C. Higher Order Mode Suppression
While the model in Fig. 4 is useful to verify ADS sim-
ulations and HFSS model accuracy (port definition, etc.),
extension of solid ground planes is not realistic. The most
common way to create a current path from the side ground to
the lower ground plane is to use via stitching. The placement
of stitching vias and their effectiveness has been analyzed
thoroughly over the years [4], [5]. However, stitching vias
change the geometry of the GBCPWG, which can affect the
transmission line performance, but can also yield HFSS errors.
A GBCPWG design with stitching vias is shown in Fig. 6.
The first row of stitching vias are strategically placed close
enough to ensure higher order modes are not excited but far
enough apart to not affect the effective relative permittivity. As
such, the distance between the edge of the gap and the nearest

Fig. 7. Simulated S-parameter comparison between the ADS, ADS-equivalent
HFSS, and the via-stitched HFSS models.
edge of the stitching via (d) is 17.72 mil or exactly 3 via radii.
The via spacing in the x-direction is 30 mil from via center-
to-center. A second row of stitching vias is placed 30 mil in
the y-direction and offset by 15 mil in the x-direction for the
far side vias and vice-versa for the near side vias.
For the via-stitched GBCPWG, it is recommended that the
spacing dbe at least greater than one via diameter to ensure
that proper via growth can take place between the top and
bottom copper using low cost standard PCB processing tech-
niques. This requirement, along with the dielectric constant
and thickness of the substrate, ultimately drives the diameter
of the via. If a high dielectric constant material and large via
diameter are chosen, the via-to-via separation (l) may become
too wide and higher order modes will be excited. To combat
this, a smaller via diameter can be chosen to reduce das well
as land push higher order modes further up the frequency
spectrum. For the given design, the spacing lis 117.43 mil.
Given a dielectric constant of 3.66, the first higher order mode
will be at 26.3 GHz, well beyond the 20 GHz passband of
interest. The HFSS model in Fig. 6 is simulated and the S-
parameters are shown in Fig. 7 along with the ADS and ADS-
equivalent HFSS simulations.
A final point needing mentioned is in regards to the high-
lighted “port vias” in Fig. 6. It is good practice to place these
half-circular vias right at the edge of the board to ensure that
high order modes at the port are not excited. In cases where
thin substrates are used, the width of the wave port might need
to become increasingly wide in order to capture a sufficient
amount of the field above the substrate. Eventually, the width
will exceed the maximum width allowed before a higher order
mode is excited on the port. This issue is exaggerated when
a thin substrate with a large dielectric constant is used in
conjunction, which is a common occurrence in modern day
circuit design to minimize circuit footprint area to reduce
SWaP. In HFSS, an error will trigger and be displayed in the
message manager thats says “the port supports an additional
propagating and/or slowly decaying mode”. This can affect
Fig. 8. Simulated S-parameter comparison between HFSS GBCPWG model
w/ and w/o port vias.
the simulation convergence and produce inaccurate results
since the energy at the port is being split between the desired
dominant mode and the undesired higher order mode, the latter
of which does not get accounted for at the port.
To illustrate this point, the port via is removed and the wave
port width is increased to L=206 mil to ensure a higher
order mode falls within the 20 GHz passband. This HFSS
model is simulated and S-parameters are shown in Fig. 8 and
compared to the original port via HFSS model. Notice the
drastic difference in simulated results by simply introducing a
higher order mode. This validates that a step-by-step/validation
process when transitioning from ADS to HFSS should be
followed rigorously. If the wide wave port model would have
been used first, the follow-on tuning scheme to optimize the
return loss across the passband would have been invalid. As
previously mentioned, if the user believes that the model is
correct, this error would not be noticed until after fabrication,
measurement, and comparison.
D. Optimized GBCPWG Design
This last subsection will focus on optimizing the validated
model in Fig. 6. In the last subsection, the placement of
the first row of stitching vias away from the gap edge (d)
was discussed in regards to higher order mode suppres-
sion. However, the placement of these vias also affects the
GBCPWG geometry and return current path, which alters the
characteristic impedance of the transmission line. Knowledge
of the electromagnetic differences between solid side grounds
(ADS) and via-stitched side grounds (HFSS) will provide a
better understanding of how to quickly tune the design for
optimal performance.
Recall that the characteristic impedance for a lossless trans-
mission line is [6]
Zo=rL
C(1)
where Land Care the per-unit length inductance and ca-
pacitance, respectively. If the first row of stitching vias are

Fig. 9. Simulated S-parameter comparison between the ADS, ADS-equivalent
HFSS, and the optimized via-stitched HFSS models.
moved farther away from the gap edge, in other words d
increases, the return current path get longer, which increases
the inductance. Furthermore, the farther away the vias, the
capacitance to ground decreases. The capacitance considered
here is the fringing capacitance between the center trace and
the through-hole vias. Therefore, the characteristic impedance
of the line will increase as a function of increasing d. This
phenomenon is the reason for the reduced S11 performance
between the ADS-equivalent HFSS simulation and the via-
stitched HFSS simulation in Fig. 7.
Given the above information, the un-tuned HFSS model in
Fig. 6 should have a higher characteristic impedance. A port
only simulation is ran and the average characteristic impedance
across the 20 GHz passband is 51.15 ohms. Fortunately, this
is very easy to compensate for by slightly increasing the
width of the GBCPWG center trace to increase the capac-
itance to ground and lower the characteristic impedance of
the transmission line. A trace width of 53.5 mil yields an
average characteristic impedance of 50.05 ohms. The width
is increased in the HFSS model and simulated. Fig. 9 shows
the simulated results of the optimized HFSS model along with
the ADS and ADS-equivalent HFSS simulated S-parameters.
Notice that the S11 performance has been corrected and is
nearly identical to the ADS-equivalent HFSS performance.
Fig. 10 shows a zoomed-in plot of the S21 performance. At
this point, an optimized and realizable GBCPWG design is
complete and ready to be used as a transmission line between
electronic components on an extended circuit.
IV. CONCLUSIONS
In this paper, a detailed analysis and series of guidelines
are provided on transitioning a design from an ADS circuit
model simulation to a HFSS finite-element-method simula-
tion. A DC-20 GHz GBCPWG design is used as a design
example throughout the paper. A major emphasis of the
paper is accurate port definition to ensure valid simulations
results, which was followed by an ADS-equivalent HFSS
model simulation. Good agreement between the simulations
Fig. 10. zoomed-n S21comparison between the ADS, ADS-equivalent HFSS,
and the optimized via-stitched HFSS models.
indicate a proper port definition and allows designers to move
forward with high confidence in their model setup. Another
important topic discussed is the need for via-stitching along
with a quick discussion on their placement to avoid higher
order mode excitation as well as their affect on GBCPWG
geometry and current return path. An understanding of these
details allow for quick tuning of the structure to optimize
the GBCPWG performance and achieve ideal results with a
realizable design that can be fabricated using standard PCB
processes. The process and guidelines offered in this paper
should allow designers, and most importantly new-comers
to the RF/microwave community, to more accurately and
efficiently generate wideband GBCPWG designs to use in
modern day circuits.
ACKNOWLEDGMENT
This work is funded by the Department of Energy’s Kansas
City National Security Campus, operated by Honeywell Fed-
eral Manufacturing & Technologies, LLC under contract num-
ber DE-NA0002839.
REFERENCES
[1] J. W. McDaniel, M. B. Yeary, H. H. Sigmarsson, J. A. Wolf, S. Garrison,
K. Byers, and M. Clewell, “Integration and miniaturization of a Ka-band
stepped frequency radar for un-manned aerial vehicle applications,” in
International Symposium on Microelectronics, vol. 2017, no. 1, October
2017, pp. 61–66.
[2] P. Hgler, F. Roos, M. Schartel, M. Geiger, and C. Waldschmidt, “Radar
taking off: New capabilities for uavs,” IEEE Microwave Magazine,
vol. 19, no. 7, pp. 43–53, Nov 2018.
[3] G. Ghione and C. Naldi, “Parameters of coplanar waveguides with
lower ground plane,” Electronics Letters, vol. 19, no. 18, pp. 734–735,
September 1983.
[4] A. Sain and K. L. Melde, “Impact of ground via placement in grounded
coplanar waveguide interconnects,” IEEE Transactions on Components,
Packaging and Manufacturing Technology, vol. 6, no. 1, pp. 136–144,
Jan 2016.
[5] T. Yuasa, T. Nishino, and H. Oh-Hashi, “Simple design formula for par-
allel plate mode suppression by ground via-holes,” in 2004 IEEE MTT-S
International Microwave Symposium Digest (IEEE Cat. No.04CH37535),
vol. 2, June 2004, pp. 641–644 Vol.2.
[6] D. Pozar, Microwave Engineering. Wiley, 2012.