Design, Fabrication and Evaluation of a MEMS-Based,
Ka-Band, 16-Element Sub-Array
Janice C. Rock, Tracy Hudson – U.S. Army Aviation and Missile Research, Development, & Engineering Center;
AMSRD-AMR-WD-UR; Redstone Arsenal, AL. 35898-5000; ph: 256-876-1426
Brandon Wolfson – Technology Service Corporation, Phase IV Systems Operation; 3405 Triana Blvd., Huntsville,
AL. 35805; ph: 256-535-2181
Daniel Lawrence – Technology Service Corporation, Phase IV Systems Operation; 3405 Triana Blvd., Huntsville,
AL. 35805; ph: 256-535-2142
Brandon Pillans,– Raytheon Space and Airborne Systems, Advanced Products Center, Dallas, TX 75243; ph: 972-
Andrew R. Brown – A. Brown Design, 46055 Bloomcrest Drive, Northville, MI 48167;
Louis Coryell – U.S. Army Communications-Electronics Research, Development and Engineering Center,
AMSRD-CER-ST-SS-TS, Fort Monmouth, NJ 07703-5203; ph 732-532-9210, x5632
email@example.com; firstname.lastname@example.org; email@example.com; firstname.lastname@example.org; b-
email@example.com; Andrew.firstname.lastname@example.org; Louis.email@example.com
Development and Engineering Center has been involved in
an Army Technology Objective (ATO) aimed at furthering
phased arrays for both tactical seekers and communication
links.1,2 The ATO has
MicroElectroMechanical Systems (MEMS) and MMIC-
based phase shifters with an overall goal of of reducing
missile seeker costs by 50% based on the missile mission.
In a collaborative effort with the Communications-
Electronics Research, Development and Engineering Center
(CERDEC), the AMRDEC has worked to improve the
maturity of Radio Frequency (RF) MEMS devices for use in
phase shifters for phased arrays.
This paper will demonstrate that RF MEMS have vastly
improved in reliability over the past few years.
Additionally, background information and most current
results of a task to implement a 16-element phased sub-
array with RF MEMS-based phase shifters will be
presented. The slat will be centered at 33.4 GHz and will
utilize ½ wavelength spacing between elements. The
individual elements will consist of Vivaldi antennas. Taylor
Weighting will be applied to lower the overall sidelobes.
The background work for this paper was presented at the
2009 IEEE Aerospace Conference. This paper includes that
information along with complete documentation of the
TABLE OF CONTENTS
2. SWITCH DESIGN......................................................2
3. PHASE SHIFTER DESIGN.........................................3
4. PHASED ARRAY T/R ASSEMBLY............................4
Aviation and Missile Research,
been pursuing both
1 U.S. Government work not protected by U.S. copyright.
2 IEEEAC paper #1085, Version 4 Updated 2009:12:1
5. LABORATORY EVALUATION..................................10
6. DISCUSSION AND CONCLUSION..............................10
In 2006, CERDEC issued a Broad Agency Announcement
requesting proposals for Affordable RF Micro Electro-
Mechanical Switch (RF MEMS) Phase Shifters for Phased
Arrays Manufacturing Technology (ManTech) Objective
(APSPA MEMS) Program. The objective is to develop a
volume production facility for reliable, low cost MEMS
phased shifters for use in phased arrays for both active and
passive missile seekers and for on-the-move SATCOM
communications systems for the Warfighter. AMRDEC has
been involved in this program since its inception and in
early Calender Year 2008 (CY2008) initiated an effort to
demonstrate the results of the CERDEC MTO program in a
16-element, Ka-band, MEMS-based electronically-steerable
linear array (slat).
A major contributor to both the cost and loss of these
system types, particularly in the passive systems, is the
packaged Monolithic Microwave
(MMIC) phase shifter. MEMS phase shifters offer a low
cost (<$10), low loss alternative to these costly components.
The successful completion of this ManTech program
should result in an overall cost reduction of up to 25% for
phased array antennas operating in the Ku, K, and Ka
frequency bands. In addition, MEMS-based phase shifting
promises to reduce the loss by 50-66% compared to MMIC-
based phase shifting.
The prime contractor, Raytheon Space and Airborne
Systems, has produced a 4-bit packaged phase shifter
demonstrating approximately 2.5 to 3.0 dB of loss. The
switch and phase shifter design are discussed in this paper.
To demonstrate the promising results of the ManTech
program, AMRDEC has implemented a task to demonstrate
electronic beam steering at Ka-band, using a 16 element
Vivaldi antenna slat array. This paper presents the results
of that task to date. The remaining laboratory evaluation of
the slat will be included in the final paper if completed in
time, and/or presented at the conference.
2. SWITCH DESIGN
In 1995, Raytheon pioneered the development of RF
MEMS technology for microwave and millimeter-wave
applications with the development of the first capacitive RF
switch . Since then, Raytheon has been involved in the
design and development of high-performance RF MEMS
for advanced phased-array applications. Raytheon has
developed an RF MEMS switch that is optimized for low
RF insertion loss, high switching speed, high-power
handling, excellent temperature stability, and long cycle
lifetime (Table 1). With support from the Army, DARPA,
ONR, and AFRL, Raytheon has demonstrated low-loss,
multi-bit phase shifters, routers, and digitally tunable filters
across the entire 0.1 to 50 GHz frequency range. A picture
of the basic switch is shown in Figure 1.
Table 1. RF MEMS switch performance summary.
RF MEMS Switch Performance
Insertion loss at 40 GHz
Isolation at 40 GHz
-55°C to +85°C
> 1011 cycles
280 × 120 μm
Figure 1 - Top-view of the individual RF MEMS switch.
Raytheon’s latest switch design achieves over 200 billion
operating cycles without failure with a switching speed
under 5 microseconds and power consumption under 1
microwatt. Innovations in switch membrane materials and
device structure now provide much greater thermal stability,
and prototype switches operate from -55°C through +85°C
with actuation voltage change of only a few volts.
Typical switch insertion loss is less than 0.1 dB at
frequencies through 40 GHz. These switches are
constructed using only metals and dielectrics, and detailed
measurements of Raytheon’s RF MEMS switches show no
observable RF nonlinearity to intercept points (TOI) as high
as +87 dBm (5,000 watts) .
Like all micromechanical systems, RF MEMS switches are
subject to damage from corrosion and contamination from
exposure to dust, air, and moisture. Encapsulating the
switches in a particle-free, hermetic package eliminates
failures due to these environmental effects. To meet
operational requirements, the package must be compact and
not degrade RF circuit performance. The challenge is to
provide this protection at very low cost in a high volume
production environment. Over the past three years,
Raytheon has developed a highly innovative wafer-level
hermetic packaging approach  that meets all requirements
for reliability, operating temperature range, yield, and cost.
This near-hermetic package (Figure 2) adds very little to the
total circuit price.
Figure 2 - Photograph of near-hermetic wafer-scale
packaged RF MEMS.
Under the Army’s Affordable
Mechanical Switch (RF MEMS) Phase Shifters for Phased
Arrays Manufacturing Technology Objective (APSPA
MEMS) Program, Raytheon is currently fabricating RF
MEMS phase shifters  in its Advanced Multilayer
Interconnect (AMI) facility, an ISO 9001-qualified
manufacturing environment that has fabricated more than
500,000 comparable circuits. The APSPA Mantech program
is a three-year, $5M effort aimed at raising the TRL and
MRL levels of Raytheon’s RF MEMS technology to 8 for
system insertion. The program is currently in its third year
with much success. Included in this effort are improvements
in reliability (1012), RF power handling (2W hot), operating
temperature range (-55 to +85°C), switching speed (<5µs)
and cost. By leveraging the Army’s RF MEMS Mantech
program investment, Raytheon is able to provide low-loss,
highly reliable switches for many applications including
low-cost phased arrays.
3. PHASE SHIFTER DESIGN
The phase shifter design is based on a class II loaded-line
topology [5, 6]. This topology was a compromise of
simplicity, low-loss, compact size, and fits well with the
equivalent model of the MEMS switch. The class II loaded
line consists of a pi-network of a switched shunt reactance,
a series transmission line of length L1eff, and impedance
Z1eff as shown in Figure 3. Full theory and design examples
of this concept is well documented in  and not detailed
here. For the implementation used in this system, the
switched reactance consists of the MEMS switch followed
by a shunt radial stub . The zero state consists of
reactive loading of the switch up-state followed by the
radial stub. The phase shifted state uses the down state
switch capacitance and the radial stub.
One of the major difficulties with a loaded line topology is
obtaining large phase shifts while maintaining bandwidth.
Typically, 45˚ is the largest state implemented with a
loaded-line topology . Higher phase shifts are typically
done by cascading 45˚ stages. However, even the 45˚ stage
proved difficult to implement and meet bandwidth
requirements with the required loaded line and the Con/Coff
ratio at Ka-band. Therefore, even multiple stages were used
for the 45˚ bit. To further reduce the number of stages
required in cascaded bits, ripple was introduced. For the
90˚ and 180˚ stages, the phase shift per section and the
overall impedance values were adjusted to provide an equal
ripple response. This has the effect of allowing reduced
number of cascaded stages while still maintaining
bandwidth when two or more stages are used. This allowed
the 90˚ stage to be implemented with as few as two stages
and the 180˚ with only 4 stages. However, the out-of-band
performance of these stages rolls off much fast then
conventional cascaded stages.
The package is based on a hermetic wafer scale package
with 20 mm2 exterior dimensions. In addition to having to
meander the phase shifter through the package interior, the
transitions in and out of the package had to be optimized for
this operating band. Due to the high capacitance of the
feedthrough line to the grounded seal ring, the feeds were
made very inductive to attempt to tune out some of the
shunt capacitance. An optimization was performed to
minimize in-band insertion loss from the package transition.
Figure 4 shows the resulting, packaged phase shifter.
Measurements of isolated package transitions show < 0.2
dB insertion loss.
The fabricated and packaged phase shifter were measured.
All data shown is from an average, fully packaged phase
shifter with no de-embedding performed. There was a
slight shift in performance from the designed values,
primarily shown in the 22.5˚ and 45˚ bits. Both of these bits
show excessive phase shift compared to the designed
values. This can be fixed in future designs and is not a
limitation of the technology. Even with the long 22.5˚ and
45˚ states, the RMS phase error at 33 GHz is only 8.8˚. The
total insertion loss at 33 GHz is 2.5 dB average with an
RMS amplitude error of 0.3 dB.
Even though the RMS phase error is high, primarily due to
the 22.5˚ and 45˚, the presented phase shifter shows vastly
reduced insertion loss compared with MMIC technologies
such as the Triquint TGP2102 5-bit phase shifter (7 dB
nominal insertion loss). The results of the laboratory
evaluation of the packaged MEMS phase shifters are
presented in Figures 5, 6 and 7, with the legend for all three
figures given in Figure 7.
L1_eff, k1_eff, Z1_eff
Figure 3 - Topology of a capacitively loaded Class II
loaded-line phase shifter.
Figure 4 - Layout of the Ka-Band phase shifter.
Figure 5 - Measured Return Loss.
Figure 6 - Measured Insertion loss of all 16 states.
Figure 7 - Measured Phase Shift.
4. PHASED ARRAY T/R ASSEMBLY
The slat assembly is composed of the microwave antenna
slat module mated to a control and interface module. A
block diagram of the slat assembly is shown in Figure 8.
The microwave antenna slat module contains all of the
assembly’s microwave circuitry as well as the circuitry
necessary to support the operation of the microwave
devices, such as DC regulators, power sequencing circuits,
switch drivers and high voltage multiplexers. The high
voltage multiplexers are used to deliver the high voltage
control signals, necessary to actuate the MEMS switches in
the phase shifters, to the appropriate inputs of the MEMS
phase shifters. The control and interface module provides
the user with a simple method of controlling and interfacing
with the microwave antenna slat module.
Figure 8 - Block Diagram of Antenna Slat Assembly.
Microwave Antenna Slat Module
The microwave antenna slat module implements 16 Vivaldi
antennas and a 16-way 20 dB Taylor weighted power
divider to drive the MEMS phase shifters at the inputs to the
antennas. A transmit/receive circuit made up of Ka-band
MMIC amplifiers and switches is utilized at the input of the
power divider to boost the transmitted/received signals. The
Vivaldi antenna was chosen for its broadband performance
and ease of implementation in a printed circuit design. The
antenna design was finalized and its performance simulated
using HFSS 3D electromagnetic modeling software. The
16-way power divider implements the 20 dB Taylor weights
by cascading unequal power dividers. The design of the
power divider was completed and modeled using
Microwave Office simulation and design software. The
circuits used to control the operational states of the Tx/Rx
circuitry as well as the phase states of the 16 MEMS phase
shifters are located on the backside of the multilayer board
used to implement the microwave circuitry. These circuits
on the backside of the microwave antenna module are
controlled by signals output from the control and interface
Vivaldi antennas were first introduced as a new class of
antennas that produce a travelling-wave through an
exponentially tapered slot . These types of antennas are
commonly referred to as tapered slot antennas (TSA) with
many possible variations on the length and profile of the
taper. A modified TSA, known as the antipodal Vivaldi
antenna, was introduced later with the benefit of direct
feeding by a microstrip line . The antipodal Vivaldi
antenna element design provides a wideband, broad
radiation pattern that is well-suited for use in a phased array
antenna. Another attractive feature of the Vivaldi antenna is
the ease with which it can be driven with a microstrip feed
network, and thus, integrated with surface-mount phase
shifter components . The basic radiating element is
shown in Figure 9. The element is fabricated from a single
substrate layer with metallization on both sides. A profile
of the substrate (5 mil thickness, εr = 3.0) is shown in
Figure 10. The element begins by exponentially tapering
the ground plane of the microstrip feed line until the width
is the same as the signal line. For a good impedance match,
the taper extends for about one-wavelength distance. At
this point the ground plane and signal lines are flared
symmetrically on opposite sides of the substrate to form a
broadband dipole radiator. Tuning of the antenna is
achieved by empirical adjustment of the flare geometry.
Figure 9 - Antipodal Vivaldi Element Design.
Figure 10 - Substrate Profile.
Antenna Modeling and Simulation
A four-element subarray of Vivaldi antennas is shown in
Figure 11. An HFSS simulation of the four-element
subarray has been constructed in order to tune the radiating
elements in the array environment. The overall size of the
simulated array has been kept small to maintain reasonable
run-times. Results in Figure 12 show the input match for
each element in the subarray. Notice the return loss is better
than 10 dB across the complete bandwidth, confirming the
broadband nature of the Vivaldi elements. Slight variations
in the resonance of each element can be attributed to the
mutual coupling and parasitic loading of different elements
in the array. Scanning of the subarray in the E-plane is
accomplished by applying a progressive phase shift across
the array. HFSS simulation for array radiation patterns at
several scan angles are shown in Figure 13. Uniform
amplitude weighting was used for the subarray simulation.
Note that scanning past 45 deg results in increased sidelobes
and a reduced mainlobe due to the roll-off of the element
Figure 11 - Four element subarray HFSS model.
Figure 12 - Input match for each element in the four-
Figure 13 - Scanned E-plane radiation patterns for the four-element subarray.
The 16-way power divider is constructed from four stages
of two-way power dividers. The first three stages are made
by simple T-junction power dividers and the final stage of
power divider utilizes unequal split Wilkinson power
dividers. The use of T-junction power dividers for the first
three stages was done to reduce design complexity and
space and to eliminate the need for isolation resistors. While
the T-junction power dividers simplify the design, they do
not provide isolation between the output ports. It was found
that the performance of the power divider could be
significantly degraded if its 16 outputs were not terminated
with a load that was well matched (Γ ≤ -20 dB) to the
design impedance of 50 ohms. In some cases the MEMS
phase shifters have been shown to have an input return loss
as high as -15 dB to -10 dB in the band of interest which
was not good enough to prevent performance degradation of
the power divider. Along these same lines, the input return
loss of the MEMS phase shifters can vary with changing
phase shift settings causing
degradation of the power divider. For these reasons, the
final stage of the power divider was chosen to be Wilkinson
power dividers. This provided enough port-to-port isolation
to allow for the imperfect loads presented by the MEMS
For the design of the power divider, 20dB Taylor one-
parameter weighting was used. The value of 20dB was
chosen to provide some reduction in sidelobe level while
making the design of the power divider more feasible. The
Taylor weighting coefficients can be found from (3) and (4)
J0 is the modified Bessel function and β is related to the
sidelobe level and can be found by finding the root of
equation (4). In (4) SLL is the desired sidelobe level in dB.
The relative output powers at the two output ports of a two-
way power divider can be defined as shown in (5). P1 and
P2 are the relative output powers at ports 1 and 2 relative to
Once the Taylor weighting coefficients are known, one can
find the values of α in (5) for each of the power dividers.
Since we have an even number of outputs and the Taylor
weighting coefficients are symmetric the value of α for the
first stage power divider will always be 0.5 and the values
of α for the cascaded power dividers at each output of the
first power divider will simply be mirrored copies of each
other. For this design the values of α at each stage are
shown in Table 2.
Table 2. Vales of α α for each power divider stage.
0.283 0.438 0.438 0.283
0.3495 0.4111 0.4516 0.4846 0.4846 0.4516 0.4111 0.3495
For a simple T-junction power divider one can use the
values of α above to find the necessary impedance that
needs to be presented to the power divider input port by
each of the two output ports to get the desired unequal
power division. The relationship between the impedances at
each side of the T-junction to the values of α is shown in
In (1, 2) Z1 and Z2 are the impedances presented to each side
of the T-junction power divider and Z0 is the characteristic
impedance of the input to the T-junction as depicted in
Figure 14 where the second stage T-junction power divider
is shown. One quickly finds that the greater the sidelobe
reduction the more difficult it becomes to design some of
the power dividers due to the unreasonably high values of
impedance needed. The high impedances can become an
issue when they lead to trace widths that are too narrow for
standard fabrication processes. The large ratio between the
two output impedances also makes the design less trivial
since one side of the T-junction will require a relatively
wide trace while the other side requires relatively very
narrow trace. This is the main reason that 20 dB Taylor
weights were chosen over those with more sidelobe level
reduction. One can help alleviate the need for unreasonably
narrow traces by decreasing the value of Z0 at the input to
each power divider stage. This is the approach that was
taken for this design as seen in Figures 14 and 15. Figure
15 shows the overall 16-way 20 dB Taylor weighted power
divider design. One can see that the final stage of power
dividers are Wilkinson power dividers. As mentioned
previously this was done to provide some channel to
channel isolation to help prevent performance degradation
due to the imperfect impedance match presented by the
MEMS phase shifters that are fed by this power divider.
Figure 14 - Second stage two-way T-junction power divider.
Figure 15 - 16-way Taylor weighted power divider.
Modeling and Simulation
The modeling and simulation for the power divider was
completed in Microwave Office. The substrate used is the
same as that used for the antenna design (5 mil thickness, εr
= 3.0). It can be seen from Figure 16 that the input
characteristic impedances for each stage were chosen
strategically to both keep the narrower line widths in the
design as a reasonable width as well as remove the need for
impedance transformers on one side of many of the power
The simulation results for the input return loss of the 16-
way power divider is shown in Figure 16. Here it is seen
that the power divider is well matched over the entire band
of interest. Also shown in Figure 16 is the insertion loss
from the input of the power divider to either of the two
center most output ports. The expected insertion loss for the
two center ports is -9.72dB in a loss-less system.
Center Port IL and Input RL
Figure 16 - Simulation results for the input loss.
From the simulation it is seen that the circuit has an
approximate substrate loss of 1dB. This is to be expected at
these frequencies. For the rest of the ports it is best to look
at the output power relative to the center most ports to
ensure that the design effectively implements the desired
Taylor weighting. Figure 17 shows this by plotting the
difference between the power at the center port and the
adjacent ports. In this figure the markers are placed at the
center of the band of interest and the insertion loss of each
port relative to the center port insertion loss of Figure 16 is
shown. The values at each marker are shown in the legend.
The marker m1 represents the port farthest from the center
whereas m7 represents the port nearest the center port.
Since the design is symmetrical we only need to look at one
side of the design.
Relative Output Power Normalized To Center Port
m2: 33400 MHz
m3: 33400 MHz
m4: 33400 MHz
m5: 33400 MHz
m6: 33400 MHz
m1: 33400 MHz
m7: 33400 MHz
Figure 17 - Plot of the difference between the power at the
center and adjacent ports.
Using the Taylor weight coefficients one can find the
expected relative output powers at each port. These are
shown in Table 3 along side of the simulated values shown
in Figure 17. In the table port numbers 1-7 are synonymous
to markers m1-7 in Figure 17 and likewise represent going
from the outermost port to the innermost port. In this table it
is seen that the calculated and measured values are closely
Table 3. Output power at each port relative to center most
port output power.
Port # Calculated
Module Layout and Controller Design
The MEMS Phased Array Assembly is made up of the
Microwave Antenna Subassembly mated to the Control and
Interface Subassembly. The division of the MEMS Phased
Array Assembly was done strategically to provide a path
forward for future developments where it may be desired to
design a 2D array. The circuit design of the Control and
Interface Subassembly could supply the appropriate signals
and voltages to multiple antenna array slats. The
architecture of the Microwave Antenna Subassembly would
require mostly mechanical upgrades to account for the
height restrictions that would be created by stacking
multiple slats to form the 2D array. The designs of each of
the two subassemblies will be discussed in more detail here.
Microwave Subassembly Design
The microwave substrate used for fabricating the phased
array antenna was designed to also provide the microstrip
transmission lines that distribute the microwave signals
between active components in the array. A 0.005” thick
Rogers RO3003 substrate was used for this application.
The substrate was designed with cutouts for each of the
active components as well as the MEMS phase shifters.
The design of the RF assembly allowed for microwave
coplanar probing of each phase shifter output. This was
accomplished by placing two small RF probe boards
between the output of each phase shifter and input to the
corresponding Vivaldi element. The microwave probe
boards are 0.050” long pieces of 0.005” thick alumina
microstrip having ground pads at one end of the substrate in
order to accommodate a ground-signal-ground (GSG)
microwave probe. The microstrip and ground pads were
spaced for a probe pitch of 150 micrometers. One probe
board allowed for phase measurement between the of each
phase shifter with respect to the common microwave input
connector. Figure 18 shows a picture of the microwave
board assembly. The initial plan was to populate an
amplified transceiver section at the input of the 16-way
power divider. Due to cost and schedule restraints it was
decided to move forward with the fully passive array in
order to achieve the primary goal of demonstrating the
ability to steer a beam using the MEMS phase shifters. The
empty spot on the lower end of the microwave substrate is
the area where the transciever circuitry could be placed.
Power Divider and Phase
Figure. 18 . Phased Array Assembly, RF Substrate and
MEMS Phase Shifters
In the Microwave Antenna Subassembly on the opposite
side of the microwave substrate is the microwave module
interface board. This board receives the voltages and control
signals from the Control and Interface Subassembly. The
voltages are used to supply power to amplifiers as well as
the high voltage switches required to control the MEMS
phase shifters. The control signals are mainly used to
command the high voltage switches to assert the provided
high voltage control to the desired MEMS phase shifter
inputs. A picture of the microwave module interface board
is shown in Figure 19.
Figure. 19 - Microwave Module Interface Board
Control and Interface Subassembly
The control and interface subassembly houses the circuitry
necessary to provide the control signals and various
voltages to the Microwave Antenna Subassembly. The
Control and Interface Subassembly also provides a means
for one to communicate with the phased array assembly
using a standard computer interface. The circuitry mainly
consists of a FPGA, a DAC, high voltage op-amps, and DC-
DC converters. A graphical user interface was developed to
interface with the FPGA and provide the user with method
to control the various states of the phased array module. The
FPGA is also used to control the DAC whose output is
buffered and amplified to the high voltage levels needed to
drive the MEMS phase shifters. This buffering and
amplification is done by high voltage op-amps. The outputs
of the op-amps are presented to the Microwave Antenna
Subassembly via the microwave module interface board. As
mentioned earlier the microwave module interface board
then provides these op-amp outputs to the high voltage
switches to set the appropriate bits of the MEMS phase
shifters and ultimately steer the antenna array beam pattern.
The DC-DC converters allow the phased array assembly to
be powered by a single 28 Volt source. Figures 20 and 21
show pictures of the two sides of the control and interface
board as well as the control and interface board mounted in
the Control and interface subassembly.
Figure. 20 - Control and Interface Board, Top and
Figure. 21 - Control and interface subassembly with the
lid on and off
Computer User Interface
A computer user interface was developed using Visual C++.
A screen shot of the Phased Array Interface (PAI) is shown
in Figure 22. One can control the MEMS Phased Array
Assembly simply by connecting to the RS232 serial port
present on the back of most PCs. The load text button is
used to load a look up table of phase shifter settings. The
phase shifter settings are then asserted by choosing the
desired beam steer angle using the slider bar and hitting the
send button. If the user enters the correct frequency of
operation in the Frequency box then each time the slider bar
is moved the angle to which the beam will be steered is
presented in the Beam Steer Angle box. Also, in the case
where the transmit and receive amplifiers are being used the
Mode buttons allow one to choose the direction of the
signals. The status window displays statements that verify
that the interface is working correctly.
Figure. 22 - Phased array assembly computer control
5. LABORATORY EVALUATION
The testing of the phased array assembly was conducted in
two stages, calibration and pattern measurement. The array
was first calibrated using a vector network analyzer and RF
coplanar probe station. S-parameter data was collected for
measurements between the microwave input connector and
the output of each phase shifter. This data was collected
over a frequency range of 31 to 36 GHz. This data was
used to compute the relative phase difference between phase
shifter settings across the array. The phase values measured
at 33, 33.5, 34, 34.5 and 35 GHz were used to create the
look up tables used by the PAI during the pattern
measurements of the array. The look up tables were created
by reading in all of the measured data into Matlab and then
using a function to find the phase shifter settings that will
lead to the closest fit possible to the desired phase slope
across the array. Figure 23 shows a plot of how this is done.
The x-axis represents each of the 16 phase shifters. The y-
axis represents the relative phase measurement in degrees.
The different lines represent the 16 different phase settings
available by each phase shifter. The legend shows which 4-
bit setting was asserted on the phase shifters for each line.
The black dashed line overlaying the phase measurements is
the ideal line for the desired phase slope. The circles near
the ideal line are the phase settings available that most
closely fits the ideal phase line. In this figure it is seen that a
22.5 degree slope was solved for.
Figure. 23 - 34 GHz phase measurements and phase
setting solver example
Figure 23 also shows how well the phase shifters work.
They are supposed to have a 22.5 degree phase shift
between each bit increment. It is seen that, for the most part,
the phase shifters do nearly apply the correct phase steps.
However, the phase shifters on channels 3 and 12 appear to
have some problems or stuck bits as shown by the sharp
discontinuities in the data for several of their phase settings.
Plots of the phase measurements at 33, 33.5, 34, 34.5 and
35 GHz are included in Appendix A.
The amplitude for each channel was also measured. These
measurements will show how well the amplitude was
tapered across the array. Figure 24 shows a plot of the
relative amplitudes measured for each channel as well as the
ideal case generated from the values shown previously in
Table 3. In this plot it is seen that a couple of the channels
are slightly off but for the most part the amplitudes follow
the desired amplitude weighting. This plot was created for
measurements made with all of the phase shifters set to the
same phase shift. Figure 25 shows the same plot but with
the phase shifters set such that the beam is steered using the
phase settings solved for and shown in Figure 23. In Figure
25, while the amplitude trend follows the desired weighting,
it is seen that significant amplitude errors are occurring. It is
believe that these errors are originating from two sources.
The first is from the amplitude variation that phase shifters
can have between different phase settings.
Figure. 24 - Measured relative amplitude vs. channel
Figure. 25 - Measured relative amplitude vs. channel
The second is from not having enough isolation between the
power divider outputs. Poor channel isolation can lead to
amplitude and phase errors that vary with phase shifter
setting. This can happen if the input return loss of the phase
shifters vary with phase setting and the reflected power then
ends up in the adjacent channels and corrupts their signal.
From other plots similar to Figure 25 it was found that the
lower channels, especially channel 6 had the most issues
with amplitude fluctuations with changing phase shifter
settings. This may not come as a surprise since channel 6
required some rework during the lab testing and debugging
phase of this program. Redesigning the power divider using
a topology that will lead to more channel isolation may help
mitigate errors due to reflections. A method of varying the
gain of each channel independently will be required to best
remove amplitude errors.
The antenna patterns for the array assembly were measured
in the compact range anechoic chamber at Building 5400.
The array assembly was mounted to a computer-controlled
pedestal for angle scanning. The antenna patterns were
measured over a 180-degree span. Pictures of the array
assembly and pedestal mount in the compact range are
shown in Figures 26 and 27.
Figure. 26 - Picture of compact range with array
assembly pointed in the +90 degree direction
Figure. 27 - Front of the array assembly on the pedestal
Antenna pattern measurements were made at 33, 33.5, 34,
34.5 and 35 GHz. For each frequency the beam was steered
to several angles. Some plots that display some of the better
measurements are shown here in Figures 28 through 32.
These plots are not representative of all radiation patterns at
all angles, but represent radiation patterns where errors were
known to be minimal. Plots of patterns at angles that
contained errors due to phase errors are not shown. The
plots of all of the measured data are shown in Appendix B.
In the plots shown in Appendix B it can be seen that the
patterns start to break down when steering past 40 degrees.
This agrees well with the simulated results of the four-
Figure. 28 - 33 GHz pattern measurements
With most of the plots shown here there are still some
noticeable side lobes and main beam deformations. It can be
shown that the channel amplitude errors have a more
significant effect than phase errors in terms of pattern
distortion. Therefore, based on earlier discussions, there
may be noticeable channel amplitude errors due to
insufficient channel isolation and the varying phase shifter
insertion loss. Thus, it can be speculated that the pattern
distortion shown in these plots are generated from these
Figure 29 - 33.5 GHz pattern measurements
Figure 30 - 34 GHz pattern measurements
Figure 31 - 34.5 GHz pattern measurements
Figure 32 - 35 GHz pattern measurements
6. DISCUSSION AND CONCLUSION
To our knowledge, this represents the first passive
electronically-steerable phased array slat populated with
packaged MEMS devices at Ka-band. The Raytheon
MEMS-based phase shifters developed under the CERDEC
MTO are significantly mature but not the only 4-bit phase
shifters known to the Government at Ka-band. They were
chosen for this task due to the collaborative environment
with CERDEC and AMRDEC.
The Ka-band linear array presented in this report
demonstrates the capability to design and fabricate a fully
integrated MEMS phase shifter phased array with angle
scanning. Additional efforts in the following areas are
needed to continue to mature and demonstrate the
capabilities of MEMS RF devices
There is an interest in redesigning the microwave
board. The items that should be considered in a
redesign are the power divider, the addition of
channel gain control and the antennas. The power
divider could be redesigned to have more channel to
channel isolation to aid in mitigating amplitude and
phase errors due to varying reflections when changing
phase shifter settings. The addition of gain control
allows one to more accurately control the relative
channel amplitudes which would help reduce sidelobe
levels. In this program it was found that the antennas
themselves were very fragile due to how thin the
substrate material was. It would require minimal
effort to develop a more rigid design.
Another area of future development is the extension
of the linear array to two dimensions. At present, the
linear array is arranged with elements in a side-by-
side configuration. If the other dimension is to be
used the elements will be in a stacked configuration
and could potentially have different performance in
this arrangement. The element patterns will probably
not be significantly affected, but there is the
possibility of more element coupling (i.e., lower
isolation) between stacked elements when compared
with the side-by-side linear array. This effect should
be investigated to evaluate the scanning performance
of the full, two-dimensional array.
Figure A.1 - 33 GHz phase measurements vs channel
number for each phase shifter setting
Figure A.2 - 33.5 GHz phase measurements vs channel
number for each phase shifter setting
Figure A.3 - GHz phase measurements vs channel
number for each phase shifter setting
Figure A.4 - 34.5 GHz phase measurements vs channel
number for each phase shifter setting
Figure A.5 - GHz phase measurements vs channel
number for each phase shifter setting
Figure B.1 - 33 GHz pattern measurements Plot 1
Figure B.2 - 33 GHz pattern measurements Plot 2
Figure B.3 - 33.5 GHz pattern measurements
Figure B.4 - 33.5 GHz pattern measurements
Figure B.5 - 34 GHz pattern measurements
Figure B.6 - 34 GHz pattern measurements
Figure B.7 - 34.5 GHz pattern measurements
Figure B.8 - 34.5 GHz pattern measurements
Figure B.9 - 35 GHz pattern measurements
Figure B.10 - 35 GHz pattern measurements
Special thanks to all members of the team for the
collaborative environment which made this possible,
including CERDEC, AMRDEC, and Raytheon. Also
thanks to the support contractors who make everything
 Goldsmith, C.L.; Zhimin Yao; Eshelman, S.; Denniston, D.;
Performance of low-loss RF MEMS capacitive switches Microwave and
Guided Wave Letters, IEEE [see also IEEE Microwave and Wireless
Components Letters] Volume 8, Issue 8, Aug. 1998 Page(s):269 – 271
 Pillans, B.; Morris, F.; Chahal, P.; Frazier, G.; Jeong-Bong Lee;
Schottky barrier contact-based RF MEMS switch, Micro Electro
Mechanical Systems, 2007. MEMS. IEEE 20th International Conference on
21-25 Jan. 2007 Page(s):167 - 170IMS 2007 – Honolulu, Hawaii
 "Near Hermetic Wafer Level Packaging Using LCP Adhesive Layer,"
Cody Moody, Brandon Pillans, Andrew Malczewski, and Francis J. Morris,
IMAPS International Conference
Scottsdale/Fountain Hills, AZ (9-12 March 2009) WP26.
 Pillans, B.; Coryell, L.; Malczewski, A.; Moody, C.; Morris, F.; Pillai,
V.; “RF MEMS Phase Shifter Manufacturing Technology,” Defense
Manufacturing Conference 2008, Orlando, FL.
 Atwater, Harry A., “Circuit Design of the Loaded-Line Phase Shifter,”
IEEE Transactions on Microwave Theory and Techniques, Vol. MTT-33,
No. 7, July 1985, pp. 626-634.
 Opp, Francis L. and Hoffman, W. F., “Design of Digital Loaded-Line
Phase-Shift Networks for Microwave Thin-Film Applications,” IEEE
Transactions on Microwave Theory and Techniques, Vol. MTT-16, No. 7,
July 1968, pp. 462-468.
 Rebeiz, Gabriel M., RF MEMS, Theory, Design, and Technology, John
Wiley & Sons, Inc., New Jersey, 2003, pp. 221-257.
 P. J. Gibson, “The Vivaldi Aerial,” in Proc. 9th Eur. Microwave Conf.,
Brighton, U.K., Sept.. 1979, pp. 101–105.
 E. Gazit, “Improved Design of the Vivaldi Antenna,” IEE Proc., Part
H, Vol. 135, No. 2, 1988, pp. 89–92.
on Device Packaging,
 S. G. Kim and K. Chang, “A low cross-polarized antipodal Vivaldi
antenna array for wide-band operation,” in Proc. IEEE Int. AP-S Symp.,
Monterey, CA, Jun. 2004, pp. 2269–2272.
Janice C. Rock was born in Birmingham, Alabama. She
has an M.S.E. in Electrical Engineering from the University
of Alabama in Huntsville. Her main area of research is in
the field of phased array antenna systems. Other interests
are in MEMS, semi-conductor antennas, and chip-level
component integration issues.
She is a research engineer with the RF Technology
Division in the Applied Sensors, Guidance, and Electronics Directorate,
Aviation and Missile, Research, Development and Engineering Center;
Redstone Arsenal, AL.
Tracy D. Hudson has earned Bachelor and Master's
degrees in electrical engineering from the University of
Alabama in Huntsville in 1988 and 1991 respectively. Dr.
Hudson also earned a Ph.D in engineering from The
Pennsylvania State University from State College, PA in
He is a Senior MEMS Research Engineer with the U.S.
Army’s RDECOM AMRDEC at Redstone Arsenal, Alabama. He has
almost 25 years experience developing component technology for Army
rotorcraft and missile systems at the U. S. Army located in Redstone
Arsenal, AL. He has publish or presented over 80 papers within his career
thus far (most being peer-reviewed journal articles). He has been
recognized for his scientific achievements on numerous occasions to
include the bestowal of the U. S. Army’s Research and Development
Award in 1991 and 2003. His current technical interests include
microfabrication process development, MEMS device prototyping (inertial,
microwave, chemical sensing, acoustic sensors, piezoelectric actuators and
materials, electro-optical materials, and power generation/management),
nanotechnology process development, and nanotechnology applications.
Dr. Hudson is also a member of The Optical Society of America, and a
Fellow of SPIE-The International Optical Engineering Society.
Dr. Brandon Pillans graduated with a Bachelor of Science
in Electrical Engineering from Texas A&M University in
1996. He received his Master of Science in Electrical
Engineering from Texas A&M University in 1998 and his
Ph.D. from the University of Texas at Dallas in 2006. He
began working at Raytheon Systems Company in 1998
performing RF MEMS switch research and development.
While in the RF MEMS group, Brandon has performed many RF MEMS
MMIC designs such as a Ka-Band phase shifter that has measured world
record performance in addition to tunable bandpass and bandstop filters and
tunable impedance matching networks. He has also worked extensively on
increasing the reliability of the capacitive RF MEMS switch, resulting in
six orders of magnitude increase in switch lifetime over that period. He has
written over 16 published articles on MEMS technology and holds three RF
MEMS patents. Brandon is currently the technical lead of the RF MEMS
group at Raytheon-Dallas.
Brandon Wolfson was born in Huntsville, Alabama in
1975. He received his B.S and M.S. degrees, both in
electrical engineering, from Auburn University, Auburn,
AL, in 1999 and 2001, respectively.
From 1999 to 2001 he worked as a teaching and
research assistant at Auburn University. Since then he has
worked for Technology Service Corporation (TSC) Phase
IV Systems Operation for eight years where he has served as a program
manager, system engineer, lead design engineer on many a RF/microwave
and multidisciplinary design programs. He has authored multiple
publications as well as been the recipient of a patent for his work.
17 Download full-text
Andrew Brown received his PhD in electrical engineering
from the University of Michigan, Ann Arbor, in 1999. His
main area of research wasapplied electromagnetics
focusing on improving component design through the
use of micromachining techniques. His current area of
focus includes, but is not limited to, microwave and
millimeter-wavetunable filters, MEMS based phase
shifters, and resonant MEMS structures.
He is currently President of A. Brown Design in Northville, Michigan, a
small company focusing on contract design and consulting services in
electrical engineering using advanced or novel technologies.
Louis A. Coryell earned the degree of Bachelor of Science in
Electrical Engineering from Newark College of Engineering,
Newark, NJ in 1968. In 1972 he earned the degree of Master
of Science in Electrical Engineering from Fairleigh
Dickinson University, Teaneck, NJ. He earned the degree of
Master of Science in Business Administration with a minor in
contract law from Fairleigh Dickinson
Rutherford, NJ in 1985.
He is Electronics Engineer in the Tactical Satellite/Airborne
Communications Branch, Satellite Communications Division in the Space
& Terrestrial Communications
Communications & Electronics Research, Development & Engineering
Center located at Fort Monmouth New Jersey. He is responsible for the
technology development for all on-the-move SATCOM Antenna systems
and component development programs, including Small Business
Innovative Research and Manufacturing Technology. His efforts also
include working with the Army Research Laboratory and Defense
Advanced Research Products Agency (DARPA) on transitioning new
technology developments into CERDEC SATCOM programs. He has over
40 years of research and development experience in the areas of satellite
communications systems, optical fiber communications systems, optical
communications systems and related components. Mr. Coryell has
authored or coauthored over 50 technical papers.
Directorate of the US Army