IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO.

10, OCTOBER 2012 4589

A 7–21 GHz Dual-Polarized Planar Ultrawideband

Modular Antenna (PUMA) Array
Steven S. Holland, Member, IEEE, Daniel H. Schaubert, Fellow, IEEE, and Marinos N. Vouvakis, Member, IEEE

Abstract—The design, fabrication, and measurement of a 16 16

dual-polarized planar ultrawideband modular antenna (PUMA)
array operating over 7–21 GHz (3:1 bandwidth) are presented.
The array is comprised of tightly coupled dipoles printed on a
grounded dielectric substrate and are excited by an unbalanced
feeding scheme that eliminates external wideband baluns and feed
organizers. The array can be assembled modularly, where each
low-profile, fully planar, low-cost tile is fabricated using standard
multilayer microwave PCB techniques. A unique solderless,
modular interconnect mates the array to a dilation fixture that
facilitates measurements using standard surface-mount assembly
(SMA) connectors and terminations. After presenting the most
critical design trends, simulation results of the final array in
infinite, infinite finite, and finite finite models are compared
with measurements. This prototype array exhibits a measured
active and close to ideal gain at broadside, and Fig. 1. Feeding topologies of tightly coupled, planar, ultrawideband dipole ar-
with low cross-polarization out to in all rays. (a) The CSA and FAA balanced feeding method requires a feed organizer
planes, showing close agreement with simulations. and an external balun at each element. (b) The PUMA feeding scheme uses un-
balanced lines and shorting vias only, allowing fully planar fabrication.
Index Terms—Antenna arrays, antenna feeds, dipole antennas,
mutual coupling, phased arrays, planar arrays.
Phased arrays offer low-profile, low-weight, electronically
I. INTRODUCTION steerable solutions that can support multiple independently
scanned beams [4]. Narrowband microstrip patch arrays have

C OMMUNICATION systems are becoming increasingly
mobile and desire low-cost, low-profile antennas that
cover multiple bands at once and can be integrated seamlessly
with platforms. The antennas in these systems must have high
(gain to noise temperature ratio) over multiple bands with
dual-polarized operation and low cross-polarization over wide
scan angles.
Reflector antennas have been used extensively in multi-band
communication systems, e.g., C/Ku- [1], Ku/K/Ka- [2], or
K/Ka-bands [3]. Reflectors offer high gain and low cross-po-
larization over a wide scan volume, but support only a single
mechanically steered beam. This steering is slow, susceptible to
the inertia of the platform, and prone to failure due to moving
parts. Additionally, these systems are inherently non-conformal
and bulky, which increases the radar cross section (RCS).
Manuscript received June 13, 2011; revised February 21, 2012; accepted May
14, 2012. Date of publication July 10, 2012; date of current version October 02,
2012. This work was supported by the Naval Research Laboratory under Grant
PG#11320000000008.
S. S. Holland was with the Center for Advanced Sensor and Communica-
tion Antennas, Electrical and Computer Engineering Department, University of
Massachusetts at Amherst, Amherst, MA 01003-9292 USA. He is now with the
MITRE Corporation, Bedford, MA 01730-1420 USA (e-mail: sholland@mitre.
org).
D. H. Schaubert and M. N. Vouvakis are with the Electrical and Computer
Engineering Department, University of Massachusetts-Amherst, Amherst, MA
01003 USA (e-mail: schaubert@ecs.umass.edu; vouvakis@ecs.umass.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TAP.2012.2207321
been developed with single-band [5]–[7] or multi-band [8],
[9] operation having single- or dual-polarization. Their printed
fabrication is low-cost, conformal, and frequency scalable up
to Q-band on soft substrates.
While these antennas can cover one or two bands, future plat-
forms desire apertures that can simultaneously cover many bands
and be able to host other systems via a so-called multifunctional
aperture [10]. One such example is a “crossed-notch” Vivaldi
array, [11], operating over 7–21 GHz with wide-scanning and
low cross-polarized radiation. However, the elements of this
array are not planar, have a relatively high profile, and require
elaborate feeding with internal hybrids to avoid resonances.
In an effort to reduce cost and profile, a number of printed-el-
ement UWB arrays have been proposed. The current sheet an-
tenna (CSA) array has demonstrated a 9:1 bandwidth, operating
up to Ku-band (2–18 GHz) [12], [13]. The fragmented aper-
ture array (FAA) [14], [15], has shown designs with 33:1 band-
widths up through X-band (300 MHz–10 GHz), and low fre-
quency long slot arrays have achieved bandwidths up to 10:1
(200–2000 MHz) [16]. For bandwidths greater than 4:1, the
latter two arrays must be backed with lossy R-card or ferrite
loaded ground planes that introduce 2–3 dB of loss, reducing
. More importantly, these printed-element arrays are
the
not easily scalable to higher frequencies since only the radi-
ating aperture is planar. Their balanced feed lines require elec-
trical shielding by non-planar feed organizers [15], as shown in
Fig. 1(a), to prevent E-plane scan resonances [17]. These “feed
organizers” preclude truly planar fabrication and modularity,

0018-926X/$31.00 © 2012 IEEE

Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
 4590 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012
Fig. 2. 16 16 dual-polarized 7–21 GHz PUMA array prototype mounted on
a measurement fixture.
and their mechanical tolerances limit fabrication to relatively
low frequencies. Moreover, their balanced feed lines must be
fed differentially from external wideband baluns that when pas-
sive are bulky and lossy, and when active are unidirectional and
handle only low power, [18].
This paper presents the design, fabrication, and measure-
ment of a modular 16 16, dual-polarized planar ultrawide-
band modular antenna (PUMA) array [19], [20], prototype,
shown in Fig. 2, operating over 7–21 GHz. The theory
and principles of operation of the PUMA array were pre-
sented in [19]. The array is fabricated as a single low-profile
thickness mm multilayer microwave
PCB, as shown in Fig. 2. The prototype cost is less than $2/el-
ement, and the PCB fabrication is mechanically robust and
scalable to higher frequencies. The dual-offset, dual-polarized
arrangement, along with the feeding scheme of Fig. 1(b), allows
for convenient split locations that further simplify construc-
tion and maintenance by allowing the array to be assembled
from modular tiles. As shown in Fig. 1(b), this array is fed
unbalanced and has additional shorting vias that eliminate feed
organizers and external wideband baluns [19].
The array performance is first optimized using infinite array
unit cell simulations that are fast and incorporate important mu-
tual coupling effects. Then, 16 and 16 16 finite array sim-
ulations are used to examine the truncation effects of the finite
array.
Measurements of the 16 16 dual-polarized prototype are
performed using a custom solderless connection to an “ex-
pander” fixture that allows for low-cost, repeatable, and reliable
array characterization. Simulations and measurements are in
good agreement, demonstrating an active VSWR less than 2.1
at broadside, and less than 2.8 out to in the E-, D-,
and H-planes. The cross-polarization level of both predictions
and measurements is approximately 15 dB at in the
D-plane. Additionally, the measured broadside element gain
closely tracks the ideal aperture-limited gain.
The remainder of the paper is organized as follows.
Sections II and III describe the design and fabrication of
the prototype array and the measurement fixture, respectively.
Section IV presents measured and simulated impedance and
radiation pattern results. The paper concludes in Section V.
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
II. PROTOTYPE ARRAY
The 16 16 dual-polarized PUMA array prototype is fabri-
cated as a single multilayer PCB, as shown in Fig. 2. This PCB
contains tightly coupled horizontally (H-) and vertically (V-)
polarized dipoles that are arranged on a dual-offset (egg-crate),
dual-polarized lattice, as illustrated in the top view of Fig. 3(a).
Along with a lack of feed organizers, this arrangement is key to
achieving modularity, as it provides convenient split planes be-
tween the dipole feed lines that allows the array aperture to be
divided into arbitrarily sized modular tiles, depicted in Fig. 3(a)
with dashed lines ( and examples are shown).
To demonstrate this concept, a second 16 \ \16 dual-polar-
ized prototype is discussed in Section II-C that assembles the
array from four 8 8 dual-polarized tiles. Finally, holes of ra-
dius (visible in Figs. 2 and 3(a)) are drilled through the
PCBs in between the dipoles to move blind angles out of the
scan volume, [19].
A. Element Design
1) Overview: As shown in the side view of Fig. 3(b), the mul-
tilayer array PCB is comprised of three dielectric layers. The
H- and V-polarized dipoles are printed onto the top and bottom
sides of a thin layer of Rogers 5880 , dielec-
tric layer 2, ensuring a robust and uniform dipole layer sepa-
ration. The superstrate and substrate, dielectric layers 1 and 3,
are comprised of Rogers 5880LZ , which has an
extremely low -axis thermal expansion and is the lowest avail-
able permittivity PTFE that supports plated vias (used to realize
the feed lines and shorting posts). Both layers are thick, with
mm , where is the wavelength at
the highest operating frequency. When thick dielectric layers are
used in multilayer PCBs, as in this array, low thermal expansion
is critical in preventing layer separation at high temperatures.
Layer 3 appropriately spaces the dipoles from the ground plane,
while layer 1 improves matching and serves as a wide angle
impedance matching (WAIM) layer to improve scanning, [21].
A detailed top view sketch of the dipole element is shown in
Fig. 3(c), where the H- and V-polarized dipoles are identical and
have the parameters listed in Table I. The dipole arms are simple
in shape with wide ends of width . Dipole ends from orthog-
onal polarizations (different layers) overlap and form parallel-
plate capacitors with a separation dictated by the middle dielec-
tric layer thickness . This simple capacitor structure avoids
complications associated with interdigitated capacitors, such as
printing tolerances on the fingers (which affects frequency scal-
ability) and potential resonances due to the size of the interdig-
itations, [19]. The dipoles have a linear taper that connects the
wide capacitor ends to the narrow feed points of width .A
pair of plated vias of radius are used to form the balanced
feed line structure. As shown in the cross-sectional sketch of
Fig. 3(b), the right-side feed line is connected directly to ground
and the left-side via is connected to the inner conductor of a 50
coaxial line through a short solderless impedance transformer
mm , detailed in Section III-B. The most impor-
tant aspect of the PUMA is the additional pair of plated vias that
connect the dipole arms directly to the ground, which mitigates
a problematic common-mode resonance that otherwise plagues
unbalanced-fed dipole arrays [19], [22]. The separation of the
 HOLLAND et al.: 7–21 GHz DUAL-POLARIZED PUMA ARRAY
Fig. 3. 7–21 GHz dual-polarized PUMA array. (a) Top view of dipole layers
(double-sided printing), showing boundaries of possible modules. (b) Cross sec-
tion of the unit cell, showing the PCB stack and the solderless interconnect be-
tween the array and the measurement fixture. (c) Top view of a unit cell dipole
layer. The inset shows the parallel plate capacitor formed between overlapping
orthogonal dipole arms.
vias from the fed and grounded feed lines are and , and
are important design parameters that are documented in [19].
The impedance transformer region just below the
fed via in Fig. 3(b) also serves as a solderless interconnect
to an “expander” measurement fixture (see Section III) that
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
4591
TABLE I
7–21 DUAL-POLARIZED PUMA ARRAY PARAMETERS
Fig. 4. Effect of decreasing the taper width on the infinite, broadside
impedance of the PUMA array.
hosts the coaxial cables. The inner conductor of the coaxial
interconnect is realized with a spring-like “fuzz button,” [23],
[24], while the outer conductor is formed by the wall of a hole
in a thick aluminum plate, which is attached to the bottom
of the array PCB. An ECCOSTOCK LoK dielectric sleeve
with an inner and outer radius of mm
and mm centers and supports the “fuzz button” in the
hole. This low-density foam is strong enough to mechanically
support the “fuzz button,” yet its low permittivity provides a
high impedance with a small outer radius ,
thereby avoiding overlapping holes of the H- and V-ports. Each
“fuzz button” presses against a circular pad of radius (con-
nected to the excited via) on one side and the central conductor
of the measurement fixture’s coaxial cable on the other. This
arrangement simplifies assembly and conveniently allows the
modular tiles to be removed and reinstalled for maintenance.
Section III-B details the interconnect fabrication.
2) Tuning: The theory of operation developed in [19] was
used to guide the optimization carried out using Ansoft/Ansys
HFSS [25] infinite array unit-cell analysis. This section outlines
the electrical design trends for a few key parameters, focusing
mainly on the impedance since other parameters, such as cross-
polarization, are affected to the second order.
The effects of varying the taper width are shown in Fig. 4
(see Fig. 3(b) for reference plane). Decreasing provides a
smoother transition to the vertical feed line vias, and reduces
the large reactive and resistive swings over the band, thereby
decreasing at mid-band and increasing at high-fre-
quencies. A compromise arises between boosting at high
 4592 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012

Fig. 5. Effect of decreasing the middle dielectric layer thickness on the in-
finite, broadside impedance of the PUMA array.

Fig. 6. Smith chart illustrating the impact of increasing superstrate thickness

on the infinite, broadside impedance loci of the PUMA array. Loci
frequencies and providing a wide enough feed point to be shown over the frequency range 8.5–17 GHz to enhance clarity. The chart is
within fabrication clearance tolerances around the feed line vias. normalized to .
The middle dielectric layer thickness is another critical
parameter since it controls the separation of overlapping par-
allel-plate capacitors at the ends of the dipoles, which in turn
strongly affects the impedance of the PUMA design, as shown in
Fig. 5. As decreases, the capacitance in the parallel-plate re-
gion increases and the first resonance (loop-mode resonance in
[19]) moves down in frequency, though at the expense of a large
swing in and at this resonance that worsens the match.
Conversely, this increased capacitive loading reduces the vari-
ation in and at mid- and high-frequencies, centering
around and around 0 , providing a good
match to standard 50 systems. For this application, the stan-
dard material thickness mil provides a good compromise
between the low frequency limit of operation and the variation
in impedance over the band.
Proper design of the superstrate thickness can improve
matching at broadside (shown in Fig. 6) and at wide scan angles.
As increases in Fig. 6, the broadside impedance locus con-
tracts considerably towards the axis, improving the Fig. 7. Smith chart plot showing the simulated impedance loci of the proto-
type array, illustrating the transformation of the impedance into (see
impedance matching over the band. The locus for mil Fig. 3(b) for impedance reference planes) via the short solderless interconnect
has the smallest variation in impedance, and hence the best of impedance . The chart is normalized to .
impedance match, but this thickness moves the dielectric sur-
face-wave (scan blindness) onset near the high frequency limit
of the band, reducing the scan volume (see [19] for discussion circular plate of radius and the ground plane (see Fig. 3(b))
on surface waves). As a compromise, mil is found and has only a minor impact on the impedance.
to provide surface-wave free operation within the desired scan
volume. B. Finite Array Design
Finally, the effectiveness of the short solder- The array prototype was designed using the methodology
less impedance matching section (interconnect) is illustrated in described in [26], which exploits progressively computationally
Fig. 7. The interconnect parameters, particularly , are chosen demanding modeling scales, i.e., , , and 16 16
to provide an impedance of . The broadside PUMA array models, to efficiently incorporate the finite array truncation
impedance at the ground plane level (see Fig. 3(b) for effects into the design process (results shown in Section IV). The
reference plane) is strongly capacitive, and has a resistance first tier simulations use infinite array (unit cell) models that are
near mid-band. The short transformer rotates the locus fast and approximate the behavior of a central element in a large
clockwise around the transformer impedance array, since they only capture mutual coupling (no truncation
until it fits inside the VSWR 2 circle and is balanced above and effects), and are useful for tuning (Section II-A). The next level
below , as shown by . A parasitic shunt capac- utilizes slower, increased memory array simulations
itance, shown dashed, is due to the annular slot formed by the to evaluate truncation effects along a single finite dimension.

Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
 HOLLAND et al.: 7–21 GHz DUAL-POLARIZED PUMA ARRAY
Fig. 8. Modular and solderless assembly of a 16 16 dual-polarized PUMA array prototype on the measurement fixture. (a) Top plate of the expander fixture and
the mounting of an 8 8 module. (b) Full array mounted on measurement fixture, with the side panel removed to show the coaxial cables exiting the top plate and
terminating in SMA connectors on the backside of the fixture.
Finally, the highest level employs computationally demanding
16 16 array simulations that account for the full finite array
truncation effects. The analysis of truncation effects is critical in
predicting the performance of elements near array edges and at
low frequency bands of UWB arrays.
C. Fabrication
First, the dipole layers are etched onto each side of the 5-mil-
thick Rogers 5880 dielectric layer 2, which is then bonded to the
top of dielectric layer 3 (Rogers 5880LZ), using a 1.5-mil-thick
Gore Speedboard C prepreg film . Vias are then
plated through both layers, thereby avoiding the use of any blind
vias. To complete the PCB, dielectric layer 1 (Rogers 5880LZ)
is then bonded to the top of this stackup using a second Speed-
board C bond film. Next, a thick ( mm (80 mil))
aluminum plate is bonded to the bottom of the array PCB with
conductive epoxy. This plate contains holes of radius
mm that align with the excited feed line vias and form the
outer conductor of the solderless impedance transformer, de-
tailed in Section III-B. Finally, the holes shown in Fig. 2 are
drilled through the PCB dielectric layers in the space between
the dipoles. In addition to their electrical benefits, the holes offer
a practical mechanical advantage by providing convenient loca-
tions to insert screws that are used to fasten the array to the top
plate of the measurement fixture, described in Section III-A.
III. MEASUREMENT ASSEMBLY
Since this array prototype operates up to K-band ( just
above 21 GHz), grating lobe onset constrains and to
less than 7 mm. To connectorize the two ports of such a small
unit-cell, expensive G3PO connectors and loads must be used.
Moreover, measurements using G3PO connectors are often un-
reliable and not repeatable, making such connectorization un-
desirable. Instead, a solderless array interconnect and an “ex-
pander” (dilation) test fixture were designed and fabricated to
allow reliable, repeatable measurements with low-cost surface-
mount assembly (SMA) connectors and terminations at a frac-
tion of the cost.
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
4593
Fig. 9. Assembly of solderless interconnect. (a.) Dielectric sleeve and gold-
plated, beryllium-copper fuzz button. (b) Bottom view of an module,
showing the aluminum plate with fuzz buttons and dielectric cylinders installed.
A. Expander Fixture
The expander test fixture is shown in Fig. 8. The flat top
plate, which also serves as a ground plane, is shown
close-up in Fig. 8(a), displaying the grid of 512 flush-mounted
50 T-Flex 405 flexible coaxial cables that align with and ex-
cite the feed lines of the array. Shown in the inset are threaded
screw holes, which align with drilled dielectric holes of the
array PCB and allow the array to be securely fastened to the
fixture. Eight alignment pins (2/module) are used to properly
align the tiles with the coaxial cables. Assembly entails placing
the array on top of the fixture and securing it with screws,
as shown in Fig. 8(b). The other end of each T-flex cable is
connectorized with an SMA connector that is mounted on the
backside plate of the fixture. The SMA connector periodicity
is large enough to accommodate SMA loads and measurement
cables. Special SMA connectors are employed that operate up
to 27 GHz without moding.
B. Solderless Interconnect
As highlighted in Section II-C, the array PCB is first modified
to accommodate the coaxial solderless interconnect with an alu-
minum plate bonded to the bottom of the PCB. The aluminum
plate has drilled holes aligned with the feed line vias that will
house the interconnect and form the outer coaxial conductor.
 4594
Fig. 10. Active broadside VSWR distribution versus frequency (vertical axis) of the
port numbering scheme. (b) Active VSWR of H-polarized elements for the
V-polarized elements for the array (left) and for comparison the infinite array (right).
Next. a gold-plated, beryllium-copper, spring-like “fuzz button”
(fabricated by Custom Interconnects, LLC) with a nominal ra-
dius of 0.25 mm is inserted into each LoK dielectric sleeve. This
fuzz button and dielectric sleeve assembly is then placed inside
each of the holes in the aluminum plate, completing the inter-
connect. The fuzz button forms the inner coaxial conductor of
the interconnect, and the dielectric sleeve provides mechanical
support. Outer conductor electrical contact between the array
and expander plate is achieved through the pressure applied by
the closely spaced screws (one per unit cell). Fig. 9(a) shows a
close-up picture of the dielectric sleeve and the fuzz button, and
Fig. 9(b) shows the bottom side of an array tile, showing the
installed dielectric sleeves and fuzz buttons. Installation of the
fuzz buttons and dielectric sleeves can be automated for large
scale production. Alternatively, elastomeric interconnects can
be used.
IV. RESULTS
The impedance, far-field patterns, cross-polarization levels,
and gain performance of the 16 16 PUMA prototype, shown
in Fig. 2, were evaluated using multiple simulation tools and
are compared to measurements, [27].
The prototype array is mounted on a finite, alu-
minum ground plane (top plate of the expander fixture) without
any absorber treatments. In contrast, all simulations use an infi-
nite ground plane. All results are referred to an unbalanced 50
interface, with one polarization excited and the other termi-
nated in 50 .
A. Impedance Performance
Impedance results will be presented in the same order as the
modeling tools employed in the design approach outlined in
Section II-B. The orthogonal polarization coupling (between H-
and V-ports) was found to be below 20 dB over most of the
band, thus will be omitted.
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012
infinite array. (a) Top view sketch of simulation model and
array (left) and for comparison the infinite array (right). (c) Active VSWR of
1) Infinite Array Simulations: Infinite array, unit-cell simu-
lations (periodic boundary conditions on all walls) were carried
out with Ansoft/Ansys HFSS [25], using PEC conductors and
realistic, lossy dielectric models. A PML absorber terminates
the top of the unit cell, situated from the array. Infi-
nite array active VSWR predictions for scanning out to
in the E- and H-planes are shown in Fig. 11(a). For this and all
of the following cases, the D-plane VSWR is omitted since it
is an approximate average of the E- and H-plane VSWRs. At
in the E-plane, the active VSWR is approximately
equal to broadside, with a maximum of 2.1 at the hump near
8.5 GHz. At in the H-plane, the VSWR reaches 2.9
near GHz.
2) Array Simulations: A dual-polarized stick
of the array was simulated using Ansoft/Ansys HFSS with the
same simulation details as in Section IV-A-1, except for the peri-
odic boundary arrangement. A top view of the simulation model
is shown in Fig. 10(a), where absorbing boundary conditions
(ABC) truncate the top of the air box and on the two narrow
sides to create a finite dimension, and periodic boundary condi-
tions (PBC) provide an infinitely periodic dimension. The model
contains 16 H- and 16 V-polarized dipoles, which are aligned
with the finite and infinite dimensions, respectively.
The impact of truncation effects on the scan VSWR is ob-
served by phasing all 16 V-polarized dipoles (for broadside and
H-plane ) or all 16 H-polarized dipoles (for E-plane
) to scan along the finite dimension. Fig. 11(b) shows
the active VSWR performance of a central element. The broad-
side VSWR has a similar shape to the infinite array simulations,
though the VSWR exhibits a ripple. This ripple is an interfer-
ence pattern due to array guided surface waves (AGSWs), which
are excited at the edges of finite arrays, [26], and are not excited
in infinite arrays. The spectrum of ASGW excitations depends
on scan angle, and is responsible for the hump in the VSWR near
15.5 GHz at in the E-plane. Additionally, the low fre-
quency hump is seen to increase until it reaches 2.5 at .
 HOLLAND et al.: 7–21 GHz DUAL-POLARIZED PUMA ARRAY
Fig. 11. Simulated and measured active VSWR versus frequency and scan
angle of the fabricated PUMA array. (a) Simulated active VSWR of infinite
array (unit cell) simulation. (b) Central element active VSWR of simu-
lation. (c) Measured central element active VSWR of the 16 16 prototype.
The active VSWR at in the H-plane is also shown in
Fig. 11(b). The VSWR contains similar ripple behavior as in the
E-plane, due to AGSWs, including an increase near 13.5 GHz
and 18 GHz. The increase in VSWR at 8.5 GHz is similar to the
infinite array H-plane results.
To get a better picture of the phenomena taking place at
each element in the array, Fig. 10(b) and (c) shows
the broadside active VSWR versus element location and fre-
quency. Fig. 10(a) shows the port numbering for the H- and
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
4595
V-polarized dipoles in the model. The left side of Fig. 10(b)
shows the broadside active VSWR at each H-polarized element
(all 16 H-polarized elements are excited), and is compared
side-by-side with the infinite array results for the same port
on the right. Oscillations in the VSWR distribution across
the elements are due to the AGSWs. The edge element H1 is
significantly affected by truncation effects, since it is only a
half-dipole, as seen in Fig. 10(a). Edge element H16 demon-
strates only minor truncation effects because it has both dipole
arms, highlighting the importance of preserving both arms of
the edge elements. Similarly, the V-polarized active VSWR
distribution is shown in Fig. 10(c) along with the respective
infinite array results on the right. Once again, only the edge
elements are seriously impacted by truncation, and primarily at
the low-end 8.5-GHz hump.
3) 16 16 Array Simulations: Finite 16 16 array simula-
tions were carried out using inhouse domain decomposition
FEM (DD-FEM) codes, [27]–[29]. These simulations are
full-wave and can capture the full finite array truncation effects.
Fig. 12 shows the active VSWR distribution over the array at
8.75, 12, and 15.5 GHz, with the H-polarized ( -directed)
elements excited. Note that for all cases, column X1 has a high
VSWR due to the half-dipole, as described in Section IV-A2.
At low frequency, in Fig. 12(a), the active VSWR increases to
3 at the edges of the array (row Y1 and Y16), and the central
element exhibits a VSWR of 2.7. A periodic oscillation is
seen in the VSWR of alternating rows and columns (particu-
larly X4, 9, and 13) due to the AGSWs that are excited and
reflect off the edges of the array, [26], which was also seen in
Section IV-A2. Similar, but less severe effects are seen at mid
and high-frequencies in Fig. 12(b) and (c).
4) Prototype Measurements: The S-parameters ( ’s) of the
16 16 dual-polarized prototype were measured in an anechoic
chamber using an Aglient PNA. The active reflection coefficient
of a center element is calculated from [26]
(1)
where is the array scan direction; are the
measured S-parameters between elements and ;
and are the number of elements and the el-
ement spacings, respectively, along the - and -dimensions;
, are the coor-
dinates; and is the free-space wavenumber. Time-gating was
used to remove most reflections from the cables in the expander
fixture. The measured active VSWR is shown in Fig. 11(c),
for broadside and scan in the E- and H-planes.
At broadside, the VSWR shows a similar overall shape as the
array results in Fig. 11(b), exhibiting a ripple at low- and
mid-band, though with a higher mid-band VSWR level. The
rapid ripple may be attributed to some residual reflections from
the cabling in the expander fixture. For scanning in the E-plane,
a VSWR hump near 15 GHz is seen in both the measured and
results. Similarly, humps near 13.5 and 18 GHz at
in the H-plane VSWR increase by approximately the
same level above the broadside VSWR as in the case,
further reinforcing the hypothesis that this increase is attributed
 4596 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012

Fig. 13. Measured central element active VSWR versus scan angle and fre-
quency. (a) E-plane scan. (b) H-plane scan.

match over the band for scanning in both planes. Overall, the
measurements show good agreement with the simulations and
demonstrate a out to in all planes.

B. Far-Field Pattern Performance

The radiation behavior of the PUMA prototype is explored
using infinite and finite array simulations and measured em-
bedded element patterns.
1) Infinite Array Analysis: Using the same simulation model
as described in Section IV-A1, the normalized co- and cross-po-
larized radiated power levels per unit cell are computed based
on Ludwig’s third definition [30] versus frequency for scan-
ning out to in the E- and D-planes and are shown in
Fig. 14 (H-plane cross-polarization is similar to the E-plane, and
is omitted).
Fig. 12. Simulated active broadside VSWR distribution at each element of the The co-polarized radiated power in both planes is nearly 0
16 16, 7–21 GHz dual-polarized PUMA array, with the H-polarized ( -di- dB over most of the band, showing a maximum mismatch loss
rected) elements excited. (a) GHz. (b) GHz. (c)
GHz. of approximately 0.8 dB when scanned to , demon-
strating good mismatch efficiency over the band. In the E-plane,
the cross-polarization is shown to be below 20 dB over most
to truncation effects. Figs. 13(a) and (b) show the measured of the band out to , though this level increases towards
E- and H-plane active VSWR of a central element versus scan the low- and high-frequency edges of the band. The increase at
angle and frequency calculated for , showing a good low frequency is due to a low-frequency loop-mode resonance

Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
 HOLLAND et al.: 7–21 GHz DUAL-POLARIZED PUMA ARRAY 4597

Fig. 14. Simulated infinite PUMA array co- and cross-polarized radiated
powers per unit cell versus frequency and scan (Ludwig’s third definition).
Both radiated powers are normalized to the input power. (a) E-plane scan.
(b) D-plane scan. The H-plane cross-polarization is similar to the E-plane, thus
is omitted.

[19] that is located just below 7 GHz and radiates cross-polar-

ized fields. The D-plane cross-polarization levels of Fig. 14(b)
are shown to be less than 20 dB out to , and less than
15 dB out to . Simulations in multiple planes, over
have revealed similar low cross-polarization be-
havior.
2) Embedded Element Patterns: The embedded element pat-
terns dictate the array’s gain variation with scan and the rela-
tive co- and cross-polarization levels. To perform this measure-
ment, the array was mounted on an azimuth positioner in the
UMass tapered far-field chamber; as highlighted previously, no
absorber was placed on or near the array or ground plane. The
embedded element patterns are found by exciting a central ele- Fig. 15. Measured and simulated central embedded element patterns at
ment and measuring the resultant radiation pattern, with all other GHz of the 16 16, 7–21 GHz dual-polarized PUMA array, showing co-
elements terminated. Results were also validated via near-field and cross-polarized patterns. (a) E-plane. (b) D-plane. (c) H-plane.
scanner measurements.
For this PUMA prototype, E-, D-, and H-plane embedded el- The patterns approach the ideal gain variation, indicating
ement patterns are shown in Figs. 15–17 for 7.5, 12, and low scan-loss in all planes out to , where deep oscilla-
20 GHz, respectively. The measurements are compared to the tory dips are seen in the E-plane patterns. A direct comparison
results of the finite array DD-FEM simulations (see of cross-polarization ratios calculated for the infinite array (see
Section IV-A-3). Note that the array is mounted on a Fig. 14) and 16 16 array simulations and the measurements are
ground plane, whereas the simulations assume an infinite summarized in Table II, showing good agreement.
ground plane. Measured and simulated co- and cross-polarized
patterns are in good agreement. Oscillations in the patterns are C. Embedded Element Gain
due to the small, finite size of the array and ground plane and will Up to now, only the relative measured power levels have been
decrease in strength as the array and ground plane sizes increase. presented, which give no indication of the actual radiation effi-

Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
 4598 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012

Fig. 16. Measured and simulated central embedded element patterns at Fig. 17. Measured and simulated central embedded element patterns at
GHz of the 16 16, 7–21 GHz dual-polarized PUMA array, showing co- GHz of the 16 16, 7–21 GHz dual-polarized PUMA array, showing co-
and cross-polarized patterns. (a) E-plane. (b) D-plane. (c) H-plane. and cross-polarized patterns. (a) E-plane. (b) D-plane. (c) H-plane.

ciency of the array. Fig. 18 shows the measured absolute broad-
side co- and cross-polarized gains of an embedded central ele-
ment. For comparison, the ideal gain of the unit cell aperture
is plotted, , where is the
unit cell area, is the active reflection coefficient, and is
the free-space wavelength. The co-polarized gain closely tracks
the ideal gain to within 2 dB; the oscillations above and below
the ideal gain curve are attributed to the finite size of the array
and ground plane. The cross-polarized gain is 20 dB below the
co-polarized gain over most of the band.
V. CONCLUSION
A low-cost 16 16 dual-polarized PUMA array prototype op-
erating over 7–21 GHz (3:1 bandwidth) was presented. The
array is low-profile, and connects directly to standard unbal-
anced 50 interfaces without external baluns or hybrids. The

Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
 HOLLAND et al.: 7–21 GHz DUAL-POLARIZED PUMA ARRAY
TABLE II
COMPARISON OF SIMULATED AND MEASURED
CO-TO-CROSS-POLARIZATION RATIOS
Fig. 18. Measured broadside embedded element absolute gain for the 16 16
dual-polarized PUMA array.
measured array prototype is fabricated as a single multilayer mi-
crowave PCB, which allows robust, fully planar fabrication and
low-cost, high-volume production. A second prototype is fabri-
cated as four tiles that demonstrate the feasibility of
modular assembly. Additionally, a unique solder-less intercon-
nect and measurement dilation fixture was specially designed
and fabricated to allow repeatable, reliable characterization of
the array without resorting to miniature G3PO connectors. Mea-
surements and simulations are in very good agreement and in-
dicate that the central elements of the prototype array maintain
a broadside over the full operating band and
for scans out to . Simulated and mea-
sured embedded element patterns indicate close to ideal abso-
lute broadside gain and cross-polarization levels below 15 dB
for scans out to in all planes over most of the band.
ACKNOWLEDGMENT
The authors would like to thank Dr. R. Kindt of the Naval
Research Laboratory for helpful discussions regarding the sol-
derless interconnect, J. Salazar for assisting with the pattern
measurements, and G. Paraschos for helping with the DD-FEM
simulations.
REFERENCES
[1] R. J. Bauerle, G. Gothard, and A. Vergamini, “A center fed multi-band
antenna for simultaneous satellite communication at C and Ku bands,”
in Proc. IEEE Military Communications Conf. (MILCOM), 2010, p.
1564.
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
4599
[2] S. Y. Eom, S. H. Son, Y. B. Jung, S. I. Jeon, S. A. Ganin, A. G. Shubov,
A. K. Tobolev, and A. V. Shishlov, “Design and test of a mobile an-
tenna system with tri-band operation for broadband satellite communi-
cations and DBS reception,” IEEE Trans. Antennas Propag., vol. 55,
no. 11, pp. 3123–3133, Nov. 2007.
[3] A. C. Densmore and V. Jamnejad, “A satellite-tracking K- and Ka-band
mobile vehicle antenna system,” IEEE Trans. Veh. Technol., vol. 42, no.
4, pp. 502–513, Nov. 1993.
[4] P. Bocon, T. McGree, and J. Renfro, “Phased array performance char-
acteristics and compliance with SATCOM military standards,” in Proc.
IEEE Military Commun. Conf. (MILCOM), Oct. 17–20, 2005, vol. 2,
p. 1712.
[5] M. E. Bialkowski and N. C. Karmakar, “A two-ring circular phased-
array antenna for mobile satellite communications,” IEEE Antennas
Propag. Mag., vol. 41, no. 3, pp. 14–23, Jun. 1999.
[6] S. Vaccaro, F. Tiezzi, D. Llorens, M. F. Rua, and C. de Oro, “Ku-band
low profile antennas for mobile SATCOM,” in Proc. Adv. Satellite Mo-
bile Syst. (ASMS), Aug. 2008, pp. 24–28.
[7] T. Lambard, O. Lafond, M. Himdi, H. Jeuland, S. Bolioli, and L. Le
Coq, “Design of a Ka-band wide scanning phased array antenna,” in
Proc. 3rd Eur. Conf. Antennas Propag. (EuCAP), Mar. 2009, p. 1247.
[8] O. Kilic, A. I. Zaghloul, E. C. Kohls, R. K. Gupta, and D. Jimenez,
“Flat antenna design considerations for satellite-on-the move and satel-
lite-on-the-pause (SOTM/SOTP) applications,” in Proc. IEEE Military
Commun. Conf. (MILCOM), Aug. 2001, vol. 2, pp. 790–794.
[9] X. Qu, S. S. Zhong, Y. M. Zhang, and W. Wang, “Design of an S/X
dual-band dual-polarised microstrip antenna array for SAR applica-
tions,” IET Microw., Antennas Propag., vol. 1, no. 2, pp. 513–517, Aug.
2007.
[10] G. C. Tavik, C. L. Hilterbrick, J. B. Evins, J. J. Alter, J. G. Crnkovich,
J. W. de Graaf, W. Habicht, G. P. Hrin, S. A. Lessin, D. C. Wu, and S.
M. Hagewood, “The advanced multifunction RF concept,” IEEE Trans.
Microw. Theory Tech., vol. 53, no. 3, pp. 1009–1020, Mar. 2005.
[11] K. Trott, B. Cummings, R. Cavener, M. Deluca, and J. Biondi, “7–21
GHz wideband phased array radiator,” in Proc. IEEE Antennas Propag.
Soc. Int. Symp., Jun. 20–25, 2004, vol. 3, p. 2265.
[12] B. Munk et al., “A low-profile broadband phased array antenna,” in
Proc. IEEE Antennas Propag. Soc. Int. Symp., Jun. 22–27, 2003, vol.
2, pp. 448–451.
[13] M. Jones and J. Rawnick, “A new approach to broadband array de-
sign using tightly coupled elements,” in Proc. IEEE Military Commun.
Conf. 2007, Oct. 2007, pp. 1–7.
[14] B. Thors, H. Steyskal, and H. Holter, “Broad-band fragmented aperture
phased array element design using genetic algorithms,” IEEE Trans.
Antennas Propag., vol. 53, no. 10, pp. 3280–3287, Oct. 2005.
[15] W. Croswell, T. Durham, M. Jones, D. Schaubert, P. Friederich, and
J. Maloney, “Wideband arrays,” in Modern Antenna Handbook, C. A.
Balanis, Ed. New York: Wiley, 2008.
[16] J. J. Lee, S. Livingston, and D. Nagata, “A low profile 10:1 (200–2000
MHz) wide band long slot array,” in Proc. IEEE Antennas Propag. Soc.
Int. Symp., Jul. 5–11, 2008, pp. 1–4.
[17] J. R. Bayard, D. H. Schaubert, and M. E. Cooley, “E-plane scan per-
formance of infinite arrays of dipoles printed on protruding dielectric
substrates: Coplanar feed line and E-plane metallic wall effects,” IEEE
Trans. Antennas Propag., vol. 41, no. 6, pp. 837–841, Jun. 1993.
[18] K. Jung, W. R. Eisenstadt, R. Fox, A. Ogden, and J. Yoon, “Broadband
active balun using combined cascode-cascade configuration,” IEEE
Trans. Microw. Theory Tech., vol. 56, no. 8, pp. 1790–1796, Aug. 2008.
[19] S. S. Holland and M. N. Vouvakis, “The planar ultrawideband modular
antenna (PUMA) array,” IEEE Trans. Antennas Propag., vol. 60, no.
1, pp. 130–140, Jan. 2012.
[20] S. S. Holland and M. N. Vouvakis, “The planar ultrawideband modular
antenna (PUMA) Array,” U.S. Patent Application 61/230,271, Jul. 31,
2009.
[21] E. G. Magill and H. A. Wheeler, “Wide-angle impedance matching of
a planar array antenna by a dielectric sheet,” IEEE Trans. Antennas
Propag., vol. AP-14, no. 1, pp. 49–53, Jan. 1966.
[22] S. S. Holland and M. N. Vouvakis, “The banyan tree antenna array,”
IEEE Trans. Antennas Propag., vol. 59, no. 11, pp. 4060–4070, Nov.
2011.
[23] D. Carter, “‘Fuzz Button’ interconnects at microwave and mm-wave
frequencies,” in Proc. Packaging and Interconnects at Microwave and
mm-Wave Frequencies IEE Seminar, 2000, pp. 3/1–3/6.
[24] R. W. Kindt and W. R. Pickles, “Ultrawideband all-metal flared-notch
array radiator,” IEEE Trans. Antennas Propag., vol. 58, no. 11, pp.
3568–3575, Nov. 2010.
[25] Ansys/Ansoft HFSS ver. 11.2 [Online]. Available: www.ansoft.com
 4600 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 10, OCTOBER 2012
[26] M. N. Vouvakis and D. H. Schaubert, “Vivaldi antenna arrays,” in
Chapter 3 in Frontiers in Antennas: Next Generation Design & En-
gineering, F. B. Gross, Ed. New York: McGraw-Hill, 2011.
[27] S. S. Holland, “Low-profile, modular, ultra-wideband phased arrays,”
Ph.D. dissertation, Dept. Elect. Eng., Univ. of Massachusetts, Amherst,
2011.
[28] S.-C. Lee, M. N. Vouvakis, and J.-F. Lee, “A non-overlapping domain
decomposition method with non-matching grids for modeling large fi-
nite antenna arrays,” J. Comput. Phys., vol. 203, no. 1, pp. 1–21, Feb.
2005.
[29] G. N. Paraschos and M. N. Vouvakis, “A design methodology for fi-
nite phased arrays and repetitive structures,” in Proc. IEEE Antennas
Propag. Soc. Int. Symp., Jun. 1–5, 2009, p. 1.
[30] A. Ludwig, “The definition of cross polarization,” IEEE Trans. An-
tennas Propag., vol. 21, no. 1, pp. 116–119, Jan. 1973.
Steven S. Holland (S’05–M’11) was born in
Chicago, IL, in 1984. He received the B.S. degree in
electrical engineering from the Milwaukee School
of Engineering (MSOE), Milwaukee, WI, in 2006,
and the M.S. and Ph.D. degrees in electrical and
computer engineering from the University of Mass-
achusetts Amherst, in 2008 and 2011, respectively.
From 2006 to 2011, he was a Research Assistant
working in the Antennas and Propagation Laboratory
(APLab), Department of Electrical and Computer
Engineering, University of Massachusetts Amherst.
Currently, he is a Senior Sensors Engineer with the MITRE Corporation,
Bedford, MA. His research interests include ultrawideband antenna arrays,
electrically small antennas, Radar systems, and HF antennas.
Dr. Holland received the Best Student Paper Award at the 2010 Antenna Ap-
plications Symposium, Allerton Park, Monticello, IL, and is a member of Tau
Beta Pi.
Authorized licensed use limited to: SOCIETY FOR APPLIED MICROWAVE ELEC ENG AND RES. Downloaded on May 28,2020 at 12:06:56 UTC from IEEE Xplore. Restrictions apply.
Daniel H. Schaubert (S’68–M’74–SM’79–F’89)
received the Ph.D. degree in electrical engineering
from the University of Illinois, Urbana-Champaign.
He is Professor Emeritus of Electrical Engineering
and former Director of the Center for Advanced
Sensor and Communication Antennas at the Univer-
sity of Massachusetts. He worked at the U.S. Army
Research Laboratory and the U.S. Food and Drug
Administration prior to joining the University of
Massachusetts in 1982. His contributions have been
mainly in the areas of antenna design and analysis.
He has patents for conformal and printed circuit antennas. Several of his
antenna designs have been used in military and civilian systems for radars,
radiometers, and communications, and he has designed low-cost antennas for
commercial cellular and local area network products. He directed the design,
fabrication and testing of antennas for the cloud profiling radar system, a
polarimetric 33-GHz and 95-GHz mobile radar, and the high-altitude wind
and rain profiler, a dual-beam 13-GHz and 35-GHz airborne radar. He led the
design efforts for several multioctave scanning array antennas, including the
first prototypes for the Thousand Element Array demonstrator of the Square
Kilometer Array project.
Dr. Schaubert was President of the IEEE Antennas and Propagation
Society, Associate Editor of the IEEE TRANSACTIONS ON ANTENNAS AND
PROPAGATION, Secretary-Treasurer of the Society, Newsletter Editor and Mem-
bership Chairman. He organized the annual Antenna Applications Symposium.
He received the Distinguished Alumni Award for the University of Illinois
Electrical and Computer Engineering Department, the IEEE John Kraus
Antenna Award, the H. A. Wheeler Prize Applications Paper Award, and the
IEEE Third Millennium Medal. He was an advisor to the European Antenna
Centre of Excellence and was a member of the executive team for the IET
Antenna and Propagation Professional Network.
Marinos N. Vouvakis (S’99–M’05) He received the Diploma degree in
electrical and computer engineering from Democritus University of Thrace
(DUTH), Xanthi, Greece, in 1999, M.S. degree from Arizona State University
(ASU), Tempe, AZ, and the Ph.D. degree from The Ohio State University
(OSU), Columbus OH, both in electrical and computer engineering.
Since 2005, he has been with the University of Massachusetts (UMass)
Amherst Electrical and Computer Engineering faculty, where he is currently
an Associate Professor. He is a member of the Center for Advanced Sensor
and Communication Antennas (CASCA) and the Antennas and Propagation
Laboratory (APLab) at UMass. His research interests are in the area of
computational electromagnetics, with emphasis on domain decomposition, fast
finite element and integral equation methods, hybrid methods, model order
reduction, and unstructured meshing for electromagnetic radiation, scattering
and microwave/millimiterwave device applications. His interests extend to the
design and manufacturing of ultra-wideband phased array systems.