Synthetic Aperture

GPS Signal Processing

Concept and Feasibility Demonstration

DoD Photo by MSGT Paul N. Hayashi

Most people relate to GNSS as a technology for positioning, navigation, and timing.
However, space weather researchers have already demonstrated the use of GPS as
a useful sensor for studies of the Earth’s atmosphere. This article introduces the
concept of applying synthetically generated, phased-array antennas for processing
GPS signals to create large antenna apertures. In turn, the narrow-beam generation
capabilities of synthetic apertures can be used to mitigate interference and jamming
and for producing high-resolution radar images passively using received GPS signals
— which raise the possibility of some interesting civil and military applications.

Andrey Soloviev
University of Florida
Frank van Graas,
Sanjeev Gunawardena
Ohio University
Mikel Miller
Air Force Research Laboratory
Aerial view of a transporter-
erector-launcher vehicle covered
with camouflage netting, during
ground launch cruise missile (GLCM)
evaluation.
S
ynthetic aperture techniques
combine data obtained from
multiple sensors — or one sen-
sor moving among multiple
locations, or both — to construct a
single image. These techniques have
been widely researched, developed, and
applied in the area of radar systems.
This article discusses efforts to
extend synthetic aperture concepts to
GPS signal processing, exploiting the
beam steering capabilities of syntheti-
cally generated phased array antennas.
In it, we will describe the fast Fourier
transform (FFT)–based method used to
simultaneously steer a synthetic array’s
beams in multiple directions. We will
also discuss the results of simulator and
flight tests to demonstrate the efficacy of
synthetic beam steering techniques for
GPS antennas.
Development of GPS-based SARs
will enable high-resolution imaging
capabilities using passive receivers of
GPS signals and allow 24-hour global
availability of imaging technology. It
represents a dual-use technology that
could support military applications such
as imaging of military ground fleet hid-
den under foliage, as well as humani-
tarian applications such as detection of
unexploded ordnance.
GPS SAR: The Concept
Large synthetic apertures allow for pro-
ducing very narrow array beams. These

www.insidegnss.com m ay/ june 200 9 InsideGNSS 37

 synthetic aperture

Synthetic beam steering A hardware-based construction has several limitations. For

(interference and jamming mitigation) instance, an increase in the physical size of the array is required
GPS satellite to narrow its beamwidth. Moreover, adjusting the phases of
GPS-based synthetic aperture
radar (passive imaging) individual antennas in hardware constrains the system’s capa-
bility to simultaneously generate multiple beams that can be
Jammer Antenna
beam
Synthetic aperture
used, for example, to track multiple satellites or to simultane-
GPS satellite
GPS signal processing ously track both direct and reflected (multipath) signals.
To overcome these limitations, we propose a synthetic array
Synthetic aperture Antenna generation scheme that uses a software GPS receiver architec-
GPS signal processing beam ture. Instead of adding new antennas to the array, the beam is
narrowed by exploiting antenna motion — that is, the array is
FIGURE 1 Applications of synthetic aperture GPS signal processing synthesized by observing an antenna at different locations over
time. Figure 3 and Figure 4 illustrate this principle for 1D and
narrow beams are two-dimensional (2D) array cases, respectively.
steered in desired Generation of synthetic GPS antenna arrays is conceptually
Antenna beam 1
δθ~ N d i rec t ions usi ng similar to synthetic aperture radar, where antenna motion is
GPS signal process-
ing techniques.
d θ0 Incoming GPS signal S(t)
d As shown in Fig-
phase delay phase delay phase delay ure 1, an array beam
d cos(θ0) 2d cos(θ0) Nd cos(θ0)
can be steered in the d d
direction of a GPS
Antenna satellite to mitigate RF RF RF RF
output front-end front-end front-end front-end
the effects of radio
FIGURE 2 Beam steering with a one-dimen- S(T0) S(T1) S(T2) S(TN)
sional phased array: phase delays are applied frequency interfer- phase delay phase delay phase delay
to individual antenna outputs to steer the ence and jamming d cos(θ0) 2d cos(θ0) Nd cos(θ0)
beam in the direction of θ0, the beam width is signals that are orig-
inversely proportional to the number of array
elements (N) inating from direc-
Antenna
tions other than the output
satellite. Steering the array beam towards reflecting objects to
FIGURE 3Synthetic generation of a one-dimensional phased array: an
record high-resolution radar images provides the foundation incoming GPS signal is down-sampled by the GPS receiver RF front-end;
for the development of GPS-based synthetic aperture radars signal samples that correspond to different spatial antenna locations
(SARs). are combined to generate a synthetic phased array; phases of individual
samples are adjusted to steer the array’s beam in the direction of the
An important consideration in using GPS-based SAR is incident angle of the GPS satellite signal θ0.
that large synthetic arrays are generated with small physi-
cal antennas utilizing platform motion and/or multiplatform
integration. As a result, the physical size of the antenna does
not act as a limiting factor. This, in turn, enables miniatur- d d
ization of the technology for applications on small platforms
such as mini-autonomous aerial vehicles (UAVs) and micro-
UAVs.
Phased array antennas have been widely employed for RF RF RF RF
antenna beam steering. In a phased array, phases of individual front-end front-end front-end front-end
antennas are adjusted to maximize the array gain in a desired Phased array
direction, while increasing the array size narrows the array signal processing
beamwidth. Figure 2 illustrates beam steering in the case of a
one-dimensional (1D) array. FIGURE 4 Synthetic generation of a two-dimensional phased array:
In the GPS domain, beam-steering techniques — both ana- Physical one-dimensional (1D) arrays are used to steer the beam in
the first dimension (beam direction is collinear to the planar surface
log beam steering and digital — have primarily been exploited that is perpendicular to the direction of platform motion); 1D arrays at
to mitigate interference. different spatial locations are applied for beam steering in the second
dimension (beam direction is collinear to the planar surface that is
Hardware versus Software GPS parallel to the direction of motion); measurements of individual anten-
nas are combined for phased array processing. Note that the 1D physical
In designing our phased array, we needed to decide whether to antenna arrays can be implemented on a single platform or can exploit
implement an architecture with a hardware or software GPS multi-platform implementations such as multiple autonomous aerial
vehicles (UAVs).
receiver.

38 InsideGNSS m ay/ june 200 9 www.insidegnss.com

used to increase the antenna aperture
in order to increase the azimuth resolu-
tion.
The synthetic array generation needs
to operate with signal samples. In par-
ticular, samples that are taken a certain
distance apart (generally, a half-wave-
length apart: d=λ/2) must be combined.
Hence, we use a software-defined GPS
receiver to generate a synthetic phased
array antenna.
The software receiver approach also
allows the generation of multiple beams
that are steered in different directions.
Instead of hardware phase adjustments,
a phase of a signal sample is adjusted to
maximize the gain in a given direction
(as seen in Figure 3).
As a result, multiple beams can be
generated simultaneously by applying
different sequences of phase shifts to the
same set of signal samples. This simulta-
neous steering in different directions can
be used, for example, for simultaneous
tracking of direct and reflected signals
such as urban and ground multipath
reflections.
In the remainder of this article, we
will first summarize previous efforts
in the area of synthetic aperture GPS
signal processing. Then we will discuss
the principles of FFT-based multi-direc-
tional beam steering and how we applied
them to develop signal processing tech-
niques for synthetic phased array GPS
antennas.
Finally, we describe the simulation
and live data test results used to verify
the beam steering methods that we have
developed. In particular, we apply actual
flight data and ground data to demon-
strate the operation of 1D and 2D syn-
thetic phased GPS antenna arrays.
Simulated data are exploited to dem-
onstrate the use of 2D synthetic phased
arrays for simultaneous tracking of
direct and multipath signals. We also
use simulated data to demonstrate GPS-
based SAR imaging.
Earlier Work
The concept of synthetic aperture signal
processing for GNSS signals has been
previously considered in both the navi-
gation domain, and in the area of radar
www.insidegnss.com m ay/ june 200 9 InsideGNSS 39
systems. (For a discussion of the former
principle, see in particular the papers
by A. Broumandan et alia, S. Draganov
et alia, and T. Pany et alia cited in the
Additional Resources section near the
end of this article. For a discussion of
the latter, see the paper by M. Chernia-
kov et alia.)
In the navigation domain, these
earlier papers describe synthetic aper-
ture GPS signal processing for a single
antenna case as well as the exploitation
of circular antenna motion to synthesize
a circular phased array.
In the work described by A. Brou-
mandan et alia, a synthetic phased
array is applied for interference mitiga-
Focusing the array beam using multiple antennas that
are mounted perpendicular to the direction of motion
to resolve the cross-track component improves the
cross-track image resolution beyond the C/A or P-
code chip duration.
tion while the paper by T. Pany et alia
discusses the application of the circular
synthetic array to suppress multipath.
S. Draganov et alia discuss the use of
the synthetic aperture technique by the
ultra-tightly coupled GNSS/INS archi-
tecture to mitigate multipath.
In their paper, Cherniakov et alia dis-
cuss the use of GLONASS signals from
the Russian GNSS system as signals of
opportunity for bi-static synthetic aper-
ture imaging. There, antenna motion
is utilized to achieve high-resolution
imaging capabilities in the direction of
motion. The cross-track resolution is
achieved through use of the GLONASS
precision (P)-code, based on publicly
available technical specifications, which
provides a ranging resolution of about
30 meters.
This article extends synthetic aper-
ture GPS signal processing for those
cases in which multiple GPS antennas are
used. Antenna motion is used to synthe-
size one-dimensional phased arrays. As
indicated in Figure 4, the second dimen-
sion is added (synthesized) through the
combining of signals received by mul-
tiple antennas mounted perpendicular
to the direction of motion.
We will introduce a computationally
efficient 2D FFT-based signal processing
algorithm to simultaneously steer the
array beam in multiple directions. As
suggested previously, the array beam can
be steered towards reflecting objects to
record high-resolution SAR images with
GPS signals.
For GPS-based SAR, the range-based
resolution of the cross-track image com-
ponent is limited by the duration of the
chip of the pseudorandom ranging
sequence: 300 meters for the C/A-code
and 30 meters for the GPS P-code.
Focusing the array beam using mul-
tiple antennas that are mounted per-
pendicular to the direction of motion
to resolve the cross-track component
improves the cross-track image reso-
lution beyond the C/A or P-code chip
duration. The approach especially bene-
fits cases where the multiplatform signal
integration can be applied to construct
large array apertures in the cross-track
direction.
FFT-Based Multi-Directional
Beam Steering
As reported by M. I. Skolnik (see Addi-
tional Resources), FFT-based multi-
directional beam steering techniques
have been previously employed for radar
applications. In our work described here,
we adopted these techniques to develop
computationally efficient methods for
multi-directional beam steering of syn-
thetic GPS antenna arrays.
The FFT-based beam steering tech-
nique processes synthetic phased array
data to construct a GPS signal image in
which each image pixel contains signal
parameter information corresponding
to a signal that is received from a par-
ticular steering angle. Figure 5 illustrates
the FFT-based signal image construction
for the 2D antenna array.
Each cell of the 2D FFT frequency
grid corresponds to a particular 2D
steering angle, where the correspon-
 synthetic aperture

Signal image
Normalized signal power

2
where R is the absolute value of the distance between the cur-
rent antenna and the first antenna of the array.
1.5 pixel For the physical array case, R is the distance between anten-
nas:
1

0.5
In (2), m is the antenna index number within the array
0
180 (starting with index zero for the first antenna) and d is the
150 angular cell 180
120
90 120 150
distance between two adjacent antennas (see Figure 2).
60 90 For the synthetic array case, R is the antenna displacement,
Elevation, deg 30 60 Azimuth, deg
0 0 30
as follows:
mapping Angular grid
(defined by FFT
parameters)
where R is the displacement vector and || is the absolute
value.
Beam steering is performed by multiplying samples S of the
incoming signal by a complex exponential:

The array output is then formed by adding individual

FFT frequency grid antenna outputs:

FIGURE 5FFT-based multi-directional beam steering: In this simulation, a

2D GPS signal image contains the received direct signal and a multipath
reflection; each image pixel corresponds to a particular 2D steering
angle derived from the FFT frequency grid; multiple signal sources (such
as direct and multipath signals) can be identified and processed based
on the signal image.
dence between frequency and angles is defined by the antenna
size and FFT parameters. (We will present detailed equations
later in this section.)
Essentially, the FFT-based approach provides a signal image
that is similar to the one constructed by a digital photographic
camera: each “pixel” in the image is defined by the “intensity”
of the signal that is received from the angular direction asso-
ciated with this pixel. In the case of multi-directional beam
steering, the “pixel intensity” is represented by a post-correla-
tion complex amplitude that contains the in-phase (I) value
(real part of the complex amplitude) and quadrature (Q) value
(imaginary part of the complex amplitude).
This signal image can be used to identify and process mul-
tiple signal sources that may include direct signals, interference
signals, and multipath reflections. Hence, this image can be
applied to simultaneously track multiple signal sources such
as direct signal and multipath reflections.
The following section describes the principles of FFT-based
multi-directional beam steering for synthetic GPS antenna
arrays.
FFT for Multi-Directional Beam Steering
To steer a phased array in the angular direction of θ0 (see Figure
2), phases of individual antennas are adjusted by Δφ, which is
defined as follows:
In (5), Sm is the output of the mth antenna, M is the num-
ber of antennas in the array, and d=λ/2 is normally chosen to
avoid beam ambiguities (i.e., a creation of multiple beams in
the angular range from 0 to 180 degrees).
For d=λ/2, equation (5) is modified as follows:
The antenna beam can be steered simultaneously in multi-
ple directions if we use the FFT mechanism to implement equa-
tion (6). FFT harmonic frequencies can be chosen to satisfy the
desired steering angular range and angular resolution.
For the case of a physical antenna array, the use of the FFT
instead of summation transforms single-direction steering into
multi-directional steering as follows:
where {θk} are steering angles and k is the FFT frequency
index.
Note that k is changing from 0 to M/2 and not from 0 to
M-1. This is due to the fact that the FFT spectrum amplitudes
with the index numbers M/2+1 to M are complex conjugates of
the spectrum amplitudes with the index numbers 1 to M/2-1.
Consequently, the former do not provide any new information
and are not considered.

40 InsideGNSS m ay/ june 200 9 www.insidegnss.com

 From equations (6) and (7), the FFT spectrum amplitude
for the kth frequency harmonic is defined as follows:
On the other hand, this spectrum amplitude can be related
to the amplitude of the phased array that is steered in the direc-
tion of θk:
From equations (8) and (9):
or:
Equation (11) defines mapping of the FFT frequencies into
steering angles for multi-directional beam steering. Thus, the
FFT that is applied to outputs of individual antennas generates
multiple beams that are steered in different directions. These
steering directions are defined by the initial steering angle θ0
and FFT harmonic frequencies as specified by equation (11).
To steer multiple beams within the angular range from 0 to
180 degrees, the FFT is first applied for θ0 = 0
This provides beam steering in the angular range from 0 to
90 degrees. To extend the steering angles to the range from 90
to 180 degrees, the second FFT is applied for θ0 = π/2
Hence, a complete FFT-based multi-direction beam steer-
ing procedure combines FFT outputs for initial steering angles
of 0 and 90 degrees:
www.insidegnss.com m ay/ june 200 9 InsideGNSS 41
Signal d d
samples
FFT1
FFT2 Synthetic
array output
FFTP
FIGURE 6 FFT-based beam steering for the synthetic array case
FFT-Based Synthetic Beam Steering for GPS
Antennas
Figure 6 illustrates an FFT-based multi-direction beam steering
method for the synthetic array case.
First, batches of signal samples are formed in such a way
that the spatial distance between first samples in the adjacent
batches is d.
Second, FFTs are applied to samples that have the same
index number within different batches: for example, to first
samples of each batch, second samples of each batch, and so
on. The total number of FFTs required equals P; where P is the
number of samples in the batch.
Third, FFT results are added together to improve the sig-
nal-to-noise ratio (SNR) for the received GPS signal. The FFT
addition operation in the synthetic array case is equivalent to
the incoming/replica signal correlation in the conventional
GPS receiver architecture. Addition of FFT outputs forms the
output of the synthetic phased array.
To reduce the computational load, the number of signal
samples in the batch can be reduced using averaging. In cases
where P≥4 and d=λ/2, distortions associated with this averag-
ing have negligible influence on the FFT-based multi-direc-
tional beam steering.
We should note that GPS carrier phase tracking must be
maintained for phased array antenna formation. Coherent sig-
nal accumulation over the entire synthetic aperture is required
to maintain carrier phase tracking capabilities.
To support coherent signal accumulation, down-sampled
GPS signals must be compensated for changes in the carrier
phase due to intermediate frequency (IF) variations and sat-
ellite motion as well as for changes in the code phase of the
pseudorandom code sequence (PRN). Equation (15) provides
the corresponding compensation expression for satellite j:
where:
tn is the discrete time within the current aperture
accumulation interval: , tn = t0 + n • Δt, Δt is the time
discrete of GPS signal down-sampling;
 synthetic aperture
f IF is the intermediate frequency, which is the frequency of
down-conversion of the incoming GPS signal to baseband;
ΔR SV (t n) is the increment in the satellite/receiver range (or
j 0.8
the line-of-sight — LOS — increment) that is due to satellite
motion;
λL is the wavelength of the GPS Link 1 (L1) carrier;
1
PRNj is the PRN for satellite j; and,
is the estimate of the code phase at the start of the aperture
accumulation, where early, prompt or late code phase estimates
can be used.
Equation (16) is applied to compute the ΔRSV term:
j
In (16), (. , .) is the vector dot product, ej is the satellite/
receiver LOS unit vector, R SV is the satellite position vector;
j
and Rrcvr is the receiver position vector.
Note that equation (16) does not compensate the PRN code
phase for antenna motion. Synthetic apertures stay in the range
from 1.2 to 6.2 meters for experimental scenarios reported
in this article (see next section), which focuses on GPS C/A
code applications. In this case, because the antenna aperture
is significantly shorter that the code chip length (equivalent
to approximately 300 meters), the signal-to-noise ratio (SNR)
processing loss caused by uncompensated antenna motion is
negligible (less than 0.2 dB).
For apertures that are comparable to the length of the C/A-
code chip or for precision (P–code) applications where the chip
duration is equivalent to approximately 30 meters, changes in
the code phase resulting from antenna motion must be taken
into account.
Future research will address modifications of synthetic
aperture algorithms that account for code phase changes due
to antenna motion. Note that antenna motion must be compen-
sated for the construction of the synthetic antennas as detailed
in the next section.
Navigation message data bits in the GPS signal must be
wiped-off to avoid energy losses due to bit transitions for those
cases where the time duration of synthetic aperture exceeds the
duration of navigation data bits (20 milliseconds). The bit wipe-
off can utilize bit estimates from GPS receiver tracking loops.
Alternatively, an energy-based bit estimation algorithm
described in a forthcoming article by A. Soloviev et alia (see
Additional Resources) or Kalman filter bit estimation routines
(see the articles by N. I. Zeidan et alia and M. L. Psiaki et alia
in Additional Resources) can be used for those cases where low
carrier-to-noise ratio (C/N0) signals are processed.
To determine the spatial distance between signal samples
we used the motion trajectory computed by an inertial navi-
gation system (INS). We applied a quarter-wavelength spatial
separation between synthetic antennas to the methods reported
in this article (see Equation [21] below). This separation cor-
responds to approximately five centimeters for the GPS L1 car-
42 InsideGNSS m ay/ june 200 9 www.insidegnss.com
Array pattern for 12 antennas
1
SV LOS angle
0.9 Physical array pattern
Synthetic array pattern (sync function)
0.7
Normalized energy
0.6
0.5
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120 140 160 180
Steering angle, deg
FIGURE 7 Simulated synthetic array pattern versus physical array pattern
rier. Thus, the INS must provide centimeter-accurate trajectory
estimates. This requirement is satisfied if INS error states are
periodically estimated by a Kalman filter that uses GPS carrier
phase measurements as filter observables.
A pattern of the synthetic phased array that is constructed
by adding outputs of individual FFTs (see Figure 6) can be
approximated by a sinc function:
where θ is the steering direction, θ* is the direction of the
incoming signal, and A is the total length of the array.
The sync function corresponds closely to the pattern of
the physical phased array for the case in which the distance
between individual antennas is equal to a half-wavelength:
Figure 7 illustrates close agreement between patterns of the
half-wavelength physical array and a synthetic array using a
simulated case where the satellite signal is received from a 45-
degree angle.
In applying equation (17) we can see that the pattern of
the synthetic array does not depend on the spatial distance
between synthetically generated antenna locations. As a result,
this spatial distance can be chosen to provide a desired angular
range for the multi-directional beam steering.
For our particular purposes, we chose a spatial distance
equal to λ/4 so as to cover the angular range between 0 to 180
degrees. In this case, the FFT amplitude that corresponds to
the kth frequency harmonic is expressed as follows:

This spectrum amplitude also serves as a synthetic array
output for steering direction θk. As a result, it satisfies the fol-
lowing equation:
Comparison of equations (19) and (20) maps FFT frequen-
cies into beam steering angles:
Equation (22) provides the final expression for the FFT-
based multi-directional beam steering for synthetic phased
arrays:
Two-Dimensional Beam Steering
For multi-directional steering in two dimensions, one-dimen-
sional FFTs for the synthetic and physical phased arrays — in
accordance with equations (22) and (14) — are combined into
a two-dimensional FFT:
In (23), m and n are index numbers of the antenna in the
physical and synthetic arrays, respectively, with corresponding
steering angles ϕ and θ, as shown in Figure 8.
FFT-Based Multi-Directional Signal Tracking
The outputs of the 2D FFT-based steering procedure that we
have described correspond to post-correlation complex ampli-
tudes of incoming signals received for different antenna steer-
ing angles. More specifically, the real part of the FFT spec-
trum represents the in-phase (I) post-correlation values and the
imaginary part represents the quadrature (Q) post-correlation
www.insidegnss.com m ay/ june 200 9 InsideGNSS 43
values for the anten-
na steering angles Synthetic array Beam direction
defined by FFT fre-
quencies as specified ϕ
by equation (23). θ
Hence, the 2D Physical array
FFT mechanism for-
mulated by equation FIGURE 8 Two-dimensional steering angles
(23) simultaneously
provides Is and Qs for multiple beam steering angles, which
can be applied for multi-directional carrier phase tracking. I
and Q values for each FFT frequency grid and its correspond-
ing steering direction are treated independently from other
frequency grids.
Multi-directional carrier phase tracking is performed by
processing I and Q values of each steering direction. Signal
parameters for different steering directions can be estimated
from I and Q values using an open-loop receiver architecture
described in the forthcoming article by F. van Graas et alia
(Additional Reources). FFT-based multi-directional tracking
will be illustrated later in the section describing simultaneous
tracking of direct and multipath signals.
So, with the foregoing theoretical background in mind, let’s
turn to the feasibility demonstration and experimental verifica-
tion phases of our discussion.
1D Synthetic Array: Flight Test Results
Synthetic generation of 1D antenna arrays was verified using
experimental data collected in actual flight environments.
Figure 9 shows the configuration of equipment on the Ohio
University Avionics Engineering Center (OU AEC) McDonald
Douglas DC-3 aircraft used for the flight test.
We selected a straight segment of the flight to demonstrate
the generation of synthetic arrays, with the aircraft traveling
at approximately 63 meters per second.
The synthetic array approach was verified using sampled
GPS data recorded during the flight test by the AEC’s soft-
ware instrumentation receiver described in the article by S.
Gunawardena et alia (Additional Resources). To implement
multi-directional beam steering in the range from 0 to 180
degrees using a single FFT — per Equation (18), antennas
in synthetically generated arrays are spatially separated by a
quarter-wavelength of the GPS L1 carrier wavelength (approxi-
mately five centimeters).
As described earlier in this article, quarter-wavelength
batches of GPS signal samples are first formed and then
processed. The motion trajectory reconstructed by the INS
is applied to determine the spatial separation of GPS signal
samples in order to combine samples that are quarter-wave-
lengths apart.
The INS motion trajectory is computed from measurements
of a low-cost inertial measurement unit (IMU). This unit’s sen-
sor errors are specified as 0.1 degree/second (one sigma) gyro
drift and one milligal (one sigma) accelerometer bias. IMU
measurements are periodically calibrated in-flight using GPS
 synthetic aperture

GPS antenna
nal prior to applying the FFT-based beam steering procedure.
Figures 10 and 11 show synthetic array patterns (with 13 and
26 elements, respectively) computed by the FFT-based steering
method for the case where the motion of the GPS satellite iden-
tified by its pseudorandom noise code — PRN 7 — is wiped off
from the incoming signal. An estimated carrier-to-noise ratio
(C/N0) for PRN 7 is 47 dB-Hz.
In the figures, the synthetic array patterns are represented
as normalized signal energy (y-axis) versus steering angle,
where the signal energy is normalized such that the maximum
energy over the entire steering angular range is equal to one.
These figures reveal patterns for cases in which the synthetic
OU AEC aperture corresponds to a 13-element array (Figure 10) and a
front-end
26-element array (Figure 11) that use half-wavelength separa-
tion between their antennas.
In the figures, the maximum energy is observed when the
antenna beam is steered towards the satellite. An energy loss
is introduced as the beam is steered away from the satellite.
Essentially, an energy loss for the angle θ is equivalent to a
suppression that would be applied to an interference signal (or
any other unwanted signal) from a given angle while the beam
is steered towards the satellite.
The beamwidth is approximated as the distance between
the points for which the energy is degraded by three decibels
from its maximum (boresight) value. Hence, the beamwidths
Data collection computer are approximately 10 degrees and 5 degrees for the 13-element
(Low-cost IMU mounted inside)
and 26-element synthetic antenna arrays, respectively.
FIGURE 9 Flight test setup
2D Synthetic Array: Ground Test Results
carrier phase in order to maintain the centimeter-accurate We use ground vehicle test data to demonstrate the generation
reconstruction of motion trajectory needed to generate the of 2D synthetic phased arrays. Figure 12 shows the experimental
synthetic phased arrays. setup in the AEC test van that was used for data collection.
For each satellite, the down-conversion frequency f IF, PRN Similar to the flight test, we implemented a straight motion
code, and satellite motion are wiped off from the incoming sig- trajectory with the van’s velocity at approximately four meters

Array pattern for 13-antenna array Array pattern for 26-antenna array
0 0
SV LOS angle SV LOS angle
Normalized signal energy, dB

Normalized signal energy, dB

-10 -10

-20 -20

-30 -30

-40 -40
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Steering angle, deg Steering angle, deg

Pattern of a one-dimensional synthetic antenna array with 13

FIGURE 10 Pattern of a one-dimensional synthetic antenna array with 26
FIGURE 11
elements generated using flight data and FFT-based multi-directional elements generated using flight data and FFT-based multi-directional
beam steering. PRN 7 satellite motion is wiped off from the incoming beam steering. PRN 7 satellite motion is wiped-off from the incoming
signal; signal

44 InsideGNSS m ay/ june 200 9 www.insidegnss.com

 synthetic aperture

Note t hat t he applied to a multipath or an interfering

RF front-end
Data collection computer
(AGNC IMU mounted inside)
FIGURE 12 Ground vehicle test setup
per second. An AEC two-channel soft-
ware instrumentation receiver was used
to record raw GPS signal samples. IMU
data were applied to reconstruct the
motion trajectory for the synthetic array
formation.
Two antennas mounted on the van
rooftop formed the array beam in the
direction perpendicular to motion.
Array aperture in the direction of
motion was synthesized by observing
these antennas at different spatial loca-
tions.
We calibrated the system’s inter-
channel and inter-antenna biases in
advance of the ground test and removed
their associated phase delays from raw
signal samples prior to implementing the
synthetic signal processing routines.
beam steering capa-
bilities in the cross-
t rack direct ion
were limited by the
number of front-end
channels (two) cur-
GPS antennas
rently available in
the software receiv-
er. This limitation
can be mitigated
by increasing the
number of front-
end channels and/
or by implementing
multi-platform signal integration tech-
niques.
Figures 13 and 14 show 2D synthetic
array patterns generated for the ground
experiment. In these figures, α is the
steering angle in the direction of motion
and β is the steering angle in the cross-
track direction. Figure 13 shows test
results for a phased array constructed
using two physical and six synthetic
antennas.
Similar to the 1D array case, the
maximum signal energy is received
when the array beam is steered in the
direction of the satellite. As the beam
is being steered away from the satel-
lite, signal attenuation occurs. At a spe-
cific angular direction, this attenuation
is equivalent to that which would be
signal coming from this direction.
As mentioned previously, beam-
steering capabilities in the cross-track
direction are currently limited by the
number of front-end channels in the
receiver. With only two channels avail-
able, therefore, the array beam in the
cross-track direction is rather wide:
approximately 30 degrees. Nevertheless,
the experimental results demonstrate
the feasibility of generating 2D arrays
by using synthetic aperture GPS signal
processing and combining signals from
multiple antennas.
Figure 14 shows the 2D synthetic
array pattern for the case where the
synthetic aperture is extended to 31 ele-
ments. This extension narrows the array
beam significantly in the direction of
motion. The beamwidth in the cross-
track direction remains unchanged.
FFT-based Multi-Directional
Beam Steering: Simulation
Results
The following section uses simulated
results to demonstrate the feasibility of
the FFT-based multi-directional signal
tracking with simulated results.
For the simulation, a multipath signal
was added to the direct satellite signal.
Multipath was simulated as a specular
reflection from a horizontal planar sur-

Pattern of 2x6 element array Pattern of 2x31 element array

180 0 180 0
-10 150
150
-20 -20
120 120
-30
α, deg

90
α, deg

90 -40 -40

60 -50 60
-60 30 -60
30
-70
0
0 0 30 60 90 120 160 180
0 30 60 90 120 160 180 β, deg
β, deg
SV location
SV location

Pattern of a two-dimensional antenna array that is comprised

FIGURE 13 FIGURE 14Pattern of a two-dimensional antenna array that is comprised
of two physical and six synthetic antennas. Satellite motion has been of two physical and thirty one synthetic antennas; PRN 4 satellite mo-
wiped off from the incoming signal from PRN4, which had a carrier-to- tion is wiped-off from the incoming signal; PRN 4 C/N0 is 49 dB-Hz
noise ratio of 49 dB-Hz.

www.insidegnss.com m ay/ june 200 9 InsideGNSS 45

 synthetic aperture

face collinear to the Y-Z plane. Figure 15 below the direct sig- Direct phase minus estimated direct phase, m
0.04
illustrates this multipath scenario. nal power.
For the FFT-based synthetic array Figure 16 shows 0.02
processing, the synthetic aperture a 2D signal image 0
corresponds to 20 physical antennas constructed using -0.02
separated by a half-wavelength. A four- the 2D FFT-based
-0.04
element physical array with half-wave- beam steering 0 0.2 0.4 0.6 0.8 1
length antenna separation was simu- method. The plot Time, s
lated to steer the beam in the direction demonstrates simul- Multipath phase minus estimated multipath phase, m
0.04
perpendicular to motion. Array beams taneous recovery of
0.02
were steered in multiple directions using the direct and mul-
2D FFTs (see Equation [23]). tipath signals. 0
For the direct signal, a C/N0 of 40 F F T - b a s e d -0.02
dB-Hz was implemented. The multipath multi-directional -0.04
0 0.2 0.4 0.6 0.8 1
power was simulated to be three decibels signal tracking was Time, s
implemented and
z SV simulated for the FIGURE 17 Simultaneous tracking of direct and multipath signal using FFT-
case of simultaneous based multi-directional beam steering
multipath tracking of direct
and multipath signals for the scenario V = 2 m/s
direct path illustrated in Figure 15. FFT-based Is
and Qs are applied to track the carrier
phase of direct and multipath signals. h = 20 m
Real and imaginary parts of FFT com- Y
vrcvr
y plex amplitudes for beams steered in the X
Receiver x direction of direct and multipath signals
are exploited as I and Q values, respec-
tively. FIGURE 18 Simulation scenario implemented to
Figure 17 illustrates the results of demonstrate GPS-based SAR imaging
Multipath reflection scenario
FIGURE 15 the simulated tracking. Carrier phase
simulated to verify multi-directional beam
steering in two dimensions tracking errors are shown as differences level. The C/N0 of the reflected signal
between true carrier phase (known in was simulated to represent an army tank
the simulation) and measured carrier hidden under a dense canopy.
4x20 antenna array
phase based on I and Q values provided Analysis of RF GPS data collected
by 2D FFT computations. in forestry areas shows that for high-
Normalized signal energy

1
0.8 Standard deviations of carrier phase elevation satellites the signal is generally
0.6 errors are evaluated as 3.2 and 4.3 mil- attenuated by 7 decibels as it propagates
0.4 limeters for the direct and multipath through the canopy. Therefore, the sig-
0.2
signals, respectively. Hence, the simula- nal attenuation of 14 decibels was intro-
0
150 180 tion results presented demonstrate the duced for the two-way propagation path
120 120 150
90 90 feasibility of simultaneous tracking of through the canopy, that is, from the sat-
60
y axis steering 30 0 0 30 60
angle, deg x axis steering multiple signals using the FFT-based ellite to the target and from the target to
angle, deg
multi-directional beam steering. the receiver.
FIGURE 16Simulated FFT-based signal image The radar cross section (RCS) of
for a two-dimensional antenna array: Direct GPS-Based SAR Imaging the target is assumed to be five square
and multipath signals are simulated; direct
signal carrier-to-noise ratio is 40 dB-Hz; Figure 18 illustrates a simulation scenario meters. For the 14-decibel signal atten-
multipath signal power is three decibels implemented to demonstrate GPS-based uation due to the propagation through
lower than the direct signal power; angles SAR imaging. the canopy and a five square meter
between X and Y axes of the antenna array
and direct signal are simulated as 90 and 46 A simulated receiver platform RCS value, the C/N0 of the signal that
degrees, respectively; angles between X and includes four GPS antennas mounted on is received at the UAV platform is esti-
Y axes of the antenna array and multipath an unmanned aerial vehicle (UAV) that mated at 17 dB-Hz.
signal are simulated as 90 and 134 degrees,
respectively; energy peaks are clearly is flying at 20 meters above the ground We simulated and processed the C/
observed for cases where the antenna beam with a 2 meters per second ground veloc- A-code component of the GPS signal.
is steered towards the direct signal and ity. The simulated scenario assumes a To process the GPS signal at a 17 dB-Hz
multipath
single point reflector located at ground level, a one-second coherent integration

46 InsideGNSS m ay/ june 200 9 www.insidegnss.com

 Focused SAR image
180
150
120
90
Y, m
60
30
0-10
-5 0 5
X, m
Target location
FIGURE 19 Simulated SAR image for the GPS-based SAR implementation
interval was implemented. Construction of SAR images was
based on the 2D FFT-based processing approach discussed in
the previous section, augmented by beam focusing techniques
discussed in the book by J. C. Curlander and R. N. McDonough
(Additional Resources) to avoid image distortions that are due
to the motion of the platform.
Figure 19 shows an example of a simulated SAR image. To
construct this image, steering angles were converted into X
and Y Cartesian coordinates using the known height above
the ground (20meters).
The GPS-based SAR image shown in Figure 19 clearly
identifies the presence of a target. Note that the image resolu-
tion in the direction of motion is achieved using the synthetic
aperture. For the cross-track direction, the image is resolved
by focusing the array beam with multiple physical antennas
mounted perpendicular to motion.
This implementation differs from the classical SAR
approach that uses a single antenna and the cross-track image
component is resolved using the carrier modulation by a rang-
ing signal. For the GPS case, the chip duration of the C/A-code
ranging modulation is 300 meters. Clearly, this resolution is
insufficient for most imaging applications. Hence, multiple-
antenna beam focusing is applied to resolve the cross-track
image component.
As shown in Figure 19, the multi-antenna focusing approach
allows for resolving the cross-track image component within
approximately three meters. In other words, the ranging resolu-
tion of the C/A code is improved by two orders of magnitude
in the simulation scenario.
Conclusions
This article has described the generation of synthetic phased
GPS antenna arrays using moving antennas. Signals received
from a GPS antenna at different spatial locations are combined
into a synthetic phased array to sharpen the array beam with-
www.insidegnss.com m ay/ june 200 9 InsideGNSS 46a
out increasing its physical size. We introduced an FFT-based
approach to simultaneously steer the array beam in multiple
directions and to perform multi-source signal tracking.
This approach can be applied for improving GPS robust-
ness to radio-frequency interference, simultaneous tracking
of multiple signal sources such as direct signal and multipath
reflections, and for recording high-resolution images utilizing
GPS signal reflections.
Future research in the area of synthetic aperture GPS signal
processing will focus on the demonstration of the GPS SAR
concept with experimental data and multi-platform signal pro-
cessing for generating large array apertures in the direction
perpendicular to motion.
10
Acknowledgment
This article is based on a presentation made at the 2009 Insti-
tute of Navigation International Technical Meeting in Ana-
heim, California.
Manufacturers
The low-cost IMU used in the AEC flight test is a Coremicro
manufactured by American GNC Corporation, Simi Valley,
California, USA.
Authors
Andrey Soloviev is a research assistant professor at
the University of Florida, Research and Engineering
Education Facility. Previously he served as a senior
research engineer at the Ohio University Avionics
Engineering Center, Athens, Ohio. He holds B.S. and
M.S. in Applied Mathematics and Physics from Mos-
cow University of Physics and Technology and a Ph.D.
in Electrical Engineering from Ohio University. Soloviev’s research interests
focus on all aspects of multi-sensor integration for navigation applications.
He is a recipient of the Institute of Navigation (ION) Early Achievement Award
and the RTCA’s William Jackson Award.
Sanjeev Gunawardena is a senior research engineer
and co-principal investigator with the Ohio Univer-
sity Avionics Engineering Center. He earned a Ph.D.
in electrical engineering from Ohio University and
was the 2007 recipient of the RTCA William E. Jackson
award. His research interests include RF systems
design, digital systems design, reconfigurable com-
puting, and all aspects of GNSS receivers and signal processing.
Frank van Graas is the Fritz J. and Dolores H. Russ Pro-
fessor of Electrical Engineering and a principal inves-
tigator with the Avionics Engineering Center at Ohio
University. A Past President of The Institute of Navi-
gation, van Graas received the 1996 Johannes Kepler
Award from the ION’s Satellite Division of the ION. His
research interests center on all facets of GPS, includ-
ing aircraft precision approach and landing, attitude determination, and sys-
tem integration.
Mikel Miller is the technical director for the Advanced Guidance Division,
Munitions Directorate, Air Force Research Laboratory, Eglin Air Force Base,
Florida. He received his Ph.D. in electrical engineering from the Air Force
 synthetic aperture
Institute of Technology
(AFIT), WPAFB, Ohio.
Since 1986, he has
focused on navigation
system R&D related to
GPS, GPS/INS integra-
tion, alterative naviga-
tion techniques including bio-inspired naviga-
tion and signals of opportunity–based navigation,
autonomous vehicle navigation and control, and
multi-sensor fusion. He is currently responsible
for directing both in-house and contracted R&D
projects advancing guidance, navigation, and
control technology for weapon systems.
Additional Resources
[1] Akos, D. M., “A Software Radio Approach to
Global Navigation Satellite System Receiver
Design,” Ph.D. dissertation, Ohio University,
August 1997
[2] Broumandan, A., and T. Lin, A. Moghaddam,
D. Lu, J. Nielsen, and G. Lachapelle, “Direction of
Arrival Estimation of GNSS Signals Based on Syn-
thetic Antenna Array,” in Proceedings of the ION
GNSS-2007, September 2007
[3] Cherniakov, M., and R.Saini, M. Antoniou, R.
Zuo, and J. Edwards, “SS-BSAR with Transmitter
of Opportunity - Practical Aspects,” in Proceed-
ings of the 3rd EMRS DTC Technical Conference,
June 2006
[4] Curlander, J. C., and R. N. McDonough, Syn-
thetic Aperture Radar: Systems and Signal Pro-
cessing, Hoboken, New Jersey, John Wiley & Sons,
1991
[5] Draganov, S., and L. Haas, “Novel Multipath
Mitigation Methods for Ultra-tightly Coupled
GNSS/INS Systems,” presented at the “for official
use only” (FOUO) session of the ION GNSS-2008,
September 2008.
[6] Farrell, J. L., “GPS/INS-Streamlined,” NAVIGA-
TION, Journal of the Institute of Navigation, Vol.
49, No. 4, Summer 2002
[7] Gold, K., and R. Silva, R. Worrell, and A. Brown
“Space Navigation with Digital Beam Steering GPS
Receiver Technology,” in Proceedings of the ION
Annual Meeting, June 2003
[8] Gunawardena, S., A. Soloviev, F. van Graas,
“Wideband Transform-Domain GPS Instrumen-
tation Receiver for Signal Quality and Anomalous
Event Monitoring,” NAVIGATION, Journal of the
Institute of Navigation, Vol. 53, No. 4, 2007
[9] Kaplan, E., and C. Hegarty (Editors), Under-
standing GPS: Principles and Applications, 2nd ed.
Norwood, Massachusetts: Artech House, 2006
[10] McGraw, G. A., and C. McDowell, and J. M.
Kelly “GPS Anti-Jam Antenna System Measure-
46b InsideGNSS m ay/ june 200 9 www.insidegnss.com
ment Error Characterization and Compensation,” in
Proc. of the ION GNSS-2006, September 2006
[11] Morton, Y. T., and L. L. Liou, D. M. Lin, J. B.
Y. Tsui, and Q. Zhou, “Broad Band Interference
Cancellation using Digital Beam Forming and a
Software GPS Receiver,” in Proceedings of the ION
GNSS-2005, September 2005
[12] Nicholson, B. W., and D. M. Upton, S. Cotterill,
J. Marchese, T. Upadhyay, and W. E. Vander Velde
“Computer Simulation of Digital Beam Forming
Adaptive Antennae for GPS Interference Mitiga-
tion,” in Proc. of the ION National Technical Meet-
ing, January 1998.
[13] Pany, T., and B. Eissfeller, “Demonstration of
a Synthetic Phased Array Antenna for Carrier/Code
Multipath Mitigation,” in Proc. of the ION GNSS-
2008, Sep. 2008.
[14] Psiaki, M. L., and H. Jung, “Extended Kalman
Filter Methods for Tracking Weak GPS Signals,” in
Proceedings of the ION GPS-2002, September
2002
[15] Rama Rao, B., and E. N. Rosario, and R. J.
Davis, “Eleven Element Beam Steering GPS Anten-
na Array in a GAS-1 CRPA Footprint,” presented at
the FOUO session of the ION GNSS-2008, Septem-
ber 2008
[16] Skolnik, M. I., Introduction to Radar Systems,
3rd ed. New York: McGraw-Hill, 2001
[17] Soloviev, A., and S. Gunawardena, and F. van
Graas, “Decoding Navigation Data Messages from
Weak GPS Signals,” to appear in IEEE Transactions
on Aerospace and Electronic Systems
[18] van Graas, F., and A. Soloviev, M. Uijt de Haag,
S. Gunawardena, “Comparison of Two Approaches
for GNSS Receiver Algorithms: Batch Processing
and Sequential Processing Considerations,” to
appear in IEEE Journal of Selected Topics in Sig-
nal Processing, Special Issue on: Advanced Sig-
nal Processing for GNSS and Robust Navigation,
August 2009
[16] Zeidan, N. I., and J. L. Garrison, “Bit Synchro-
nization and Doppler Frequency Removal at Very
Low Carrier to Noise Ratio Using a Combination of
the Viterbi Algorithm with an Extended Kalman Fil-
ter,” in Proceedings of the ION GPS/GNSS-2003,
September 2003