Você está na página 1de 9

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 62, NO.

1, JANUARY 2013

221

A Low-Cost High-Performance
Digital Radar Test Bed
Hasan S. Mir, Senior Member, IEEE, and Lutfi Albasha, Senior Member, IEEE

AbstractThis paper describes the design of a dual-channel


S-band digital radar test bed. The test bed combines stretch
processing with a novel and cost-effective hardware architecture
that enables it to achieve an in-band dynamic range of 60 dB over
600 MHz of instantaneous bandwidth. The dual digital receiver
channels allow for adaptive digital beamforming which can be
used to mitigate a directional source of interference. Experimental
test and verification results are presented to demonstrate system
performance.
Index TermsDigital beamforming, digital radar, high
dynamic range radar, high-resolution radar, S-band transceiver,
stretch processing.

I. I NTRODUCTION

MONG THE various characteristics of a radar system,


it is particularly desirable to have high range resolution,
high dynamic range, and multiple receiver channels. The significance of each of these attributes is as follows:
1) High range resolution is the ability to resolve two targets
that are closely spaced in range, allowing for distinction between scatterers originating from separate targets
or from a larger extended target. Compared to lowresolution profiles in which a target may be confined to a
single range gate, high-resolution profiles provide information on the spatial distribution of target reflectivity [1],
[2]. This can be useful in distinguishing closely spaced
targets or in determining the features of an extended
target.
2) Dynamic range is defined as the difference (in decibels)
between the weakest detectable target and the strongest
target that can be detected without driving the system into
nonlinearity. High dynamic range is important for operation in high-clutter environments such as littoral regions,
wherein a weak target could be potentially masked by a
large clutter reflection [3], [4].

Manuscript received February 19, 2012; revised May 15, 2012; accepted
May 17, 2012. Date of publication September 7, 2012; date of current version
December 12, 2012. This work was supported in part by a Faculty Research
Grant from the American University of Sharjah FRG11-III-18. The Associate
Editor coordinating the review process for this paper was Dr. Mark Yeary.
The authors are with the Department of Electrical Engineering, American
University of Sharjah, Sharjah, UAE (e-mail: hmir@aus.edu; lalbasha@
aus.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/TIM.2012.2212497

3) Multiple receiver channels or an array of receivers enables beamforming, which is essentially a spatial filtering
operation that can be used to emphasize signals originating from a particular angular region in space while
rejecting signals originating from outside the region. This
operation is useful in practical scenarios, wherein the
radar must be able to reject a jammer from a certain
angular location in order to maintain a track on a target
of interest [4].
Much work has been conducted in the theory of obtaining
high-resolution target scattering profiles (see [5][9] and the
references therein). Among the earliest works in this area,
Rihaczek [5] examined the fundamental tradeoff between localization in time and localization in frequency, arriving at the
now well-known conclusion that the optimal radar waveform
for generating a high-resolution profile is the one that is simply
matched to the environment. It therefore follows that a waveform with wide bandwidth and, hence, coarse localization in
frequency will have high range resolution or fine localization
in time (a simple rule for determining the range resolution is
the expression c/2B, where c is the speed of light and B is the
waveform bandwidth). As such, high-resolution radar systems
transmit wideband waveforms to interrogate the environment.
The receiver architecture generally consists of a filter whose
response is matched to the transmitted waveform. By processing the received signal through this matched filter, a correlation
operation known as pulse compression is performed. The output
from the matched filter, known as the target range profile, is of
primary interest since targets are declared to exist at the time
delays/range gates where the matched-filter output exceeds a
certain value.
The theoretical body work has also been implemented in
a variety of radio frequency (RF) test beds. Holloway et al.
[10] describe a W-band frequency-modulated continuous-wave
(FMCW) radar for airborne situational awareness. A digitaldirect-synthesizer chip generates the transmitted and local oscillator signal, providing a highly linear and robust platform for
supporting multiple missions. In [11], a radar design methodology is discussed in the context of simulation tools that generate
code for eventual implementation on a field-programmable gate
array. Such tools provide a robust platform for the design of
high-fidelity waveform generators and tuning of functions such
as sensitivity time control. Crain et al. [12], [13] discuss upgrading the National Weather Radar Testbed with a phased array antenna, allowing for focused observation of severe weather
phenomena. Multiple digital channels are built using commercial off-the-shelf components, with a focus on synchronization

0018-9456/$31.00 2012 IEEE

222

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 62, NO. 1, JANUARY 2013

across channels, making the receiver well suited for applications such as the direction-of-arrival estimation. Nelander and
Toth-Pal [14] describe a modular design for a maritime radar.
The radar operates at S-band and also incorporates a
semiconductor-based power amplifier. The system has two
receiver channels that cover the linear frequency modulated
(LFM) waveform bandwidth of 20 MHz. Yu et al. [15] describe
the design of a radar receiver. Although the transmitter phase
noise and other nonlinearities limit the dynamic range, the system can operate at wide bandwidths. Adrian [16] describes the
evolution of radar technology to the multimission paradigm. In
particular, the ability of adaptive beamforming (ABF) for jamming mitigation, high dynamic range for weak signal detection,
and wide bandwidth for target classification are emphasized.
Derham et al. [17] describe a novel multistatic radar system, in
which there are several (distributed) transmitters and receivers.
Signal-to-noise ratio (SNR) improvement in target amplitude
of up to 18 dB was demonstrated due to the systems ability
to view a target from multiple perspectives. George et al. [18]
present an approach to improve spurious-free dynamic range
(SFDR) using a novel compensated Kaiser window-based technique. Results are presented that show that this method can
achieve an SFDR of 24 dB over a 1-GHz bandwidth. Lin et al.
[19] address the problem of isolation between the transmitter
and the receiver in FMCW radar. A novel scheme is proposed
that uses loop adaptation, allowing for the precise control of
the leakage cancellation. Gupta et al. [20] describe a modular
RF receiver based on superconductor technology. Such systems
operate at a very low temperature (around 4 K) and in a
low magnetic flux environment. As such, a hybrid temperature
heterogeneous technology paradigm is proposed wherein the
superior speed of superconductor electronics is complemented
by the increased complexity of CMOS processors at ambient
temperature.
Digital radar has recently been proposed as a new paradigm
that offers several advantages [3], [4], [21][24]. For example,
because a digital radar employs an analog-to-digital converter
(ADC) behind each receiver element, beamforming or other
adaptive signal processing operations can be performed using
digital hardware. The ability to form multiple simultaneous
beams can be used to quickly scan a spatial region of interest
or perform various tasks such as anti-air warfare of missile
defense. Additionally, digital radar offers the added potential
of improved dynamic range by moving the digitization stage
closer to the array face.
The existing literature is generally based on using the
matched-filter receiver architecture. An inherent problem when
designing a digital radar using a matched-filter receiver architecture is that the usage of wideband waveforms (for high
range resolution) necessitates high-speed ADCs due to the
Nyquist theorem. This approach as such not only generates
large amounts of data but also reduces the overall system
dynamic range, thereby offsetting the improvement gained by
moving the digitization stage closer to the array face (Li et al.
[22], [23] develop a quantitative analysis of this tradeoff in the
context of digital radar). It should be noted that the matched
filter produces a target range profile over all unambiguous range
gates. In general, however, since distant targets may not be of

interest, only a specified range window needs to be examined.


As such, a technique known as stretch processing [25] can be
applied. This technique reduces the required sampling rate so
that only range gates of interest are obtained, producing smaller
amounts of raw data and allowing for improved dynamic range.
In the context of maritime surveillance radars, the S-band
and X-band are the primary frequencies that are used. While
the higher X-band frequency allows for systems with very fine
range resolution, it also makes such systems very susceptible
to rain and snow. The S-band is therefore preferred for longrange surveillance applications. Thus, this paper describes the
design of an S-band radar that meets the competing demands
of future maritime systems to have both high range resolution
(for applications such as target identification) and high dynamic
range (for operation in high-clutter environments such as littoral regions). The system has dual channels and uses stretch
processing. While there are many factors associated with the
design of an actual fielded system (such as packaging, cooling,
etc), the purpose of this test bed is to better understand the challenges associated with the practical implementation of a system
that has both high dynamic range and high range resolution.
The test bed as such described in this paper simultaneously
achieves a dynamic range in excess of 60 dB over an instantaneous bandwidth in excess of 600 MHz. Moreover, the digital
channels give the system an ABF capability that can be used to
maintain gain on a signal of interest while suppressing an interfering source such as a jammer. Experimental test and verification results are provided as an example of system performance.
A number of novel concepts and attributions are presented in
this paper. These include a system architecture that enables high
in-band experimentally measured dynamic range of 60 dB over
a wide instantaneous bandwidth of 600 MHz. In comparison to
previous work such as [4], this new system was designed to be
far more cost-effective yet maintaining the same level of high
performance. Furthermore, the new upconverter architecture
presented in this paper employs a more efficient design that uses
far less components through better active intercircuit matching.
Finally, an ABF architecture that can effectively suppress wideband jamming is enhanced and developed in this work.
This paper is organized as follows. Section II describes
the setup and processing chain of the test bed. Section III
describes in detail the hardware architecture and associated
design considerations. Section IV discusses the experimental
results.
II. D IGITAL R ADAR T EST B ED OVERVIEW
As the speed of an ADC increases in order to accommodate
wider bandwidth waveforms, the corresponding increase in the
noise floor bandwidth results in lowering the overall system
dynamic range. As mentioned in Section I, stretch processing
can be used to overcome this dilemma.
A. Stretch Processing
Consider a transmitted waveform sTX (t) and a local oscillator waveform sLO (t), both of which are LFM chirps with
identical duration T and rate k, where k is defined as the ratio

MIR AND ALBASHA: LOW-COST HIGH-PERFORMANCE DIGITAL RADAR TEST BED

223

of the chirp bandwidth to the chirp duration. These waveforms


may be expressed as
 

k 2
(initial)
t + fTX
sTX (t) = cos 2
t
2
 

k 2
(initial)
sLO (t) = cos 2
t + fLO
t
2
(initial)

(initial)

where fTX
and fLO
are the respective initial frequencies. Both waveforms are generated locally at the radar.
The signal sTX (t), after being reflected from a target, can
be expressed as sTX (t ) for some target delay . Mixing
sTX (t ) with the local oscillator waveform sLO (t) and
bandpass filtering result in a tone with frequency


(initial)
(initial)
.
(1)
fTX
k + fLO
This sequence of operations is known as stretch processing.
It can be seen from (1) that stretch processing results in
a tone whose frequency is proportional to the target range.
The bandpass filter that follows the mixing with sLO (t) is
(initial)
(initial)
centered at fLO
fTX
. The selected width of this filter
corresponds to the target range window that is selected based
upon the range of interest and ADC capabilities. By confining
the target range of interest to a relatively narrow frequency
band, stretch processing accommodates wideband waveforms
without the need for high-speed ADCs. Additionally, since a
target at a certain range corresponds to a tone with a certain
frequency, the range profile is extracted simply as the Fourier
transform, which is digitally approximated by the discrete
Fourier transform (DFT), of the received signal. Thus, pulse
compression for a stretch-processed system is accomplished via
a simple fast Fourier transform (FFT), an efficient numerical
implementation of the DFT.

Fig. 1. System setup of digital radar test bed.

Fig. 2. Block diagram of processing chain.

linearity, gain, and noise figure. The desired total system gain of
4 dB agrees well with the measured data that were fed into the
digitizer. The different stages of the upconverter and receiver
show gain and noise figure levels that were possible to build in
the laboratory as well.
An approximate component cost breakdown is shown in
Table I. Through intercircuit matching as shown in Fig. 3, the
number of components, and hence the system cost, is kept low.
The upconverter and receiver required customized design and
integration. Each component is discussed in the next section in
more detail.

III. D IGITAL R ADAR H ARDWARE D ESCRIPTION


A. AWG

B. System Setup
Fig. 1 shows a laboratory photograph of the test bed. A highlevel block diagram of the test bed architecture is shown in
Fig. 2, from which it can be seen that the test bed consists four
major components.
1) The arbitrary waveform generator (AWG), which serves
as a source of a baseband high-fidelity chirp signal as an
input to the upconverter.
2) The upconverter, which shifts the baseband chirp to
S-band. It consists of a transmission channel (TX
channel) and a downconverting reference channel (LO
channel).
3) The receiver, which shifts the S-band input to baseband.
The test bed has two identical receiver channels (RX 1
and RX 2).
4) The ADC, which serves as a sink that samples the output
of both receiver channels.
The source (AWG) and sink (ADC) are classified as test and
measurement equipment and were commercially procured.
Fig. 3 shows the system lineup parameters, which include
the upconverter and receiver components and the cascaded

The Agilent M9331A AWG was selected as the source


primarily because of its high vertical resolution (10 b), high
sample rate (1.25 GHz), and low spurious product levels
(> 65 dB). These specifications allow the device to create a
high-fidelity wideband input signal to the upconverter. This
device has two output channels which are used to generate the
signals sTX (t) and sLO (t) that are input to the upconverter as
described hereinafter.

B. Upconverter
The frequency upconversion of a signal is accomplished
through the use of mixers. The process of mixing, however,
introduces spurious intermodulation products. These spurious
products may fall inside the frequency band of interest and, if
strong enough, may corrupt the desired signal. Thus, for the
preservation of the dynamic range, an important design issue is
the mitigation of intermodulation products from the mixers and,
hence, the choice of a suitable frequency plan. Additionally,
the tuning of the power levels at various points in the system
(through the use of low-noise amplifiers (LNAs) or pads) is

224

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 62, NO. 1, JANUARY 2013

Fig. 3. System lineup parameters.


TABLE I
C OMPONENT C OST B REAKDOWN

necessary in order to maintain a linear operating characteristic


of the system.
As discussed in Section II-A, stretch processing mixes the
received chirp signal (a delayed version of the transmitted
signal sTX (t)) with a demodulating chirp signal sLO (t). The
generation of these two respective signals is accomplished
through the TX channel and LO channel of the upconverter,
with the architecture as shown in Figs. 4 and 5.
A multistage upconverter design that enables spurious free
regions of operation was developed in view of the fact that
single stage upconversion cannot sufficiently reject in-band
intermodulation products and IF/LO feedthrough. The upconverter employs an efficient design that uses only a single mixer
and doubler. It should be noted that two separate upconverters
(as shown in Figs. 4 and 5) are needed since some surveillance
systems require that the signal paths of sTX (t) and sLO (t) be
separated in frequency (selected as 760 MHz for the current test
bed in view of minimizing distortion due to intermodulation
products).
The input waveforms from the AWG to both upconverter
channels are LFM chirps that span the frequencies of 200
500 MHz (300-MHz instantaneous bandwidth) with a power
level of 0 dBm. The first stage of upconversion is performed
by mixing the LFM chirp with an oscillator at a power level
of 3 dBm, followed by bandpass filtering to select the lowside (difference) frequencies. An LNA is then used so that
the input to the doubler has a power level of around 3 dBm.
The doubler results in increasing the instantaneous bandwidth
to 600 MHz. The doubler output is then bandpass filtered
in order to reject any out-of-band distortion so that only
S-band signals (30303640 MHz for the TX channel and 3800
4400 MHz for the LO channel) remain. The filter output level
is then adjusted so that the TX channel output power is 7 dBm
and the LO channel output power is 13 dBm. Since these two
channels are input to the receiver, such an adjustment ensures
that the drive levels to the receiver are appropriate in order to
maintain linearity.

C. Receiver
Fig. 6 depicts the operations that are performed at the
receiver. The output of the TX channel and that of the LO
channel are mixed. The mixer output is bandpass filtered in
order to reject spurious signals as well as the high-frequency
chirp generated as part of the mixing operation. A combination
of an LNA and a pad is used so that the resulting signal has
a power level of 3 dBm. This signal is then mixed with an
830-MHz oscillator (with a power level of 3 dBm) in order
to downconvert the center frequency of 760 MHz to the IF
frequency of 70 MHZ. A bandpass filter centered at 70 MHz
with a bandwidth of 16 MHz is then applied to reject the highside (sum) frequencies. A final combination of an LNA and a
pad is used to increase the output signal power level to 4 dBm,
which is sufficiently high for the final digitization stage.
D. ADC
The output of both receiver channels is fed to the Agilent
M9210A. This device is a digital scope with dual channels,
10-b resolution, and up to 4-GS/s sampling rate. This device can
be used for the real-time display of the receiver output, while
the 512-MB onboard memory can be used to store sampled
data that can be analyzed/processed offline. The source and
sink are both modular cards housed in a Peripheral component
interconnect eXtensions for Instrumentation (PXI) chassis (Agilent M9018A). The chassis provides an interface that allows
the AWG and ADC cards to be controlled via a PC.
IV. E XPERIMENTAL R ESULTS
The specifications of the AWG and ADC were selected in
order to preserve the dynamic range afforded by the upconverter
and receiver design. In order to assess the composite system
dynamic range and the system noise floor, the AWG was used
to generate a 100-s LFM chirp with a 300-MHz instantaneous
bandwidth. This signal was fed to both the TX channel and LO
channel of the upconverter, resulting in a signal with 600 MHz
of instantaneous bandwidth at the output of each channel. These
two outputs were then input to the receivers, following which
the output of the receivers was fed to the digital scope (ADC).
The output of the soft front panel of the digital scope is shown
in Fig. 7. The output is shown in the frequency domain since
frequency corresponds to range for a stretch-processed system
(as discussed in Section II-A). It should be noted that, for a
system using stretch processing, the frequency resolution f is

MIR AND ALBASHA: LOW-COST HIGH-PERFORMANCE DIGITAL RADAR TEST BED

Fig. 4.

Block diagram of upconverter architecture (TX channel).

Fig. 5.

Block diagram of upconverter architecture (LO channel).

Fig. 6.

Block diagram of receiver architecture.

Fig. 7.

Digital radar test bed output for 600-MHz instantaneous bandwidth.

simply the inverse of the pulse duration T . This can be verified


by noting that, from (1)

f = k =

1
T

225

where is the time resolution. Thus, converting time resolution to range resolution R yields

kc

=
2
2T

kR =

c
.
2T

226

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 62, NO. 1, JANUARY 2013

TABLE II
D IGITAL R ADAR T EST B ED C HARACTERISTICS

which is then applied to the data in order to reject interference.


Consider a sequence of K narrow-band snapshots recorded at
the receiver, {xk }K1
k=0 . One method to design the ABF weight
is to seek a vector w that minimizes the beamformer output
energy subject to maintaining gain in the signal look direction
[26], i.e.,


2
arg min wH xk 

s.t. wH v(, f ) = 1.

(2)

Now, define the sample covariance matrix as


R x = xk xH
k

Recalling that k = B/T


c
B
R =
T
2T

R =

c
2B

K1
1
xk xH
k .
K

(3)
(4)

k=0

which is the usual range resolution equation.


As can be seen from Fig. 7, the output contains a strong tone
at the IF frequency of 70 MHz. The band of 6278 MHz is at a
higher level due to the final bandpass filter in the receiver. Fig. 7
also shows that the in-band dynamic range is approximately
60 dB, while the out-of-band dynamic range is around 90 dB.
The SNRs at the system input and system output were also measured to compute the noise figure, which was found to be 11 dB.
The results show that this test bed is capable of achieving a high
in-band dynamic range (60 dB) over an instantaneously wide
bandwidth (600 MHz or a range resolution of approximately
0.25 m). The high dynamic range is specifically indicative of the
fact that all blocks are operating under linear conditions, well
below the P1dB compression point defined by the component
manufacturer.
A summary of the salient characteristics of the digital radar
test bed is given in Table II.
A. ABF
A key characteristic of a sensor array is its directiondependent behavior. That is, for the same transmitted signal
from a source, the signal received on the array will differ depending upon the orientation of the array relative to the source.
The nature of this orientation-dependent behavior, which is
governed by the element-to-element time delay of the source
wavefront, can be exploited for various purposes. For example,
a radar must also be able function in environments where
sources of interference, such as jamming, are present. Using
an operation known as beamforming, the beam pattern can
be shaped to null interference while maintaining gain on the
desired signal.
In traditional narrow-band phased array radars, the time
delay of a signal incident on the array may be approximated
as a phase shift. Thus, spatially adaptive weights may be
computed for each channel in order to create a spatial null in
the jammer direction while maintaining gain in the direction
of interest. Consider a signal with frequency f arriving at
angle on an M -element array. The array response vector
will be denoted as v(, f ). For narrow-band signals, spatialonly ABF may be used to reject interference. Samples from
each channel are used to form an adaptive weight vector,

Then, (2) can be cast using the Lagrangian




J(w) = wH Rx w + wH v(, f ) 1 .
Solving J(w)/w = 0 yields the solution for the optimal
weight vector

1
H 1
.
w = R1
x v(, f ) v(, f ) Rx v(, f )

(5)

Note that the center frequency f is fixed since the narrow-band


assumption is made.
Suppose now that the received data are wideband, as is
the case for the test bed described in this paper. Because the
timebandwidth product is not significantly smaller than one
for wideband systems, the array response exhibits a frequencydependent behavior known as dispersion. In the context of
jammer mitigation, dispersion necessitates the inclusion of
temporal adaptivity in addition to spatial adaptivity, leading to
a class of techniques known as spacetime adaptive processing
(STAP) that can mitigate dispersive effects. In the subband
STAP architecture, the data from each channel are first divided
to into multiple narrow subbands. The subbands corresponding
to the same frequency bin are grouped so that narrow-band
ABF may be applied to each bin. Thus, the processor in (5)
may be used for wideband data if subbanding is first applied to
{xk }K1
k=0 and the weight vector is suitably recomputed for each
subband. This process is illustrated in Fig. 8. Note that each
subband will have a different target steering vector v(, f ) since
the value of f is changing with each subband. One important
advantage of stretch processing is that the subbanding procedure is essentially free from a computational standpoint. In a
conventional matched-filter system, a filter bank would be necessary to divide the received data into subbands. With stretch
processing, a natural timefrequency mapping exists since time
cells correspond to frequency bins. Thus, subbanded data can be
generated by simply grouping time-domain samples according
to the desired subband width, as depicted in Fig. 9.
Because the test bed has two digital receiver channels, it is
capable of performing digital beamforming. In order to test the
ABF capability of the system, cables were used to connect the

MIR AND ALBASHA: LOW-COST HIGH-PERFORMANCE DIGITAL RADAR TEST BED

Fig. 8.

ABF processing architecture.

Fig. 9.

Timefrequency mapping concept.

upconverter and receivers, thus simulating the received signal


from a target, for which 30 000 samples were collected from
the ADC. This corresponds to the 250-MHz sampling of one
pulse with a 20-s guard interval. Note that the actual pulse
bandwidth is 600 MHz due to the frequency doubling in the
upconverter. Hence, normal matched filtering would require at
least a 1.2-GHz sampling according to the Nyquist theorem.
However, the use of stretch processing has allowed for the
usage of a sampling frequency that is nearly 1/10 of the Nyquist
rate. Simulation of a jammer was performed using MATLAB.
The correlation matrix for the data collected on the two digital
receiver channels was estimated using (4). The beamforming
weights were then computed for each subband using (5), thus
implementing the architecture in Fig. 8. The collected data, consisting of a pulse contaminated by a jammer on both channels,
was then processed using the architecture in Fig. 8. The range
profile was computed using the DFT (no windowing was used).
As shown in Fig. 10, without further processing, the jammer
completely obscures the presence of the target, and only noise
appears to be present. This also shows the inability of a single
receiver channel to detect a target in the presence of jamming.
With two channels, however, an adaptive spatial filter (ABF)

227

Fig. 10. Target range profile with and without ABF.

can be digitally implemented. As can be seen in Fig. 10, with


ABF, the target is clearly detectable against the background,
with a very pronounced peak.
The test bed presented in this paper presents a low-cost
solution for researching the issues associated with fielding a
large-scale digital array radar. The simplicity of the test bed

228

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 62, NO. 1, JANUARY 2013

architecture coupled with its use of readily available commercial components provides a robust platform that simplifies
hardware debugging and enables easy extension to a larger
number of channels.
The basic limitation of the current test bed is that the
frequency plan is fixed and that waveform bandwidth cannot
exceed that of the RF filters, whose passband is currently fixed
at 600 MHz as per the test bed design specification. However,
the general architecture presented in this paper can be modified
for higher bandwidths and/or different frequency plans.
To accommodate higher bandwidth waveforms, the RF bandpass filters simply need to be replaced with filters with a wider
passband. It should be noted, however, that increasing the bandwidth will lead to spurious intermodulation products falling
inside the bandwidth of interest, thus reducing the effective
dynamic range. The presented architecture has been designed
so as to place all intermodulation products outside of the signal
bandwidth in order to preserve the dynamic range. The frequency plan of the system can also be modified relatively easily
by the changing oscillator source frequencies and changing the
center frequency of the RF filters.
Apart from modifying the system bandwidth and/or frequency plan, the system architecture can also accommodate
different waveform types. Various other waveforms such as
nonlinear frequency modulation, which has lower sidelobes
compared to LFM, can be used.
Lastly, even though the presented architecture is designed for
stretch processing, it can be easily converted to a conventional
matched filtering system by simply replacing the LO channel of
the upconverter with a conventional sinusoidal oscillator source
and removing the IF filters at the receiver. This would enable the
usage of any desired waveform whose bandwidth is compatible
with the system, including variants such as frequency jump
burst waveforms.
V. C ONCLUSION AND F UTURE W ORK
A dual-channel digital radar test bed that operates at S-band
and uses stretch processing has been constructed.
This test bed possesses a number of novel features. The
system architecture was shown to have a high in-band experimentally measured dynamic range of 60 dB over a wide
instantaneous bandwidth of 600 MHz. The system is also
very cost-effective yet has been shown to preserve a highlevel performance. By employing intelligent active component
matching, the new upconverter architecture uses a minimal
number of components.
The dual digital receiver channels were used to demonstrate
that the system can effectively suppress a source of jamming
to reveal a target that would otherwise be obscured. The low
assembly cost of the test bed is also a notable advantage. The
modular design allows it to be easily extended to a larger
number of receiver channels. Such modularity and high level
of performance can also be leveraged to develop and test signal
processing algorithms (STAP, channel equalization, etc.). These
features, coupled with the digital radar hardware, can provide a
performance baseline for future fielded digital radar systems.
The low cost of this system coupled with the high level of

measured performance has provided a platform for system


miniaturization at the printed circuit board and chip integration
levels.

R EFERENCES
[1] D. Mensa, Wideband radar cross section diagnostic measurements,
IEEE Trans. Instrum. Meas., vol. IM-33, no. 3, pp. 206214, Sep. 1984.
[2] T. Tice, An overview of radar cross section measurement techniques,
IEEE Trans. Instrum. Meas., vol. 39, no. 1, pp. 205207, Feb. 1990.
[3] D. J. Rabideau, R. J. Galejs, F. G. Willwerth, and D. S. McQueen, An
S-band digital array radar testbed, in Proc. IEEE Int. Symp. Phased Array
Syst. Technol., Oct. 2003, pp. 113118.
[4] H. Mir, A. Myne, and D. Spinuzzi, A high dynamic range, wideband
digital array radar testbed, in Proc. Int. Radar Conf., Sep. 2009, pp. 15.
[5] A. Rihaczek, Radar signal design for target resolution, Proc. IEEE,
vol. 53, no. 2, pp. 116128, Feb. 1965.
[6] L. Graham, Synthetic interferometer radar for topographic mapping,
Proc. IEEE, vol. 62, no. 6, pp. 763768, Jun. 1974.
[7] A. Jain and I. Patel, Ground-to-air imaging radar for RCS measurements, IEEE Trans. Instrum. Meas., vol. 41, no. 6, pp. 951956,
Dec. 1992.
[8] D. Wehner, High-Resolution Radar. Norwood, MA: Artech House,
1994.
[9] S. Blunt, K. Gerlach, and T. Higgins, Aspects of radar range superresolution, in Proc. IEEE Radar Conf., Apr. 2007, pp. 683687.
[10] R. Holloway, W. Weedon, B. Houshmand, and A. Roll, Next-generation
W-band radar testbed, in Proc. IEEE Radar Conf., Apr. 2007, pp. 6571.
[11] L. Prakasam, T. Roy, and D. Meena, Digital signal generator and receiver design for S-band radar, in Proc. IEEE Radar Conf., Apr. 2007,
pp. 10491054.
[12] G. Crain, M. Yeary, C. Kidder, A. Zahrai, G. Zhang, R. Doviak, R. Palmer,
T. Yu, M. Xue, Y. Zhang, Q. Xu, and P. Chilson, Multi-channel conversion of the National Weather Radar Testbed receiver, in Proc. IEEE
Radar Conf., May 48, 2009, pp. 15.
[13] M. Yeary, G. Crain, A. Zahrai, R. Kelley, J. Meier, Y. Zhang, I. Ivic,
C. Curtis, R. Palmer, T. Yu, and R. Doviak, An update on the multichannel phased array weather radar at the National Weather Radar
Testbed, in Proc. IEEE Radar Conf., May 2327, 2011, pp. 971973.
[14] A. Nelander and Z. Toth-Pal, Modular system design for a new S-band
marine radar, in Proc. Int. Radar Conf., Oct. 1216, 2009, pp. 15.
[15] J. Yu, M. Reynolds, and J. Krolik, An indoor S-band radar receive array
testbed, in Proc. IEEE Radar Conf., May 1014, 2010, pp. 712717.
[16] O. Adrian, Future surface radar technology: From air defence to
air and missile defence, in Proc. IEEE Radar Conf., Apr. 1720, 2007,
pp. 4954.
[17] T. Derham, S. Doughty, K. Woodbridge, and C. Baker, Design and evaluation of a low-cost multistatic netted radar system, IET Radar, Sonar
Navigat., vol. 1, no. 5, pp. 362368, Oct. 2007.
[18] K. George, C. Chen, and J. Tsui, Extension of two-signal spurious-free
dynamic range of wideband digital receivers using Kaiser window and
compensation method, IEEE Trans. Microw. Theory Tech., vol. 55, no. 4,
pp. 788794, Apr. 2007.
[19] K. Lin, Y. Wang, C. Pao, and Y. Shih, A Ka-band FMCW radar front-end
with adaptive leakage cancellation, IEEE Trans. Microw. Theory Tech.,
vol. 54, no. 12, pp. 40414048, Dec. 2006.
[20] D. Gupta, D. Kirichenko, V. Dotsenko, R. Miller, S. Sarwana,
A. Talalaevskii, J. Delmas, R. Webber, S. Govorkov, A. Kirichenko,
I. Vernik, and J. Tang, Modular, multi-function digital-RF receiver
systems, IEEE Trans. Appl. Supercond., vol. 21, no. 3, pp. 883890,
Jun. 2011.
[21] Y. Wu and J. Li, The design of digital radar receivers, IEEE Aerosp.
Electron. Syst. Mag., vol. 35, no. 41, pp. 3541, Jan. 1998.
[22] Z. Li, L. P. Ligthart, P. Huang, W. Lu, and W. F. van der Zwan, Tradeoff between sensitivity and dynamic range in designing digital radar receivers, in Proc. Int. Conf. Microw. Millim. Wave Technol., Apr. 2008,
pp. 13681371.
[23] Z. Li, L. P. Ligthart, P. Huang, W. Lu, W. F. van der Zwan, E. P. Lys,
and O. A. Kraznov, Design considerations of the RF front-end for high
dynamic range digital radar receivers, in Proc. Int. Conf. Microw., Radar,
Wireless Commun., May 2008, pp. 14.
[24] O. Adrian, From AESA radar to digital radar for surface-based applications, in Proc. IEEE Radar Conf., May 2009, pp. 15.
[25] W. J. Caputi, Stretch: A time-transformation technique, IEEE Trans.
Aerosp. Electron. Syst., vol. AES-7, no. 2, pp. 269278, Mar. 1971.
[26] H. Van Trees, Optimum Array Processing (Detection, Estimation, and
Modulation Theory, Part IV). Hoboken, NJ: Wiley, 2002.

MIR AND ALBASHA: LOW-COST HIGH-PERFORMANCE DIGITAL RADAR TEST BED

Hasan S. Mir (M06SM10) received the B.S.


(cum laude), M.S., and Ph.D. degrees in electrical engineering from the University of Washington,
Seattle, in 2000, 2001, and 2005, respectively.
From 2005 to 2009, he was with the Air Defense Technology Group at the Massachusetts Institute of Technology Lincoln Laboratory, Lexington.
Since 2009, he has been an Assistant Professor with
the Department of Electrical Engineering, American
University of Sharjah, Sharjah, UAE.

229

Lutfi Albasha (M95SM04) received the Ph.D.


degree from the University of Leeds, Leeds, U.K.
He spent many years working for Sony Corporation and Filtronic (RFMD) Semiconductor Foundry.
He received several industrial and academic awards
from Sony Corporation and the IET. He has served
on the boards of governmental, commercial, and
academic bodies and on the committees of several
international conferences in wireless electronics. He
has authored over 45 papers and is the Editor of
the IET Microwaves Journal. His research interests
are in microelectronic mixed-signal and RFIC systems. He has been with
the Department of Electrical Engineering, American University of Sharjah,
Sharjah, UAE, since 2007, where he is currently an Associate Professor.

Você também pode gostar