Professional Documents
Culture Documents
Orthogonal Waveforms
Michael S. Davis Aaron D. Lanterman
Gregory A. Showman Georgia Institute of Technology
Georgia Tech Research Institute Atlanta, Georgia USA
Atlanta, Georgia USA
Coherent multiple-input multiple-output (MIMO) radar is architecture that has been used in radar and other systems for
a natural extension of the phased array antenna that has decades. With this philosophical approach, a remarkable num-
been used by radar systems for decades. This tutorial uni- ber of concepts described in the MIMO radar literature can be
fies concepts from the literature and provides a framework illuminated.
for the analysis of an arbitrary suite of MIMO radar wave- An active radar system emits electromagnetic energy to
forms. A number of gain patterns are introduced, which probe its environment. A MIMO radar transmits independent
quantify the antenna performance of a MIMO radar, and waveforms from a number of spatially separated radiating ele-
the impact of the waveform characteristics (e.g., range sid- ments and observes the returns from a set of spatially diverse
elobes) is discussed. receive elements. The environment is considered as a system
where the inputs are the transmitted waveforms and the outputs
are the echoes observed by the receivers. This is, in general, a
I. INTRODUCTION MIMO system. There has been a long history of exploiting mul-
Since the turn of the century, a great deal of research has in- tiple degrees of freedom on receive [2]; the novelty of MIMO is
vestigated the utility of applying multiple-input multiple-output to exploit similar degrees of freedom on transmit.
(MIMO) techniques to enhance the performance of radar sys- Of course, the idea of using multiple radiating elements on
tems. Although MIMO has been demonstrated to dramatically transmit is not new [3]. Phased array antennas have been an
improve the capacity of communications systems in multipa- enabling technology for many systems in support of a variety
th-rich environments [1], and MIMO technology has been in- of radar missions. Rather than construct a prohibitively large
cluded in multiple communications standards, the adoption of antenna to support long range detection of small targets, many
MIMO in operational radar systems has been limited. As will be small antennas are employed, each of which transmits an identi-
discussed, this may be attributed to radar systems using wave- cal signal (up to a phase shift). This phase progression is chosen
forms and antennas to interact with their environments in a to steer a high-gain beam in a particular direction. This beam
dramatically different way than communications systems. Still, may be rapidly resteered through electronic scanning, which
some specific radar applications have been identified where a enables multiple target tracking performance that would be im-
MIMO radar may outperform a more traditional system. possible for a mechanically scanned antenna.
This paper seeks to establish a framework for understand- Although a phased array uses multiple elements on transmit,
ing the performance benefits and potential drawbacks of MIMO the transmitted waveforms are perfectly correlated and there-
radar systems. MIMO radar is realized by transmitting indepen- fore provide no additional degrees of freedom. In contrast to the
dent waveforms, but in many ways the waveforms are simply general MIMO radar, phased arrays operate as a single-input
a means to an end. We view MIMO radar as essentially an an- multiple-output (SIMO) system. Phased array radars typically
tenna technology since it is a generalization of the phased array provide receive degrees of freedom by digitizing multiple, spa-
tially-diverse receive channels, hence they are multiple-output.
We refer to the special case of a radar with a single transmit
Authors' addresses: M. Davis and G. Showman, Sensors waveform and a single receive channel as a single-input single-
and Electromagnetic Applications Laboratory, Georgia Tech output (SISO) radar system.
Research Institute, Atlanta, GA. E-mail: (mike.davis@gtri. The idea of transmitting multiple, uncorrelated waveforms
gatech.edu). A. Lanterman, School of Electrical and Computer
has been explored over the years. For example, in [4], it was
Engineering, Georgia Institute of Technology, Atlanta, GA.
Manuscript received September 5, 2013; ready for publication observed that a radar system transmitting two orthogonal wave-
December 23, 2013. forms and using a single receiver provides bearing estimates
DOI. No. 10.1109/MAES.2014.130148. that are equivalent to a traditional radar transmitting a single
Review handled by P. Willett and B. Himed. waveform and using two receivers. (An analogous result in the
0885/8985/14/ $26.00 © 2014 IEEE communications literature [5] demonstrated that the diversity
order of the same two configurations is equivalent if the trans- ent derivation explicitly deals with the possibility of spatially
mitted waveform is an orthogonal space-time block code.) An colored interference and proposes an alternative formulation
experimental radar that transmitted “orthogonal” waveforms for dealing with correlated transmit signals. We also explicitly
was described in [6]. identify a signal processing architecture and spatial weights that
It was not until the benefits of MIMO for communications can be applied to form beams in desired directions.
were clear that the idea of transmitting uncorrelated waveforms A key result described in the MIMO radar literature is that
in radar received consistent attention. Using the language of the transmit beampattern is characterized by the correlation be-
MIMO communications, researchers have described two broad tween the transmitted signals [13]. This correlation impacts the
classes of MIMO radars: those with widely-separated antennas MIMO radar ambiguity function, which was presented in [14].
that, as in the communications application, seek to exploit spa- Indeed, the correlation properties of the waveforms transmitted
tial diversity to mitigate target fading [7] as opposed to those by a MIMO radar determine the characteristics of the system.
with closely-spaced antennas that permit coherent processing These characteristics are captured by considering the quantities
[8, 9]. presented in this paper.
The class of MIMO radars that use widely-separated anten- Although radar and communications systems are based on
nas is often referred to as “statistical MIMO radar” because similar physical phenomena and are described by similar math-
such systems seek to exploit the random fluctuation of target ematics, there are fundamental differences, particularly the role
reflectivity as a function of aspect angle. This fluctuation causes of the antenna subsystem. A discussion of this is presented in
spatial decorrelation of the target returns, which precludes co- Section II. This is followed by a derivation of a MIMO radar
herent processing. As a result, statistical MIMO is often referred signal processing chain, described in Section III. We highlight
to as noncoherent MIMO, while the term coherent MIMO is the importance of the correlation among the transmitted signals
restricted to systems where the antennas are sufficiently close and develop a formulation for the gain of a MIMO radar an-
(perhaps on the same platform) to limit target decorrelation. tenna. These observations are used in Section IV to compare the
The following discussion is limited to the latter case of coher- traditional phased array with a MIMO radar that is transmitting
ent MIMO radar. The case of noncoherent MIMO with widely- orthogonal waveforms. In Section V, some quasi-orthogonal
separated antennas has been discussed extensively in the radar waveforms are described and their performance is examined.
community in the context of multistatic radar systems [10]. Just This paper presents significant refinements and substantial ex-
as coherent MIMO may be considered a natural extension of tensions to the ideas presented in [15].
the phased array, statistical MIMO generalizes the concept of
bistatic radar [11].
The goal of this paper is to establish a methodology for
II. MIMO SYSTEMS: RADAR AND COMMUNICATIONS
evaluating the potential of a suite of waveforms to enable a In both radar and communications systems, electromagnetic en-
MIMO radar to effectively perform its mission. This provides ergy is radiated. This signal interacts with the environment, and
the capability of determining if MIMO techniques are appro- the resulting electromagnetic field is observed by a receiver.
priate for a given radar system. The variety of operational ra- The goal of a communications system is to estimate the param-
dars that exist today is tremendous. MIMO can dramatically eters of the input signal, e.g., the sequence of message symbols
improve the performance of some, but others will benefit from used to generate it in spite of any environmental effects. In ra-
traditional phased array configurations. Further, this framework dar, the goal is to infer some property of the environment based
can inform the design of waveforms, which are critical to the on knowledge of the transmitted waveform.
realization of a MIMO radar system. In either case, an input signal is applied to some system, and
The first task is to characterize the gain of the antenna em- the response of this system to this input is observed. Radar and
ployed by a MIMO radar by determining the MIMO radar be- communications systems are typically modeled as linear. First,
ampattern. A similar approach is taken in [12], though the pres- consider a SISO system. Let x(t) be the complex-baseband rep-
resentation of the input signal that is transmitted on a carrier channel matrix is not well conditioned [17]. In a sense, by using
frequency ωc. If the system is linear, the output of the system multiple channels on transmit, a MIMO communications sys-
y(t) may be written in terms of the input signal and the system tem can exploit available spatial diversity. In many cases, two
impulse response h(t) as communications antennas placed on the order of a wavelength
apart can observe completely independent channel realizations.
Now, consider the case of a coherent MIMO radar and a single
(1)
target. By definition, the antenna elements are spaced so closely
that they observe the same target reflectivity and the only dif-
where v(t) represents receiver noise.
ference will be a phase shift that is related to the target angle.
In communications, the system impulse response h(t) de-
Statistical MIMO radar, in which the elements are separated
scribes the channel between the transmitter and receiver, which
widely enough to provide independent realizations of target fad-
may involve a direct path contribution and/or multipath. For a
ing, is based on the same observation as MIMO communications.
monostatic radar system, where the transmitter and receiver are
However, the connection between coherent MIMO radar tech-
colocated, the relevant impulse response describes the scatter-
niques and MIMO communications is more elusive. Transmit
ers in the environment and may be called the range profile.
precoding approaches to MIMO communications may seem re-
For both communications and monostatic radar cases, the im-
lated to coherent MIMO radar, but they are more closely related
pulse response at a particular lag t, may be formed from returns
to the traditional phased array approach. Just as the transmit pre-
from a number of scatterers. The returns from these scatterers all
coder of a MIMO communications system uses its knowledge of
arrive after the same delay, but they may correspond to differ-
the multipath environment to maximize the signal-to-noise ratio
ent angles-of-arrival q. In this case, we may write the impulse
(SNR) at the receiver, the phased array forms a beam to maxi-
response in terms of the angle/delay reflectivity profile a(t, q) as
mize SNR for a particular target location [18]. We will instead
see that the approach of a coherent MIMO radar will be to op-
(2) timize other properties of the radar antenna at the cost of SNR.
While it would seem natural that MIMO radar and MIMO
communications would possess a great deal of similarity, the
In many radar applications, the goal is to estimate this angle/
parallels in the context of coherent MIMO radar are limited. A
delay (or, equivalently, angle/range profile). However, in com-
coherent MIMO radar is not seeking to exploit spatial diversity
munications, the only relevant parameter is the aggregate chan-
to ensure that at least one of the “links” is available. Instead,
nel response h(t). The fundamental difference between the com-
just like the phased array, the goal is to use coherent process-
munications problem and the radar problem is the role of the
ing to provide a more capable antenna for the radar system.
channel impulse response: in communications it is a nuisance
The following analysis of MIMO radar signal processing will
factor that must be estimated to establish a link, but in radar,
demonstrate how this is accomplished and provide a method to
estimating the “channel response” and, further, the underlying
assess its performance.
reflectivity profile a(t,q) is essential.
We now turn to the MIMO case, where a set of input sig-
nals, is used and a set of output signals, are
observed where M is the number of transmitted signals, and N
III. MIMO RADAR SIGNAL PROCESSING
is the number of received signals. We extend the model in (1)
to the MIMO case:
A. THE MIMO VIRTUAL ARRAY
The advantage of a MIMO radar transmitting orthogonal wave-
forms over a traditional phased array is often explained by con-
(3)
sidering the virtual array [19]. If an array of M elements is used
on transmit and an array of N elements is used on receive, the
where H(t) is the N × M MIMO channel matrix that describes corresponding virtual array of the system consists of M N vir-
the impulse response of the M N channels of the MIMO system. tual elements if the M waveforms are perfectly orthogonal.
If the multipath environment described by the channel First, consider a single transmit element and a single receive
matrix H(t) is suitably rich and sufficient information about element. This may be considered as a pseudobistatic pair, which
the channel is available, then tremendous gains in the capac- allows us to invoke the monostatic-bistatic equivalence theo-
ity of the communications channel are provided by MIMO. rem: the signal observed by the bistatic pair is well approxi-
These M N channels may be exploited by sending redundant mated by the signal observed in the monostatic case where the
information over independent channels via space-time cod- transmit and receive elements are located directly between the
ing. Alternatively, if channel state information is available to bistatic elements [20]. As the bistatic angle increases, variation
the transmitter, diversity may be exploited through precoding, in radar cross section (RCS) between the bistatic pair and the
which is analogous to beamforming on transmit [16]. monostatic equivalent will become apparent, but this effect is
In either approach to MIMO communications, it is assumed negligible in the coherent MIMO case.
that some level of diversity is provided by having access to the If orthogonal waveforms are transmitted, then each of the
M N channels. Indeed, capacity gains are limited if the MIMO M N virtual phase centers may be processed independently. In
Figure 1.
Virtual arrays corresponding to same physical array using either phased array configuration or transmitting orthogonal waveforms.
Figure 2.
Virtual arrays corresponding to MIMO Nyquist array.
the phased array case, where the waveforms are perfectly cor- spectively. If the array elements are uniformly spaced, then the
related from element to element, only data corresponding to a beampattern is the familiar Dirichlet kernel of Fourier analysis,
single transmit phase center can be processed, providing only N which is also called the “aliased” or “periodic” sinc function:
virtual phase centers. This is illustrated in Figure 1.
The MIMO virtual array is an extension of the concept of
the coarray for coherent imaging described in [21], which con- (6)
siders an imaging system that scans a transmit and receive beam
simultaneously to estimate a reflectivity profile over a number
of transmit-receive events. The MIMO virtual array, however, The virtual array is obtained by convolving the transmit and
corresponds to a single transmit pulse or dwell. receive aperture functions:
A well-known result of antenna theory is that the aperture il-
lumination function g(x) and the antenna beampattern are Fou-
(7)
rier duals of one another. For example, if {xT/m} is the set of
locations of the elements in the transmit array, then the aperture
illumination function for the (unweighted) transmit array is The virtual array consists of elements located at {xT/m + xR/n: m
= 1, …, M; n = 1, …, N}. The virtual array is the convolution
of the aperture illumination functions, so the two-way beampat-
(4)
tern of a MIMO radar with orthogonal waveforms is the Fourier
transform of its virtual array. The length of the virtual array dic-
We can similarly construct the aperture illumination function tates the angular resolution, and any taper applied by repeated
for the receive array: elements can improve sidelobe performance.
We now consider the MIMO virtual array for a couple of
physical arrays and show how they lead to an improved beam-
(5)
pattern or can improve angular resolution. First, consider the
case shown in Figure 1, where a critically sampled, uniform
where {xR/n} is the set of element locations for the receive array. linear array (ULA) is assumed. While the orthogonal waveform
The Fourier transform of the aperture illumination functions case has a slightly longer virtual array, which predicts slightly
(4) and (5) are the array transmit and receive beampatterns, re- improved angular resolution, the dominant effect is to provide
Figure 3.
MIMO radar signal model. In this illustration, M = N = 3.
an effective antenna taper. This is evident from the nondistinct generality, assume that the return is observed with zero delay;
phase centers, which admit a triangular coarray. equivalently, we consider the output of the radar signal proces-
Alternatively, consider the configuration where the transmit sor in the center of the range bin containing a target. We revisit
and receive elements are spaced so that the virtual phase centers this assumption in Section V and investigate its impact. For
form a ULA of M N distinct phase centers. This configuration clarity, we also assume that the signal is narrowband relative to
has been called a Nyquist array [22] because it is critically sam- the size of the array. Consequently, the data observed by each
pled in the spatial sense. This will be M times longer than the array element will be identical up to a phase shift.
virtual array generated by a phased array of N receive elements, Let xm(t) be the waveform emitted by element m of the trans-
suggesting an M-fold improvement in angular resolution. This mit array. The signal observed by element n of the receive array
array configuration and its corresponding MIMO virtual array from a target at an angle q0 with a (complex-valued) backscatter
are presented in Figure 2. coefficient a may be written as
However, these improvements provided by transmitting or-
thogonal waveforms come at a cost. The phased array forms a
(8)
high-gain beam on transmit that improves SNR by a factor of M.
This is not the case when orthogonal waveforms are transmitted.
The costs and benefits of transmitting orthogonal wave- where vn(t) is the observation noise, am(q0) and bn(q0) are the
forms or perfectly correlated waveforms can be best under- phase shifts on transmit and receive, respectively, which cor-
stood by considering an approach to spatial beamforming. respond to the target angle q0. If, for example, the array used
This will also allow us to consider cases in between these two on transmit is a ULA with interelement spacing d, the transmit
extremes. A signal model and beamforming approach will be phase shifts corresponding to an angle q would be
developed. This is followed by an extension of standard an-
tenna metrics to the MIMO case, which will elucidate the rela-
(9)
tion between SNR, resolution, and beampattern for a specified
set of waveforms.
for m = 1, …, M, where c is the speed of waveform propagation.
Of course, steering vectors may be specified for arbitrary array
B. MIMO RADAR SIGNAL MODEL configurations:
Consider a MIMO radar that transmits M (generally) indepen-
dent waveforms. Each signal may be emitted by a separate ra-
(10)
diating element such as a reflector antenna or a horn antenna.
Alternatively, an array divided into a number of subarrays (pos-
sibly overlapped) could be used where each subarray acts as where Rm is the range to the target from element m.
an independent radiating element. The only requirement is that
These phase shifts may be arranged into the transmit and receive
each signal be emitted from a distinct phase center. The signals
steering vectors a(q) and b(q). Note that and are
will reflect off of scatterers in the environment, and the echoes
considered to be column vectors. While (8) is the data observed
will be observed by N receive elements. The elements used on
by a single receiver, we may use this steering-vector notation to
receive may or may not be the same as those used on transmit.
compactly write the data observed by the MIMO radar, again
The operation of a MIMO radar is illustrated in Figure 3.
corresponding to a single target at angle q0, as
We first construct a model for the data observed by a MIMO
radar that is due to a single point scatterer. This can be extended
to more complex scenarios because of linearity. Without loss of (11)
(12)
(14) If the noise vector v(t) consists of only thermal receiver noise
and no external noise sources are present, then the receiver
noise should be spatially white and Rv = IN, where IN is the N ×
The integral of the matrix is understood to be computed ele-
N identity matrix.
ment-wise. Each element of this matrix is recognized as the in-
The signal model in (15) describes the data observed by a
ner product of two of the transmitted signals. Each element of
MIMO radar given a target at an angle q0. Each element of the
the matrix in (14) is the zero-lag term of the cross-correlation
length-M N vector is the voltage signal observed if the target
function of two of the transmitted signals. The diagonal ele-
is in the center of the range bin of interest. This data vector
ments correspond to the autocorrelation functions of the wave-
is characterized by the MIMO steering vector in (17) and the
forms.
MIMO interference covariance matrix in (18). Note the funda-
The N × M data matrix Z(q0) given in (13) may be vector-
mental importance of the MIMO signal correlation matrix Rx,
ized by stacking its columns. By exploiting properties of the
which figures prominently in the observed data. We will see
Kronecker product operator this may be written as
that this matrix characterizes the performance of a MIMO radar.
(15)
C. THE MIMO SIGNAL CORRELATION MATRIX
From the definition of the MIMO signal correlation matrix Rx in
where the MIMO steering vector s(q), which corresponds to an
(14), we see that this M × M matrix describes the zero-lag term
angle q, is given by
of the cross- and autocorrelations of the M transmitted wave-
forms {xm(t)}. So, for any set of waveforms, we can compute
(16) the corresponding correlation matrix.
As previously mentioned, a phased array is a special case As we see from the signal model given in (13), the rank
of MIMO radar. In this case, each element transmits an identi- of the signal correlation matrix determines how observable the
cal signal up to a phase shift from element to element. Let x0(t) channel matrix H(q) is. Of course, in the degenerate case when
be the (scalar-valued) radar waveform that is common to each no energy is transmitted (Rx = 0), no portion of the channel ma-
element. To steer a beam in a direction , the appropriate phase trix is observed. The other extreme is the full rank case, where a
progression is applied if the transmitted signals are full-rank observation is available. The phased array case, where
the signal correlation matrix is rank-1, provides an observation
of a single subspace of the channel matrix.
(19)
A third case will also be of interest, namely the phased array
that uses a spoiled beam on transmit [23]. Instead of forming
where a* denotes the complex-conjugate (without transpose) of a narrow, high-gain beam, a phased array may trade some of
a vector, a. The corresponding signal correlation matrix is its transmit beamforming in return for illuminating a larger set
of angles. This may be accomplished by applying a nonlinear
phase progression or an amplitude taper across the aperture. An
(20)
alternate method is to transmit out of a single subarray. The
term “spoiled” indicates that the phased array intentionally de-
where we assume that the steering vector a* was normalized so grades coherence on transmit and accepts a loss in peak gain to
that and that the waveform x0(t) was normalized to form a wider beam.
have total energy of
(23) Observe that the average power of the output noise when the
weights w are applied is
Note that the signal correlation matrices Rx/PA and Rx/^ have
been normalized so that the trace (the sum of the diagonal ele- (26)
ments) of the correlation matrix is unity in both cases:
where Re is the spatial covariance matrix of the interference
(24) after matched filtering, which was given in (18). Similarly, the
signal power in the beamformer output is
mize the noise power, given in (26), yet preserve signals from
the desired direction q. This leads to the constrained optimiza- (32)
tion problem
(33)
for some arbitrary scalar g ≠ 0.
The weights will be unique if (and only if) the matrix, Re, is
invertible. When this is the case, the optimal weights are given This represents the (voltage) gain on a target at angle q0 rela-
by . tive to the noise as is evident from (25). If the optimal weights
However, for the interference covariance matrix Re to be in- to steer the beam in a direction q are employed, the resulting
vertible, Rv and Rx must be invertible. The spatial noise covari- MIMO array factor is
ance matrix Rv will generally be invertible due to the inevitable
presence of thermal noise in the receiver, but we have already
encountered a noninvertible signal correlation matrix, namely (34)
the rank-1 matrix Rx/PA of (20), which is the signal correlation
matrix of the phased array. Of course, this is not a problem in
the degenerate case where there is a single transmit element (M where we employ the notation f(q0; q) to imply that this is the
= 1), but this “phased array” is also using a set of waveforms array factor for a beamformer steered to a particular angle q.
that are mutually orthogonal. Observe that the first quotient in (34) is the transmit array factor
We cannot solve the system of equations in (29) precisely while the second quotient is the standard receive array factor. A
when the vector s(q) does not lie in the column space of the similar expression for the array factor is derived in [12] for the
matrix Re. However, inspecting the structure of the MIMO case of spatially white receiver noise (Rv = IN) using an alternate
steering vector in (17) reveals that valid steering vectors but equivalent formulation of the optimal spatial weights.
satisfy this requirement. So, while we cannot find a unique We are often interested in the gain in SNR for a target at
set of optimal spatial weights, optimal weights do exist. The angle q0, when a beam is steered to a particular angle q. This
minimum-energy weights that achieve the maximum SNR are gain is the magnitude-squared of the array factor and includes
of the same form as those in the case where the interference the subarray gain. If, for a target at angle q0, each transmit sub-
covariance matrix is invertible. This can be derived via the array has identical gain GTX(q0), and each receive subarray has
pseudoinverse. The optimal weights for the direction q are identical gain GRX(q0), the MIMO gain is
proportional to
(30)
(35)
The output of the optimal linear beamformer for a direction q to
a single target at an angle q0, has an SNR of
case, this extends to the transmit gain through the MIMO sig-
IV. THE PHASED ARRAY VS. ORTHOGONAL WAVEFORMS
nal correlation matrix Rx. Indeed, this matrix fully characterizes
the behavior of a MIMO radar. This highlights the fact that, The MIMO signal correlation matrix Rx characterizes the per-
although MIMO radar is enabled by flexibility in waveform formance of the radar antenna in terms of its transmit gain for a
generation, it is inherently an antenna-based technology. given set of waveforms {xm(t)}. We have also established that a
phased array and a radar using orthogonal waveforms represent
the two extremes of MIMO radars. The phased array has the
F. ANTENNA PATTERNS rank-1 signal correlation matrix Rx/PA given in (20). The sig-
The performance of an array antenna for use in a radar system nal correlation matrix for orthogonal waveforms Rx/^ given in
is well quantified by considering three gain patterns: the steered (23), is a full-rank matrix. Using our framework, we are able to
response, the beampattern, and the (angular) point spread func- determine the MIMO steering vector, interference covariance,
tion [27]. These describe the ability of the data collected by the and optimal spatial weights for each case. These quantities are
system to be used to digitally form beams in desired directions summarized in Table II.
with desired properties. Recall that the phased array seeks to steer its transmit beam
in some direction . This results in a transmit gain (neglecting
CC The steered response G1 (q) quantifies the ability of the
subarray gain) for a target at angle q0 of
array to digitally steer a beam in a direction q.
TABLE I
Description of Antenna Patterns. G (q 0; q) is the Gain for a Target at an Angle q 0 when the Beam is Digitally
Steered in a Particular Direction q.
size arbitrary receive beams. Exploiting the orthogonality of correlated waveforms, an improvement in gain in this direction
the transmit waveforms, a MIMO radar uses matched filters to is provided. The spoiled phased array or the radar using orthog-
isolate the contribution of each transmitter to provide a number onal waveforms has a steered response identical to the subarray
of spatially-diverse channels that are linearly combined to syn- pattern because no array gain is realized on transmit.
thesize arbitrary transmit beams. By considering the transmit beampattern, it is apparent that
Two things are evident when we compare the transmit gains the radar using orthogonal waveforms is able to synthesize a
of these two extreme cases of MIMO radar in (36) and (37). beampattern with the same mainlobe width and same sidelobe
First, the phased array provides additional transmit gain for tar- performance as the phased array. We recall from the steered re-
gets with angle . Indeed, if M transmit elements are used, sponse that the overall gain is not as high as the phased array,
the phased array can provide an improvement in SNR by a fac- but from this we see that the beampattern is preserved.
tor of M when compared with the orthogonal waveform case. The point spread function demonstrates that the orthogonal
However, this transmit beamforming gain provided by the waveform case possesses superior angular resolution perfor-
phased array is only applicable to targets in the direction where mance compared with the phased array. In fact, neither the full
the transmit beam was steered, . Because the phased array lacks phased array or the spoiled phased array are able to provide any
transmit degrees of freedom, it is unable to use digital process- resolution on transmit. Note that a similar point spread func-
ing to resteer its transmit beam. We see that the phased array tion is presented in [12], where it is (somewhat misleadingly)
provides improved gain in a particular direction, but a MIMO referred to as a beampattern.
radar provides a digital beamforming capability on transmit by To summarize, by transmitting orthogonal waveforms,
using orthogonal waveforms. a MIMO radar is able to achieve the wide area coverage of a
spoiled phased array while preserving its beampattern. It is
also able to provide improved angular resolution. Note that, in
A. A COMPARISON OF GAIN PATTERNS this case, a ULA was employed and the correlation among the
These concepts are made more concrete by considering the be- waveforms was varied. More novel configurations may be used
ampattern, steered response, and angular point spread function where orthogonal waveforms are able to operate in a sparse,
of the phased array and orthogonal waveforms. An example of irregular configuration.
each of these is presented in Figure 5. The (transmit) subarray
pattern and the patterns of the spoiled phased array are also pre-
B. A DISCUSSION OF SEARCH RATES
sented for reference.
From the steered response, we see that the phased array In many phased array radar systems, a high-gain, directional
is able to provide superior gain for the direction in which the beam dwells in a particular location for some coherent process-
beam was formed (in analog) on transmit. By transmitting M ing interval (CPI). The length of the CPI is chosen to meet some
TABLE 2
Signal Correlation Rx
use a single CPI that was M times as long as that used by each
phased array dwell. In both cases, the search rate and SNR on
any target will be equivalent.
This is consistent with the standard search-radar equation,
which demonstrates that the search rate of a radar is fundamen-
tally limited by its power-aperture product [30]. Indeed, the
SNR constraints on search rate are seen to be unaffected by the
choice of operating an array as a phased array or using orthogo-
nal waveforms.
This demonstrates that the “total” SNR of a coherent MIMO
radar is unaffected by the correlation matrix in many cases. How-
ever, as we saw before, using orthogonal waveforms provides
many benefits. This includes improved angular resolution and an
improved beampattern compared with the spoiled configuration.
Compare this with the case of statistical, i.e., noncoherent
MIMO radar. For a noncoherent MIMO dwell, M radar systems
simultaneously illuminate some area/volume. It is not clear that
the CPI duration could be decreased by a factor of M. Also, for
a volume-surveillance radar, the volumes illuminated by each
radar may not completely overlap, representing an additional
inefficiency. Careful bookkeeping is required when benchmark-
ing the performance of a noncoherent MIMO radar and compar-
ing it with the case where each radar operates independently
and data fusion occurs at a higher level, e.g., track-level fusion.
onstrate improved detection performance that is provided by a the response in other range bins, so we introduce a lag term t,
MIMO GMTI [32]. which allows us to capture the response of each matched filter
While the angular point spread function of the array used by to a shifted version of each of the other waveforms. The MIMO
a GMTI system is critical, it is of less importance to a synthetic signal correlation matrix for a lag t is given by
aperture radar (SAR) system, which seeks to form high- resolu-
tion imagery of the ground by synthesizing a synthetic aperture
(38)
much larger than its physical aperture through platform motion.
In this case, the physical aperture does not provide the resolu-
tion necessary to resolve targets in the along-track dimension. This can obviously be extended to include mismatches in Dop-
Instead, the role of the antenna is to act as a spatial filter to re- pler frequency [14, 38]. Following a similar development to that
ject returns that would otherwise be Doppler ambiguous for the above, the response of the M matched filters to the signals from
radar’s along-track sampling rate. In this case, the key figure of the N receivers given a target at an angle q0 may be written as
merit is the antenna beampattern.
This is an extension of the Vernier array approach that uses
multiple receive channels to improve area coverage rates in (39)
along-track sampling-limited applications [33]. This view of
MIMO SAR was presented in [34]. An analogous approach ap-
where the filtered noise matrix is
plies to the related technology of synthetic aperture sonar (SAS)
[24]. Other discussions of MIMO SAR are presented in [35],
[36], and [37]. (40)
(43)
A. THE MIMO RANGE RESPONSE
Up to this point, the analysis has tacitly assumed that the range
to the target was known so that the outputs of the matched fil-
ters in (13) may be sampled at the peak of the target response.
Compare this with the array factor in (34). This is interpreted as
In practice, a radar signal processor generates the sampled re-
the (voltage) gain on a target at an angle q0 provided by a beam-
sponse of the matched filter at a rate corresponding to the ex-
former steered to (t,q). The response of the linear beamformer
pected range resolution. Each sample is the (complex) voltage
that is steered to an angle q, and a lag t to a single target at an
associated with a range bin. We now extend the model devel-
angle q0, as
oped above in (15) for z(q0), which corresponds to the peak of
the matched filter response to a target at an angle q0 to include a
lag term t, which is relative to the peak of the response.
Observe that each of the elements of the matrix defined by
(14) is the inner product of a pair of transmitted waveforms. (44)
This describes the response of each waveform to the matched
filter constructed for each of the other waveforms sampled
at the peak of the autocorrelations. We are now interested in
For fixed angles q and q0, this may be considered the range
response of the waveforms. This is analogous to the waveform
autocorrelation, which describes the range response of the
waveform to the matched filter that is applied by a SISO radar
system.
Observe that the contribution of the receive array is equiva-
lent regardless of which range bin is considered. However, the
transmit array factor varies as a function of range as well as
angle.
For a moment, let us consider the SIMO case where the ma-
trix Rx is simply the scalar-valued autocorrelation function of
the transmitted waveform. This allows us to separate the varia-
tion in range, which is due to the signal autocorrelation func-
tion, from the variation in angle, which is related to the similar-
ity between the steering vectors.
However, in the MIMO case, unless the waveforms are Figure 6.
orthogonal, the range response can vary as a function of both Spectrograms of an up-chirp and a down-chirp.
angle and range in a nonseparable manner. This is alluded to in
[38], where the impact of range straddle loss is also discussed. may indirectly impact the detection performance of a MIMO
Of course, this latter effect is a problem encountered by any MTI radar because the characteristics of the cross-correlation
radar system [39]. energy may impact the cancellation ratio [40].
An interpretation of the result in (43) is that the cross-cor- The range response of a MIMO radar depends on the trans-
relation functions of the transmitted waveforms interfere with mit signal correlation matrix, as seen in (44). Instead of merely
each other in different ways depending on the angle of arrival. contending with the autocorrelation sidelobes of a single wave-
This impacts beamforming and direction finding [38] and the form, a MIMO radar will also need to operate in the presence of
cancellation of clutter for GMTI radar [40]. sidelobes arising from the cross-correlation of the waveforms.
REFERENCES
[1] Foschini, G., and Gans, M. On limits of wireless communications in a
Figure 7. Correlation properties and range response of an up- fading environment when using multiple antennas. Wireless Personal
chirp and a down-chirp. Communications, 6 (1998), 311–335.
[2] Southworth, G. Certain factors affecting the gain of directive anten-
or noise-like waveforms for these cases. In some applications,
nas. Proceedings of the Institute of Radio Engineers, 18, 9 (Sept.
it may be possible to design waveforms that minimize cross-
1930), 1502–1536.
correlation over certain regions of interest (e.g., over a certain
[3] Silver, S. Microwave Antenna Theory and Design (Radiation Labora-
range interval) or to design mismatched filters that suppress sid-
tory Series). New York: McGraw-Hill, 1949.
elobe energy [41, 42].
[4] Messer, H., Singal, G., and Bialy, L. On the achievable DF accuracy of
two kinds of active interferometers. IEEE Transactions on Aerospace
and Electronic Systems, 32, 3 (July 1996), 1158–1164.
VI. CONCLUSION [5] Alamouti, S. A simple transmit diversity technique for wireless com-
Coherent MIMO radar is a natural extension of the phased ar- munications. IEEE Journal on Selected Areas in Communications, 16,
ray concept. Just as digital beamforming on receive provides 8 (Oct. 1998), 1451–1458.
additional degrees of freedom that improve radar performance, [6] Dorey, J., and Garnier, G. RIAS, Radar à impulsionét antenna synthé-
transmitting independent waveforms provides further flex- tique. L’Onde Electrique, 69, 6 (1989), 36–44.
ibility. Like any new idea proposed to improve a decades-old [7] Fishler, E., Haimovich, A., Blum, R., Cimini, J. Chizhik, D., and Va-
technology, MIMO radar has not been without its critics [43]. lenzuela, R. Spatial diversity in radars - Models and detection perfor-
Indeed, many radar applications will not benefit from MIMO mance. IEEE Transactions on Signal Processing, 54, 3 (Mar. 2006),
and performance may be degraded by transmitting uncorrelated 823–838.
waveforms. In situations where performance is limited only by [8] Bliss, D., and Forsythe, K. MIMO radar and imaging: Degrees of
thermal noise, little improvement will likely be provided by freedom and resolution. In Conference Record of the Thirty-Seventh
MIMO. However, when performance is limited by other fac- Asilomar Conference on Signals, Systems and Computers, 2003, vol.
tors, e.g., multiplicative noise in SAR [34], clutter in GMTI 1, Nov. 2003, pp. 54–59.
[9] Li, J., and Stoica, P. MIMO radar with colocated antennas. IEEE Sig- [31] Melvin, W. A STAP overview. IEEE Aerospace and Electronic Sys-
nal Processing Magazine, 24, 5 (Sept. 2007), 106–114. tems Magazine, 19 1 (Jan. 2004), 19–35.
[10] Chernyak, V. Fundamentals of Multisite Radar Systems. Boco Raton, [32] Kantor, J., and Davis, S. K. Airborne GMTI using MIMO techniques.
FL: CRC Press, 1998. In 2010 IEEE Radar Conference, May 2010, pp. 1344–1349.
[11] Willis, N. Bistatic Radar. Raleigh, NC: SciTech Publishing, 2005. [33] Kock, W. Extending the maximum range of synthetic aperture (holo-
[12] Bekkerman, I., and Tabrikian, J. Target detection and localization us- gram) systems. Proceedings of the IEEE, 60, 11 (Nov. 1972), 1459–
ing MIMO radars and sonarsIEEE Transactions on Signal Processing, 1460.
54, 10 (Oct. 2006), 3873–3883. [34] Davis, M. S., Showman, G., Pechner, D., and Dai, L. Improving SAR
[13] Fuhrmann, D., and San Antonio, G. Transmit beamforming for MIMO image quality and area coverage through MIMO operation. In Pro-
radar systems using signal cross-correlation. IEEE Transactions on ceedings of the 57th Annual Meeting of the MSS Tri-Service Radar
Aerospace and Electronic Systems, 44, 1 (Jan. 2008), 171–186. Symposium, Monterey, CA, June 2011.
[14] San Antonio, G., Fuhrmann, D., and Robey, F. MIMO radar ambiguity [35] Krieger, G., Gebert, N., and Moreira, A. Multidimensional waveform
functions. IEEE Journal of Selected Topics in Signal Processing, 1, 1 encoding: A new digital beamforming technique for synthetic aperture
(June 2007), 167–177. radar remote sensing. IEEE Transactions on Geoscience and Remote
[15] Davis, M. S. MIMO radar. In Melvin, W. L., and Scheer, J. A. (Eds.), Sensing, 46, 1 (2008), 31–46.
Principles of Modern Radar: Advanced Techniques. Raleigh, NC: Sci- [36] Rennich, P. Four-platform distributed MIMO radar measurements and
Tech Publishing, 2012, pp. 119–145. imagery. In 2009 IEEE Radar Conference, May 2009.
[16] Biglieri, E., Calderbank, R., Constantinides, A., Goldsmith, A., Paul- [37] Ender, J., and Klare, J. System architectures and algorithms for ra-
raj, A., and Poor, H. MIMO Wireless Communications. New York: dar imaging by MIMO-SAR. In 2009 IEEE Radar Conference, May
Cambridge University Press, 2007. 2009, pp. 1–6.
[17] Bliss, D., Forsythe, K., Hero, A., and Yegulalp, A. Environmental is- [38] Friedlander, B. Effects of model mismatch in MIMO radar. IEEE
sues for MIMO capacity. IEEE Transactions on Signal Processing, Transactions on Signal Processing, 60, 4 (Apr. 2012), 2071–2076.
50, 9 (Sept. 2002), 2128–2142. [39] Cann, A. Range gate straddling loss and joint probability with partial
[18] Li, J., and Stoica, P. The phased array is the maximum SNR active correlation. IEEE Transactions on Aerospace and Electronic Systems,
array. IEEE Signal Processing Magazine, 27, 2 (Mar. 2010), 143–144. 38, 3 (July 2002), 1054–1058.
[19] Li, J., Stoica, P., and Zheng, X. Signal synthesis and receiver design [40] Rabideau, D. MIMO radar waveforms and cancellation ratio. IEEE
for MIMO radar imaging. IEEE Transactions on Signal Processing, Transactions on Aerospace and Electronic Systems, 48, 2 (Apr. 2012),
56, 8 (Aug. 2008), 3959–3968. 1167–1178.
[20] Kell, R. On the derivation of bistatic RCS from monostatic measure- [41] Ma, C., Yeo, T. S., Tan, C. S., Qiang, Y., and Zhang, T. Receiver design
ments. Proceedings of the IEEE, 53, 8 (Aug. 1965), 983–988. for MIMO radar range sidelobes suppression. IEEE Transactions on
[21] Hoctor, R., and Kassam, S. The unifying role of the coarray in aper- Signal Processing, 58, 10 (2010), 5469–5474.
ture synthesis for coherent and incoherent imaging. Proceedings of the [42] Hua, G., and Abeysekera, S. Receiver design for range and Dop-
IEEE, 78, 4 (Apr. 1990) 735–752. pler sidelobe suppression using MIMO and phased-array radar.
[22] Forsythe, K., and Bliss, D. MIMO radar waveform constraints for IEEE Transactions on Signal Processing, 61, 6 (2013), 1315–
GMTI. IEEE Journal of Selected Topics in Signal Processing, 4, 1 1326.
(Feb. 2010), 21–32. [43] Daum, F., and Huang, J. MIMO radar: Snake oil or good idea? IEEE
[23] Kinsey, R. Phased array beam spoiling technique. in Digest of 1997 Aerospace and Electronic Systems Magazine, 24, 5 (May 2009), 8–12.
IEEE Antennas and Propagation Society International Symposium, [44] Frazer, G., Abramovich, Y., and Johnson, B. Multiple-input multiple-
vol. 2, July 1997, pp. 698–701. output over-the-horizon radar: Experimental results. IET Radar, So-
[24] Davis, M., and Cook, D. MIMOSAS: Improving SAS performance nar, and Navigation, 3, 4 (Aug. 2009), 290–303.
with transmit diversity. In Proceedings of the 4th Underwater Acous- [45] Steyskal, H., Schindler, J., Franchi, P., and Mailloux, R. Pattern syn-
tic Measurements Conference, June 2011. thesis for TechSat21 - A distributed space-based radar system. IEEE
[25] Applebaum, S. Adaptive arrays. IEEE Transactions on Antennas and Antennas and Propagation Magazine, 45, 4 (Aug. 2003), 19–25.
Propagation, AP-24, 5, (Sept. 1976), 585–598. [46] North, D. An analysis of the factors which determine signal/noise dis-
[26] Johnson, D. The application of spectral estimation methods to bearing esti- crimination in pulsed-carrier systems. Proceedings of the IEEE, 51, 7
mation problems. Proceedings of the IEEE, 70, 9 (Sept. 1982), 1018–1028. (July 1963), 1016–1027.
[27] Johnson, D., and Dudgeon, D. Array Signal Processing: Concepts and [47] Donoho, D. L. Compressed sensing. IEEE Transactions on Informa-
Techniques. Upper Saddle River, NJ: Prentice-Hall, 1993. tion Theory, 52, 4 (Apr. 2006), 1289–1306.
[28] Hassanien, A., and Vorobyov, S. Phased-MIMO radar: A tradeoff be- [48] Yu, Y., Petropulu, A., and Poor, H. MIMO radar using compressive
tween phased-array and MIMO radars. IEEE Transactions on Signal sampling. IEEE Journal of Selected Topics in Signal Processing, 4, 1
Processing, 58, 6 (June 2010), 3137–3151. (Feb. 2010), 146–163.
[29] Cheston, T. C. Phased arrays for radars. IEEE Spectrum, 5, 11 (Nov.
1968), 102–111.
BIOS
[30] Scheer, J. The radar range equation. In Richards, M., Scheer, J., and
Holm, W. (Eds.), Principles of Modern Radar: Basic Principles. Ra- Michael S. Davis (M'01-SM'12) is a Senior Research Engineer
leigh, NC: SciTech Publishing, 2010, ch. 2. at the Georgia Tech Research Institute (GTRI), where he has