Ultra Wideband Multiple-Input Multiple-Output Radar

Hammad A. Khan, Wasim Q. Malik, David J. Edwards, Christopher J. Stevens

Department of Engineering Science, University of Oxford
Parks Road, Oxford OX1 3PJ, UK.
Email: {hammad.khan, wasim.malik, david.edwards, christopher.stevens}@eng.ox.ac.uk

Key Words: Ultra Wideband (UWB), MIMO, radar.

ABSTRACT innocuous electromagnetic operation due to their low

transmitting power. UWB radar has been a subject of research
The utilization of ultra wideband (UWB) signals enables for a long time [3, 4]. But many areas are still being
the radar designer to solve the most important problems of researched due to the benefits of UWB radars [5]. This paper
radar target observation. The extremely wide bandwidth explores the use of UWB signals together with a spatially
enables greater information to be obtained due to high time distributed multiple transceiver scheme in radars. Multiple
resolution and the frequency dependence of the scattering transceivers (or multiple signaling) stem from the use of
centers over this large bandwidth. Increase in the radar’s spatial diversity in communication systems. In a
signal bandwidth can improve radar performance by providing communication system, four diversity schemes have been in
better range measurement accuracy, improving the target use: time diversity, frequency diversity, polarization diversity
identification and tracking capability, improving radar and space diversity. Similar schemes can be used for radars.
immunity to passive interference, and enhancing radar Time is a limiting commodity in a target tracking radar and
countermeasure against narrowband electromagnetic signal hence not an efficient technique. The use of polarization has
interference. Recently there have been many advances in been shown to have its benefits in UWB radars [6], but its
multiple-input multiple-output (MIMO) antenna systems in capabilities are limited by the scattering properties of the
communications. These diversity systems have been shown to target. Use of multiple frequency bands can be used in
have the potential to dramatically improve the performance of narrowband radars but it requires the use of further bandwidth,
the communications systems. Unlike the traditional which would otherwise be used for improving the range
beamforming approach, which uses highly correlated signals resolution. The use of spatially distributed multiple
of an array of transmitting or receiving antenna elements to transceivers is limited only in terms of the usage of more
collimate a beam towards a certain direction in space, MIMO space for implementation, but it improves the information
capitalizes on the independence between signals from different extraction from the target by viewing the target from multiple
transmitters and on the diversity of target scattering to angels. Coupling the multiple transceiver scheme with UWB
improve the information received from the response. signals yields fine range resolution along with the benefits of
Motivated by the advances and benefits of MIMO in multi-aspect imaging.
communications and advantages of using UWB signals, this Radars with more than one transmitting or receiving
paper presents the experimental investigation of UWB-MIMO antenna elements have been built and tested widely [7]. An
radars. The analysis of such radars has been carried out to array of radar elements is used with highly correlated signals
demonstrate its promising features in terms of better target to form a controllable beam, which can be collimated in a
identification and improved signal to noise ratio (SNR). certain direction to scan the desired space. Such radars are
known as phased array radars and the process is called
1. INTRODUCTION beamforming [8]. Beamforming uses highly correlated signals
at the radar antenna elements. Instead of beamforming, which
There are four metrics which determine the performance views one aspect of the target at a time, some work is being
of a radar: reliability of detection, accuracy of target parameter done in multiple-input multiple-output (MIMO) radar [9-12].
estimation, resolution of multiple targets and ambiguity of MIMO radar uses spatially distributed multiple transceivers. It
target estimates [1]. Signal to noise ratio (SNR) affects the capitalizes on the independence of transmitted and received
reliability of detection and accuracy of estimation. The signal signals and on the diversity of target scattering by viewing the
bandwidth affects all performance measures. Signal waveform target from multiple aspects. It is known that only very small
affects estimation ambiguity and repetition of waveforms changes in viewing angle causes large fluctuations in radar
affects detection, estimation and ambiguity [2]. The utilization cross section (RCS) because the reflected energy depends
of ultra wideband (UWB) signals improves the target heavily upon the range as well as the receive look-angle of the
resolution and accuracy of parameter estimation. UWB signals target [2]. This implies that while maximizing the received
are less prone to passive interference due to their large energy in radars, imaging the target with different orientations
bandwidth and have low probability of intercept and and appropriately combining the received signals could lead to
Fig. 1. Radar and target in the anechoic chamber (schematic
Fig. 2. MIMO antenna configuration. T and R denote
plan view).
different transmitters and receivers respectively.
better information from the target. MIMO radar is inspired
mainly by the synthetic impulse and aperture radar (SIAR) Three transmit and three receive antenna positions on the
[13]. In SIAR, the sparse array antenna elements emit antenna block were positioned as shown in Fig. 2. A
orthogonal waveforms, utilizing a time-space beamforming; minimum separation of antennas was maintained in order to
improving the angular resolution by the use of sparse arrays. avoid cross coupling and ensure the decorrelation of the
SIAR achieves the advantages of wideband radar using channels. For analysis, same simulated UWB signal was sent
narrowband signals at individual antenna elements. But it over all the channels to simplify the signal processing. The
suffers from the same performance degradation that the target was a metallic cylinder of height 30 cm and diameter 9
phased array radar is prone to: fading due to target’s cm. Three sets of measurements were performed in the
orientation fluctuation. Basic theoretical models of MIMO anechoic chamber. The first experiment sought to image the
radar have shown its advantages over traditional beamforming target alone. The second was the cylinder in the presence of
approach [9, 10]. clutter and the third measurement was without any target to
This paper presents the initial experimental results and correct for any unwanted reflections. Only one set of
analysis of UWB-MIMO radar. A set of measurements of transmitter and receiver was connected to the VNA at a time
UWB-MIMO radar signals was conducted using a vector and the positions of the antennas were changed to measure
network analyzer (VNA). A metallic target (a cylinder) was data for all permutations of transmitters and receivers. The
imaged with the radar in an anechoic chamber. The target clutter-based measurements are not included in this paper and
response was analyzed to compare MIMO radar with its will be a part of another publication with details of the clutter
variants. Results show that the MIMO radar outperforms its rejection algorithms used.
counterparts and provides better target identification showing In performing VNA based measurements, the VNA acts
the promise in this new architecture of radars. The results also both as a transmitter and a receiver. The antennas used in the
show that much more insight is needed to find out why some measurements at the transmitter and the receiver ends are
channels of the MIMO radar system perform better than the Vivaldi antennas [14].
others. The aim of the experiments was to analyze initial The channel was sounded over a frequency range of 1.99
measurements for UWB-MIMO radar in a non-multipath and GHz to 10.6 GHz in accordance with the FCC rules for UWB
less dense environment using simple signal processing tools communications and systems [15] and the amplitude of
and to establish the advantages achieved with UWB-MIMO transmission (S21 parameter of the VNA) was measured. The
radar. The initial experimental work has set the stage for signals were combined assuming a stationary channel.
investigation into an unexplored and promising new radar
architecture. 3. MIMO PROCESSING

2. EXPERIMENTAL SETUP The above experiment gave the transfer functions of three
systems: anechoic chamber without any target, anechoic
Fig. 1 shows the plan view of the experimental set-up. chamber with the target and anechoic chamber with target in
Both the target and the radar antennas were housed in an the presence of clutter. As noted above, the clutter-based
anechoic chamber and the antenna cables were connected to a measurements are not included in this paper. Since the data
VNA outside the chamber. Antennas and the target were was measured on one transmit-receive system at a time, using
supported by 4 m high resting blocks. The resting blocks were highly directional antennas, the signals must be combined in
separated by 2.85 m. This distance was the range of the radar. an appropriate manner for the MIMO system in order to get
 Fig. 3. Normalized Vivaldi antenna azimuth pattern. Fig. 4. Ultra-short pulse with the required frequency range of
1.99 – 10.6 GHz. The basic pulse shape is derived from a
full benefits of the multiple transceivers. There are well- Gaussian function.
established methods of combining signals in communications
systems [16]. The sidelobes of the resulting pulse were windowed using a
Gaussian window [8] to obtain the shape shown in Fig. 4. The
3.1 Antenna Radiation Pattern –10 dB frequency range for this pulse is 1.99 to 10.6 GHz in
agreement with the regulation requirements of UWB [15].
Fig. 3 shows the antenna azimuth pattern of the Vivaldi Let H ( f ) be the (corrected) frequency response of the target
antennas used in this work. The 5 dB power of the antenna is obtained by the VNA and X ( f ) be the transfer function of the
within 30o of both the azimuth. Elevation pattern is not
pulse as shown in Fig. 4. Then the target response, Y ( f ) to
relevant to the analysis as all the antenna elements are in the
same plane. This information is important when combining the this pulse is given as
signals at the receiver. The angle of view of the antenna Y ( f ) = H ( f ). X ( f ) (1)
element can be of great significance due to variation in the And the time response is given as
power received at various look-angles.
y (t ) = ∑ Y ( f )e − j 2π ft (2)
all f
3.2 Combining Techniques
The target’s response to this pulse is given in Fig. 5 using
The most popular diversity combining schemes are EGC and MRC. The response using ECG gives an erroneous
Selection Combining (SC), Equal Gain Combining (EGC) and value of 3.05 m (6.10/2 from the graph) for the range. The
Maximal Ratio Combining (MRC) [16]. In EGC, all the error in this result is around 7%. This clearly does not achieve
received signals are co-phased before being added. The extra the expected range resolution of UWB radars. The high error
phase accrued by each transmit-receive path is calculated can be due to the same SNR addition used in EGC. The
based upon the excess path length traversed by the signal Vivaldi antenna azimuth pattern (Fig. 3) reveals that different
measured with respect to the distance between the two resting receive look-angles have different SNR for the received
blocks shown in Fig. 1. In MRC, the weights of the signal are signals. SNR has thus to be weighed before combining the
proportional to the SNR at the receiver before they are co- signals at the receivers. This type of combining is MRC gives
phased. While MRC only offers an advantage of about 1 dB a better range estimate of 2.90m (1.8% error). The target
over EGC in a communication channel [16], it might have an response in both the figures reveals two distinct and close
effect of considerably improving the range accuracy due to the peaks, which can be used to find out information about the
variation of antenna power with the receive look-angle (Fig. target, e.g. the its cross sectional size.
3). Before carrying out the analysis, target frequency response The horizontal axis in Fig. 5 represents the distance, d,
from any unwanted reflections off the anechoic chamber walls corresponding to the time, t, where the receiver scans for
was corrected for. Channel transfer functions for different signals all the time and d=ct; c being the speed of light. A
channels in the MIMO radar were measured using the VNA. strong signal at time t comes from the transmitted signal that
Simulated UWB pulses were then used in Matlab® to find the has been reflected from an object at a distance d/2 from the
target response. Fig. 4 shows the simulated pulse of 0.10 ns radar. The vertical axis in the figure represents the amplitude
duration used to generate the required frequency range of 1.99 of the received signals in dBs. The threshold of the returned
to 10.6 GHz. This pulse was obtained by using a Gaussian signals is set to be 20 dB below the maximum. This is shown
function. For additional pulse shaping, unwanted frequencies as black dotted line in the figures. The rest of the response is
in the transfer function of the pulse were suppressed and the considered as noise. There is no signal above the threshold
final time domain pulse obtained by inverse Fourier transform. until after 5.95m for ECG and 5.7m for MRC.
 Table I
Correlation coefficients of the different channels that are
set up in the MIMO radar. An x indicates that the correlation
is between the same channel and hence not meaningful. (Note
that the right upper triangle of the table is redundant).

T1R1

T1R2

T1R3

T2R1

T2R2

T2R3

T3R1

T3R2

T3R3
T1R1 x
T1R2 .19 x
T1R3 .22 .33 x
T2R1 .49 .45 .47 x
T2R2 .56 .20 .51 .56 x
Fig. 5. Target response using equal gain combining (ECG) T2R3 .51 .19 .56 .52 .56 x
Fig.
and5.maximal
Target response using a) (MRC).
ratio combining Equal gain combining
Threshold (ECG
signal is ) T3R1 .17 .27 .73 .49 .37 .57 x
and b) Maximshown
al ratioby
combining (MRC). Both
a dotted line at –70 dB.figures show a T3R2 .23 .21 .68 .47 .36 .57 .83 x
T3R3 .48 .24 .55 .44 .44 .58 .54 .54 x
3.3 Correlation Coefficients

The MIMO radar sets up communication channels in the coherent integration was used which increases the SNR at the
anechoic chamber described in this paper. Every transmit- receiver by a factor equal to the number of pulses integrated
[2]. The figure reveals the advantage of multi-aspect radar by
receive antenna pair has a channel set up between them. This
gives rise to nine (ideally) independent channels. In a multiple observing the variation in range estimates and SNR with
transceiver system, such as this one, it is vital that the received different transceivers. It should also be noted that pulse
integration is not ideal in practice due to receiver
signals (and the communication channels) are independent of
each other [9]. This independence is measured by computing inefficiencies and some loss of energy occurs in integrating
the correlation coefficient between the signals received at the returned pulses due to their different times of arrival and
coherent integration becomes difficult to realize [2].
different receivers or between different communication
channels. The correlation coefficient between two signals (or
4.2 MIMO vs. SIMO and MISO
channels) x and y can be calculated as
E[ xy ] − E[ x]E[ y ] In Fig. 7, the MIMO radar is compared with a single-input
(3)
( E[ x ] − E[ x] )( E[ y ] − E[ y ] )
2 2 2 2 multiple-output (SIMO) radar and multiple-input single-output
(MISO) radar. The figure shows that there is no significance
where E[.] is the expected value of a variable. Here x and y difference in the information content between MISO and
are assumed to be random variables. A correlation value of SIMO radars. Fig. 6 and Fig. 7 also reveal the promising
0.7 is used as the threshold below which the signals are advantages of MIMO radar for enhanced SNR, better range
considered independent [17]. Table I shows the correlation accuracy, and the possibility of obtaining more target
coefficients between all the nine channels set up in this information. The last improvement stems from the highly
experiment. It is observed that apart from one anomaly detailed signal peaks of MIMO. Given a better representation
(between T1R3 and T3R1), the correlation values are well of the target responses (using time-frequency analysis), this
below the 0.7 threshold, hence showing the validity of the information from the MIMO radar can be used to find out the
implemented multiple transceiver scheme. target size and orientation if its shape is known or vice versa
and fairly hidden targets can also be detected and identified.
4. COMPARATIVE RESULTS These issues will be addressed in a later publication.

UWB radars mostly use single transmit-receive systems 5. CONCLUSIONS

or multi-static single transmit-multiple receive systems [3].
Some systems use multiple pulses at the transmitter and pulse This paper has presented experimental and simulation
integration at the receiver to enhance the SNR. This section results of the preliminary work on UWB-MIMO radar systems
compares different approaches with the proposed MIMO with three transmit and three receive antennas. The MIMO
radar. radar has been compared with its variants – the pulse
integration radar and the MISO and SIMO radars. The results
4.1 MIMO vs. Single Transmit-Receive System show the promising features of UWB-MIMO radar, which
will be explored further in terms of clutter rejection and target
Fig. 6 shows the comparison of MIMO radar with pulse identification in dense environments before realizing the full
integration by two single transmit-receive systems. Ideal potential of these types of radars.
 Fig. 6. Comparison of MIMO radar and pulse integration. Fig. 7. Comparison of target responses from MIMO, MISO
Two different pulse integration transceivers are used. and SIMO radars. Threshold signal is shown by a dotted line
Threshold signal is shown by a dotted line at –70 dB. at –70 dB.

ACKNOWLEDGEMENTS on Adaptive Sensor Array Processing. Lexington,

Massachusetts, 2003.
Hammad Khan’s work is supported by the Rhodes [13] C. Baixiao, L. Hongliang, W. Yajun, and W. Jun,
scholarship. "Analysis and experimental results on sparse array
synthetic impulse and aperture radar," in Proc. 2001 CIE
REFERENCES International Conference on Radar, Oct. 2001, pp. 76-80.
[14] P. J. Gibson, "The Vivaldi aerial," in Proceedings of the
[1] W. M. Siebert, "A radar detection philosophy," IEEE 9th European Microwave Conference. Brighton, UK,
Trans. on Information Theory, vol. 2, pp. 204-221, 1956. 1979.
[2] M. I. Skolnik, Introduction to Radar, 3rd ed: McGraw- [15] "Authorization of Ultrawideband Technology, First
Hill International Editions, 2001. Report and Order, Federal Communications Commission,
[3] J. D. Taylor, Ultra-wideband Radar Technology: CRC February 14, 2002," Federal Communications
Press, 2000. Commission, Washington, DC, USA FCC 02-48, 14
[4] L. Y. Astanin and A. A. Kostylev, Ultra-wideband Radar February 2002.
Measurements-Analysis and Processing: IEE, 1997. [16] D. G. Brennan, "Linear diversity combining techniques,,"
[5] I. J. Immoreev and J. D. Taylor, "Future of Radars," in Proc. IRE, vol. 47, pp. 1075-1101, 1959.
2002 IEEE Conference on Ultra Wideband Systems and [17] G. J. Foschini and M. J. Gans, "On the limits of wireless
Technologies, 2002, pp. 197-199. communications in a fading environment," Wireless Pers
[6] Y. Xu, "Random noise polarimetry: theory and Commun, vol. 6, pp. 311-335, 1998.
applications towards subsurface probing." PhD Thesis:
University of Nebraska-Lincoln, 1997.
[7] A. Farina, Antenna Based Signal Processing Techniques Hammad Khan was born on 15th April 1980
in Gujranwala, Pakistan. He received his B.S.
for Radar Systems. Norwood, MA: Artech House, 1992.
degree in Electronic Engineering from GIK
[8] S. Haykin, J. Litva, and T. J. Shepherd, Radar array Institute of Engineering Sciences and
processing, 1st ed. New York: Springer-Verlag, 1993. Technology, Topi, Pakistan, in 2002. He
[9] E. Fishler, A. Haimovich, R. Blum, L. Cimini, D. worked as a Software Engineer and a Project
Chizhik, and R. Valenzuela, "MIMO radar: An idea Management Consultant at Techlogix
whose time has come," in Proc. of the IEEE Int. Conf. on Pakistan Ltd. for a year before joining the
Radar. Philadelphia, PA, April 2004. Communications Research Group at the University of Oxford, UK
[10] E. Fishler, A. Haimovich, R. Blum, L. Cimini, D. where he is working towards a DPhil degree.
Chizhik, and R. Valenzuela, "Spatial diversity in radars- Mr. Khan is a member of IEEE, IEEE Communications Society
and IEE. He has received a number of scholarships and awards
models and detection performance.," IEEE Trans. on
including the Rhodes scholarship to study at Oxford. His research
Signal Processing, 2004, In Press. interests include radar theory and applications, electromagnetics and
[11] D. Rabideau, "Ubiquitous MIMO digital array radar," in ultra wideband systems.
Proc. of the Asilomar Conference on Signals, Systems and
Computers, November 2003. Wolfson College, Linton Road Oxford OX2 6UD, UK.
[12] P. Bidigare, "MIMO Capacity of Radar as a
Communication Channel," in Eleventh Annual Workshop [email protected]