Академический Документы
Профессиональный Документы
Культура Документы
net/publication/252019876
CITATIONS READS
37 2,410
3 authors, including:
Ard-Jan de Jong
21 PUBLICATIONS 158 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Ard-Jan de Jong on 09 June 2016.
Abstract—Multiple input/multiple output (MIMO) radar system The easiest way to obtain orthogonal waveforms is time
performance benefits from the capability to simultaneously division. At each time only a single transmitter transmits. The
transmit and receive multiple orthogonal waveforms. For pulse receivers receive the signals which correspond to a single
radars fitting orthogonal waveforms have been developed. These transmitter. The path from the transmitter to target to receiver
waveforms are however not necessarily suitable for frequency- is associated with the received signal. In time division, the
modulated continuous-wave (FMCW) radar. signal corresponding to a whole array (with many
The major benefit of the FMCW radar principle is that the transmitters) requires multiple pulses, thus reducing the
bandwidth of the beat signal is generally much smaller than the effective pulse repetition frequency (PRF). Therefore, time-
signal bandwidth, relaxing sampling requirements. Preferably,
division multiplexing might not be adequate for operations
this benefit should be preserved when applying orthogonal
waveforms in an FMCW MIMO radar configuration. In this
requiring high PRF.
paper, orthogonal waveforms compatible with the FMCW radar For frequency division there is no overlap in frequencies
principle will be discussed and tested by experiment or transmitted; each transmitter occupies at least a bandwidth in
simulation. the spectral domain. Simultaneous reception of the different
waveforms requires an extended bandwidth to be sampled.
I. INTRODUCTION This may lead to a combined bandwidth of all transmitters
The operations of the army have shifted in recent years wider than can be sampled by present-day analog-to-digital
from an open environment with regular forces to an urban convertors.
environment with irregular opponents. The challenges in an Code division requires no increase in sampling rate. All
urban environment are maintaining situation awareness, transmitters transmit at the same center frequency with similar
detection of small targets such as improvised explosive bandwidths. For pulsed radar systems the matched filter is
devices, and identification of friendly forces, insurgents, and used for filtering the different waveforms. For frequency-
neutral civilians. For the navy the operational environment modulated continuous-wave (FMCW) radar this approach is
becomes more complex also, since operations shift to littoral not possible.
areas and harbors.
In the following section, orthogonal waveforms that are
Multiple input/multiple output (MIMO) operation compatible with FMCW radar principle will be discussed. In
improves radar performance in challenging environments. Section III, the performance of the proposed waveforms will
MIMO radar combines the benefits of the floodlight concept, be assessed with the aid of measured data and simulations.
longer integration times, and frequency and spatial diversity of The paper is completed with the conclusion in Section IV.
multiple transmit and receive antennas to improve detection,
tracking, and recognition of extended objects [1]. II. ORTHOGONAL WAVEFORMS FOR FMCW RADAR
The key ingredient of MIMO radar operation is that As opposed to pulse radar systems, FMCW radar systems
multiple orthogonal waveforms can be used simultaneously, transmit a continuous wave. The frequency of the transmitted
where each waveform is associated with a single transmitter. waveform is linearly modulated. Targets present in the
Orthogonal waveforms are defined as waveforms with very antenna beam scatter a fraction of the transmitted signal back
low cross correlation. Ideally, the cross correlation should be in the direction of the radar. After a certain time delay this
zero for all time delays and Doppler shifts. Although some backscattered signal is captured by the radar receiver.
codes exist with very low autocorrelation in time, such as
In the receiver, the backscattered signal is demodulated
Barker-codes, and codes exist with very low cross correlation
with a nondelayed replica of the transmitted signal (i.e. the
when time synchronized, such as Walsh-Hadamard and Gold
reference signal) and subsequently low-pass filtered. The
sequences, codes which are orthogonal in time and Doppler
resulting signal is called the beat signal. The frequency of the
are rare.
beat signal is a measure for the range to the target. One major
∆fb
{ ( )}
s R (t ) = exp i ⋅ 2π fc t + πγ t 2 . (5)
5. Fourier transform back to beat frequency domain. Figure 3. The beat spectrum of a single receiver when three transmitters are
In order for the RVP compensation to align the data in switched-on with 2 MHz beat-frequency offset. The y-axis is in dB scale.
time correctly with the known BPSK code, the response of the
target should be sufficiently narrow after the first Fourier The tests were carried out in the anechoic room at TNO
transform. However, a narrow response in the frequency with a corner reflector in front of the radar. Due to the limited
domain yields a low suppression of the other semi-orthogonal size of the anechoic room, the reflector could be placed at a
codes. If the target response becomes too wide, the maximum range of 7 m from the radar. As a consequence, the
compensation of the BPSK code needs to take into account the response to the reflector and the array crosstalk are in adjacent
influence of the RVP compensation. The cross correlation range resolution cells (recall that the range resolution of
between received signals with different codes but the same EMERALD is 3 m). The large peaks as seen in Figure 3 are
beat frequency originates from the original bandwidth of the actually due to the array crosstalk. The actual responses to the
BPSK code. corner reflector are much smaller.
The response to the corner reflector is visible in Figure 4.
III. RESULTS This figure shows a detailed look of the peak around 4 MHz.
The experiments to verify the feasibility of the proposed As can be seen, the response to the reflector is indeed very
orthogonal waveforms have been conducted with EMERALD close to the array crosstalk. As a result, the response to the
[3], [4]. EMERALD is an X-band FMCW radar equipped with reflector is corrupted by phase noise associated with the array
an active electronically scanned array. The signal bandwidth is crosstalk. It was therefore not feasible to perform e.g. beam
50 MHz (3 m range resolution). forming to show experimentally that coherency between the
transmitters is preserved when applying beat-frequency
In EMERALD, the FMCW waveforms are generated division. In simulations the coherence was present and beam
digitally with the aid of a direct digital synthesizer (DDS). forming could be applied.
130
120
Energy T2; ∆B = 4.8 MHz
120
115
Array Crosstalk
110 110
Power [.]
Response to Reflector
Power [.]
105
100
100
90
95
80
Energy T3; ∆B = 9.6 MHz
90
85 70
0 1 2 3 4 5 6 7 8 9 10
Beat frequency [MHz]
3.6 3.7 3.8 3.9 4 4.1 4.2 4.3 4.4
Beat frequency [MHz] Division by sweep-rate offset
Figure 4. A detailed look at the peak around 4 MHz; the response to the 130
reflector is much smaller than the crosstalk. The y-axis is in dB scale. Focused Response
(close to zero-frequency)
120
C D
Figure 6. Phase code with fixed number of samples per symbol (red)
compared to a code with variable number of samples per symbol (blue).
D. Ambiguity Functions
Figure 7. Ambiguity functions: A) autocorrelation, B) cross correlation
The different orthogonal waveforms are also compared by with frequency offset, C) cross correlation with chirp-rate offset, and D) with
investigation of the ambiguity function. In Figure 7A the different BPSK modulation.
autocorrelation is shown for a single sweep for a stationary
target at 75 m (sweep time of 1 ms and bandwidth of The proposed waveforms that are orthogonal in the beat
150 MHz). As is expected for a FMCW signal the range and frequency domain lead to increased sample rates, as compared
radial velocity are ambiguous. In Figure 7B the correlation to the single transmitter case. However, compared to for
with the response from a transmitter with a 2 MHz beat instance FDMA, the required increase in sample rate is
frequency offset is shown. The signal is suppressed by 80 dB. relatively small. Chirp-rate and code division do not
Figure 7C shows the correlation of the response with signal necessitate higher sample rates but they have the disadvantage
which has a difference in sweep rate of 5.3 GHz/s. The signal that the energy associated with orthogonal transmissions is
is suppressed by 40 dB, similar to the expected noise spread over a wide band, effectively increasing the noise floor.
reduction (fluctuations in the cross correlation are very small
compared to the color scale). Finally, Figure 7D shows the REFERENCES
cross correlation between the response with a different BPSK [1] B. J. Donnet and I. D. Longstaff, “MIMO Radar Techniques and
code (100 bits with between 50 and 150 samples per bit). The opportunities,” in Proc. EuRAD, Manchester, U.K., September 13 – 15,
suppression is again noise-like, about 35 to 40 dB. 2006, pp. 112 – 115.
[2] J. J. M. de Wit, W. L. van Rossum, and F. M. A. Smits, "Principal
IV. CONCLUSION scatterer detection in buildings with SAPPHIRE," in Proc. EUSAR,
Aachen, Germany, June 7 – 10, 2010.
Orthogonal waveforms compatible with the FMCW radar
[3] C. M. Lievers, W. L. van Rossum, A. P. M. Maas, and A. G. Huizing,
principle have been proposed and their performance has been “Digital beam forming on transmit and receive with an AESA FMCW
evaluated using measured data and simulations. When radar,” in Proc. EuRAD, Munich, Germany, October 8 - 12, 2007,
applying orthogonal waveforms in FMCW radar, it is pp. 47 – 50.
important that the main benefit of the FMCW radar principle, [4] W. L. van Rossum, C. M. Lievers, A. P. M. Maas, and A. G. Huizing,
i.e. the relative low sample rate, is preserved. “Suppression of sidelobe scatterers in an AESA FMCW radar,” in
Proc. IEEE Radar Conf., Rome, Italy, May 26 – 30, 2008, pp. 1167 –
1171.