Introduction
The safety of drivers, passengers, and pedestrians has been of major interest in the last few decades. To this end, radar sensors are considered an essential means to detect not only other vehicles, pedestrians, or bicyclists but also the entire road environment. Notably, many commercial radar sensors in the automobile industry use microwave and millimeter-wave technology, which offer the advantages of low fabrication cost and high reliability [
1,
2]. The easy and relatively low-cost availability of several radar channels on one chip, made it possible to establish a new system and antenna configuration featuring one or more transmit and receive channels, each connected to an antenna element. For many years, microstrip patch antennas have been the most popular radiators used in automotive radar systems due to their compact size, low profile, and directional radiation pattern. Moreover, they cover low to moderate bandwidths of frequency bands for radar applications. In addition, they are easy to integrate, since they can be directly printed onto a PCB, and do not require additional components or multiple manufacturing steps.
A typical vehicle radar uses frequency-modulated continuous wave (FMCW) signals in the millimeter band of 76–81 GHz. The frequency of the FMCW signal, also called a chirp signal, increases or decreases linearly during a beat cycle time. The slope of the modulation frequency with regard to the beat time period determines radar performance [
3,
4].
This indicates that short-range radars (SRRs) with wide-angle array antennas used in automobiles must be designed to operate within the millimeter band of the 77–81 GHz [
5,
6]. In particular, when array antennas with beamwidths greater than 150 degrees are used, a wide field of view (FoV) radar (WFR) for blind spot detection serves to prevent object non-recognition problems. Notably, the detectable range of a radar, also called the FoV, has the greatest effect on the half-power beamwidth (HPBW) of an antenna. In general, the antenna of a wide-angle radar requires a 4 GHz bandwidth in the 77–81 GHz range, along with a coverage of 150°.
In this paper, we design and implement a radar system for short-range detection that covers the 77–81 GHz bandwidth and has a wide beamwidth of 150°. Furthermore, to measure and verify the beamwidth of the proposed radar module, a trihedral corner reflector with an RCS of 180 m2 is used as the reflector target, while target detection is processed using the CA-CFAR (cell averaging constant false alarm rate) algorithm.
Design of Wide Beamwidth Microstrip Patch Antenna
Fig. 1(a) and 1(b) present the structures of a conventional microstrip comb line array antenna and the proposed microstrip patch array antenna, respectively. The structure in
Fig. 1(a) was created by connecting 10 open stubs with a feed line, while that in
Fig. 1(b) has a shorted patch structure, likely a planar inverted-F antenna. Both were designed using a comb line structure as well as a 10-by-1 array structure on the dielectric substrate of RO3003 (
ɛr = 3.0), maintaining a thickness of 5 mil using AWR AXIEM.
Fig. 2 compares the E-plane radiation pattern simulation results at 79 GHz obtained using AWR AXIEM for the antennas presented in
Fig. 1. It is observed that the 3-dB beamwidth is 165.1° and 70.4° for the proposed antenna and the conventional antenna, respectively. Therefore, the 3-dB beamwidth of the former is 94.7° wider than that of the latter.
Table 1 summarizes the dimensions of the design parameters, where
Wa and
La indicate the width and length of each radiator, respectively, while
Li,i+1 is the distance from
i-th radiator center to (
i + 1)-th radiator center. Furthermore, the
S11 of the designed antenna was found to be less than −7.5 dB in the operating frequency band of 77–81 GHz, as shown in
Fig. 3.
Fig. 4 illustrates the simulated radiation patterns of the designed antenna, while its radiation characteristics are summarized in
Table 2. The designed antenna achieved a peak gain of more than 8 dBi and a 3-dB beamwidth of more than 160°.
Table 3 presents a comparison of the gain, polarization, and HPBW of the E-plane and H-plane of the proposed antenna and other antennas proposed in existing research [
5] and [
6] as well as Texas Institutes (TI) AWR1642BOOST. The proposed antenna outperformed the other designs in terms of HPBW.
Radar Module Design
Fig. 5 presents the structure of the radar system introduced in this paper. The signal processing module named as the chip of AWR1642 [
7], is composed of an integrated microprocessor control unit and a digital signal processor, which help discriminate the incoming signals from four receive channels and the outgoing signals from two transmit channels. The additional subsystem composed of the antennas and RF front end is connected to the four input ports and the two output ports of the signal processing module.
Fig. 6 presents photographs of the entire radar module. The sensor chip and antennas are included in the sensor module, as shown in
Fig. 6(a), which is fabricated on the dielectric substrate of RO3003 (
ɛr = 3.0). In this study, we used AWR1642 as the sensor chip.
Fig. 1(b) shows the structure of the wide beamwidth antennas. Two transmit antennas and four receive antennas were connected to the output ports of the transmit channels and the input ports of the receive channels of the radar sensor chip, respectively. The separated distance between the receive antennas was half-wavelength, and that between the transmit antennas was two wavelengths to be included in the four receive channels. The control unit, shown in
Fig. 6(b), consisted of an XDS110 [
8], a USB port, a mode setting switch, a DC power supply, and so on. Notably XDS110 is a new class of debug probes for TI embedded processors, which was connected to the target board through a TI 20-pin connector and to the host PC via high speed USB2.0. The control unit was fabricated on the dielectric substrate of FR4 (
ɛr = 4.4).
Radar Module Fabrication and Measurement
Measurements pertaining to the fabricated wide FoV radar in
Fig. 6(c) were obtained using a trihedral corner reflector, as depicted in
Fig. 7. Since the dimensions of the reflector were fixed by the inner sides
l, the radar-cross-section (RCS) was calculated using
Eq. (1). In this study, the inner side
l = 159.89 mm.
The detected power
Pr received by the radar [
9–
11] was obtained using the following equation:
where Pt is the power of the radar transmitter, G refers to antenna gain, and R indicates the distance between the radar and the corner reflector target.
Fig. 8 outlines the detected power calculated using
Eq. (2) and the measured power of the trihedral corner reflector in terms of distance. It is observed that the power decreases in inverse proportion to the fourth power of the distance.
Furthermore, solving
Eq. (2) to obtain the RCS gives the following equation:
where Pr is the measured power at distance R, c is the speed of light, and f is the frequency.
Using the power measured according to distance and
Eq. (3), the RCS with regard to frequency was calculated, as in
Eq. (1).
Fig. 9 compares the RCS estimated using
Eqs. (1) and
(3), showing that the RCS calculated for a distance of 5 m is almost identical to the value calculated by
Eq. (1).
Fig. 10(a) depicts a block diagram of the CA-CFAR process [
12]. Noise samples were extracted from reference cells around the cell under test (CUT). Subsequently, guard cells were placed adjacent to the CUT to prevent signal components from leaking into the reference cell, which could adversely affect noise estimation. The CA-CFAR threshold,
CZ, was determined from the length of the guard cell, the length of the reference cell, and the bias value. Successful detection in the CUT was declared when:
where
C refers to CFAR bias, calculated based on the desired false alarm rate as follows [
13]:
where
M signifies the length of the reference cells and
Pfa is the false alarm rate. Therefore, cells with amplitudes more than the CA-CFAR threshold were considered detections, as shown in
Fig. 10(b).
Fig. 10(b) presents the detected results at a 0° front view. Two peak values are observed one at 0 m and the other at 5 m in the test distance. The peak at 0 m is the DC component generated by the down-conversion process in the radar module receiver, while the peak at 5 m represents the actual detection of the target.
Fig. 10(b) also shows that the CA-CFAR threshold exhibits high shoulders beside the peaks. This finding can be attributed to the length of the guard cells as the length of the guard cells increases, so does the distance between the shoulders centered on the peak. In addition, as the length of the reference cells decreases, the fluctuation of the threshold becomes the same as the input signal. This indicates that it is necessary to appropriately adjust the length of the guard cells, the length of the reference cells, and the false alarm rate
Pfa according to the target’s situation. For instance, in
Fig. 10(b), the CA-CFAR uses two guard cells and 10 reference cells with
Pfa = 10
−2.
The proposed radar sensor module generated the FMCW waveform at 77–81 GHz, and downconverted the reflected signal to 5 MHz IF (intermediate frequency). The IF was digitized by an ADC, which sampled at 5,000 ksps, and the sampling data was collected in DAQ. The radar configurations used for data collection are shown in
Table 4. To enhance the echo signal level, one target with a trihedral corner reflector was moved in front of the proposed radar and the AWR1642BOOST. Subsequently, the trihedral corner reflector was rotated more than 140° to compare the characteristics of the two radar sensor modules.
Fig. 11 illustrates the radiation pattern of the wide FoV radar in
Fig. 6. The test environment of the radar module for measuring antenna characteristics was an open site. To determine whether the objects were being detected, one reflector was placed 5 m in front of the radar, after which the radar sensor was rotated by 5° from −90° to +90°.
In
Fig. 11, the radiation pattern of the designed radar module exhibits a ripple characteristic, which can be attributed to its fabrication on a finite ground plane. The surface wave of the fabricated antenna was reflected by the edge of the finite ground plane, thus causing the ripple pattern. Despite the manufactured radar module exhibiting some ripple characteristics,
Fig. 11 shows that the radar module achieved a radiation pattern of 150°.
The graph presented in
Fig. 12 the measurements obtained by the wide FoV radar proposed in this paper with those of the AWR1642BOOST. Notably, AWR1642BOOST is an evaluation board for millimeter-wave sensors manufactured by TI. To compare the two radars, measurements were conducted in the measurement environment shown in
Fig. 11. The AWR1642BOOST detected the object from −70° to +65°, and beyond which it failed to show accurate detection. In contrast, the wide FoV radar detected the object at all tested angles. Overall, the proposed radar was about 15° wider than the AWR1642BOOST.
Conclusion
This study implements an SRR system for broadband and wide-angle detection in millimeter-wave applications owing to its advantages, such as good performance, simplicity, and low power handling. In addition, a short-range radar for application in vehicles was designed and fabricated in the millimeter band of 77–81 GHz. To conduct measurements, a trihedral corner reflector was utilized. Notably, the RCS measured at a distance of 5 m was consistent with the theoretical value. To verify whether objects were indeed being detected, CA-CFAR processing was employed. A reflector was placed 5 m in front of the radar, after which the radar sensor was rotated by 5° from −90° to +90°. The measurement results showed that the detection performance improved in the position and azimuth angles. Moreover, the radar exhibited good detection performance at a view angle of 150° as well as a broad bandwidth of 4 GHz. Therefore, the proposed radar module is applicable for use in blind spot detection sensors in autonomous vehicles.