A ovel Range Detection Method for 6GHz LFMCW Radar Yizhong Wu,YingBao, Zhiguo Shi, Jiming Chen and Youxian Sun Department of Control Science and Engineering, Zhejiang University Email:{yzwu, jmchen, yxsun}@iipc.zju.edu.cn Department of Information Science and Electronic Engineering, Zhejiang University Email:{bao62, shizg}@zju.edu.cn Abstract The linear frequency-modulated continuous-wave (LFMCW) millimeter-wave radar has been widely used in automotive industry applications. The motivation of this paper is to propose a segmented range detection (SRD) method on the foundation of plentiful mature research on the target detection of LFMCW radar. According to the SRD method, the detection range (about m in the scene of vehicular collision warning) is divided into several segments and triangular modulation waveform which consists of several different frequencies is designed. Compared with typical digital signal processing method, the SRD method can availably decrease the bandwidth of the IF signal, therefore the in-band noise will be significantly reduced and the signal to noise ratio (SR) improved. Additionally, different range resolutions can be achieved and average range resolution improved. The SRD method is also validated through multiple outdoor experiments with the use of 6 GHz LFMCW millimeterwave radar system designed by our research group. The efficiency of the SRD method is demonstrated in terms of ranging accuracy and range resolution after the analysis of experiment results. I. ITRODUCTIO The number of private cars is increasing year by year and the market demand for automotive radar equipment is growing rapidly all over the world. Under the assumption that the instantaneous value of the distance to the front car in the same lane and the relative velocity of ego vehicle can be precisely acquired and displayed real time, the driving safety will be visibly boosted [1]. Millimeter-wave radar is really a better choice for its less sensitivity to environmental impacts, finer resolution and higher accuracy, compared with other sensing devices, such as infrared sensors, ultrasound sensors, image-processing sensors, and so on [2]. The millimeterwave automotive radars at 24 GHz, 6GHz and 77 GHz are commercially available on the market. The method of measuring the distance to the target primarily depends on which type of radar system we adopt [3]. Most of millimeter-wave radar systems for automotive application are based on the frequency-modulated continuous-wave (FMCW) technique, which has very long history [4]. Study on further performance increase of FMCW radar, such as reduction of the system noise, improvement of the radar resolution, and so on, is still a hot topic [] [6] [7] [8] [9] [1]. It is important to design a proper signal processing algorithm for FMCW radar, because the signal attenuation in millimeter-wave band is significant. Generally speaking, the performance of a radar system mainly depends on the waveform design of radar modulation signal and the corresponding signal processing algorithm. The modulation signal of triangular wave or saw-tooth wave is most frequently used in a linear frequency-modulated continuous-wave (LFMCW) radar system. At early stages the signal processing algorithm is to use a analog filter whose frequency band is very narrow. With respect to the analog filtering technique, although the in-band noise is smaller, the complexity and cost of relevant hardware design are high, furthermore, the flexibility of the system is insufficient. With the fast development of digital signal processing, it is now dominant that the range can be easily detected by calculating the beat frequency of the intermediated frequency (IF) signal through the utilization of fast Fourier transform (FFT) operation. But with this method, the IF processing circuit has a rather large noise bandwidth, resulting in performance degradation. The aim of this paper is to propose a segmented range detection (SRD) method which combines the idea of the above-mentioned analog and digital method. With the use of SRD, on one hand, the signal to noise ratio (SR) of the system can be significantly improved; on the other hand, the range resolution can also be improved when the range from radar sensor to target is short. The proposed SRD was tested and evaluated using the 6 GHz millimeter-wave radar system designed by our research group. The remainder of this paper is organized as follows: Section II describes the basic LFMCW radar principle. Section III gives the details of the novel SRD method. Section IV briefly introduces the radar system for experiments and presents experiment results. Finally, Section V concludes the paper. II. LFMCW RADAR PRICIPLE The well-known LFMCW radar principle is summarized in this section. The frequency of the transmitting signal and the received signal (the latter is the former delayed by a certain time lag) versus time is shown in Fig. 1. The IF frequency generated by mixing the transmitted and received signal (also referred to as the intermediate frequency signal) is shown in 978-1-4244-374-6/1/$. 21 IEEE
f frequency B f f IF frequency transmitted signal f IF T reflected signal time time Fig. 1. LFMCW radar waveform: frequency of the transmitted and received signal vs. time; the corresponding beat (IF) frequency vs. time Fig. 1. As seen from Fig. 1, the triangular wave frequencymodulated RF signal is transmitted from the transmit antenna and the received signal is actually an attenuated and delayed replica of the transmitted signal. To obtain the IF signal, part of the transmitted signal is coupled into a mixer as the local oscillator (LO) to down-convert the signal received by the receive antenna after reflection from the target. The ramp slope μ of a LFMCW chirp is defined as μ = B (1) T m /2 where B denotes the triangular wave modulation bandwidth of RF signal, T m is the repetition period of the triangular wave. The round trip delay time for signal propagation between the radar and the target is τ =2R/c (2) where R denotes the range from radar to target, c is the velocity of electromagnetic wave in free space. As shown in Fig. 1, we can derive the IF frequency f IF by the following equation: f IF = μτ (3) Substituting Equation (1) and (2) into (3) we obtain: f IF = 4B c f mr (4) where f m =1/T m is the repetition frequency of the triangular wave. The range to the target R is found by rearranging equation (4): R = c f IF () 4Bf m Thus, when the parameters c, B, and f m are known, the range of the target R can be obtained by calculating Equation () subject to that the IF frequency f IF can be obtained by FFT operation to the time-domain signal from the mixer. III. SEGMETED RAGE DETECTIO METHOD The performance of range detection of a radar system depends on the waveform design of the modulation signal in the transmitter and the corresponding signal processing algorithm to a large extent. In order to achieve lower in-band noise, higher ranging accuracy, finer range resolution, the SRD method is proposed with the integration of the advantages of fixed IF analog method and typical digital method with FFT operation. Prior to the description of the proposed SRD method, some fundamental concepts of the radar performance are discussed for further understanding the proposed method. 1) Radar performance metrics: According to the radar equation, the power obtained from the receiving antenna is: P r = P tg t G r λ 2 σ (4π) 3 R 4 (6) where P r is receive signal power, P t is transmitted power, G t and G r denote the gain of the transmitting and receiving antenna, λ is the wavelength of transmitted signal, σ is radar cross section of an object, R is distance betwen radar sensor and object. The SR of the IF signal can be described as: P t G 2 SR = P r /LkT BF = λ 2 σ (4π) 3 R 4 (7) LkT BF where L is the loss of RF module, k is Boltzmann s constant, T is the reference temperature which equals to 298. K, B is the noise bandwidth of IF signal, F is single side band (SSB) noise figure. From Equation (7) it is obvious that the in-band noise in the system is proportional to the noise bandwidth of the IF signal. In order to have a higher SR value, smaller noise bandwidth is necessary. Another performance parameter of LFMCW radar is the range resolution. According to Equation (), the range resolution ΔR = c Δf IF (8) 4Bf m where Δf IF is the minimum distinguishable frequency of the IF signal and determined by the sampling frequency of the analog-to-digital converter (ADC) and the number of FFT points. Subject to that the RF signal bandwidth, the ADC sampling frequency and the number of FFT points are fixed, we can see from Equation (8) that the range resolution is inversely proportional to f m, which is the frequency of the triangular wave for modulation. A. Fixed IF method with narrow-band analog filter When the analog signal is dominant previously, the signal processing algorithm for LFMCW radar is to use a very narrow band analog filter, such as acoustic surface wave filter. When obvious signal can be watched in the window of the output of the narrow band analog filter, the target range can be detected
Fig. 2. modulation waveform of one detection repetition period from the instantaneous frequency of the triangular wave and the center frequency of the filter. Owing to the small bandwidth of the filter, the in-band noise will be very small, but the hardware complexity and cost are high, and the flexibility of such a system is insufficient. It is difficult and complex to generate numerous triangular waves whose frequencies are continuously-varying. For system using fixed IF method, the frequency of the triangular wave for modulation is often tuned by hand. Furthermore, it is difficult to detect the range of a target in a fast and realtime manner, which is stringently required in the automotive collision-warning scenario. B. Digital signal processing method with FFT operation As the development of digital signal processing, it is now dominant that in LFMCW radar the range can be easily detected by calculating the frequency of the IF signal using FFT operation. When using the digital signal processing method with FFT operation, the frequency of the IF signal varies proportionally with the range of the target. In order to detect all the targets over the specified detection range, the IF circuit must have enough bandwidth for all the signal components in frequency domain corresponding to the detection range to pass. Hence, the bandwidth of IF circuit is comparably larger compared to the fixed IF method with narrow-band analog filter. Two consequent results occur. The first is, bigger IF bandwidth means more noises and lower SR, which is indicated in Equation (7). The second is, because the strength of the reflected signal from targets at different distances is inversely proportional to the signal strength in a nonlinear way, the IF signals of short distance target will flood the signals of long distance target, which make the detection of long distance target formidable. C. Proposed segmented range detection method From the above analysis, we can see that both of the abovementioned methods have their benefits and drawbacks. In this subsection, we try to combine these two methods to produce a new range detection method. The key idea of the proposed SRD method is to divide the whole detection range into a few segments, and each segment is detected by the use of triangular wave with certain frequency. That means different range segment is detected using different frequency triangular wave for modulation. As indicated in Equation (4), the frequency of IF signal is determined by the frequency of triangular signal for modulation and the target distance. In the typical signal processing method with FFT operation, the frequency of triangular signal is constant. Thus the IF bandwidth is linearly proportional to the maximum detection range. In the proposed SRD method, the detection range is segmented. Triangular waveforms with larger frequency are used for shorter target distance while triangular waveforms with smaller frequency are used for longer target distance. In this way, the IF signal bandwidth can be minimized and the in-band noise is reduced, resulting in a higher SR and longer detection distance. An illustration of time-domain modulation waveform of one detection repetition period is depicted in Fig. 2. In the illustration example, according to the SRD method, the whole target detection range has been segmented into short distance zone (1 m-22 m), middle distance zone (22 m- m) and long distance zone ( m-1 m). The corresponding frequencies used are f m1 =3.37 khz, f m2 =1. khz and f m3 =.67 khz respectively. When the detection range is 1 m to 1 m, it s enough that the IF circuit in our 6GHz millimeter-wave radar has its IF pass-band from 18 khz to 4 khz [11], one third of that using typical signal processing method. Also note that, as shown in Fig. 2, with regard to each range segment, periods of triangular waveform are used with the purpose of accumulation so as to further improve the SR of IF signal. The details of the proposed SRD method will be described in Algorithm 1 in consideration of three segments used in this illustration example. The outer loop is for the three segments, while the inner loop is for the accumulation of periods of triangular waveform to further improve the SR of IF signal. Algorithm 1 Illustration example of the proposed SRD procedure 1: for i=1:3 do 2: for i=1: do 3: -Select triangular wave frequency f m1=3.37 khz 4: -Calculate 124-point FFT : end for 6: for i=1: do 7: -Select triangular wave frequency f m2=1. khz 8: -Calculate 124-point FFT 9: end for 1: for i=1: do 11: -Select triangular wave frequency f m3=.67 khz 12: -Calculate 124-point FFT 13: end for 14: end for The proposed SRD method can improve the range resolution. Equation (8) points out that the range resolution is inversely proportional to the frequency of the modulation triangular wave. Given a certain IF circuit, the proposed SRD method can have different range resolution relative to different detection segment. In the illustration example, the range resolution in the short distance zone is approximately times smaller than that in the long distance zone. But for the typical case as described in section III-B, the achievable range resolution is as that in the long distance zone by the use of the proposed SRD method.
3 2 1 3 2 1 3 2 1 Fig. 3. The cart equipped with the 6GHz millimeter-wave radar system digital to analog converter DAC DSP triangular wave digital signal processor DAQ VCO mixer coupler transmit antenna receive antenna target vehicle x 1 x 1 x 1 Fig.. Spectrum of the IF signal with a single corner reflector still at 1m: f m1 =3.37 khz, f m2 =1. khz, and f m3 =.67 khz. 6 4 3 2 6 4 3 2 6 4 3 2 PC data acquisition card amplifier band pass filter 1 1 1 Fig. 4. Block diagram of the 6 GHz LFMCW radar system IV. EXPERIMET RESULTS A. Experiment setup To evaluate the performance of the proposed method, we prepare a cart on which a 6GHz millimeter-wave radar sistem is mounted as shown in Fig. 3. A functional block diagram of the radar system is shown in Fig. 4. The RF signal is generated with the combination of a voltage controlled oscillator (VCO) and a digital-to-analog converter (DAC) in the 6GHz band covering approximately 4MHz bandwidth. The DAC under the control of a digital signal processor (DSP) operates as the triangular wave signal generator. Part of the transmitted signal is coupled and used as local oscillator (LO) signal for the mixer in the receiver. In order to provide highest sensitivity and increase mixer isolation, separate transmit and receive antennas are used. The received signal is fed into the mixer, whose output is filtered, amplified by the IF circuit [11], digitized by a data acquisition card (DAQ) from I Instruments and transferred to a PC. The spectrum of the IF signal is analyzed using 124-point FFT in PC. B. Experiment results Experiments concerning stationary targets were conducted to evaluate the effectiveness of the proposed SRD method with the aid of the 6GHz millimeter-wave radar prototype. We used a corner reflector as the target located at specific distances in a playground to avoid multipath. For simplicity, three groups of experiments were carried out with one single x 1 x 1 x 1 Fig. 6. Spectrum of the IF signal with a single corner reflector still at 3m: f m1 =3.37 khz, f m2 =1. khz, and f m3 =.67 khz. corner reflector positioned at the different distance of 1m, 3m and 6m. When the corner reflector located at the distance of 1m in front of the radar prototype, the saved data is imported to the PC and the spectrum of the IF signal can be plotted by using MATLAB, as shown in Fig.. Since the distance of 1m falls into the short distance zone, theoretically one peak can be found in the resulting spectrum corresponding to the triangular modulation frequency f m1 =3.37 khz, but there should be no obvious peaks in the resulting spectrum corresponding to the triangular modulation frequency f m2 =1. khz and f m3 =.67 khz. This is verified in Fig., Fig. and Fig.. It can be seen from Fig. that the IF frequency is 184. khz, and the target range is calculated as 1. m. In the case of the corner reflector at the distance of 3m, the resulting spectrum of the IF signal is depicted in Fig. 6. Since the distance of 3m drops into the middle distance zone, in theory there is one peak in the resulting spectrum corresponding to the triangular modulation frequency f m2 =1. khz, and no obvious peaks could be found in the resulting spectrum corresponding to the triangular modulation frequency f m1 =3.37 khz and f m3 =.67 khz. This theoretical analysis
1 x 1 1 x 1 1 x 1 Fig. 7. Spectrum of the IF signal with a single corner reflector still at 6m: f m1 =3.37 khz, f m2 =1. khz, and f m3 =.67 khz. 1 x 1 7 6 4 3 2 1 2 4 6 x 1 Fig. 8. Comparison of the spectrum of the IF signal from two different experiments employing the proposed SRD method and the typical digital signal processing method with FFT operation respectively. is verified by comparison among Fig. 6, Fig. 6 and Fig. 6. As shown in Fig. 6, the IF frequency is 239.3 khz, and the target range is calculated as 29.9 m. Considering that a single corner reflector at the distance of 6m, which drops into the long distance zone, the resulting spectrum of the IF signal is depicted in Fig. 7. It can be seen that one peak exists in the resulting spectrum corresponding to the triangular modulation frequency f m3 =.67 khz, and no obvious peaks visible in the resulting spectrum corresponding to the triangular modulation frequency f m1 =3.37 khz and f m2 =1. khz. Additionally, the IF frequency is 213.3 khz, and the target range is 9.19 m. In the above experiment, it can be found that the proposed SRD method is effective in detecting targets in different distance zone by using triangular waveforms of different frequencies repeatedly. In the following experiments, the effect of the in-band noise on the detection performance is considered. We let a single corner reflector situated at the distance of 6m in front of the millimeter-wave radar. The spectrums of the IF signal from two different experiments employing the proposed SRD method and the typical digital signal processing method with FFT operation respectively are represented in Fig. 8. It can be clearly seen that the spectrum in Fig. 8 is seriously deteriorated because of the strong system noise and it is hard to determine accurate IF frequency. The comparison experiments demonstrate that the SRD method can significantly decrease the sensitivity to the noise compared to the typical digital signal processing method with FFT operation. V. COCLUSIOS In this paper a novel method named SRD for LFMCW radar is presented. It can offer higher range resolution and better noise performance when compared with the typical digital signal processing method with FFT operation. Experiment was carried out to verify the efficiency of the proposed SRD method by using a 6 GHz LFMCW millimeter-wave radar. Theoretically the proposed SRD method can be applied to all LFMCW radars and used in various radar applications. ACKOWLEDGMET The research is supported in part by SFC-Guangdong joint Project grant no. U733; SFC grant nos. 6814, 673621 and 6974122; 863 High-Tech Project no. 27AA4121. REFERECES [1] H. Zhang and K. Wu, Three-frequency principle for automotive radar system, in Proceedings of IEEE Radio and Wireless Conference, 24, pp. 3 318. [2] C. Wang, R. Rong Qian, and X. Sun, Low cost k-band FMCW radar modules for automobile application, in Proceedings of IEEE International Workshop on Radio-Frequency Integration Technology,, pp. 3 6. [3] S. Honma and. Uehara, Millimeter-wave radar technology for automotive application, Mitsubishi Electric Advance, vol. 94, pp. 11 13, 21. [4] D. Luck, Frequency modulated radar. McGraw-Hill, 1949. [] I. Matsunami, Y. akahata, K. Ono, and A. Kajiwara, Power delay profile matching for vehicle target recognition, in Proceedings of IEEE 7th Vehicular Technology Conference Fall, 29, pp. 1 6. [6] M. Abou-Khousa, D. Simms, S. Kharkovsky, and R. Zoughi, Highresolution short-range wideband FMCW radar measurements based on music algorithm, in Proceedings of International Instrumentation and Measurement Technology Conference, Singapore, 29. [7] H. Ko, K. Cheng, and H. Su, Range resolution improvement for fmcw radars, in Proceedings of European Radar Conference, 28, pp. 32 3. [8] A. Haderer, F. R., J. Schrattenecker, C. Wagner, and A. Stelzer, A fixed if 77-ghz FMCW radar sensor, in Proceedings of Asia-Pacific Microwave Conference, 28, pp. 1 4. [9] M. Abousetta and D. Cooper, oise analysis of digitised FMCW radar waveforms, IEE Proceedings - Radar, Sonar and avigation, vol. 14, no. 4, pp. 29 2, 1998. [1] E. Hyun, S. Kim, C. Park, and J. Lee, Automotive FMCW radar with adaptive range resolution, in Proceedings of the 28 Second International Conference on Future Generation Communication and etworking Symposia, 28, pp. 13 133. [11] Y. Bao, Z. Shi, and K. Chen, Intermediate frequency circuit design for a 6ghz LFMCW radar, in Proceedings of 21 International Conference on Microwave and Millimeter Wave Technology, 21.