DSRC using OFDM for roadside-vehicle communication systems Akihiro Kamemura, Takashi Maehata SUMITOMO ELECTRIC INDUSTRIES, LTD. Phone: +81 6 6466 5644, Fax: +81 6 6462 4586 e-mail:kamemura@rrad.sei.co.jp, maehata@rrad.sei.co.jp Summary We propose to apply DSRC using OFDM to roadside-vehicle communication systems as a means to overcome frequency selective fading, shadowing and high-speed hand-over difficulties. We optimized the symbol rate per subcarrier of DQPSK which is a transmission parameter of the OFDM system. Our goal is to achieve BER of less than 10-5 under the conditions of transmission rate 2.048 Mbps, carrier frequency 5.8GHz, vehicle speed 180km/h, simulcasting of several roadside transmitters, maximum delay of less than 1µsec and utilize forward error correction (FEC). We developed the transmitter and receiver of the DQPSK-OFDM system in the 5.8GHz band and measured the BER performance to show that OFDM system was applicable to roadsidevehicle communication systems. Introduction For future Intelligent Transport Systems (ITS), cruise assist and so on have been developed. It requires continual communication as opposed to spot communication in vehicle road side communications. There are a number of problems to solve to meet these requirements. The most important problems are: (1) interference by delayed waves, (2) shadowing effects, and (3) a method for high-speed hand over. In Japan, ARIB takes a leading part, and the configuration of ITS and measurement of the BER performance using various modulation schemes have been studied. We measured the BER performance of DSRC using OFDM as one of the modulation schemes to solve these problems. The OFDM system has been adopted in European Digital Audio Broadcasting(DAB) and Integrated Services Digital Broadcasting (ISDV) because of its robustness against frequency selective fading caused by delayed waves on propagation paths. This modulation is a multicarrier modulation scheme, where data symbols are transmitted in parallel on multiple subcarriers. This technique narrows the subcarrier bandwidth and achieves tolerance against fading. This modulation utilizes the orthogonality relation of frequency with frequency spacing of the subcarrier and the symbol rate per subcarrier equally, and arranges the carrier by minimum frequency spacing. The spectrum efficiency is higher as a result. The guard interval reduces ISI and ICI to utilize this orthogonality relation for the delayed wave in the guard interval. In this paper, we apply the OFDM system to roadside-vehicle communication systems, and optimize the symbol rate per subcarrier and introduce a method of estimated BER performance. We also measured the BER performance with a transmitter and a receiver we developed in the 5.8GHz band. DSRC/SR-DSC We study the Dedicated Short Range Communication/Single Radio frequency-dividing Sub- Carrier (DSRC/SR-DSC) as a means to realize high speed hand-over systems. This 1
DSRC/SR-DSC is shown in fig.1. In this system, the roadside transmitter transmits a subcarrier group which represents a data zone along the road, and these groups are arranged in frequency domain. The roadside transmitters in a data zone simulcast the identical data to reduce shadowing effects. All roadside transmitters are synchronized. The receiver on the vehicle has the equivalence of several receivers on the basis of post-detection because DFT operation of the receiver can demodulate several subcarrier groups simultaneously. When a vehicle passes into a data zone, the receiver on the vehicle demodulates the signal which sums up several waves from roadside transmitters in a data zone. When a vehicle passes a boundary between data zones, the receiver demodulates the signal of the two data zones, and extracts suitable data zones (subcarrier groups) in accordance with RSSI or the number of corrected symbols by forward error correction (FEC). Since the handover in DSRC/SR-DSC is equivalent to selective diversity as stated above, high-speed handover can be achieved only by switching the subcarrier group, not changing the frequency of the local oscillator. Symbol rate per subcarrier We consider these three most important problems (frequency selective fading, time selective fading, spectrum efficiency) while optimizing the symbol rate per subcarrier of DQPSK in the OFDM system. First, to study frequency selective fading, two paths with some delay between them which induce notches by 1 / τ [Hz] spacing in the frequency domain were assumed. These notches decrease the BER performance. It is known that it is necessary to make the symbol rate per subcarrier less than 1 / ( 2πτ ) [Hz] to reduce this effect. As to time selective fading, the random FM noise caused by the receiver movement induces interference between subcarriers. It is important to widen the frequency spacing to more than a doppler spread between each subcarrier. Because the frequency spacing of the subcarriers and the symbol rate per subcarrier is equal, the symbol rate per subcarrier ( R [sps] ) must satisfy the following relation in order to reduce the effect of the above-mentioned two kinds of fading simultaneously. f d <<R << 1 2πτ (1) fd:maximum doppler frequency Next, the guard interval duration tg [sec] of the OFDM system must be longer than the delay time of the delayed wave to reduce ISI and ICI effects. t g τ (2) However, it consequently decreases the effective transmission rate compared with the case of no guard interval. The bandwidth of (t s +t g )/t s times is therefore needed to realize no degradation of the effective transmission rate on condition that t s is FFT symbol duration and equals 1/R. This leads to the deterioration of spectrum efficiency. In this paper, since it should be no less than 80%, the symbol rate per subcarrier must satisfy the following expression. R < 1 4tgg (3) The parameters of the system model are summarized in table1. 2
Table 1. Modulation DQPSK-OFDM Transmission frequency 5.8GHz Transmission rate 2.048Mbps Number of subcarriers 12 24 Frequency spacing between subcarriers 93.3kHz 44.7kHz FFT symbol duration t s 10.7 s 22.4 s t g /(t s + t g ) 3/35 3/67 Multicarrier modulation system We developed a transmitter and a receiver of the conventional DQPSK-OFDM system. In the transmitter, the FEC of the Reed-Solomon encoder is applied to input data and the output of FEC is applied to the interleave, differential encoding. Then inverse discrete fourier transform is performed. The guard interval is added and orthogonal modulation is performed. The output of this modulation is converted to 5.8GHz by an analog block, and it is transmitted from an antenna. The transmitted signals are expressed as the following. where is the differential encoded. N is the number of carriers. 1 0 (4) (5) (6) The transmitted signal is propagated through the channel h( τ ; t ), and additive white Gaussian noise n(t) is added at the analog front-end, and the orthogonal demodulation is done after frequency down-conversion. This received signal is given in the following expression. By calculating the correlation of the guard interval portion with this received signal, the correlator detects a carrier frequency error between the transmitter and the receiver and the FFT symbol timing, and the Discrete Fourier Transform (DFT) and the Auto-Frequency Control(AFC) of the analog blocks work on the basis of the calculation result. Reed-Solomon decoding is performed on the received signal y(t), DFT, differential decoding, and deinterleaving. We next introduce an estimation method of BER performance. This estimation method requires the probability density function pdf (SINR, power) of the instantaneous power and the averaged signal to interference plus noise ratio (SINR) considering the interference between subcarriers, and calculates the BER performance from eq. (8). (8) We next derive SINR to estimate the BER performance where the channel fluctuation cannot be neglected and propagation delay can. Under these assumptions, the propagation path is expressed in the following equation. (7) 3
where g(t) is a complex Gaussian random process with a zero mean and has an autocorrelation function with the following equation. Substituting eq. (4), (9) with (7), and performing Fourier Transform, the following equation is provided. (9) (10) (11) The next equation is provided from the correlation of z im and z (i-1)m with a ij being equal to a (i-1)j. Interference between subcarriers is given with the equation.[1] (12) Assuming the directivity of the receiving antenna is isotropic, the mean received power s, and autocorrelation function R ( τ ) if ( fd τ << 1 ) satisfies the following equation. Substituting eq. (14) with eq. (13), and simplifying the equation. It can be rewritten as eq.(16) when we utilize the relationship that the FFT symbol duration and the symbol rate per subcarrier is equal. (16) This equation represents the interference between subcarriers shown by the ratio of the frequency spacing of the subcarriers to a doppler spread. This relation has been described considering the time selective fading. This interference increases with the number of subcarriers, but is almost saturated with 10 or more subcarriers. Considering this, the interference of more than 10 subcarriers is expressed as: (17) SINR is given from eq. (17) as follows. Optimizing the symbol rate per subcarrier In this section, we optimized the symbol rate per subcarrier by the BER performances of the suggested computation model and actual measurement results of the developed DQPSK- (18) (13) (14) (15) 4
OFDM system. In simulation, we used the rayleigh fading channel, and substituting its probability density function of power[2] into eq. (8), and eq. (19) is provided. We first focus on the BER that is irreducible when SINR is large. the Irreducible Bit Error Rate (IBER) is only dependent on fdts because the SINR has a limiting value from eq.(18) as the SNR becomes large. We can calculate the IBER by substituting this limiting value into eq. (19), it was found that the IBER becomes 10-2 when fdts is equal to 0.025. Using an appropriate error correction code, we can improve the BER performance from 10-2 to 10-5 sufficiently. So, we decided on a limiting value of 10-2 as the IBER.We found that the allowable fdts is less than 0.025, and confirmed that the actual measurement led to the same result as well. We next investigated the effect of frequency selective fading under two-path rayleigh channels on the assumption that two waves are transmitted by two roadside transmitters. In this case, the IBER performance is dependent on the level difference D/U and delay time τ between the two waves. We measured the IBER by using the combination of various D/U, τ / ts.[3] Measurement results revealed that the IBER became worse when D/U was smallerand τ / ts was larger. So in assuming the worst case that D/U = 0[dB], we measured IBER vs τ / ts in detail, and found that IBER was 10-2 when τ/ts was 0.06. We found that the allowable τ / ts was less than 0.06. In summary, we found that fdts < 0.025 and τ / ts < 0.06 reduce the effect of time selective fading and frequency selective fading simultaneously. The next expression was concluded when we substituted fd and τ into transmission rate 2.048 Mbps, carrier frequency 5.8GHz, vehicle speed 180km/h, simulcasting, maximum delay < 1µsec 17 sec t s 26 sec (20) when we utilized the relationship between the FFT symbol duration and the symbol rate per subcarrier, the symbol rate per subcarrier satisfying the following inequality. 38ksps R 60ksps (21) The expression satisfied the spectrum efficiency of eq. (3) simultaneously. BER performance In a data zone The configuration of simulcasting is shown in fig. 2. In this figure, two transmitters are synchronized by sending the local frequency to PLL of RF block and transmitting data which is PN9 through optical fiber. In the transmitter, the FEC of the Reed-Solomon encoder is applied to input data and the output of FEC is applied to the interleaving, differential encoding. Then inverse discrete fourier transform is performed. The guard interval is added and orthogonal modulation is conducted. The output of this modulation is up-converted to 5.8GHz by an analog block, and it is transmitted from an antenna. In a receiver, the orthogonal demodulation is done after frequency down-conversion at the analog front-end, and DFT, differential decoding, de-interleaving, Reed-Solomon decoding are performed. We next assumed a data zone composed of only two roadside transmitters nearest to the vehicle. Figure 3-(a) indicates that the simulcasting improves the profile of the received power under shadowing. Figure 3-(b), (c) measured the BER performances with FEC under shadowing by a truck in the single-transmitter case and simulcasting case under the condition that the vehicle passes slowly at 20km/h to acquire the detailed received power. (19) 5
From fig.3 (b), (c), we confirmed the improved BER performance under shadowing around 50[m] from antenna 1. We next measure the BER performance in a data zone using a fading simulator to simulate the case where the vehicle passes at 50km/h, 100km/h. Figure 4-(a) shows the channel model of fig. 2, and figure 4-(b) indicates the delay time, the level difference and the doppler spread which considers the elevation angle of an observed vehicle when using the channel model of fig.4-(a). Fig.4-(c) shows the BER performance under two-path rayleigh channels. From this graph, we found that the BER performance became almost 10-3 before FEC, independently of vehicle speed. When using the error correcting code RS(63,43), it was confirmed that the BER performance improves from 10-3 to less than 10-8 at 100km/h and from 10-3 to less than 10-5 at 50km/h. In Fig.4-(c) we found that BER performance at 50km/h is inferior to that at 100km/h using FEC. When a vehicle moves slowly, the average fade time duration of receiving power lengthens, causing a long burst error. So, this requires a larger size interleave to spread these burst errors. In this experiment we use an interleave that has a fixed size not considering vehicle speed. The size of the interleave is therefore considered insufficient to spread the burst errors at 50km/h and FEC cannot be utilized. From this result we believe that further improve the BER performance at low speed is needed. Conclusions We developed the transmitter and receiver of the DQPSK-OFDM system in the 5.8GHz band, and measured the BER performance in both indoor and outdoor environments. As a result, we confirmed the effect of simulcasting against shadowing, and showed that the OFDM system was applicable to roadside-vehicle communications systems. Acknowledgements The authors also would like to thank the members of the road-vehicle communications technology group in the Study Group for Efficient Use of the Radio Spectrum for Vehicle Communication Systems for their earnest discussions and for installing the measuring system. The authors also wish to express their sincere appreciation to our Corporate Adviser Mr. Nobuo Yumoto for his valuable advice. Reference [1] Okada, M., Hara, S.and Morinaga, N.,"Bit Error Rate Performances of Orthogonal Multicarrier Modulation Radio Transmission Systems," IEICE Trans. Commun., Vol. E76-B, No2, pp.113-119, Feb. 1993. [2] Inoue, T., Karasawa, Y.,"Theoretical Analysis on Level Variation of Wideband Signals under Multipath Fading Environments," IEICE Technical Report, AP98-40,Aug. 1998. [3] Maehata, T., Masaharu Imai.,"DSRC using OFDM for roadside-vehicle communication system" IEEE, 51st VTC proceedings,may.2000. 6
Figure1. System model of DSRC/SR-DSC Figure 2. Configuration of simulcasting 3 - (a). Profile of received power BER BER 3-(b) single-transmitter case 3-(c) double-transmitter case Figure 3. Effect of simulcasting 7
Doppler frequency [Hz] level difference [db] delay time [µsec] 4-(b). Doppler frequencies, level difference and the delay time between the two paths transmitted from ANT1, ANT2. 4-(c) BER performance Figure 4. BER performance in a data zone using a fading simulator to simulate a vehicle passing at 50km/h, 100km/h 8