Research Article  Open Access
Shingo Yoshizawa, Hiroshi Tanimoto, Takashi Saito, "Data Selective Rake Reception for Underwater Acoustic Communication in Strong Multipath Interference", Journal of Electrical and Computer Engineering, vol. 2017, Article ID 5793507, 9 pages, 2017. https://doi.org/10.1155/2017/5793507
Data Selective Rake Reception for Underwater Acoustic Communication in Strong Multipath Interference
Abstract
In underwater acoustic communication (UAC), very long delay waves are caused by reflection from water surfaces and bottoms and obstacles. Their waves interfere with desired waves and induce strong multipath interference. Use of a guard interval (GI) is effective for channel compensation in OFDM. However, a GI tends to be long in shallowwater environment because a guard time is determined by a delay time of multipath. A long GI produces a very long OFDM frame in several seconds, which is disadvantageous to a response speed of communication. This paper presents a method of keeping good communication performance even for a short GI. We discuss influence of intercarrier interference (ICI) in OFDM demodulation and propose a method of data selective rake reception (DSRake). The effectiveness of the proposed method is discussed by received signal distribution and confirmed by simulation results.
1. Introduction
Remotely operated underwater vehicle (ROV) and autonomous underwater vehicle (AUV) are widely used in current marine surveys [1, 2]. Wireless communication is an important underlying technology in remote control and information gathering for ROV and AUV. Since light and electromagnetic waves have large attenuation in seawater, use of sound waves is suitable for long range communication. Underwater acoustic communication (UAC) has been studied for a long time as well as radio communication. For instance, a communication unit of singlesideband amplitude modulation (SSBAM) was developed in the 1950s. Digital modulation schemes of spread spectrum [3, 4], OFDM [5–7], and MIMO [8, 9] have been studied in recent studies.
Demodulation is affected by multipath interference and Doppler in UAC, which degrade communication performance. Doppler compensation has been discussed in [10–13]. We focus on the problem of multipath interference in this paper. Very long delay waves are caused by reflection from water surfaces and bottoms and obstacles. Their waves interfere with desired waves and induce strong multipath interference. For mitigation of multipath interference, OFDM with a guard interval (GI) (also named as a cyclic prefix (CP)) is adopted. As far as a delay time of multipath is less than a guard time, influence of delay waves can be expressed by channel coefficients for every frequency bin. These channel coefficients can be estimated and equalized by frequency domain equalization (FDE). Effectiveness of OFDM using a GI has been verified by sea trials in [5–7].
The drawback of using a GI is decrease of communication efficiency because a GI itself is redundant. In shallowwater environment, a long GI is required when a guard time is determined by a delay time of multipath. The delay time ranges from several milliseconds to 100 milliseconds in underwater acoustic propagation, being dependent on surrounding environments. In the sea trial presented by Berger et al. [7], the GI and FFT length were set to 48 ms and 491 ms. OFDM frame duration runs up to 5.4 seconds, which would be undesirable in terms of a response speed of communication.
This paper presents a method of keeping good communication performance even for a short GI. Strong multipath interference is assumed in our study, where arrival times of large delay waves exceed a guard time. First, we discuss the influence of interblock interference (IBI) and intercarrier interference (ICI) in received signal distribution. Although IBI always interferes with demodulation, ICI can be suppressed by taking an appropriate FFT window timing. Next, we propose a new idea of data selective rake reception (DSRake) according to the above discussion. DSRake takes multiple fingers by changing FFT window timing for every OFDM block. The best finger with the least ICI is selected by checking data errors for all fingers. With regard to QPSK modulation, the mitigation of ICI has an impact on avoiding error floor in BER performance. This paper discusses OFDM as communication scheme. As for single carrier frequency domain equalization (SCFDE), we briefly report it in [14].
This paper is organized as follows. Section 2 discusses the influences of IBI and ICI by received signal distribution. Section 3 proposes DSRake for the mitigation of ICI. Section 4 reports simulation results evaluating DSRake in strong multipath interference. Section 5 summarizes our work.
2. Received Signal Distribution
2.1. OFDM Model
We discuss the influences of IBI and ICI by received signal distribution. Theoretical symbol error rates (SERs) of PSK and QAM can be obtained by probability density function (PDF) when we observe received signal amplitudes in noisy propagation channels. We use a basic OFDM model illustrated in Figure 1. In the transmitter side, all transmitted data are set to zero, given by (). denotes a block number for OFDM blocks. is an subcarrier index for OFDM subcarriers. A transmitted symbol becomes after BPSK modulation. The transmitted symbol is converted into 1 or by multiplying random patterns of in the scramble block, which becomes . A timedomain signal block is given by after IFFT operation, where is a discrete sample time. A transmitted signal is expressed by after GI insertion and parallel to serial conversion. We presuppose that this GI is given by a cyclic prefix.
In the receiver side, a received signal block of is obtained by cutting out a received signal of by a FFT window having a rectangular shape. A frequency domain signal block is given by after FFT operation. is obtained by multiplying the random patterns of used in transmitter side. Received data of are obtained after BPSK demodulation. We set lengths of a data block, GI, and OFDM block to , , and .
We use a twopath channel model consisting of direct and delay waves. A relation between transmitted and received signals is expressed aswhere is a propagation channel coefficient () for the delay wave and is an arrival time difference between direct and delay waves. denotes noise signal component determined by a metric of the carrier to noise ratio (CNR).
Figure 2 shows an OFDM frame structure and timing positions for FFT windowing. This figure shows the case of receiving only a direct wave. When timing synchronization is perfect, their positions are the same of those of data blocks, not overlapping with GIs. The block boundary is emphasized between OFDM blocks.
The received signal distribution for a 30dB CNR is shown in Figure 3. We set a data block length and a guard time to and , respectively. The signal distribution for the received BPSK symbols of is plotted. The total number of received BPSK symbols is . In BPSK demodulation, a symbol error occurs when has a negative value. All the signals in Figure 3 locate around 1, which indicates the errorfree demodulation of .
2.2. Influence of Interblock Interference (IBI)
Let us consider the influence of IBI as a long delay wave overlaps with a direct wave. Figure 4 shows the relations between direct and delay waves where their arrival time differences of and . The propagation channel coefficient is set to for a delay wave. IBI happens due to the collision of different data blocks for direct and delay waves.
(a) , , 30dB CNR
(b) , , 30dB CNR
The received signal distributions for Figure 4(a) are shown in Figure 5. In Figure 4(a), ()th block of the delay wave exactly overlaps with th block of the direct wave in the FFT window period. The received symbol of can be introduced from the following equations, omitting the noise component of .Since and are random patterns consisting of 1 or , (6) gives . This signal distribution can be observed in Figure 5(a). Although the signal values of do not concentrate on 1, all of them are positive. A symbol error does not occur in Figure 4(a).
(a) , , 30dB CNR
(b) , , 30dB CNR
In Figure 4(b), ()th block of the delay wave is slightly deviated from th block of the direct wave. can be introduced bywhere we apply from circular shift property. The signal values of range from 0.3 to 1.7 as shown in Figure 5(b). This case also does not induce a symbol error.
The IBI does not take a symbol error as long as a high CNR condition is kept as for this observation. The same phenomenon would be observed even in QPSK transmission. Improvement of SNR using antenna arrays is practical rather than keeping a high CNR, where Zheng presented MRC diversity in SIMOOFDM as a measure against insufficient guard interval in [15].
2.3. Influence of Intercarrier Interference (ICI)
Let us consider the influence of ICI by giving another arrival time difference of . The relation between direct and delay waves is shown in Figure 6. Different from Figure 4, ()th data block and th GI of the delay wave overlap with th data block of the direct wave. This signal distribution is shown in Figure 7. Since some of have a negative value, a symbol error occurs.
We introduce as well as Section 2.2. First, the received signal of is given byWe decompose a received signal of the delay wave into and as shown in Figure 6. Their functions are given bywhere we apply = and from circular shift property. can be replaced with . is given by can be expressed asThe received signal distribution of (15) would be almost the same as that of (10) if and are excluded. and can be expressed by using inverse discrete Fourier transform (IDFT) and DFT asThe interferences of (16) and (17) are added for every subcarrier, which corresponds to ICI. Assuming that the average amplitude for the OFDM transmit signals after IDFT is (i.e., calculation within the square bracket in (16)), the average of deviations caused by and is roughly calculated asThese deviations would be observed by comparing the received signal distributions in Figures 5 and 7. The difference between Figures 4 and 6 is whether a block boundary is included within a FFT window.
2.4. Adjustment of FFT Window
The ICI can be avoided by changing FFT window timings, whose adjustment is illustrated in Figure 8. The time positions of FFT windows have been shifted by 40 samples ahead. The block boundaries for the delay wave are not included for their FFT windows. Although this adjustment induces a phase rotation after FFT operation in frequency domain, the phase rotation can be detected and compensated by FDE. The received signal distribution after the FFT window adjustment is shown in Figure 9, where the phase rotation can be compensated before descramble. This distribution looks like Figure 5(b) owing to the ICI avoidance.
3. Data Selective Rake Reception (DSRake)
The ICI avoidance is achieved when the arrival time of delay wave is perfectly known. Note that arrival times of individual delay waves are almost unknown in the actual environment. We introduce an OFDM rake reception as an alternative method, whose scheme is shown in Figure 10. Since the arrival times ( and ) and magnitude ( and ) of delay waves are unknown, we take multiple FFT window timings for OFDM demodulation, that is, rake fingers.
Original rake reception itself is used as path diversity in spread spectrum [16]. In general, OFDM and rake reception for path diversity are not compatible. Received symbols in rake fingers have high correlation with each other as far as multipath delay time is less than a guard time. The improvement of received SNR is little considering increase of computational complexity in demodulation. We use the rake reception to find the best rake finger that is not affected by ICI so much. It does not aim at path diversity. The selection of rake fingers is achieved by checking data errors after demodulation, where the proposed scheme of data selective rake reception (DSRake) is shown in Figure 11. In the transmitter side, cyclic redundancy check (CRC) codes are inserted in binary data before forward error correcting (FEC) coding. In the receiver side, multiple OFDM demodulators accept received signals in rake fingers and output decoded data blocks. The best data block having no error is selected as final data by observing the CRC results in the data selection unit. If all fingers have data errors, the final data are generated by merging all decoded data in bit level.
(a) Transmitter
(b) Receiver
DSRake would not be adopted in general OFDM systems such as IEEE802 WLANs and LTE in RF communication due to considerable increase in computational complexity. Note that the bandwidth of UAC is much narrower than that of RF. The increase of computational complexity for UAC does not become a problem from the viewpoint of implementation in RF. The overhead of CRC is trivial because its length is enough for 16 bits (CRC16) in typical usage.
DSRake belongs to selection combining (SC) in diversity combining. Maximal ratio combining (MRC) should be discussed as another method. The alternative scheme of MRC rake reception (MRCRake) is shown in Figure 12. The received symbols in rake fingers are synthesized after OFDM demodulation. Generally, a diversity gain of MRC is higher than that of SC. However, MRCRake is inferior to DSRake in terms of the mitigation of ICI. The synthesis of rake fingers takes in undesirable received symbols affected by ICI and the effect is limited. The superiority of DSRake will be confirmed by our simulation in the next section.
(a) Transmitter
(b) Receiver
4. Simulation
4.1. Channel Model
As an example of underwater acoustic propagation, we use two channel models measured in a swimming pool. The delay profiles were measured on the condition of horizontal link where one transmitter and four receiver hydrophones horizontally face each other. The location of hydrophones is drawn in Figure 13. The pool length and width are 25 m and 13 m and the water depth is 1.2 m. The distances between transmitter and receiver hydrophones are 8 m and 20 m. The space of four hydrophones is 5 cm.
The delay profiles for 8 m and 20 m distances are shown in Figures 14 and 15. A direct wave is located at 0 on the time axis and has normalized magnitude of 0 dB. Delay waves are expressed by individual values of relative magnitude and delay time. Several clusters of delay waves are periodically observed around 30 to 35 ms, 65 to 70 ms, and 97 to 102 ms in Figure 15. These clusters come from several round trip reflections at the side walls. The delay waves of more than −10 dB (i.e., less than 10 dB in desired to undesired signal ratio (DUR)) range from 0 seconds to 35 ms. Since we set a guard time to 12.8 ms in our simulation, the delay waves beyond the GI induce IBI and ICI. If a guard time is more than 110 ms (i.e., more than 20 dB DUR), the influences of IBI and ICI would be small. However, we must keep in mind that a long GI is undesirable in terms of a response speed of communication.
Summary of the delay profiles is reported in Table 1. The 20 m distance shows larger values in average delay time and RMS delay spread than the 8 m distance. The results of average delay time and RMS delay spread are different among receiver channels to some extent. The signal correlation among received antennas would not be very high as having different propagations. Space diversity using antenna arrays is effective to improve a received SNR in this case. RMS delay spread is helpful in the determination of a GI length as long as the magnitude of delay waves is exponentially decaying. However, the magnitude of delay waves does not always fade as time goes on as shown in Figures 14 and 15. Even though the RMS delay spread is less than the GI length, the strong interference of delay waves should be considered.
(a) 8m distance  
 
(b) 20m distance  

4.2. Simulation Parameters
The simulation parameters are enumerated in Table 2. The baseband OFDM signals with a frequency band of kHz to kHz are modulated by a carrier wave of 50 kHz. Onetap frequency domain linear equalization based on MMSE criterion is used in channel equalization. The GI length is set to 12.8 ms, corresponding to 256 samples in baseband domain. Two training data blocks are added to the beginning of an OFDM frame, where the frame format is shown in Figure 16. The two long training fields (LTFs) are used for channel estimation. Since the LTFs are located at the head of frame, they do not have the influence of IBI and ICI. The number of rake fingers is set to 64 for DSRake and MRCRake. We have used convolutional coding with a coding rate of 1/2. The transmit data rate is about 13.3 kbps considering the overhead of LTFs and GIs. Although the overhead of CRC codes (CRC16) might be counted for DSRake, this overhead is very small (less than 2%).

We apply space diversity using array antennas for the mitigation of IBI. The scheme of OFDM space diversity is shown in Figure 17. Space diversity combining based on MRC is performed after channel equalization. Space diversity combining and OFDM rake reception of DSRake or MRCRake are compatible. The diversity block is inserted into the OFDM demodulation units in Figures 11 and 12.
4.3. Simulation Results
Bit error rates (BERs) for the 8 m and 20 m distances are plotted in Figures 18 and 19. We have evaluated the schemes of single channel reception (average of four channels), space diversity, DSRake, and MRCRake. Both DSRake and MRCRake are given by the combination of space diversity and rake reception. The single channel reception has the BER floor of due to strong multipath interference. The space diversity decreases the BER floor from to as shown in both figures. The influence of IBI would be decreased by space diversity combining to some extent. DSRake and MRCRake show further improvement of decreasing BER floor. DSRake is clearly superior to MRCRake from the BER results. The ICI mitigation contributes to the improvement of communication quality rather than taking path diversity. DSRake can eliminate a BER floor for the 8 m distance and decrease by up to for the 20 m distance. The effectiveness of DSRake in strong multipath interference has been observed from this simulation.
5. Conclusion
This paper presents a new method of OFDM rake reception in strong multipath interference. Very long delay waves beyond GI induce IBI and ICI. The influence of IBI and ICI is discussed by received signal distribution. Regarding ICI, we reported that the ICI avoidance can be achieved by changing FFT window timing. According to the idea of ICI avoidance, we have proposed DSRake as one of rake reception techniques. Original rake reception is used for obtaining path diversity. However, our rake reception aims at the mitigation of ICI. We have explained that selection combining by DSRake is superior to maximal ratio combining by MRCRake. The effectiveness of DSRkae has been confirmed by the simulation results based on actual underwater propagation models. In our future work, we will investigate communication performance of DSRake when Doppler effect is added.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
The authors would like to thank the staff of Kitami City Board of Education. This work was supported by JSPS KAKENHI Grants nos. 16K18099 and 15K06048.
References
 L. E. Freitag and J. A. Catipovic, “A signal processing system for underwater acoustic ROV communication,” in Proceedings of the 6th International Symposium on Unmanned Untethered Submersible Technology Technology, pp. 34–41, June 1989. View at: Google Scholar
 J. Borden and J. Dearruda, “Long range acoustic underwater communication with a compact AUV,” in IEEE OCEANS, October 2012. View at: Publisher Site  Google Scholar
 G. Loubet, V. Capellano, and R. Filipiak, “Underwater spreadspectrum communications,” in Proceedings of the MTS/IEEE Conference OCEANS, pp. 574–579, October 1997. View at: Google Scholar
 K. G. Kebkal and R. Bannasch, “Implementation of a sweepspread function for communication over underwater acoustic channels,” in Proceedings of the MTS/IEEE Conference and Exhibition OCEANS, vol. 3, pp. 1829–1837, October 2000. View at: Google Scholar
 C. M. Anil, Underwater acoustic communications in warm shallow water channels, Thesis, Doctor of Philosophy, National University of Singapore, 2006.
 F. Frassati, C. Lafon, P. Laurent, and J. Passerieux, “Experimental assessment of OFDM and DSSS modulations for use in littoral waters underwater acoustic communications,” in Proceeding of the Europe Oceans, vol. 2, pp. 826–831, June 2005. View at: Publisher Site  Google Scholar
 C. R. Berger, J. Gomes, and J. M. F. Moura, “Seatrial results for cyclicprefix OFDM with long symbol duration,” in IEEE OCEANS, June 2011. View at: Publisher Site  Google Scholar
 S. Roy, T. M. Duman, V. McDonald, and J. G. Proakis, “Highrate communication for underwater acoustic channels using multiple transmitters and space—time coding: receiver structures and experimental results,” IEEE Journal of Oceanic Engineering, vol. 32, no. 3, pp. 663–688, 2007. View at: Publisher Site  Google Scholar
 P. Bouvet and A. Loussert, “An analysis of MIMO–OFDM for shallow water acoustic communications,” in Proceedings of the IEEE OCEANS, pp. 1–5, September 2011. View at: Publisher Site  Google Scholar
 B. S. Sharif, J. Neasham, O. R. Hinton, and A. E. Adams, “A computationally efficient doppler compensation system for underwater acoustic communications,” IEEE Journal of Oceanic Engineering, vol. 25, no. 1, pp. 52–61, 2000. View at: Publisher Site  Google Scholar
 B. Li, S. Zhou, M. Stojanovic, L. L. Freitag, and P. Willett, “Multicarrier communication over underwater acoustic channels with nonuniform Doppler shifts,” IEEE Journal of Oceanic Engineering, vol. 33, no. 2, pp. 198–209, 2008. View at: Publisher Site  Google Scholar
 C.H. Hwang, K.M. Kim, S.Y. Chun, and S.K. Lee, “Doppler estimation based on frequency average and remodulation for underwater acoustic communication,” International Journal of Distributed Sensor Networks, vol. 2015, Article ID 746919, 8 pages, 2015. View at: Publisher Site  Google Scholar
 A. E. Abdelkareem, B. S. Sharif, C. C. Tsimenidis, and J. A. Neasham, “Compensation of linear multiscale doppler for OFDMbased underwater acoustic communication systems,” Journal of Electrical and Computer Engineering, vol. 2012, 16 pages, 2012. View at: Publisher Site  Google Scholar
 S. Yoshizawa, H. Tanimoto, and T. Saito, “SCFDE vs OFDM: performance comparison in shallowsea underwater acoustic communication,” in Proceedings of the IEEE International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS '16), p. 5, October 2016. View at: Publisher Site  Google Scholar
 L. T. Phuc, Y. Zheng, and Y. Karasawa, “A simplified propagation channel model for evaluating MRC diversity characteristics in SIMO OFDM with insufficient guard interval,” in Proceedings of the United States National Committee of URSI National Radio Science Meeting (USNCURSI NRSM '16), usa, January 2016. View at: Publisher Site  Google Scholar
 J.H. Son, E.H. Jeon, K.M. Kim, D.W. Lee, and T.D. Park, “Alternative approach for combination of fingers in underwater acoustic communication,” International Journal of Distributed Sensor Networks, vol. 2016, 16 pages, 2016. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2017 Shingo Yoshizawa et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.