CI/OFDM Underwater Acoustic Communication System

The underwater acoustic channel (UAC) is one of the most challenging environments to be encountered for the communication. Because of the absorption of the signal, the path loss depends on the signal frequency (Berkhovskikh & Lysanov, 2003; Jensen et al., 2011). Multipath transmission causes intersymbol interference (ISI), and it extends over tens to hundreds of milliseconds according to the communication distance (Stojanovic & Preisig, 2009). Since the velocity of sound in water is about 1500m/s, any relative motion includes the transmitter or receiver and even surface waves will cause non-negligible Doppler effects, including shifting and spreading. All these phenomena dramatically limit the data rate achievable and the performance of the communications. The bandwidth is very limited, and the system is actually a broadband communication system because the center frequency of the signal is always at the same order of the bandwidth (Stojanovic, 1996; Stojanovic, 2007; Stojanovic & Preisig, 2009).


Introduction
The underwater acoustic channel (UAC) is one of the most challenging environments to be encountered for the communication. Because of the absorption of the signal, the path loss depends on the signal frequency (Berkhovskikh & Lysanov, 2003;Jensen et al., 2011). Multipath transmission causes intersymbol interference (ISI), and it extends over tens to hundreds of milliseconds according to the communication distance (Stojanovic & Preisig, 2009). Since the velocity of sound in water is about 1500m/s, any relative motion includes the transmitter or receiver and even surface waves will cause non-negligible Doppler effects, including shifting and spreading. All these phenomena dramatically limit the data rate achievable and the performance of the communications. The bandwidth is very limited, and the system is actually a broadband communication system because the center frequency of the signal is always at the same order of the bandwidth (Stojanovic, 1996;Stojanovic, 2007;Stojanovic & Preisig, 2009).
In this chapter, we introduce a new multi-carrier modulation into the UAC channels which is called Carrier Interferometry OFDM (CI/OFDM) (Nassar et al., 1999;Nassar et al., 2002aNassar et al., , 2002b. Compared with OFDM, the CI/OFDM has a low PAPR characteristic and inherent frequency selective combining, which makes it a very attractive signaling scheme in frequency selective fading channels Wiegandt & Nassar, 2003;Wiegandt et al., 2004 ).
The chapter is organized as follows. In Section II, the characteristics of CI signal are analyzed. Two algorithms are proposed in Section III. Details are focused on the PAPR performance, and new algorithms to complete the modulation and demodulation of the www.intechopen.com CI/OFDM. The configuration of the CI/OFDM underwater acoustic communication system is presented in Section IV. Furthermore, the key algorithms including synchronization, channel estimation and equalization are described. In Section V, Performance results for different field tests are summarized. Conclusions are drawn in Section VI.

CI/OFDM signals 2.1 The theory of the CI/OFDM
In CI/OFDM transmitter, after serial to parallel transform, information symbols are modulated onto all the N parallel subcarriers and then added linearly together to get the output signal . As shown in Fig.1    . Assume perfect synchronization, the received signal is first projected onto the N orthogonal carriers, multi-carrier demodulation and phase offsets remove are carried out after that. This leads to the decision vector The first part of (6) is the desired information symbol which is randomly faded by factor i  , and the second part is the interferences of the other N-1 information symbols which modulated on the same carrier.
Different combining strategies are employed to help restore orthogonality between subcarriers. In AWGN channel, the optimal combining is equal gain combining (EGC). After performing 1 0 N i i Cr     , interferences are close to zero. While in frequency selective channel, different combining strategies are used to get combining gains, for example, the maximum ratio combining (MRC), the minimum mean square error combining (MMSEC) (Itagkai & Adachi, 2004). After combining, the signals are sent to the detector.

www.intechopen.com
As presented to date, the implementation of original CI/OFDM is complicated, and it is important to note that the receiver is designed for detecting only one information symbol. Although CI/OFDM had been proved that it could improve BER performance by exploiting frequency diversity and depress the PAPR simultaneously, its implementation was complicated and only conceptual transmitter and receiver models had been given in the literature. Fig. 2. The CI/OFDM receiver for the kth symbol

The characteristics of the CI signal
The baseband CI signal  is given bellow where N is the number of the subcarriers and f  is the interval of the subcarriers. It is obviously that CI signal is a periodic signal. Simulation results are shown in Fig. 3 We rewrite the discrete CI signal as bellow It is clearly that the discrete CI signal is the result of sampling a rectangle pulse with the sampling rate f  in the frequency domain. It has constant amplitude (CA).
Based on the analysis of the CI signal, two novel algorithms for CI/OFDM modulation and demodulation are presented in this chapter. www.intechopen.com It is obviously that (9) www.intechopen.com Since i is the index of subcarrier, the columns of the matrix are corresponding to different subcarriers, that is subcarrier subcarrier subcarrier subcarrier (11) implies that the coefficient of the IDFT can also be corresponding to another IDFT (Xu et al., 2007a(Xu et al., , 2007b Fig.4 shows a block diagram of the proposed system. At the transmitter side, the input data is first mapped into a baseband constellation. Then the data sequence is converted to parallel and enters the first IDFT to perform CI spreading. After that, the second IDFT is used to implement orthogonal multi-carrier modulation. Parallel data is first transformed to serial data, then the complex base-band signal is then up-converted to the transmission frequency, and the real part of the signal is sent out to the channel. In the receiver, the signal is first down-converted to the base-band. Serial to parallel transformation followed by orthogonal multi-carrier demodulated which completed by the first discrete Fourier transform (DFT). Then CI code de-spreading is implemented by the second DFT and finally, the phase constellation of the data is extracted ).

Single-carrier algorithm
The CI spread code in (10) is similar to the polyphase codes (Heimiller, 1961). Polyphase codes were proven to have good periodic correlation properties, the sequence is , 01 kp  is a primitive Pth root of unity, the sequence has zero periodic correlation except for the peaks at 22 0, ,2 ,... ip p  .
Polyphase code were proven to be a constant amplitude , zero autocorrelation (CAZAC) sequence (Heimiller, 1961). According to the characteristics of the CAZAC sequence, if i u is a CAZAC sequence, then i u , where u denotes complex conjugation, is also a CAZAC sequence (Milewski, 1983). Note that the orthogonality, periodicity, constant amplitude and zero autocorrelation are not changed, it suggests a new way of thinking about constructing new CI signals.
In this chapter, the new CI signals are complex conjugations of primary CI signals , which can be written as ( 1 ) Then, the CI/OFDM signal is mathematically characterized by the following equation Fig.5 shows the proposed system using the single-carrier algorithm. In the transmitter, the input data is mapping into a baseband constellation. Then the data sequence is converted to parallel and enters the first DFT to perform CI spreading. After that, the IDFT is used to implement orthogonal multi-carrier modulation. The complex baseband signal is then upconverted to the transmission frequency. In the receiver, the signal is down-converted to the baseband. After serial to parallel transformation, the signal is first demodulated by DFT, Then CI de-spreading is implemented by the IDFT and the phase constellation of the data is extracted. Of course, the two algorithms have their own problems. As in the multi-carrier method, the physical concept is not very clear, since there are two cascaded IDFT in the transmitter which may cause confusion about the transformation between the frequency domain and time domain. On the other hand, in the single-carrier method, filter should be well designed to compress the bandwidth of the output signal.
As shown in fig.6, the performance of CI/OFDM system is verified under AWGN channel. We replace the IDFT by the inverse fast Fourier transform (IFFT) due to the efficiency of the algorithm. It is obviously that there is no difference between the conceptual CI/OFDM and the two algorithms proposed in this chapter.

www.intechopen.com
Since lower PAPR is the most important characteristic of CI/OFDM, and the algorithms presented here are somewhat different from the theoretical realization of the CI/OFDM, it is reasonable for us to verify the PAPR performance based on these two algorithms. Fig. 7 shows the simulation result. A conclusion can be drawn that the two algorithms presented in this chapter have the same PAPR and BER performance as the conceptual CI/OFDM, and lower complexity which make it applicable to engineering.

Configuration of the CI/OFDM underwater acoustic communication system
Based on the aforementioned algorithms, two CI/OFDM underwater acoustic communication systems are proposed. Simplified block diagrams of the proposed systems are shown in Fig.8 and Fig.9. We also replace the IDFT by the IFFT due to the efficiency of the algorithm.

www.intechopen.com
As shown in Fig.8, based on the multi-carrier algorithm, the input data is first coded by Low Density Parity Check Codes (LDPC). After baseband mapping, the data sequence is converted to parallel and enters the first IFFT to perform CI spreading. Pilot signals are inserted before the second IFFT. The second IFFT is used to implement orthogonal multicarrier modulation. A cyclic prefix and postfix are also appended to the data sequence as guard intervals in order to combat the ISI induced by the multi-path delay spread in the UWA channel. The complex base-band signal is then up-converted to the transmission frequency and the real part of the signal is sent out to the UWA channel by the transducer.
In the receiver, the signal is first down-converted to the base-band. Then the signal is demodulated by the first FFT. Channel estimation is performed to track the channel response and compensations of the signal are performed. Then CI code de-spreading is implemented by the second FFT, and finally, the phase constellation of the data is extracted. Fig. 8. The block diagram of the system based on multi-carrier algorithm Fig.9 shows a simplified block diagram of the proposed system based on single-carrier algorithm. In the transmitter, the input data is first encoded by LDPC and then mapping into a baseband constellation. The data sequence is converted to parallel and enters the first FFT to perform CI spreading. After that, the IFFT is used to implement orthogonal multicarrier modulation. A cyclic prefix is also appended as a guard interval to the data sequence in order to combat the inter ISI induced by multipath delay spread in the selective fading channel. In addition, a pilot signal is appended for the purposes of channel estimation in the receiver. The complex baseband signal is then up-converted to the transmission frequency. In the receiver, the signal is down-converted to baseband. The signal is first demodulated by FFT, and diversity combining scheme is employed as frequency-domain equalization where the combining weights are estimated by the pilot signal. Then de-spreading is implemented by the IFFT and the phase constellation of the data is extracted. Finally, the data is mapped back to the binary form, and a soft LDPC decoding is performed.
We here focus on the multi-carrier algorithm and explain the key technologies used in the underwater acoustic communication system.

Synchronization
We use Linear Frequency Modulation (LFM) (Rihaczek, 1969;Shaw & Srivastava, 2007) signal to get coarse synchronization and CI complex spreading sequence to get accurate synchronization and fractional frequency offset estimation. Fig. 9. The block diagram of the system based on single-carrier algorithm

Coarse synchronization
The expression for LFM signal is given as where A is the amplitude of the signal, T is the width of the signal, 0 f is the carrier frequency, u is the gradient of the instantaneous frequency which is called chirp rate, 2 uB T   . () rect  is a rectangle function, defined as Since the ambiguity function of LFM signals is wide in Doppler axis (Rihaczek, 1969), it is highly tolerant of the Doppler shift which makes it useful in mobile wireless communication systems.

Fine synchronization and fractional frequency shift estimation
Two identical CI complex sequences are used as fine synchronization signals and sliding correlator is applied at the receiver side to obtain the correlation peak. CI complex sequences is given by  (Hlaing et al. 2003;Ren, 2005), the cross-correlation function of the two sequences is where the L is the length of the sequence. Since the sequences are identical at the transmitter, the impact of the channel is assumed to be same to the two sequences, (18) where f  is the fractional frequency shift,  is the phase offset caused by frequency shift, T is the period of the synchronization signal which is equals to the CI/OFDM signal. Fig. 10 shows the sliding correlation peaks of LFM signal and CI complex sequence at the receiver side in AWGN channel.

Channel estimation and equalization
When the pilot is a CAZAC sequence, it was proven that in the presence of noise, the mean square error of the channel response estimation is minus (Milewski, 1983

CI complex sequence
CI complex sequence was given in equation (18). It is easy to prove the CAZAC feature of the CI complex sequence. Fig. 11 gives simulation results of the CI complex sequence.

CHU sequence
CHU sequence is a polyphase code with a periodic autocorrelation function (CHU, 1972). It was proven that the CHU sequence can be constructed for any code length. When N is even, CHU sequence is defined as where M is an integer relatively prime to N. Fig. 12 shows the amplitude-frequency and autocorrelation results of CHU sequence. Fig. 13 and Fig.14 show the channel impulse respond estimated by CI complex sequence and CHU sequence under AWGN and 4-path Rayleigh channel.

Frequency domain equalization
In the conceptual model of CI/OFDM, the information symbols are simultaneously modulated on multi-subcarrier which makes CI/OFDM inherently having the frequency diversity. At the receiver side, frequency combining may be used to improve the performance of the system.
We still focus on the multi-carrier algorithm which makes use of the properties of IDFT/DFT, such as the linearity and the circular shift. Since every parallel input signal of the second IDFT is the summation of information signals, which are spread by CI signal, the frequency diversity combining should be at the end of the first DFT module at the receiver side.At the receiver, after orthogonal multi-carrier demodulation, the output of the first FFT is where [1, ] iN  is the number of the subcarrier, i H is the transition function of the subchannel and k a is the information symbol, According to Fig. 15, the input signals at the second FFT module which can be expressed as where i w is the gain of the ith subcarrier determined by different combining strategies.
Since i R includes all information about transmitted symbols, the second FFT not only is applied for decomposing the CI spreading into the subcarrier components, but also is used to complete the frequency combining. Several frequency combining strategies such as EGC, the MRC, orthogonal restoration combining (ORC) and MMSEC are considered in frequency selective channel (Itagkai & Adachi, 2004). Computer simulation results are shown in Fig. 16. In CI/OFDM system, though both combing strategies are sensitive to the inter-carrier interference (ICI), the MRC performance is worse than EGC. It is because that in MRC, the phase changes of the signals are lost, which is very important to CI/OFDM. Since ORC equalization perfectly restores frequency non-selective channel but produces the noise enhancement, its performance is better than MRC and EGC but worse than MMSEC. As SNR increases, the BER performance of ORC gets better, but MMSEC equalization provides the best performance. AS we know, the MMSEC not only restores frequency non-selective channel but also minimizes the noise.

Experimental pool
The experimental pool in Xiamen University is an un-censored pool which size is 430(L)x320(W)x200(H)cm.
Channel estimation was carried out, 13kHz single carrier signal is sent at the 30ms intervals. The sample rate of A/D is 160kHz. The transmitted and received signals are shown in Fig.  17. We can see that key features of the channel are multi-path transmission with low noise because of the stationary water in the pool. The maximum delay is about 19 ms which magnitude is under 3% of the maximum one.

Results of experimental pool
In this experiment, LFM signal was used to get coarse synchronization of the frame and CI complex sequence was applied to achieve accurate synchronization and fractional frequency shift estimation (Xu et al., 2008). SNR estimation algorithm is borrowed from the work (Ren, 2005) with synthesis of CI complex sequence. CI complex sequence is used as pilot to estimate the impulse response of the channel. Frequency-domain equalization ORC is used at the receiver to improve the performance of the system. Table 1 shows the parameters of the CI/OFDM system.
The results are shown in

Shallow water in Wuyuan Bay of Xiamen
A CI/OFDM underwater acoustic communication experiment was conducted on Dec. 12, 2008 in Wuyuan Bay of Xiamen, China. The distance between the transmitter and the receiver was 1000m, and they were deployed at 3 m and 2 m below the sea-surface, respectively. The average depth of the water is 4. 5m which was changed with the tide. The channel probing signals include two kinds. One is a single carrier signal which was transmitted at the 100ms intervals repeatedly. The other is a sweeping signal which frequency is from 9kHz to 21kHz. It was also be transmitted once every 100ms.  As shown in Fig. 18, multi-path signals is visible during the 100ms intervals though the amplitude of the multi-path signal is about 2% of the maximum one. The amplitude of the 14KHz signal was changed after 100ms. Fig. 19 denotes the different amplitudes of the sweeping signals which reveal the time-varying and frequency-varying features of the acoustic underwater channel.  Table 3 is the parameters of CI/OFDM underwater acoustic communication system. Table 4 is the BER performance of the system without and with frequency selective combining. The BER performance was significantly improved due to the frequency diversity combining by the cost of decrease of the data rate.

Shallow water in Baicheng water
Two CI/OFDM underwater acoustic communication experiments were conducted on Dec. 17, 2009 andDec. 31, 2009 in Baicheng water of Xiamen, China, respectively. The transmitter was deployed at 2.5m above the sea-floor in 5.5m deep water and the receiver was deployed at 9m below the sea-surface in 16m deep water. The horizontal distance between the transmitter and receiver were 2000m and 5000m, respectively.
The channel probing signals used in these two experiments were same sweeping signals, with frequency from 8KHz to 16KHz. The time interval between different frequencies was 30 ms. Fig.20 is the received probing signal at the short range (2000m). The sea condition was calm but vessels passed through the water frequently. The Frequency-varying feature is different from the feature in Wuyuan Bay. In this experiment, the amplitudes of the lower and higher frequency were faded significantly. The ambient noise was much higher in this underwater channel. From the enlarged map of received signal of 14KHz and 14. 5KHz, it is easy to see that the amplitudes of the strongest arrival changed with time and frequencies. The high level of noise made it difficult to distinguish multi-path signals from the ambient noise.
Note that there is no apparent impulse interference and the amplitudes of multi-path signals are much smaller than that of the main path. The Frequency-varying feature is different from the feature in short range at the same water domain. In this experiment, the ambient noise was much higher than that in the short range.

Results of experiment III
Based on the channel probing results, we concluded that the channel in Baicheng water was worse than that in Wuyuan Bay. In experiments, 4-fold frequency diversity and (2,1) LDPC were applied in CI/OFDM systems (Bai et al., 2009) in order to guarantee the BER performance of the system. System parameters are same in Table 3 except that the comb pilot pattern is used instead of the block pilot. The results of using (2,1) LDPC is the performance improvement and the decrease of the data rate. was deeply frequency fading in the bandwidth of the signal. It might explain the reason of BER performance degradation even though the 4-fold frequency diversities were used.

Conclusion
In this chapter, we first proposed two algorithms which simplify the modulation and demodulation of the conceptual CI/OFDM. Secondly, based on these algorithms and jointed with synchronization, channel estimation and equalization, we constructed CI/OFDM underwater acoustic communication systems. In the end, a number of experiments were carried out in the experimental pool and shallow waters in Xiamen of China to verify the performance of the system. Field results are as followed: The BER of the uncoded CI/OFDM underwater acoustic communication system at the data rate 4.97kbps is lower than 6 41 0   in the experimental pool (experiment I) and 2 1.5 10   in Wuyuan Bay in Xiamen (experiment II). When 4-fold frequency diversity is applied, the data rate is 1.24kbps and the BER performance of the system is lower than 3 61 0   in experiment II.
The BER of the uncoded 4-fold frequency diversity CI/OFDM acoustic communication system at the data rate 1.24kbps is lower than 2 41 0   and 2 71 0   in experiment III at Baicheng water in Xiamen. The BER of the coded frequency diversity CI/OFDM acoustic communication system at the data rate 620bps is almost zero in the short range and the long range.

www.intechopen.com
A few problems exist in the system. The most important one is that the inherent frequency diversity of CI/OFDM did not play its due role in the system. According to our analysis, the orthogonality between the different symbols modulated on the same subcarrier will be destroyed if the phases were changed when signals transmitted in the channel. It means that the intersymbol interference cancels out the diversity combining gain. Future researches should focus on the optimization of the algorithms in order to take advantage of the inherent frequency diversity and other realizations based on the conceptual CI/OFDM.

Acknowledgment
This work was supported by the Fundamental Research Funds for the Central Universities of China 2010121062.