WO2009106880A1

WO2009106880A1 - Acoustic multi-carrier modulation communication scheme and device therefor

Info

Publication number: WO2009106880A1
Application number: PCT/GB2009/050176
Authority: WO
Inventors: John Domokos; Christopher Nigel Smith; John Joseph Spicer
Original assignee: Roke Manor Research Limited
Priority date: 2008-02-29
Filing date: 2009-02-23
Publication date: 2009-09-03
Also published as: GB2457967B; GB0803774D0; GB2457967A; GB0813559D0

Abstract

An acoustic multi-carrier modulation communication scheme comprises a plurality of narrowband carriers (60 to 68) in individual communication channels, wherein each channel is separated by a guard band (69) of frequency dependent bandwidth, the frequency of which is proporti onal to the mean of sum of the carrier frequencies of each channel adjacent to th at guard band. An acoustic multi-carrier modulation communication device comprises a transmitter (73 to 75) and a receiver (76 to 78). The transmitter comprises a signal generator to generate a plurality of narrowband carriers (60 to 68) in individual communication channels (60 to 68). Each channel is separated by a guard band (69) of frequency dependent bandwidth, the frequency of which is proportional to the m ean of sum of the carrier frequencies of each channel adjacent to that guard band. The receiver comp rises individual receivers (76 to 78) for each carrier frequency.

Description

COMMUNICATION SCHEME AND DEVICE

This invention relates to an acoustic multi-carrier modulation communication scheme and device, in particular, one giving improvements to an orthogonal frequency division multiplex (OFDM) modulation scheme for shallow water acoustic communications .

Acoustic communication in shallow water is very challenging, particularly over long distances. Impairments over a 10 Km acoustic channel restrict the communication to a few hundred bauds, particularly under extreme weather conditions. The main impairments are time spread due to multipath propagation and Doppler. The shallow water acoustic channel suffers from severe multi-path effects, Doppler spread and signal delay variation. These impairments vary in time significantly. OFDM modulation copes well with multi-path effects, but still does not solve the Doppler problem. Furthermore, the tracking of the severe signal delay variation is a major problem in most communication systems.

OFDM modulation is well known for its robust performance in radio frequency (RF) communication. Previously, the performance of an OFDM modulation scheme was investigated for the acoustic environment as described by Dean Thomas, John Domokos," Investigation of advanced algorithms for acoustic communication in shallow water" 2nd SEAS DTC Technical Conference-Edinburgh 2007 (Thomas et al). The symbol rate was optimized for a simple channel model proposed by Xueyi Geng, Adam Zielinski, "An Eigenpath Underwater Acoustic Communication Channel Model", Oceans '95 (Zielinski et al). This model predicts a delay spread of 36 ms and a Doppler spread of 3.6 Hz peak to peak (p-p) at 11.5 KHz carrier frequency. However, this OFDM system cannot cope with the simultaneous presence of these impairments. At best, at the optimal symbol rate of 8 Hz, maximum 1.8 Hz p-p Doppler spread could be tolerated.

In accordance with a first aspect of the present invention, an acoustic multi- carrier modulation communication scheme comprises a plurality of narrowband carriers in individual communication channels, wherein each channel is separated by a guard band of frequency dependent bandwidth, the frequency of which is proportional to the mean of sum of the carrier frequencies of each channel adjacent to that guard band. The present invention provides an improvement to the OFDM system and a robust frequency diversity scheme.

Preferably, a pilot signal is transmitted at the centre of the guard band between the communication channels. Preferably, the bandwidth of the guardband is greater than or equal to the expected Doppler spread.

Preferably, the bandwidth of the guardband is scaled with the carrier frequency and the pilot signal is positioned relative to the centre frequency of the carriers.

Preferably, signal delay variation in time is estimated based upon an instantaneous phase difference between adjacent pilots, and the estimated delay variation is subtracted from the signal delay.

This provides a substantially constant delayed signal.

Preferably, the carrier phase is estimated based upon the average phase between adjacent pilots and the carrier phase is de-rotated by the average phase using a complex multiplier.

Preferably, data signals on the communication channels are filtered, modulated to their respective carriers and summed for transmission.

Preferably, for reception data signals on each carrier are down-converted, filtered and demodulated by a separate receiver. Preferably, the receivers are tuned to different frequencies to provide diversity.

Preferably, the number of different frequencies is proportional to the carrier frequencies.

Preferably, the signals are filtered using root raised cosine filters. In accordance with a second aspect of the present invention, an acoustic multi- carrier modulation communication device comprises a transmitter and a receiver; the transmitter comprising a signal generator to generate a plurality of narrowband carriers in individual communication channels, wherein each channel is separated by a guard band of frequency dependent bandwidth, the frequency of which is proportional to the mean of sum of the carrier frequencies of each channel adjacent to that guard band; and wherein the receiver comprises individual receivers for each carrier frequency.

Preferably, the transmitter comprises a filter, modulator and up-converter. Preferably, each receiver comprises a down-converter, filter and demodulator. Preferably, the filters comprise root raised cosine filters. Preferably, the transmitter further comprises a pilot signal generator, for generating at least one pilot signal for transmission between the communication channels.

Preferably, each receiver further comprises a tuner, whereby the frequency at which the receiver receives may be varied.

An example of an acoustic communication multi-carrier modulation scheme and device according to the present invention will now be described with reference to the accompanying drawings in which:

Figure 1 illustrates an example of a frequency plan for a scheme according to the present invention;

Figure 2 illustrates an acoustic multi-carrier communication device according to the present invention;

Figure 3 is a block diagram of a quadrature phase shift keying (QPSK) transmitter for use in the scheme of the present invention; Figure 4 is a graph showing transmitted and received spectrum for four channels with 4Hz Doppler spread;

Figure 5 is a block diagram of a receiver for use with the scheme of the present invention;

Figure 6 is a graph of receive raised root cosine (RRC) filter response, used in the transmitter and receiver of Figs. 3 and 5;

Figure 7 is a block diagram a revised channel model used in the simulation in figures 8 to 11;

Figure 8 shows graphs of raw bit error rate (BER) against signal to noise ratio (SNR) for a single channel; Figure 9 is a graph of BER against pilot frequency at IHz Doppler spread;

Figure 10 is a graph of raw bit error rate against SNR with frequency diversity; and,

Figure 11 is a graph of orthogonal frequency division multiplex and single channel QPSK BER against Doppler spread.

Doppler spread affects the OFDM signals as described below. The OFDM signal comprises 2^k carriers which are generated by the IFFT algorithm. The key aspect of this scheme is that the symbol rate is the reciprocal of the sub-carrier frequency spacing. Under this condition, the sub-carriers are orthogonal. The OFDM signal is demodulated using the fast Fourier transform (FFT) algorithm, i.e.

where n is the index for the input sequence, N is the size of the FFT, and k represents the demodulated carriers. Providing that all carriers are spaced at 2π/N intervals, the "bins", X[k], are isolated from each other.

Due to the Doppler spread the transmitted signal is widened by an additional bandwidth (B), which leaks into the adjacent bins. With an ideal receiver, at a carrier spacing of 2π/N, the spectra of the received signal in the k^th bin is therefore X[k_rx]_ldeal = X[Jc] ₊ B (2) where B is the double sided Doppler spread and X[k_ra] is the received signal. The noise from both, the upper and the lower sidebands adds to the channel.

In the case of the FFT demodulation used in the OFDM receiver, the non- orthogonal components (B) in all other k bins will also add to the received signal, so

N X[KΛ_0FDM = X[k] + 2± ^B (3) k=\ -X IΛJ

Equation 3 shows that in addition to the leakage from the adjacent channels, further leakage from other bins are also added to the signal, but these are diminishing by the sinc(k) envelope.

An improved transceiver architecture according to the invention has similarities with OFDM, in that a large number of carriers are generated. The symbol rate is low compared with the delay spread, so that the carriers suffer only flat fading, hence there is no need for equalisers.

One aspect of the present invention is that the N carriers are transmitted and received independently. In the transmitter, the channels are individually Root Raised Cosine (RRC) filtered, modulated to their respective carriers and then these are summed. In the receiver each carrier is down-converted, filtered and demodulated with separate receivers.

This method requires a lot more computation resources than the simple FFT/

IFFT process, but now the carrier spacing and even the modulation scheme can be arbitrarily scaled and optimised to the properties of the channel at every frequency. In the acoustic channel the Doppler spread is proportional to the carrier frequency, that is

B ∞ B₀ ^ (4) c_s where Bo, is the double sided Doppler spread at fo frequency, c_s is the speed of sound and V is the velocity of the vehicle.

Fig. 1 illustrates the signals over a range of frequencies, where F_p are the pilot frequencies, F_c are the carrier frequencies 60 to 68, with centre frequencies 70, 71 for FcI and FcN illustrated, the frequency separation Δf = 1/R, (where R is the symbol rate) and B is the Doppler spread. In the frequency plan of Fig.1, frequencies FcI ...FcN 60 to 68, range from a minimum frequency at L to a maximum frequency at M. Pilot tones 1 to 12 are transmitted between every channel (FcI to FcN). The spacing between the channel centre 70, 71 and the pilot (FpI, Fpn) is determined by the expected Doppler spread . At low frequencies, where the Doppler spread is very small (Bl) compared to the bit rate, the pilot (1 to 4) is placed close to the channel edge. In this case the excess bandwidth (occupied by the pilot) is very small and hence the throughput is large. With quadrature phase shift keying (QPSK), in an ideal case, 2 bits/s/Hz can be achieved. At the maximum frequency (point M), the pilot (9 to 12) is placed further away (Bn) to accommodate the higher value of Doppler spread. In the system of the present invention, the Doppler spread can extend up to half the symbol rate (R) in which case the pilot symbols may occupy a full channel. A guard band Gb 69 is provided between adjacent channels. The bandwidth of the guardband is frequency dependent.

Another aspect of the present invention is that the spacings between the pilots and the channel centres are scaled with the carrier frequencies because the Doppler spread also scales by the frequency (fo) as shown by equation 4.

An ideal coherent QPSK receiver was constructed. The delay spread for the channel (Tc) is 36 ms, hence the coherence bandwidth (1/Tc) is ~28 Hz. The system is designed for a maximum Doppler spread of B=4 Hz. The symbol rate is chosen to be R=8Hz, so that 1/Tc > R > B (5)

Fig. 2 is a simplified block diagram of an acoustic multi-carrier communication device according to the present invention. A plurality of transmitters 73, 74, 75 illustrated for carrier frequencies FcI to FcN 60, 61, 68, provide inputs to a summer 25 and the output 80 of the summer is input to an acoustic channel 26. The output 81 from the acoustic channel is split in splitter 82 and input to respective receivers 76, 77, 78 for the frequencies FcI to FcN. Fig. 3 shows a simplified block diagram of the transmitter stage. The input I/Q symbols 20 are filtered with a root raised cosine finite impulse response (FIR) filter 21. The channel separation is 8Hz. Every second channel carries a pilot tone 23. The pilot tones are added 22 to the channel at a level of -3dB_ch. Signals 20 output from the summer 22 are modulated 72 to their respective carriers Fc 60 to 68. In this simulation up to four identical channels 24 are used to investigate the benefit of frequency diversity. These are added 25 to the modulated signal and the resultant signal 80 is input to the acoustic channel 26.

Fig. 4 shows the transmitted spectrum (A) and the received spectrum (B) for 4 channels with 4 Hz p-p Doppler spread. It can be seen that the signals arriving at the receiver are badly impaired by both the Doppler and by the noise arising from the sub- path contribution. Magnitude of the signal is given in dB over the range -40 to +30. The frequency is given in Hz from -150 to +150, where dF = 31.25e^"3 Hz. The spectrum shown is overlay wlO: CxFFT: 20Log[FFT] ImRx + JReRX, wl 1 : CxFFT: 20Log[FFT] ReTX + jlmTX. Fig. 5 shows a simplified receiver block diagram. In this example, there are four identical receivers tuned to different frequencies to provide diversity (not shown). The incoming signals 30 from the acoustic channel 26 are downconverted through the multiplier 41 using the centre frequency of the carrier from the carrier frequency generator 40 and are applied to the pilot recovery 31 , in which the signals are shifted 34, 35 by ± 8 Hz, i.e. the pilot frequencies +Fp and -Fp 32, 33. then low pass filtered 36, 37 with a 2 Hz bandwidth FIR (128 taps). Tuning to the desired frequency for each receiver is achieved using the carrier frequency generator 40 and the multiplier 41. The channel filter 43 is a 128 tap RRC filter operating at 8 Hz symbol rate and 16 times over-sampled. The roll-off factor is α=0.3 to accommodate the ±2Hz Doppler spread at the maximum frequency. Because of the large stop-band rejection, and selectivity, this filter provides improved signal to noise (S/N) rejection compared with the FFT demodulation used in the original OFDM receiver. The frequency response is shown in Fig. 6 illustrating how the magnitude in dB tails off with frequency over the range 0 to less than -4OdB.

An important aspect of the present invention is that the delays of the filters 36, 37 and 43 are identical so that the signals when applied to the subsequent blocks are synchronous. This is necessary, because in the acoustic environment the phase and the delay of the channel varies rapidly in time.

The phase at the channel centre is estimated, as the average of the upper (+Fp) and the Lower (-Fp) pilots, 44. The signal is conjugate multiplied 45 with the estimated phase which in turn results in the demodulated signal. Yet another aspect of the present invention is the compensation of the signal delay variations caused by the channel. Group delay (GD) is defined as the phase variation over frequency, that is: GD= dΦ/dω (5)

Due to the simultaneous presence of the upper and lower pilots, the group delay in the channel can be readily calculated, 46 and the instantaneous delay variations are removed 47. The signal delay at the input of the decimator 48 is therefore substantially constant even though the channel is time -variant.

The diversity combiner is controlled by the Signal to Noise Ratio of the received channels (Not shown in Fig.5). This technique achieves substantial frequency diversity improvement under fading conditions. The frequency spacing for the diversity is 48 Hz. This is adequate to achieve the full benefit from the diversity for the given channel model.

Currently, there are no commercially available channel models that represent the shallow water acoustic environment accurately. In the original work (Thomas et al) the acoustic channel was represented by a Rice fading model as proposed by Zielinski et al. In this model, 8 dominant eigen-paths were used and cascaded with their respective subeigen-path models. The primary path represents the dominant mode of propagation. The sub-path represents the non-specular reflections and the local inhomogeneities in the medium. The Doppler spread was modelled with static frequency offsets in each primary path. Given that the Zielinski model represents a Rice fading channel, it might be too optimistic. Some authors proposed the Rayleigh fading channel - an example for this is described in Abolfazl Falahati, Bryan Woodward, Stephen C. Bateman, "Underwater Acoustic Channel Models for 4800 b/s QPSK Signals", IEE Journal of Oceanic Engineering Vol.16, NoI, 1991.

In the present simulation, the static frequencies of the Zielinski model are replaced with noise sources, which are band-limited to their appropriate Doppler spread values. This method eliminates the cyclic response of the static Doppler shifts but represents a harsher environment because the fading profile is now Rayleigh and the spread applies to both the magnitude and the phase of the signals in each ray.

A block diagram of the revised channel model is shown in Fig.7. In the primary channel the bandwidth of the noise source 50 is selectable from 0 to 3Hz. The noise source passes through filter 51 and is multiplied 52 with the signals 80 on the primary path from the input node. The magnitude of the attenuation in attenuator 53 and the values of delay 54 for the eight most significant primary paths and for 10Km path length in sedimentary sea bed scenario are listed in Table 1. The delay values and the transmission losses of Table 1 vary over time. Therefore, these values can be viewed as a 'snapshot' of the propagation condition at an arbitrary instant.

The sub-Eigen path represents the effects of non-specular reflections and the local inhomogeneities such as wave motion and localised turbulences of the medium. The model is very similar to that of the primary path. The noise 55 is band limited to 29 Hz using the filter 56. The band limited output is multiplied 57 with the signal from the primary path and added back 58 to the channel. The total noise power in the sub- Eigen path is -1OdBc in this model. The signals from the 8 primary and secondary paths are combined in the summer 59. For the BER simulation, additional white Gaussian noise is introduced in the system by blocks 27, 28. The processed signals 30 are output from an output node 83 to the receiver stage.

Table 1

The following simulation results are based on the assumption that the Doppler shift is removed. This is can easily be implemented in practice using the pilot tones. Fig.8 shows the raw BER versus SNR for various Doppler spread values for a single QPSK channel. In this simulation the pilot frequency is not scaled with Doppler spread but kept at 8 Hz. The graphs shown are for a static channel, IHz, 2Hz, 4Hz, and 6Hz spread. In this simulation the pilot frequency is not scaled with Doppler spread but kept at 8 Hz. Above 12 dB SNR the BER does not improve noticeably. This means that in this region the sub-path noise dominates. At lower Doppler spreads, the pilot tones can be moved closer to the channel. This, not only increases the throughput (as the pilots occupies less bandwidth), but also improves the accuracy of the phase estimation. This in turn reduces the degradation due to multipath. Fig.9 demonstrates this effect.

Further improvements are achievable by scaling the filter shape to the expected Doppler spread at lower frequencies. A sharper roll-off RRC filter and a narrower FIR pilot filter would reduce the BER even further. It is anticipated that the raw BER could be reduced to 0.02 with optimised filters.

Fig. 10 shows the simulation results for frequency diversity in the Rayleigh fading channel. In this plot the symbol rate is 8 Hz and the Doppler spread is 4Hz. Without frequency diversity the BER is 0.076 which is unusable. The simulation shows that substantial improvement is achieved with diversity; the 2 channel BER is 0.029 and for the 4 channel BER is 0.009 at 20 dB SNR.

Finally, the performance of the original differentially encoded OFDM system of Thomas et al is compared with the proposed QPSK transceiver using the Rayleigh fading channel. Fig.11 compares the residual BER for the two systems without frequency diversity. The OFDM system degrades gradually. The proposed modulation performs much better up to 4 Hz and then the two systems converge. The OFDM system was differentially encoded and there were no guard-bands between the carriers to accommodate the Doppler spread. The multi-carrier modulation scheme of the present invention addresses the impairments in the shallow water channel using a large number of narrow band carriers. However, the carriers are independently processed in the transceiver. The improvement, compared to the OFDM case, is due to a number of features including the use of guard-bands between the carriers to accommodate the Doppler spread. Sharp RRC filters are used to provide high selectivity and large stop-band attenuation and the filtering and the frequency plan is scaled to carrier frequency i.e. matched to expected Doppler spread. The number of the frequency diversity channels is proportional to the Doppler spread. The scheme of the present invention gives rise to performance results at 8 Hz symbol rate, with 4 Hz Doppler spread in a Rayleigh fading channel, a raw BER of 10 ² for the 4 times diversity case. The predicted raw throughput in worse case is 0.25 b/s/Hz and in best case can be close to 2 b/s/Hz.

Claims

1. An acoustic multi-carrier modulation communication scheme comprising a plurality of narrowband carriers (60 to 68) in individual communication channels (60 to 68), wherein each channel is separated by a guard band (69) of frequency dependent bandwidth, the frequency of which is proportional to the mean of sum of the carrier frequencies of each channel adjacent to that guard band.

2. A scheme according to claim 1, wherein a pilot signal (1 to 12) is transmitted at the centre of the guard band (69) between the communication channels (60 to 68).

3. A scheme according to claim 2, wherein the bandwidth of the guardband (69) is greater than or equal to the expected Doppler spread.

4 A scheme according to claim 3, wherein the bandwidth of the guardband (69) is scaled with the carrier frequency and the pilot signal (1 to 12) is positioned relative to the centre frequency (70, 71) of the carriers (60 to 68).

5. A scheme according to any preceding claim wherein signal delay (54) variation in time is estimated based upon an instantaneous phase difference (46) between adjacent pilots (1 to 12), and the estimated delay variation (47) is subtracted from the signal delay.

6. A scheme according to any preceding claim wherein the carrier phase (60 to 68) is estimated based upon the average phase (44) between adjacent pilots (1 to 12) and the carrier phase is de-rotated by the average phase (44) using a complex multiplier (45).

7. A scheme according to any preceding claim, wherein data signals (20) on the communication channels (26) are filtered (21), modulated (72) to their respective carriers (60 to 68) and summed (25) for transmission

8. A scheme according to any preceding claim, wherein for reception, data signals (20, 30) on each carrier (60-68) are down-converted (40, 41), filtered (43) and demodulated (45, 47, 48) by a separate receiver (76 to 78).

9. A scheme according to claim 8, wherein the receivers (76 to 78) are tuned (40, 41) to different frequencies to provide diversity.

10. A scheme according to claim 9, wherein the guard band (69) is proportional to the carrier frequencies (70 to 71).

11. A scheme according to any of claims 7 to 10, wherein the data signals (20, 30) are filtered using root raised cosine filters (21, 43).

12. An acoustic multi-carrier modulation communication device comprising a transmitter (73 to 75) and a receiver (76 to 78); the transmitter comprising a signal generator to generate a plurality of narrowband carriers (60 to 68) in individual communication channels (60 to 68), wherein each channel is separated by a guard band (69) of frequency dependent bandwidth, the frequency of which is proportional to the mean of sum of the carrier frequencies of each channel adjacent to that guard band; and wherein the receiver comprises individual receivers (76 to 78) for each carrier frequency.

13. A device according to claim 12, wherein the transmitter (73 to 75) comprises a filter (21), modulator and up-converter (72).

14. A device according to claim 12 or 13, wherein each receiver (76 to 78) comprises a down-converter (40, 41), filter (43) and demodulator (45, 47, 48).

15. A device according to claim 13 or 14, wherein the filters (43) comprise root raised cosine filters.

16. A device according to any of claims 12 to 15, wherein the transmitter (73 to 75) further comprises a pilot signal generator (23), for generating at least one pilot signal (1 to 12) for transmission between the communication channels (60 to 68).

17. A device according to any of claims 12 to 16, wherein each receiver (76 to 78) further comprises a tuner, whereby the frequency at which the receiver receives may be varied.