CHARACTERIZATION OF AN ACOUSTIC COMMUNICATION CHANNEL WITH PSEUDORANDOM BINARY SEQUENCES

Transcription

1 CHARACTERIZATION OF AN ACOUSTIC COMMUNICATION CHANNEL WITH PSEUDORANDOM BINARY SEQUENCES P. A. van Walree a and G. Bertolotto b a TNO, Oude Waalsdorperweg 63, P.O. Box 96864, 2509 JG The Hague, The Netherlands b FINCANTIERI Cantieri Navali Italiani S.p.A., Via Cipro 11, Genova, Italy Contact author: Paul van Walree, The Netherlands Organisation for Applied Scientific Research TNO, Underwater Technology Group, Oude Waalsdorperweg 63, P.O. Box 96864, 2509 JG The Hague, The Netherlands. Facsimile: paul.vanwalree@tno.nl Abstract: The joint European project "UUV Covert Acoustic Communications" explores methods for underwater communication at low signal-to-noise ratios. The first phase of the project focuses on characterization of the communication channel. Sea trials were conducted in two littoral environments in September The first site is near the island of Bornholm, in the Baltic Sea, and the second site close to the Norwegian coast in the vicinity of the city of Bergen. Part of the trials time was devoted to pseudorandom binary sequences, broadcast in the frequency band between 2.1 and 5.6 khz. This type of waveform combines time delay resolution and Doppler resolution in a versatile probe signal. The present paper examines the evolution of the channel impulse response, together with instantaneous and time-averaged scattering functions. Motion of the transmitter, which is towed by a surface vessel, is found to be the dominant cause of Doppler spreading. Finally, a direct indication of the quality of the communication channel is obtained by demodulating the binary sequences with a communications receiver. Keywords: Channel characterization, pseudonoise, underwater communications. 1. INTRODUCTION The project UUV Covert Acoustic Communications (UCAC) investigates underwater communication at low SNR between a mother ship and an unmanned underwater vehicle 609

2 (UUV). Three sea trials are planned in the course of the project. The first sea trials, conducted in September 2006, focused on characterization of the communication channel. To this end a series of probe signals were broadcast with a towed projector, and recorded on a vertical hydrophone array featuring 128 elements over a 40-m aperture. An overview of the first UCAC sea trials is given by Jans et al. [1]. The present paper describes a probe signal type known as a pseudorandom binary sequence (PRBS). The emphasis is on the description of various processing options and types of information that can be extracted from this probe signal. A systematic comparison between the two experiment sites is not pursued, and there are no comparisons with propagation modelling predictions. A parallel paper addresses chirp probe signals, and also provides more information about the sea trials and run geometries [2]. 2. PSEUDORANDOM BINARY SEQUENCES The PRBS waveforms broadcast during the first UCAC sea trials are described by N 2 f t C gt mt nmt n1 M s( t) cos c m (1) m1 with a band-limited root raised cosine pulse 2 T cos 2t / T g( t) 2 2 T 16t (2) as the elementary pulse g (t). T denotes the bit duration and f c the carrier frequency. The summation over m in Eq.(1) creates a pseudorandom code of M bits C m [-1, 1]. This sequence is simply repeated N times. PRBSs are very similar to a binary phase-shift keyed (BPSK) communication signal, but with a repeated sequence of pseudo random bits also known as chips in spread-spectrum communications rather than meaningful information bits. Maximum-length (m-)sequences are selected for the spreading code C. The periodic autocorrelation function of an m-sequence of length M is binary valued and assumes values of 1 or M. For a PRBS probe signal, the output of a filter matched to the m- sequence is a continually updated measurement of the channel impulse response. Table 1 presents the parameters of the two PRBSs considered in this paper. For both PRBSs the number of sequences N is chosen so as to achieve a total probe signal duration of about 30 s. The bandwidth of 3500 Hz equals twice the bit rate and contains 100% of the energy of the band-limited waveform defined by Eqs.(1)-(2). The sequence length for the probe signal labelled PN1 in Table 1 allows the tracking of a relatively fast changing channel. PN1 probes the acoustic channel every 150 ms, which yields approximately 7 measurements per second. The penalty for the fast tracking capability is the inability to monitor impulse responses longer than 150 ms, and that the filter gain may not overcome long-range propagation losses. PN2 is designed to handle longer impulse responses and can be used over longer ranges, but it cannot follow rapid changes as it probes the channel only once every 1.2 seconds. 610

3 A PRBS combines a good range resolution with a good Doppler resolution, its ambiguity function approaching a double Dirac function in the delay-doppler plane [3]. The price that is paid comes in the form of ambiguous clutter away from the nominal Doppler value. Owing to their Doppler sensitivity, PRBSs require filtering with a bank of frequency shifted (resampled) replicas of the underlying m-sequence. A measure of the Doppler resolution as a fraction of the sound velocity c is given by R D c MT f. c (3) The filter gain drops when the Doppler shift varies by more than R D within the duration MT of a single sequence, which sets a limit to the longest sequence that is still useful under given kinematic conditions. Signal ID PN1 PN2 Bandwidth (Hz) Carrier frequency (Hz) Bit rate T (s -1 ) Sequence length M Sequence duration MT (s) Number of sequences Signal duration (s) Delay time resolution T (ms) Doppler resolution R D (m/s) Table 1: Parameters of the PRBS probe signals. Fig. 1. Temporal evolution of the impulse response, exemplified by two received PN1 signals with seven updates per second. Left: Baltic Sea. Right: North Sea. 611

4 3. TIME EVOLUTION OF THE IMPULSE RESPONSE A received probe signal is matched filtered with a bank of Doppler shifted replicas of the transmitted m-sequence. The filter branch with the highest gain is selected for further processing. A single replica suffices for filtering a 30-s signal as long as drifts of the nominal Doppler shift are much smaller than the Doppler resolution mentioned in Table I. Successive impulse responses are detected in the filter output and stacked in order to visualize the time evolution. Two examples are shown in Fig.1. The example for the Baltic Sea reveals fading and reappearance of multipath arrivals on a timescale of several seconds. In contrast, the North Sea illustration shows two reasonably stable paths followed by a rapidly fluctuating multipath arrival at a 20-ms delay. Note that the blue speckles between 15 and 20 db in the second graph are not due to noise, but to reverberation of previous m-sequences. 4. INSTANTANEOUS DELAY DOPPLER SPREAD An indication of the delay-doppler spread is obtained by correlating a portion of a received signal with the bank of Doppler-shifted replicas. Two examples of this processing are shown in Fig.2 for a tow ship sailing away from the receiver at a speed of ~3m/s. The filter output is interpreted as the delay Doppler spread averaged over the duration of an m-sequence. For short sequences the measurement approaches a instantaneous picture of the doubly spread channel. The Baltic Sea example reveals a number of arrivals with approximately the same Doppler shift. It also shows that the ambiguous clutter occurs everywhere but at the value of the nominal frequency shift. In contrast, the North Sea example reveals a scenario with a significant Doppler spread. Such an instantaneous frequency spreading has only been observed for transmission over short range, in this case up to a few km at a water depth of 200 m, which leads to the hypothesis that the spreading is due to a mixture of direct paths and bottom or surface reflected rays, for which the arrival angles at the receiver differ markedly. Fig.2. Instantaneous delay Doppler spread, determined for the PN2 waveform. Left: Baltic Sea; 38-km range. Right: North Sea; 3-km range. 612

5 5. SCATTERING FUNCTIONS The measured impulse responses h(, t), with the delay time and t absolute time (cf. Fig.1) are converted to a delay-doppler power distribution on the assumption of wide sense stationary uncorrelated scattering [4, 5]. The so-called scattering function P s is computed as a function of delay time and frequency f according to P s, f h(, t) h (, t t)dt exp 2i f tdt (4) In practice the double integral is performed as a series of autocorrelations of h(, t) over the available observation window of 30s, followed by a fast Fourier transform with respect to the time difference t. P s is a stochastic quantity that describes the statistics of the channel in terms of time delay and frequency shift. Scattering functions extracted from sea trials data are instrumental in the design of a communication channel simulator [5]. Two examples of scattering functions are shown in Fig. 3. Both probe signals were received over a range of 52 km in the Baltic Sea. The left panel corresponds to a sailing transmitter ship and the right panel to a stationary transmitter ship, using thruster control to remain in position. The two scattering functions in Fig. 3 differ mainly in the frequency dimension, with a significantly larger Doppler spread registered for the scenario with a sailing tow ship. (Notice that the nominal Doppler shift has been removed from the data.) Fig.3. Scattering functions of received PN1 signals. Left: Baltic Sea, sailing tow ship. Right: Same environment and configuration, but with a stationary transmitter vessel. 6. DEMODULATION A PRBS is a binary phase-shift keyed waveform and as such it can be fed to a communications receiver. Demodulation of a received probe signal provides a direct indication of the channel s suitability to support acoustic communication links. The 613

6 receiver of choice is the multichannel decision-feedback equalizer shown in Fig.4. K hydrophone channels are selected from the total of 128, resampled to remove the nominal Doppler shift, brought to complex baseband, and downsampled to 2 samples per symbol. The receiver further features K fractionally spaced feedforward filters and a single, baudspaced feedback section. The filter updates its tap coefficients at the symbol rate under the minimum mean-square error criterion. The first m-sequence is used to train the equalizer, and the remaining symbols are estimated in decision-directed mode. A second-order phase locked loop (PLL) is included in the structure to compensate for residual Doppler shifts. e iâ PLL CH 1 LPF + T /2 downsampler feedforward filter CH 2 LPF + T /2 downsampler feedforward filter + decision device + å out CH K LPF + T /2 downsampler feedforward filter feedback filter ð 2 i fc t e LMS Fig.4. Multichannel adaptive equalizer with an integrated PLL. All filter taps are jointly updated with the least mean squares algorithm. Fig.5. Measured impulse response and equalized symbols at 1750 bps. An equalized BPSK symbol constellation is shown in Fig. 5 for a PN2 reception in the Baltic Sea, transmitted over a range of 54 km. This scattering function in the left panel of Fig.3 is representative of the communication channel. With the transmitter towed near the channel axis, the impulse response reveals a crescendo of multipath arrivals characteristic 614

7 of sound trapped in a channel [6]. The severe time dispersion stretches over more than 200 symbols. At an input SNR of 14.0 db, the intersymbol interference is too intricate to be resolved for a single hydrophone channel. Sixteen hydrophones are therefore selected, evenly distributed over the available aperture of 40 m. The equalizer is operated with no less than feedforward taps and 100 feedback taps, resulting in adequately resolved BPSK constellation clouds at an output SNR of 13.7 db. A typical picture of residual Doppler is shown in Fig.6 for a North Sea reception. The white markers 205 in total represent direct measurements of the carrier phase, obtained by unwrapping the phase angle of the most energetic multipath arrival of the measured h(, t). A second opinion is provided by the phase offset as estimated by the PLL upon demodulating the signal. The two approaches compare well, however with the PLL lagging somewhat behind. A cyclic component is observed with a period of ~5s, corresponding to motion of the tow ship on the waves. Tow ship accelerations transferred onto the acoustic source are ultimately responsible for most of the observed Doppler spreading. The contribution of the medium itself is smaller and hard to quantify in the absence of absolutely immobile transmitters and receivers. 10 Phase offset (radians) Direct measurement Digital PLL Time (s) Fig.6. Residual Doppler for a North Sea reception of PN1. 7. CONCLUDING REMARKS This paper illustrated several types of information that can be extracted from pseudorandom binary sequences. Specifically, the time evolution of the channel impulse response was considered, short-term and long-term scattering functions were determined, and a communications receiver was applied to the recorded data. Both the instantaneous and long-term delay-doppler measurements revealed a Doppler spread with a large contribution of tow ship motion. In the former case the spread occurred for transmission over a short range, with sound arrival angles differing strongly between direct paths and bottom or surface reflected sound. In the latter case, concerning long-range transmission, the chief cause was motion of the transmitter ship on the waves, transferred onto the 615

8 transmitted sound via the towed projector. Instantaneous delay-doppler measurements do not reveal significant Doppler spreading under these conditions. Considering the successful BPSK signalling scheme under challenging conditions, one may be inclined to think that a more covert scheme operating at a low data rate might be feasible at a reduced input SNR. Indeed, as the symbol rate decreases the signal becomes less sensitive to time spreading, and spread-spectrum gain may be used to overcome a low input SNR. However, as the symbol rate decreases the rate of filter parameter updates is also reduced, rendering the signal increasingly sensitive to Doppler effects and time variability [7]. Future papers will report on the design and performance of covert modulations developed within the UCAC project. 8. ACKNOWLEDGEMENTS The work described in this publication was done under a multinational, three-year project aimed at developing and demonstrating long-range covert acoustic communication with unmanned underwater vehicles (UUVs) in coastal waters. This project under the EUROPE MOU ERG No1 is known under the name RTP "UUV - Covert Acoustic Communications". The project partners are: Kongsberg Maritime AS (Norway); Fincantieri (Italy); Reson A/S (Denmark); TNO Defence, Security and Safety (Netherlands); Patria Systems (Finland); and Saab Underwater Systems AB (Sweden). The Federal Armed Forces Underwater Acoustics and Marine Geophysics Research Institute (FWG) has been commissioned by the Federal Office of Defence Technology and Procurement (BWB) (Germany). Subcontractors are the national defence research establishments of Sweden (FOI) and Norway (FFI), Cetena (Italy), the University of Genova (also Italy), and the Technical University of Delft (the Netherlands). REFERENCES [1] W. Jans, I. Nissen, F. Gerdes, E. Sangfelt, C.-E. Solberg, and P. van Walree, "UUV covert acoustic communications preliminary results of the first sea experiment," in Techniques and technologies for unmanned autonomous underwater vehicles (AUVs) - a dual use view, SCI-182/RWS-016, Eckernförde, Germany, [2] W. Jans, I. Nissen, and F. Gerdes, "Characterization of an acoustic communication channel. UCAC I sea trial, propagation loss, noise floor and impulse response," submitted for the UAM2007 conference. [3] G. Jourdain and J. P. Henrioux, "Use of large bandwidth-duration binary phase shift keying signals in target delay Doppler measurements," J. Acoust. Soc. Am. 90, (1991). [4] P. A. Bello, "Characterization of randomly time-variant linear channels," IEEE transactions on communication systems 11, (1963). [5] G. Bertolotto, T. Jenserud, and P.A. van Walree, "Initial design of an acoustic communication channel simulator," Proc. UDT Europe 2007, Naples, Italy. [6] R. J. Urick, Principles of underwater sound, 3rd ed., Peninsula Publishing, pp [7] M. Stojanovic, J. G. Proakis, and J. A. Catipovic, "Performance of high-rate adaptive equalization on a shallow water acoustic channel," J. Acoust. Soc. Am. 100, (1996). 616