Acoustic Channel Characterization in the Baltic Sea and in the North Sea

Transcription

1 Acoustic Channel Characterization in the Baltic Sea and in the North Sea H. S. Dol*, F. Gerdes**, P. A. van Walree*, W. Jans** and S. Künzel** * TNO Defence, Security and Safety, P.O. Box 96864, 2509JG The Hague, The Netherlands (Henry.Dol@tno.nl) ** Federal Armed Forces Underwater Acoustics and Marine Geophysics Research Institute (FWG), Klausdorfer Weg 2-24, Kiel, Germany (FrankGerdes@bwb.org) Abstract This paper reports results from the joint European project UUV Covert Acoustic Communications (UCAC), which aims at the establishment of a covert communication link between an Unmanned Underwater Vehicle (UUV) and a distant mother platform using acoustic telemetry. To this end, sea trials were carried out in 2006 and 2007 in two different areas in the Baltic and Norwegian North Sea. During the sea trials, acoustic and environmental data were obtained for the characterization of the acoustic channel particularly the transmission loss and the channel impulse response with the purpose of evaluating different (covert) acoustic communication schemes. A selection of these data is analyzed in the present paper. I. INTRODUCTION To achieve the objective of long-range covert acoustic communication through littoral waters, knowledge about the characteristics of the acoustic channel and the influence of the environment is required. Environmental aspects include transmission loss due to spreading and absorption, ambient noise, multipath propagation, Doppler and time variability due to the moving platforms and due to variability of the medium itself. System performance and its information throughput depend strongly on the signal distortions that may result from these environmental conditions. In order to understand and evaluate the performance of the communication algorithms developed within the UCAC project [1], several aspects of channel characterization are treated in the present paper. The channels under examination were probed during two sea trials in 2006 and 2007, respectively, conducted at three sites for both trials: east of the Danish island Bornholm in the Baltic Sea (site A, see Fig. 1), in the Norwegian North Sea west of the city of Bergen (site B), and in the nearby Bjørnafjord (site C). In this paper, only the first two sites (A and B) are considered. The paper begins with a description of the experimental procedure, the transmitted signals and the data processing. Next, results including transmission loss and multipath are presented, first for the Baltic Sea and then for the North Sea. A comparison between 2006 and 2007 is performed. Both trials took place in the same time of the year, but the weather conditions were different. In 2007, the conditions were windier than in 2006, resulting in an increase of wind-induced waves, and consequently, movement of transmitter and receiver. After discussion of these results, the paper is concluded with a summary of our findings. Figure 1. Geographic overview of the measurement locations of the UCAC sea trials in 2006 and 2007 (sites A, B and C). II. EXPERIMENTAL PROCEDURE A. UCAC Sea Trials The first sea trial, UCAC I, was carried out in 2006 between the end of August and mid September in the Baltic and North Sea. The second sea trial, UCAC II, took place in the same areas almost exactly one year later (Fig. 2). For characterization of the acoustic channel, Chirp (Linear Frequency Modulation, LFM) and Pseudo-Random Binary Sequence (PRBS) probe signals were broadcast by two powerful omni-directional transducers (ITC4008 and ITC4009). These were mounted together in a transducer frame, which was towed away or towards a stationary receiving system by S/V Ocean Surveyor (Baltic, 2006), HMS Belos (Baltic, 2007) or HU Sverdrup II (North Sea 2006 and 2007). Motion and depth of the deployed source frame, as well as the water depth below the ship, were recorded on board of the transmitting vessel. The receiving system was a vertical hydrophone chain deployed from FS Planet, which anchored at a fixed position. For UCAC I, a 128-hydrophone vertical array (called VAIII) with an aperture of about 38 m was used /08/$ IEEE

2 For UCAC II, a hydrophone chain named NESSY using either 6 (Baltic Sea) or 8 hydrophones (North Sea) was applied. The spacing of the NESSY hydrophones was 10 m. Both the receiving and transmitting vessels were equipped with GPSsynchronized trigger units for synchronizing the acoustic data recordings during the runs. A sound velocity profile (SVP) was measured with a Seabird CTD probe on board of FS Planet normally every hour, and with a second probe on board of the transmitting vessel before the start and after the end of the runs. The acoustic data set was complemented with meteorological data and surface motion data from a waverider buoy. In between inbound and outbound runs, covert acoustic communication experiments were performed with a different transmitter. The prototype UCAC acoustic modem was used to broadcast several covert modulation schemes [2]. B. Chirp Probe Signal Analysis of Transmission Loss The determination of Transmission Loss (TL) was based on the processing of a series of Chirp signals (LFM sweeps), called an A-frame. The A-frame pulse train consisted of 20 1-s long narrow-band LFM sweeps with frequency-dependent bandwidths of 8 to 25 percent of the center frequency and four 1-s long broad-band LFM sweeps. The center frequencies of the narrow-band LFMs ranged from 800 to 1800 Hz and 2100 to 5600 Hz. The A-frame was sent every 60 seconds (UCAC I) or every 160 seconds (UCAC II). The received LFM signals were analyzed using standard matched-filtering procedures, which yielded both impulse responses and signal levels as function of distance between source and receiver and as a function of 20 center-frequencies. The signal levels were transformed into TL values using the transmitter characteristics [3]. For the matched-filtering procedure, two methods were implemented and checked against each other. In the first method, the received energy level was obtained by adding up the amplitude values of the five largest matched-filter peaks. This means that the contribution of the five strongest multipaths was taken into account. The transmission-loss data presented in this paper have been obtained with this method. For comparison, a second method was implemented. For this method, all matched-filter peaks with amplitudes below a certain threshold were set to zero. The threshold value was chosen such that the correlation noise was removed. Then a de-correlation was performed to obtain the times series of the signal but without the noise. The time-series signal was integrated to obtain its energy. For propagation situations with many equally important multipaths, the second method yields slightly larger energy levels, which are considered to resemble more closely the real (in an energy-detect sense) energy levels of the signals. If the propagation is dominated by one, two or three strong multipaths, then both methods produce almost identical results. The transmission-loss data presented in this paper have been obtained with the first method. The differences with the results of the second method are insignificant for the purposes of this paper. B07 / B23 HU Sverdrup II A05 / A28 HMS Belos FS Planet FS Planet Figure 2. Run geometry for the Baltic Sea (top) and the Norwegian North Sea (bottom) during UCAC II. The black arrows point towards the position of FS Planet, that is, of the receiving hydrophone chain. The red lines indicate the tracks along which the acoustic source was towed by either HMS Belos (Baltic) or HU Sverdrup II (North Sea). The UCAC I run geometry was very similar. C. PRBS Probe Signal Analysis of Multipath and Temporal Variability The PRBS waveforms discussed in this paper were transmitted every 480 seconds. The relatively long repetition interval arises because the A-frame and PRBS signals were only two of several other waveforms that were transmitted at regular intervals. The broadcast PRBS waveforms are repeated m-sequences as described by [4]. Table I presents the parameters of PRBS signals for both UCAC I and UCAC II. The sequence length for the pseudonoise waveform given in Table I allows the tracking of a relatively fast changing channel. The signal probes the acoustic channel every ms, which yields 7 8 measurements of the impulse response per second, allowing it to follow rapid changes of the channel. The penalty for the fast tracking capability is that this signal is not suited to monitor impulse responses longer than ms, and that the filter gain may not overcome longrange propagation losses. For the PRBS signals, the number of sequences N was chosen so as to achieve a total probe signal duration of about 30 s. For the UCAC I signal, the frequency band is Hz, with a center frequency of 3850 Hz. For UCAC II, the frequency band and carrier frequency are

3 and 5000 Hz, respectively. Consequently, the vertical array VAIII of UCAC I was no longer suitable as it has an upper limit of 6000 Hz. Therefore, only the NESSY hydrophone chain, which covers a wider frequency range, was deployed during the second sea trial. TABLE I PARAMETERS OF THE TRANSMITTED PRBS SEQUENCES Signal ID UCAC I UCAC II Bandwidth (Hz) Carrier frequency (Hz) Chip rate T 1 (s 1 ) Sequence length M Sequence duration MT (s) Number of sequences N Signal duration NMT (s) Measurements of temporal variability touch upon the essence of communication channel characterization. Timevariable impulse responses affect the viability of particular modulations and influences the computational complexity of receiver algorithms. The working method is as follows. One of the hydrophone channels is selected. The mean Doppler shift is completely removed from the signal by re-sampling. Subsequently, the re-sampled signal is matched-filtered with a zero-doppler replica of the transmitted sequence. During the UCAC sea trials, this condition was always met for the applied probe signal. Successive impulse responses are detected in the filter output and stacked to obtain a complexvalued matrix of measured impulse responses h as a function of time t and time delay τ: h = h(t,τ). The measured impulse responses are further processed to obtain the following results, displayed by the panels in Figs. 4, 5, 7 and 8. Time evolution This panel shows the matrix of stacked impulse responses h(t,τ) (waterfall plot). Delay-Doppler spread This panel shows the Fourier transform of the impulse responses according to: ( 2πi f t) h( t, )d, S( f, τ ) = exp τ t (1) where the integral runs over the available observation period of approx. 30 s. The spreading function S gives the deterministic distribution of signal power as a function of delay time and frequency shift f. Power delay profile This panel gives the received signal intensity as a function of the time delay: I delay 2 ( τ ) = S( f, τ ) df, (2) where the integral runs over the frequency shift. The result equals the integral of h(t,τ) 2 over the observation window. Doppler power spectrum 2 I ( f ) = S( f, τ ) dτ. (3) doppler This panel reveals the received signal intensity as a function of frequency shift. The range of frequencies over which the Doppler power spectrum is (essentially) nonzero is known as the Doppler spread. Strictly speaking, (2) and (3) should be defined in terms of the scattering function, which is the stochastic equivalent of the spreading function. However, the difference is negligible for the shown examples. The applied PRBS analysis tool further delivers multipath phase measurements and the temporal coherence [4], but these are not shown in the present paper. III. BALTIC SEA The experiments in the Baltic Sea were performed east of Bornholm, at approx. 55 latitude and 15 longitude, with the anchoring positions of FS Planet being the same for UCAC I and UCAC II. Of the many runs performed, we will here compare only Run A05 of the first sea trial with Run A28 of the second sea trial. Both runs were outbound runs (increasing distance with time) along the same track. The speed of the transmitting vessels (S/V Ocean Surveryor for UCAC I, and HMS Belos for UCAC II) was about 5 knots. A05 covered ranges from about 2.5 km to 55 km, and A28 covered ranges from about 5 km to 52 km (Fig. 3, middle and bottom panels). The water depth along the track was about 80 m. The significant wave heights were about 0.3 m for most of Run A05 (sea state 2). In 2007, the weather was rougher with wave heights of about 1 m for Run A28 (sea state 3). The observed oceanographic conditions were typical for Baltic Sea summer conditions. Below a well-mixed and solarheated isothermal surface layer with low salinity, we observe a steep drop of the temperature (thermocline). Below the depth of minimum temperature, the temperature and salinity increase steadily towards the seafloor. The stagnant water column is stably stratified because of the dominant effect of salinity on density, which over-compensates the effect of the increasing temperature, which would cause the density to decrease. The important result for the acoustics is the existence of a pronounced sound channel. It is well known that a sound channel acts as a waveguide in which acoustic signals can travel long distances with little attenuation when the sound source is located within the sound channel. For UCAC I, when the sound channel axis was at m, the source was towed at about 35 m. Because we wanted to have comparable sound propagation conditions and because the sound channel axis was at m for UCAC II, the source was towed at a depth of about 40 m in It should be understood that the source depth was not exactly constant during a run but depended somewhat on tow speed and currents. During Runs A05 and A28, the source depth slowly varied by about ±1 m. On the

4 other hand, the depth and extent of the sound channel was also quite variable in As previously mentioned, the vertical array VAIII was used as receiver in Of its 128 hydrophones, 7 were selected for the transmission-loss analysis. These cover the depth range from about 14 m to 48 m. In 2007, the longer NESSY chain was able to cover the depth range from 10 m to 60 m (Fig. 3). Fig. 3 shows transmission loss as a function of distance and hydrophone depth for Runs A05 and A28 for sub-pulses 10 and 20, respectively (note that the vertical axis is upside down). This corresponds to LFMs with center frequencies of 3400 Hz and 5400 Hz. Note that the left panels for Run A05 display data from 7 hydrophones, whereas the right panels for Run A28 display data from 6 hydrophones. Beware that particularly large spikes do not represent real changes of the transmission loss but are rather related to changes in the receiver attenuation settings, which were not accounted for at exactly the right instant of time. For easier identification of such spikes, the solid disks indicate the samples at which attenuation settings changed. For both runs, we observe the typical increase of the transmission loss (TL) with distance. The TL is largest for the hydrophones located outside the sound channel. In fact, inside the sound channel, the TL is between 20 and 25 db smaller than outside the sound channel. This holds true for both runs, for both frequencies and also for pretty much all propagation distances. We further note that the TL is almost independent of the exact depth as long as the hydrophones are anywhere inside the sound channel. Because the sound-velocity gradient at the upper boundary of the sound channel was very sharp in 2006 (Fig. 3, upper left panel), the hydrophone at 19.2 m, which usually was inside the sound channel, was sometimes above the sound channel. Then the TL increased abruptly towards the TL curve of the hydrophone at 13.8 m, which was always above the sound channel. Beginning with Run A28, we note that for the hydrophones within the sound channel the TL for the LFM at 3400 Hz follows approximately a 16 log 10 (R) law between 5 km and 20 km range (R). Between 20 km and 35 km, it stays almost constant and increases again beyond 35 km. The behavior is similar for the LFM at 5400 Hz. The shape of the TL curves and the dependence of the TL on the position of the hydrophones relative to the sound channel is very similar to the observations made for Run A05. Also the absolute values of the TL agree well between Runs A05 and A28, particularly for propagation inside the sound channel. If we take, for example, the data for sub-pulse 10 (LFM from 3200 Hz to 3600 Hz), we find at a distance of 5 km TL values of about 55 db for Run A05 and 60 db for Run A28. At a distance of 20 km, the TL values are 70 db for Run A05 and 72 db for Run A28. Similar agreement is observed for other frequencies. For the hydrophones above the sound channel, there is less agreement between Run A05 and Run A28, particularly for shorter propagation distances. For example, at a distance of 10 km, the TL for Run A05 is 90 db and 100 db for Run A28. On the other hand, for propagation distances beyond 20 km the TL curves for A05 and A28 become increasingly similar. We now turn our attention to the (normalized) impulse responses based on the PRBS probe signals and the temporal variability of the multipath. The results of Runs A05 and A28 are compared for three propagation ranges: a short range (approx. 5 km), a medium range (approx. 25 km) and a long range (approx. 50 km), see Fig. 4. The recordings of the hydrophones at 30 and 40 m depth are used for Runs A05 and A28, respectively. The first observation is that the impulse responses all have their maximum intensity at, or close to, the end. This agrees with the fact that the direct path along the sound channel axis takes the longest time as the sound velocity has a minimum at the channel center (both transmitter and receiver are near the sound channel axis). The second observation is that the time spreading increases with the range. This is in agreement with the idea that most multipath contributions remain captured in the sound channel but are more spread out in time with increasing range due to the variation of the sound velocity across the channel. The third observation is that the time spreading was significantly smaller in 2007 than it was in The most probable explanation is that the sound channel was weaker in Fig. 3 shows that the thermocline was much steeper in 2006 with colder water directly below. The sound-velocity profiles in Fig. 3 show that the sound velocities at the channel boundaries were approximately the same for both trials, but that the sound velocity at the center was much lower in In fact, the velocity difference between the center and the lower boundary was roughly a factor 2 lower in 2007 (approx. 15 m/s versus approx. 30 m/s in 2006). Consequently, the signal was less spread in time (and distance) due to the lower gradients. Fig. 5 shows waterfall plots of the time evolution of the impulse response at the longest range, giving an idea of the temporal variability that seems to be qualitatively comparable for the two sea trials in the Baltic Sea. The Doppler spread, on the other hand, was significantly larger in 2007 due to increased relative motion of transmitter and receiver (platform kinematics) caused by more wind-induced wave activity. The observed wave period of approx. 5 s is clearly recognized in the approx. 0.2 Hz peak separation in the Doppler power spectrum. Of particular influence is the motion of the tow ship, which leads to pitch and roll movements of the transducer frame. This explanation of the increased Doppler spreading is supported by the observation that the Doppler spreading was smaller in a stationary run (Run A29, not shown; the transmitting vessel applied dynamic positioning). The increase of the Doppler spreading is also partly due to the higher frequency band used in Note that there are two contributions to the measured Doppler variance: multipath fading and platform motion. The first effect is connected to the channel coherence time, the second is not.

5 Figure 3. Temperature, salinity, density and sound velocity profiles, and TL curves obtained by matched-filtering for Run A05 of UCAC I (left) and for Run A28 of UCAC II (right). The middle and bottom panels show the TL based on the LFM from 3200 to 3600 Hz and from 5200 to 5600 Hz, respectively. The depth of the hydrophones is given in the legend. The source depth was about 35 m for A05 and 40 m for A28.

6 Figure 4. PRBS channel impulse responses at short (approx. 4 km; left), medium (approx. 27 km; center) and long range (approx. 53 km; right) in 2006 (top) and 2007 (bottom). Both sea trials were performed at approximately the same time of the year, that is in late summer, in the same area of the Baltic Sea. Figure 5. Temporal variability in the Baltic Sea during Run A05 (2006; top) and Run A28 (2007; bottom) at long range (approx. 53 km). The left panels depict the impulse response (time evolution) and the center panels its Fourier transform. The Doppler power spectrum in the right panels results from the center panels by integration over delay time.

7 IV. NORTH SEA The trial area west of Bergen was at about 60 latitude and 5 longitude. Again, we will compare data from UCAC I with those from UCAC II. Runs B07 and B23 have been chosen for this purpose. The water depth in this area was approximately 200 m. The source was towed at a depth of about 60 m for both runs. The vertical array used for Run B07 covered a depth from 26 m to 66 m, whereas the NESSY chain used for Run B23 covered a range of almost twice the size, that is, from 10 m to 90 m. Therefore, only the data from hydrophones between 30 and 70 m can be compared. The significant wave heights were about 0.5 m during Run B07 (sea state 2), while they were about 1 m the year after (sea state 3). For Run B07, the sound velocity profiles (Fig. 6) consist of a shallow surface layer (until approx. 10 m), followed by a downward-refracting layer in which the velocity decreases almost linearly with depth. At about 80 m, the sound velocity has a minimum and then increases slowly because of increasing pressure. The result is a sound channel that is very weak compared to the one found in the Baltic Sea. The soundvelocity profile of Run B23 looks quite different from that of B07. Its profile is more irregular and there is a sharp increase in sound velocity at a depth of about 130 m. However, we emphasize that in the North Sea the sound-velocity profiles did not only vary from year to year, but also from day to day or even from one hour to the other. The skewed sound channel between 80 and 130 m did almost completely disappear after a few hours. The reason for the instability is the relatively strong tidal currents that may advect water masses of differing characteristics through the trial area. A discussion of this effect is beyond the scope of this paper. In contrast to the situation in the Baltic Sea, we do not find a large difference between the TL curves for the various hydrophones. The reason is, of course, the absence of a strong sound channel with large sound-velocity gradients. Over the depth range covered by the hydrophone chain, the soundvelocity profile was downward refracting. With the acoustic source being located at 60 m depth, the consequence is that the largest TL is found for the topmost hydrophone and the smallest TL for the deepest hydrophone. The higher the frequency, the more pronounced this effect becomes. Due to the larger transmission losses in the North Sea water column, when compared to the Baltic Sea, the maximum ranges are significantly smaller in the North Sea. A transmission loss of about 110 db is obtained at a range of approx. 35 km in the North Sea, while such losses are only obtained in the Baltic Sea at more than 50 km for transmission through the surface layer. Probably more than twice this range could have been obtained for signals travelling inside the Baltic sound channel. For UCAC II, we note that for large distances the TL for the 90 m hydrophone is smaller than the TL for the other hydrophones. This could be related to the fact that the soundvelocity profile shows an increase at 130 m, so that a small sound channel is formed. During UCAC I, the sound-velocity profile was different but the VAIII antenna did not reach to depths of 90 m. Therefore, we cannot compare the TL from Run B07 (UCAC I) and B23 (UCAC II) for depths larger than 70 m. For the depth range from 30 m to 70 m, the TL curves show comparable values for the two runs. However, a curious difference is that the TL increases quite steadily with increasing distance for Run B23, whereas for Run B07 we find the increase to have slopes that vary with the propagation distance. We are not able to provide a conclusive explanation for the variation of TL data of Run B07. The main hindrance is that for this run we do not have precise information on the depth of the acoustic source so that we cannot rule out the possibility that changes of the depth of the transducer are responsible for the variations of the TL. The other, more likely, explanation is that the tidal currents caused the sound-velocity profile to vary during the course of the run. The absence of TL variations for Run B23 would then indicate the absence of variations of the sound-velocity profile during Run B23. Again, this is difficult to prove with the available environmental data. The temporal variability is again considered for three ranges: a short range (approx. 5 km), a medium range (approx. 20 km) and a long range (approx. 30 km), see the PRBS results in Fig. 7. Due to the absence of a (strong) sound channel at the transmitting and receiving depth of 60 m, the strongest arrivals are obtained near the beginning of the temporal power distribution, presumably corresponding to the direct (or least reflected) path. The typical shape of the impulse response thus has its maximum in front, in contrast to the Baltic situation where the maximum is at the end. The time spreading is expected to decrease with range in the North Sea as paths with multiple bottom and/or surface reflections damp out. In practice, a clear trend has not been observed. In 2006, the impulse response consisted of many distinct peaks (multipaths) at the short range, which were largely gone at the medium range, but somehow a significant time spread reappeared at the long range. In 2007, almost no time spreading was observed at the short range, increasing significantly at the medium range, to fall back a little bit at the long range. As said before, the North Sea acoustics were very variable and differed from ping to ping. A consistent difference between 2006 and 2007 appears to be the significantly lower signal-to-noise ratio in the North Sea in 2007, especially for the long range, see Figs. 7 and 8. Analysis of noise spectra (not shown here) revealed a broadband but intermittent spectral energy distribution, with spiky fluctuations corresponding to clicking and crunching sounds that were clearly heard during the trial. These sounds were probably caused by straining of FS Planet s anchor line because of larger wave action in Nevertheless, an impulse response can still be recognized between the ambient noise. Finally, the Doppler spectra in Fig. 8 show significant and comparable Doppler spreads in 2006 and Since the

8 Doppler spread measures approximately 2 Hz, and since the time evolution panels reveal stable multipath arrivals with a fading time of 10 s or longer, the Doppler spread is mostly due to relative motion of transmitter and receiver. V. CONCLUSIONS For characterization of the acoustic channel, Chirp (LFM) and Pseudo-Random Binary Sequence (PRBS) probe signals were broadcast by omni-directional transducers and recorded by vertical arrays covering between 40 and 90 m of the water column, depending on configuration and trial area. The horizontal range was varied from a few kilometers up to more than 50 km by towing the source away from and towards the stationary receiving platform. The influence of the environment on the acoustic propagation characteristics was studied by executing the same experimental procedures at three different locations (only two were discussed here) under different meteorological circumstances. In the Baltic Sea, the relatively stagnant and shallow waters were characterized by a stable stratification with an isovelocity surface layer on top of a well-developed sound channel, which acts as a waveguide for acoustic signals. Consequently, within the sound channel signals can travel long distances with little attenuation. A notable feature of propagation through the sound channel is that rays deviating from the direct path, which is the route of the strongest arrival, propagate through regions with sound velocities larger than that in the center of the sound channel. Hence, they arrive earlier than the main arrival. This results in significant time spread that increases with the propagation distance (Fig. 4). The time spread also increases with increasing difference in sound velocity between the center and the boundaries of the sound channel. This is seen by comparing the sound-velocity profiles and the measured impulse responses from the first and second sea trial in the Baltic Sea (Fig. 4). In the Norwegian North Sea, the water column showed a thin iso-velocity surface layer with constant sound velocity above water showing sound velocity that first decreased down to a depth of roughly 100 m before it increased again because of increasing pressure. A pronounced channel was absent. Hence, best receptions were made at the deepest hydrophones due to the downward-refracting water column (Fig. 6). An important difference between the 2006 and 2007 sea trials was the weather, as there was more wind and the waves were higher in The less favorable weather conditions during the second sea trial had an adverse effect on platform stability and noise levels. The increased relative motion of source and receiver translated into a somewhat larger Doppler spread in the Baltic Sea (Fig. 5), while the noise levels were significantly higher in the North Sea (Figs. 7-8). However, overall, the propagation characteristics of the underwater environment were qualitatively the same for the two sea trials. Necessary adaptations to ensure comparable transmission ranges were primarily the depths of the transmitter and receiver, which followed from the measured sound-velocity profiles. ACKNOWLEDGMENTS 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, P. van Walree, UUV covert acoustic communications preliminary results of the first sea experiment, in Techniques and technologies for unmanned autonomous underwater vehicles a dual use view, RTO Workshop SCI-182/RWS-016, Eckernförde, Germany, [2] P. van Walree, E. Sangfelt, G. Leus, Multicarrier spread spectrum for covert acoustic communications, Proceedings of Oceans '08, Quebec, Canada. [3] W. Jans, I. Nissen, F. Gerdes, Characterization of an acoustic communication channel UCAC I 2006 sea trial, propagation loss, angular distribution and impulse responses, Proceedings of Underwater Acoustic Measurements UAM2007, Heraklion, Crete. [4] P. A. van Walree, T. Jenserud, M. Smedsrud, A discrete-time channel simulator driven by measured scattering functions, IEEE J. Sel. Areas Commun., submitted 2008.

9 Figure 6. TL curves obtained by matched-filtering for Run B07 of UCAC I (left) and for Run B23 of UCAC II (right). The middle and bottom panels show the TL based on the LFM from 3200 to 3600 Hz and from 5200 to 5600 Hz, respectively. The source depth was about 60 m for both B07 and B23.

10 Figure 7. PRBS channel impulse responses at short (approx. 5 km; left), medium (approx. 18 km; center) and long range (approx. 32 km; right) in 2006 (top) and 2007 (bottom). Both sea trials were performed at approximately the same time of the year (late summer) in the same area of the North Sea. Figure 8. Temporal variability in the Baltic Sea during Run B07 (2006; top) and Run B23 (2007; bottom) at long range (approx. 32 km). The left graphs depict the impulse response (time evolution) and the center graphs its Fourier transform. The Doppler power spectrum in the right graphs results from the center graphs by integration over delay time.