Assessment of HF Sonar Performance from a Surface Ship

Transcription

1 Assessment of HF Sonar Performance from a Surface Ship Mark V. Trevorrow DRDC Atlantic Research Centre Defence Research and Development Canada Scientific Report DRDC-RDDC-2016-R053 April 2016

2 IMPORTANT INFORMATIVE STATEMENTS This work was conducted under DRDC Project 01CC (torpedoes and torpedo defence). DRDC and DFO jointly supported use of the CCGS Vector for these sea-trials. Template in use: (2010) SR Advanced Template_EN (051115).dotm Her Majesty the Queen in Right of Canada, as represented by the Minister of National Defence, 2016 Sa Majesté la Reine (en droit du Canada), telle que représentée par le ministre de la Défense nationale, 2016

3 Abstract This Scientific Report presents a quantitative analysis of sea-trial data assessing the target detection performance of a High-Frequency (HF, 100 khz) horizontally-oriented, multibeam sonar from a surface ship. Forward-looking target detection tests were conducted in two different locations, featuring upward- and downward-refracting sound speed profiles under sea-states from 0 to 4. Ship wake effects were evaluated in aft-looking tests with a towed target at ship speeds of knots. The HF sonar output was calibrated so that absolute reverberation levels and acoustic target strengths could be assessed. In forward-looking tests the targets were detected at ranges up to 580 m in low sea-states, but detection range was reduced to roughly m under higher sea states. In the aft-looking sonar tests the target was detectable up to 450 m range. In both cases signal to noise ratios of db (range dependent) were routinely observed. Near-surface sound speed gradients were found to produce important acoustic propagation effects in both cases. Significant ping-to-ping variability in both target echo strength and background reverberation levels was observed in all tests. A HF sonar performance modeling tool generated predictions in reasonable agreement with the sea-trial results. Significance to Defence and Security A specific assessment of the performance of HF (~100 khz) active sonars for torpedo detection and tracking in the vicinity of a ship was planned under the DRDC Torpedo Defence project. The sonar utilized in these sea-trials has some of the characteristics of systems which might be employed for close-range target detection and tracking purposes from a ship. Similar sonars have been previously developed for forward-looking obstacle avoidance and harbour surveillance purposes. DRDC-RDDC-2016-R053 i

4 Résumé Le présent rapport contient une analyse quantitative des données d essais en mer pour évaluer la performance de détection de cibles au moyen d un sonar multifaisceaux haute fréquence ( 100 khz), orienté horizontalement, à bord d un navire de surface. Des essais de détection frontale de cibles ont été réalisés à deux endroits différents, révélant des profils de vitesse du son, à réfraction vers le haut et à réfraction vers le bas, dans des états de mer de 0 à 4. Les effets de sillage du navire ont été évalués lors d essais de détection arrière, sur une cible remorquée, à des vitesses de 6 à 7,5 noeuds. La sortie du sonar HF a été étalonnée de manière à ce que les niveaux de réverbération absolue et l indice de réflexion acoustique puissent être évalués. Lors des essais de détection frontale, les cibles ont été détectées à une distance pouvant atteindre 580 m lorsque la mer était calme, mais la portée de détection était réduite à près de 150 à 300 m lorsque la mer était agitée. Pendant les essais de sonars à balayage arrière, la cible était détectable jusqu à une distance de 450 m. Dans les deux cas, on a couramment observé un rapport signal-bruit de 10 à 25 db (en fonction de la distance). Des gradients son vitesse, près de la surface, ont eu d importants effets sur la propagation acoustique dans les deux cas. On a également observé une variabilité «ping-à-ping» importante du niveau de l écho des deux cibles et du niveau de réverbération à chaque essai. Un outil de modélisation du rendement du sonar HF a généré des prévisions montrant une concordance raisonnable avec les résultats obtenus lors des essais en mer. Importance pour la défense et la sécurité Une évaluation spécifique du rendement des sonars actifs HF (~100 khz) pour la détection et la poursuite de torpilles à proximité d un bâtiment a été prévue dans le cadre du projet de défense antitorpilles de RDDC. Le sonar utilisé pour ces essais en mer possède certaines caractéristiques des systèmes qui pourraient être employés à des fins de détection et de poursuite de cibles à courte distance d un navire. Des sonars similaires ont déjà été mis au point à des fins d évitement d obstacles frontaux et pour la surveillance des zones portuaires. ii DRDC-RDDC-2016-R053

5 Table of Contents Abstract i Significance to Defence and Security i Résumé ii Importance pour la défense et la sécurité ii Table of Contents iii List of Figures v List of Tables viii 1 Introduction Sea-Trials, Instrumentation, and Analysis Approach Overview of Sea-Trials HF Multibeam Sonar Sonar Data Analysis Methods Forward-Looking Sonar Sonar Data Examples Saanich Inlet at Low Sea-State Strait of Georgia at Sea-State Strait of Georgia at Sea-State Summary Results Transmission Loss Effects Variability of Target Echoes Propeller Cavitation Effects Aft-Looking Sonar Sonar Data Examples Fixed-Range Towing Target Pull-Ins Target Detection During S-Turns Wake Backscatter and Extinction Target Echo Variability Modeling HF Sonar Performance Summary Discussions References Annex A HF Sonar Calibrations A.1 Sonar Self-Noise Levels A.2 Clipping Levels A.3 Transmitter Source Levels DRDC-RDDC-2016-R053 iii

6 A.4 Transmitter Horizontal and Vertical BeamWidths A.5 Echo Strength from Reference Target Spheres A.6 Receiver Bandwidth List of Symbols/Abbreviations/Acronyms/Initialisms iv DRDC-RDDC-2016-R053

7 List of Figures Figure 1: Map showing the two sea-trial locations (from 2 Figure 2: Figure 3: Comparison of measured sound speed profiles from Saanich Inlet and Strait of Georgia locations Sketch of (upper) forward-looking and (lower) aft-looking sonar geometry. Surface waves are omitted for clarity Figure 4: Sonar targets (left). Steel sphere with chain ballast, (right) Sharko towbody.. 5 Figure 5: Figure 6: Figure 7: Figure 8: Figure 9: Figure 10: Figure 11: Figure 12: Figure 13: Figure 14: Photograph of the SM-2000 sonar mounted on the sonar strut in the aft-looking orientation. The strut pivots outboard and down to lock into a vertical position. 6 Forward-looking HF sonar image to 600 m range, 1711Z April 9 in Saanich Inlet. Wind speed = 2.5 knots. Range rings are in 120 m increments. Arrow denotes location of big target sphere, suspended at 4 m depth Comparison of target echo and background intensity (db re sonar units) vs. range for approach run against the big sphere starting 1710Z April 9 in Saanich Inlet. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping and noise thresholds Apparent Target Strength (db re m 2 ) vs. range for approach run against the big sphere starting 1710Z April 9 in Saanich Inlet. Dashed line shows sonar clipping threshold Forward-looking HF sonar image to 400 m range, 1735Z April 11 in Strait of Georgia. Wind speed = 12 knots. Range rings are in 80 m increments. Arrow denotes location of large target sphere, suspended at 4 m depth Comparison of target echo and background intensity (db re sonar units) vs. range for run against the big sphere starting 1734Z April 11 in Strait of Georgia. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping and noise thresholds Forward-looking HF sonar image to 200 m range, 2108Z April 12 in Strait of Georgia. Wind speed = 16 knots. Range rings are in 40 m increments. Arrow denotes location of large target sphere suspended at 1 m depth Comparison of target echo and background intensity (db re sonar units) vs. range for run against the big sphere starting 2108Z April 12 in the Strait of Georgia. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping threshold Summary of maximum detection ranges from all forward-looking sonar tests with big and small spheres in Saanich Inlet and Strait of Georgia, April Best fit line to data where wind > 6 knots is shown Comparison of averaged Reverberation Level (db re 1 µpa) vs. range for runs against the big sphere in Saanich Inlet and Strait of Georgia on April 9, 11, and DRDC-RDDC-2016-R053 v

8 Figure 15: Comparison of apparent TS (db re m 2 ) vs. range for big sphere on April 9, 11, and 12 under wind speeds of 2.5, 12, and 16 knots, respectively. Data are 10-pt. adjacent averaged. Dashed line shows estimated true TS of db. Target depths were 4 m on April 9 and 11 and 1 m on April Figure 16: Figure 17: Figure 18: Figure 19: Figure 20: Figure 21: Figure 22: Figure 23: Figure 24: Figure 25: Figure 26: Figure 27: Figure 28: Frequency distribution of apparent TS over all ranges for detection run against the big sphere starting 1710Z April 9 in Saanich Inlet. Best fit Gaussian distribution (mean = -8.1 db, width = 14.8 db, r 2 = 0.926) shown in red line.. 20 Forward-looking HF sonar image to 600 m range, 2118Z April 9 in Saanich Inlet during period of ship acceleration from 7.5 to 8 knots. Range rings are in 120 m increments. Arrows denote bands believed to be blade rate modulations. 21 Summary of Sharko depths during aft-looking HF sonar tests April 13 in Saanich Inlet. Black lines show approximate -3 db limits of sonar vertical beam Aft-looking HF sonar image to 200 m range, 1828Z April 13 in Saanich Inlet. Range rings are 40 m increments. Arrow denotes location of Sharko target, at 16 m depth Temporal variation in target echo and background intensity for stationary tow at 220 m range in ship wake starting 1618Z April 13 in Saanich Inlet. Ship speed during this period was 7.2 knots Apparent target strength (db re m 2, black squares) and SNR (db, red circles) vs. range from all fixed-range towing tests, April 13 in Saanich Inlet Comparison of averaged target, background, and SNR vs. range from six separate pull-in runs in arbitrary sonar units. Dashed lines show -20log 10 [r] and -30log 10 [r] variations Apparent target strength (db re m 2 ) estimated from average of six separate pull-in runs. Red line shows best fit line (intercept = db, slope = db/m) Aft-looking HF sonar image to 300 m range during S-Turn, 1626Z April 13 in Saanich Inlet. Range rings are 60 m increments. Arrow denotes location of Sharko target Apparent target strength (db re m 2 ) vs. time following Sharko towed target at 210 m range through series of S-Turns, 1624Z-1630Z April 13 in Saanich Inlet Comparison of in-wake and non-wake reverberation level vs. range profiles at two speeds, taken near 1800Z April 13 in Saanich Inlet Distributions of apparent target strength during fixed-range towing at 220 m range 1618Z April 13. Best fit Gaussian distribution (mean = db, width = 4.0 db, r 2 = 0.994) shown in red line Predicted sonar SE vs. target range and depth for the April 9/10 Saanich Inlet conditions, with assumed wind speed of 5 knots. Target TS = -14 db vi DRDC-RDDC-2016-R053

9 Figure 29: Figure 30: Figure 31: Figure 32: Figure A.1: Figure A.2: Figure A.3: Figure A.4: Figure A.5: Figure A.6: Figure A.7: Predicted sonar SE vs. target range and depth under April 11 Strait of Georgia conditions, with assumed wind speed of 10 knots. Target TS = -14 db Predicted sonar Pd vs. range at 4 assumed wind speeds for a 1 m deep, -14 db target under Strait of Georgia conditions Predicted sonar SE vs. target range and depth under April 13 Saanich Inlet conditions, with assumed wind speed of 2 knots. Target TS = -19 db. Black line shows approximate target depth vs range Comparison between measured wake RL vs. range, taken 1758Z April 13 in Saanich Inlet, with ESPRESSO prediction at wind speeds of 10, 15, and 20 knots Comparison of sonar self-noise intensity (db re sonar units) at various sonar range settings. TVG given in Eq.(A.1) with -92 db and -82 db offsets Comparison of sonar self-noise intensity (db re sonar units) at various gain offsets for the 600 m sonar range setting Measured SM-2000 vertical beam-pattern (normalized), with reference curve based on simple line-array with length of 66 mm Measured SM-2000 horizontal beam-pattern (normalized) at medium and high transmit power Analytic prediction of Tungsten-Carbide sphere backscatter target strength vs. frequency for two sphere diameters Example target echo strength parameters extracted from a 90 sweep using the 60 m sonar range setting. Target sphere was located at 14.3 m range Measured variation in receive response vs. frequency for default high-resolution 600 m setting DRDC-RDDC-2016-R053 vii

10 List of Tables Table 1: SM-2000 sonar settings used in these sea-trials. Acoustic and sample resolution based on sound speed of 1480 m/s Table 2: ESPRESSO input parameters for the SM-2000 sonar models Table 3: ESPRESSO predictions of SM-2000 sonar detection ranges Table A.1: Measured sonar amplitudes and estimated transmit source levels Table A.2: Summary of SM-2000 sonar calibration parameter, using Eq.(A.2) viii DRDC-RDDC-2016-R053

11 1 Introduction This work explores the capabilities and limitations of high-resolution, hull-mounted sonar for detection and tracking of near-surface targets. The intended Naval applications generally require that targets must be detected and tracked at horizontal ranges up to roughly 500 m around the ship with spatial resolution of order 1 m and update rates near 1 Hz. In order to achieve high spatial and temporal resolution, use of relatively High Frequency (HF: 10 to 100 khz) multibeam sonar systems is required. Such sonars have been previously utilized for seabed mapping, obstacle avoidance, and harbour surveillance purposes. The use of HF sonar in this manner is subject to significant environmental constraints. Seabed reverberation and clutter is an obvious problem in shallow waters. Additionally, previous work has suggested that the performance of high-resolution active sonars near the ocean surface may be strongly limited by sea-state conditions and ship wakes [1, 2]. HF sonars are highly sensitive to the presence of small ( µm radius) air bubbles, which are created by both natural processes (e.g., due to surface wave-breaking) and in ship wakes. These bubbles generate both strong acoustic backscattering, which raises the background reverberation levels, and strong acoustic extinction, which reduces the target echo strength. Several additional physical effects also complicate the target detection process, namely: acoustic refraction by near-surface sound speed gradients, interference between direct and surface-reflected acoustic paths, and modulation and possible shadowing of surface-reflected acoustic paths by surface waves. All of these effects are accentuated in the near-horizontal, small surface grazing-angle geometry of this sonar application. A focused sonar detection performance sea-trial was conducted in order to assess the impact of these difficulties. These sea-trials were conducted in relatively deep waters in order to isolate the near-surface constraints. A new 90 khz multibeam active sonar was acquired specifically for these tests. This sea-trial was held April 8 to 14, 2015 based on the ship CCGS VECTOR operating in B.C. south coast waters [3]. This one-week sea-trial was a collaboration between DRDC Atlantic Research Centre (Dept. of National Defence) and Institute of Ocean Sciences (IOS), which is part of the Dept. Fisheries and Oceans (DFO). These tests took advantage of a unique retractable sonar strut mounted on the CCGS VECTOR. The HF multibeam sonar was mounted in both forward- and aft-looking configurations. Drifting and towed acoustic targets were constructed and deployed for these tests. This Scientific Report is focused on quantitative sonar performance, investigating sonar signal and noise characteristics under a variety of conditions. Specifically tests were conducted under sea-states from 0 to 4 and with both upward- and downward-refracting sound speed profiles. Ship wake effects were evaluated at ship speeds up to 7.5 knots. The HF sonar output was calibrated so that absolute reverberation levels and acoustic target strengths could be assessed. Finally, a sonar performance modeling tool was validated against some of the sea-trials results. DRDC-RDDC-2016-R053 1

12 2 Sea-Trials, Instrumentation, and Analysis Approach This section briefly describes the sonar sea-trials and HF multibeam sonar, followed by details on the post-analysis approach. A detailed summary of the sea-trials is presented in [3]. The brief descriptions herein are intended to provide context for the analysis of results and sonar modeling to be reported in the following three sections. Sonar results in the post-sea-trial report [3] were based solely on sonar data replay without any follow-on quantitative analysis. Following the sea-trials, acoustic calibration measurements were performed on the HF multibeam sonar in order to verify the sonar characteristics, quantify performance limitations, and provide quantitative acoustic output. The HF sonar calibration results are reported in Annex A. 2.1 Overview of Sea-Trials The sea-trials were conducted at two locations on the southern coast of British Columbia, as shown in Figure 1. The Saanich Inlet site was just west of the IOS facilities at Patricia Bay, B.C. Saanich Inlet was chosen for its calm (wind < 10 knots), relatively deep water ( m) conditions with minimal shipping traffic. The limited fetch prevented development of any significant wave height, particularly for wind directions across the inlet. Tests were conducted in Saanich Inlet on April 9, 10, and 13. The southern Strait of Georgia operating area was northeast of Galiano Island in waters m deep. The Strait of Georgia location featured higher winds (typically 8 16 knots) and Sea-States (SS 3 to 4, significant wave heights 1 2 m), providing a more demanding test of the forward-looking sonar geometry. Tests were conducted in the Strait of Georgia on April 11 and 12. In the Strait of Georgia location the wind direction was from the SW, and so seas were fetch-limited in the lee of Galiano Island. The distance from shore was greater than 5 km. Figure 1: Map showing the two sea-trial locations (from 2 DRDC-RDDC-2016-R053

13 Water property measurements were conducted at various locations in Saanich Inlet, Strait of Georgia, and the Canadian Gulf Islands during these sea-trials in support of a DFO monitoring program. Figure 2 shows a comparison of sound speed estimated from Conductivity-Temperature-Depth (CTD) measurements at the sea-trial sites. The two sea-trial locations showed distinctly different near-surface (< 20 m depth) water properties. The Saanich Inlet location exhibited nominally downward-refracting conditions whereas the Strait of Georgia location had upward-refracting conditions. All three Saanich Inlet profiles were taken at the same location, within a few kilometers of the location of the sonar tests. In Saanich Inlet there is some variation in the near-surface sound speed profiles, but they are essentially identical below 40 m (profiles extend down to > 200 m). Meteorological measurements were made with a ship-based system, corrected for ship speed and heading (see [3] for details). No independent measurement of Sea State (e.g., measurement of significant wave height) was conducted. Figure 2: Comparison of measured sound speed profiles from Saanich Inlet and Strait of Georgia locations. The sea-trials ship, CCGS Vector, is 40 m length overall with displacement of 515 tonnes and top speed of 12 knots. These sea-trials took advantage of a retractable sonar strut mounted on the ship s starboard side, allowing deployment of the HF multibeam sonar at a depth of 3.5 m in either a forward- or aft-looking horizontal orientation. A Kongsberg Mesotech SM-2000 sonar system operating at 90 khz was utilized for these sea-trials (to be described in greater detail in Section 2.2 below). Ship position, speed, and heading (from Global Positioning System GPS) as well as sonar heading (from a gyrocompass) were recorded along with the raw sonar data. The experimental geometry is sketched in Figure 3. Because of the near-horizontal sonar geometry there will be generally both direct and surface-reflected acoustic paths from sonar to target (and back). Both cases will experience strong backscattered reverberation from surface waves and sub-surface bubble clouds. In the forward-looking tests the target depth was 1 or 4 m, putting it within or close to the bubble layer. In the aft-looking tests the target was generally below the ship wake. In both cases the surface-reflected path is subject to scattering and absorption by bubbles relative to the direct path. In the presence of surface waves, the reflection point will change its location depending on the local surface slope. Because the experimental sites were chosen in relatively deep waters any seabed reflections can be ignored. DRDC-RDDC-2016-R053 3

14 Figure 3: Sketch of (upper) forward-looking and (lower) aft-looking sonar geometry. Surface waves are omitted for clarity. The forward-looking sonar tests examined the detectability of drifting near-surface targets at low and medium sea states. These tests were conducted in both Saanich Inlet and the southern Strait of Georgia. The targets were two hollow steel spheres, 91 cm and 71 cm diameter (denoted big and small, respectively), ballasted to submerge them just below the surface (see Figure 4, left). The target depth was usually 4 m, except for all tests on April 12 in the Strait of Georgia where it was 1 m. Sonar data was recorded as the ship made a series of approach runs near the targets at differing approach headings relative to the wind/sea. The typical Closest Point of Approach (CPA) was between 15 m and 50 m. The typical ship speed during these approach runs was 3 to 5 knots under sea state conditions ranging from 0 to 4. Over the 4 days of forward-looking tests a total of 40 runs were made against one or more targets in each run. The sea-trial summary report [3] made predictions of the rigid Target Strength (TS) of the two target spheres; specifically db and db for the small and large spheres, respectively. While these predictions were believed to be approximately valid, it was expected that errors would arise from the assumption of rigid scattering and echo contributions from the steel chain ballast (see Figure 4, left). Therefore this report will utilize the sonar calibrations to produce in situ estimates of the target strengths as part of the overall assessment. The aft-looking sonar tests were designed to test target detection in the vicinity of the ship wake. The aft-looking sonar tests were only conducted under calm conditions in Saanich Inlet on April 13. For these tests the SM-2000 sonar was simply rotated to face aft on the sonar strut. Unfortunately, for reasons unknown the sonar strut was deployed on April 13 with a 9 counter-clockwise rotation. This posed no difficulty in conduct of the sea-trials or post-processing because the wake and towed target remained fully within the sonar 90 horizontal aperture. 4 DRDC-RDDC-2016-R053

15 Figure 4: Sonar targets (left). Steel sphere with chain ballast, (right) Sharko towbody. A newly designed acoustic target, nicknamed Sharko (see Figure 4, right), was towed at distances between 20 and 450 m behind the ship at speeds of 6 to 7.5 knots. Sharko was 2.41 m in length and 41 cm in diameter (excluding fins) with a weight of approximately 135 kg (in air). It was internally freely-flooding. Sharko had an internally-recording pressure sensor to determine its depth. At tow speeds of 6 to 7.5 knots the typical Sharko depth was between 6 and 25 m, dependent on cable scope. Two types of tests were conducted: fixed-range, where the target was held at a constant cable scope for periods of 5 minutes, and target pull-ins, where the tow-rope was reeled-in at a speed of 2 m/s (3.9 knots) starting from approximately 220 m range. Sharko carried internally two 30 cm diameter target spheres (SonarBell), each of which was specially designed to have Target Strength (TS) near -6 db at frequencies near 10 khz. The TS at 90 khz was not known, but was expected to be closer to the well-known geometric TS (e.g., from [4]) of 10 log 10 [radius 2 /4] = db (re m 2 ). It was expected that backscatter contributions from other parts of the Sharko tow-body (e.g., a stainless steel bulkhead plate immediately aft of the nose cone) would make some contribution at these frequencies. Owing to time and logistical constraints, independent Sharko TS measurements have yet to be conducted. In the forward-looking sonar tests there were two additional targets, an empty oil drum and an acoustic bubble measurement device called the Resonator. The barrel was only used on April 9, and hence did not provide a comparison between sites. The Resonator was routinely observed in the sonar data, however its echo strength was found to be highly variable and inconsistent. Neither of these is used in this present work. 2.2 HF Multibeam Sonar DRDC acquired a new HF multibeam active sonar, a Kongsberg Mesotech SM-2000 system, for these sea-tests. The SM-2000 system consists of a sonar head, a power and data telemetry interface unit, and a sonar processing, display, and data storage computer. The sonar head was mounted on a sonar strut over the side of the ship (see Figure 5). The strut allowed deployment of the sonar head at 3.5 m depth in either a forward-looking or aft-looking orientation. It was found that the sonar strut with the SM-2000 head remained stable at ship speeds up to 8 knots under calm sea conditions. The sonar head was connected to dry-end electronics (inside the ship s DRDC-RDDC-2016-R053 5

16 science lab) via a 30 m umbilical cable. The processor was also setup to accept and record inputs from both a gyrocompass (TSS Meridian Surveyor) and GPS. The SM-2000 sonar operates at 90 khz, providing 128 overlapping 1.8 wide beams covering an angular aperture of 90. The vertical beamwidth (full width to -3 db) is 13.5 on both transmit and receive. The sonar operates by transmitting over the full horizontal aperture (90 ) then beam-forming the received echoes. The maximum operating range is 800 m; reduced range settings were used to decrease the ping interval and improve range resolution. The nominal transmit source level (given by the manufacturer) is 205 db (re 1 µpa at 1 m). The sonar allows some control over transmit power, pulse lengths, receiver time-varying gain, and receiver bandwidth. These tests utilized sonar range settings between 200 m and 800 m, pulse lengths between 0.5 and 1 ms, and pulse intervals between 0.5 and 1 s. Laboratory measurements on the sonar as part of the acoustic calibrations (see Annex A) generally confirmed the manufacturer specifications. Figure 5: Photograph of the SM-2000 sonar mounted on the sonar strut in the aft-looking orientation. The strut pivots outboard and down to lock into a vertical position. During all these sea-trials the SM-2000 was configured to provide maximum transmit power and best range-resolution for a given range setting. The sonar parameters corresponding to the various range settings are shown in Table 1. Sonar range settings from 200 to 600 m were in the sea-trials. The varying receiver bandwidth is presumably set to minimize internal noise limitations in the sonar. This bandwidth constraint also slightly reduced the effective range resolution, e.g., at the 600 m setting the 800 Hz bandwidth corresponds to an effective pulse length of 1.25 ms and hence range resolution of m. A fixed sound speed of 1480 m/s (an average value based on CTD measurements, see Figure 2) was used for all sonar processing and range calculations. 6 DRDC-RDDC-2016-R053

17 Range Setting (m) Table 1: SM-2000 sonar settings used in these sea-trials. Acoustic and sample resolution based on sound speed of 1480 m/s. Ping Rate (Hz) Pulse Length (ms) Acoustic Resolution (m) Sample Resolution (m) Receiver Bandwidth (Hz) The SM-2000 Time-Varying Gain (TVG, in db) was set to the factory default of 20 log 10 [r] r, (1) where r is the range in metres and 1 is the acoustic absorption, having the value db/m. Note that this value of 1 under-predicts the expected in situ value, which should be close to db/m under the sea-water conditions encountered in this trial (using relations found in [5]). A detailed post-calibration of this sonar was performed, reported in Annex A, which included verification of this TVG function. 2.3 Sonar Data Analysis Methods The SM-2000 processor recorded all the raw sonar data along with GPS position and gyrocompass heading data in a single file. Only the raw, unprocessed sonar element data was recorded. The SM-2000 then regenerates the sonar images upon playback, thereby providing some flexibility to adjust beam-forming parameters. The manufacturer provided (at special request) a MATLAB script to perform the beam-forming. The sonar data were extracted from the raw sonar recordings using a two-stage process. Firstly the raw receiver stave data vs. time from a single ping (herein called a snapshot) were extracted and used in a beamforming algorithm to produce a matrix of sonar amplitude vs. range and beam-angle. At this stage the amplitudes were in arbitrary sonar units, to be converted to acoustic units later. These were stored in separate snapshot files. Then various follow-on processing steps (e.g., beam-wise averaging) were performed on these snapshot files. A fixed sound speed of 1480 m/s was used to calculate sonar range. The sonar beams were assumed to be equally spaced at 90 /128 beams = increments, with zero angle in the center of the sonar aperture and positive angles to starboard. The range increment was dependent on the sonar range setting. The maximum sonar amplitude after beam-forming was near 2000 (sonar units), presumably derived from a raw A/D resolution of 12-bits (±2048 units). This implies a maximum sonar echo level near 66.2 db (re sonar units). DRDC-RDDC-2016-R053 7

18 An analysis routine was written to extract estimates of the target sphere and Sharko echo strength and Signal to Noise Ratio (SNR) from successive ping snapshots. This was performed on a sequence of snapshot files using appropriate initial values for target range and bearing. The initial target location was determined through visual inspection of the sonar images. In each snapshot the maximum target echo amplitude in the vicinity of the previous estimate was found. This was assumed to be an estimate of the target echo amplitude. No thresholding was applied. Then the background amplitude was averaged over a region 20 samples in range by 10 samples in angle (excluding the central peak due to the target). This corresponds to a region roughly 5 m in range by 7 in angle. The peak and averaged background echo-amplitudes were stored along with ping number, range, and bearing. A display of the sonar snapshot with the target region highlighted was used to visually confirm the target tracking process. Estimates showing significant deviation in range and/or bearing from the overall trend of the approach run were manually removed. The acoustic calibration information was used to estimate target strengths and reverberation levels. The backscattered acoustic Target Strength (TS, db re 1 m 2 ) is an intrinsic property of the target, and is generally dependent on frequency and incidence angle. For the sphere or Sharko targets, the apparent TS was calculated using a sonar equation approach [4], i.e., TS = 20 log 10 [A] + K + 40 log 10 [r] + 2 r - TVG(r), (2) where A is the target echo amplitude (sonar units) and K is the calibration coefficient (db). This latter parameter is a catch-all that includes sonar source level, sensitivity, fixed gains, analog-to-digital conversion, and beamforming calculations. Estimated values of K (mildly dependent on sonar range setting) can be found in Annex A, Table A.2. Note that Eq.(2) implicitly utilizes a spherical spreading acoustic propagation model and only includes losses due to seawater absorption. Variations in the measured TS due to non-spherical acoustic propagation and bubble-induced losses will generate range-dependent variations in the estimated TS. Non-spherical acoustic propagation and bubble-induced losses are to be expected in this work due to sea-surface reflections, acoustic refraction due to sound speed gradients, and bubble-induced losses. Here the term Apparent TS will be used to refer to this measured TS which includes acoustic propagation effects. By searching for data which do not have appreciable non-spherical acoustic propagation (e.g., at short range) estimates of the true TS can be made. The above TS equation also assumes no sonar beam deviation losses. This assumption is considered generally valid because in these tests the target elevation angles relative to the sonar were always near zero, and thus less than the ±6.5 vertical beamwidth of the sonar. The exception to this occurs in the aft-looking tests where the Sharko target was towed at depths up to 25 m, creating depression angles up to 5. Where vertical Beam Deviation Loss (BDL) corrections are required, a simple line-array model [4] can be used, hence: BDL( ) = 20 log 10 [sin(x)/x], where x = ( L/ ) sin( ) (3) where is the vertical angle, L is the array effective vertical length (66 mm), and is the acoustic wavelength ( m). This is for one-way (either transmission or reception) loss. Additionally, the calibration results (Annex A) showed a significant reduction in sonar sensitivity near the outer 8 DRDC-RDDC-2016-R053

19 edges of the beam, hence data from the outer 10 of the sonar horizontal aperture were ignored in any quantitative analysis. The sonar Reverberation Level (RL, db re 1 µpa) can be estimated using a similar sonar equation approach, i.e., RL = SL + 20 log 10 [A] + K - TVG(r). (4) where SL is the sonar Source Level (db re µpa at 1 m) verified by the calibrations (see Annex A). This RL is independent of any geometric spreading or beam-pattern corrections. Occasional interference due to the ship s navigational Doppler speed sensor was observed. This sensor operated at 100 khz using a broadband, structure pulse. The signature of this in the SM-2000 data was an isolated, high-intensity ring, approximately 4 m in width, with highest intensity centered on the bearing intersecting the ship s bow. Depending on the SM-2000 range setting (and hence ping rate) this interference ring occurred at varying range in the images. Any contamination from this in the target and background intensity estimates was either avoided or manually removed. Echoes from the ship s hull were consistently observed in the forward-looking tests. These dominated the port side of the sonar image up to a strong highlight 25.5 m forward and 4.5 m to port, believed to be due to the breaking bow wave. In areas of stronger sea-surface reverberation, an acoustic shadow created by the ship s hull was also observed. This shadowing was not a problem for forward-looking target detection because the target spheres were usually kept on the ship s starboard side (same as the sonar). Any beam-wise averaging avoided use of shadowed regions. The ship s hull was not visible in the aft-looking sonar tests. Occasionally a direct seabed-reflected echo was observed. This appeared as a narrow band in all sonar beams at a range equal to the local water depth (approx. 180 to 220 m). This was more commonly observed in Saanich Inlet. Seabed echoes did not interfere with target measurements, but do show up in averaged background data. On occasions when the ship accelerated, strong acoustic interference attributed to propeller cavitation was observed. This usually disappeared when the ship speed reached a steady-state speed. This did not pose any operational limitations on the data collection, and was avoided in the data processing by only choosing data from steady-state ship operations. Note that this type of interference may (in future) be a problem for higher-ship-speed applications. This will be investigated in Section 3.4 below. DRDC-RDDC-2016-R053 9

20 3 Forward-Looking Sonar This section will present an overview of the HF sonar target detection results. Recorded sonar data was replayed and analysed following methods described in Section 2.3. Several example data sets will be presented below, followed by summary results and investigation of particular features of the data. Based on a review of the sonar imagery there were a number of qualitative observations which generally applied to both the forward- and aft-looking sonar tests: Under low wind and sea-state conditions the targets were detectable at ranges up to 580 m. The targets were generally detected and followed using their temporal and location consistency from ping to ping. On a single-ping basis targets were frequently difficult to detect due to the presence of strong background clutter with similar intensity and spatial scale. Target echoes were generally 1 2 m in range dimension by approximately 1 2 in angular width. As the wind and sea-state increased, targets were detectable only at shorter ranges, in some cases at ranges less than 200 m. While this suggests wind speed is a significant factor in target detectability, the effect of near-surface acoustic propagation enhancement was also found to be important. Qualitatively the sonar performance appeared to be generally reverberation limited, with only minor evidence for flow-noise limitations on the sonar. In cases with very low background reverberation, at longer ranges the sonar background was dominated by internal electronic noise enhanced by the sonar TVG. A key characteristic of all the sonar data was strong small-scale variability, both spatially within each snapshot and temporally from ping to ping. This might be described as fluctuations or scintillation. Both target echoes and the background reverberation exhibited this variability, even under calm conditions in Saanich Inlet. Under higher sea-state conditions (e.g., in Strait of Georgia) these fluctuations appear to be modulated by surface wave effects. 3.1 Sonar Data Examples In this sub-section, three examples covering different sea-states and locations will be examined. These examples will focus on the big target sphere. The results from the small target sphere were quantitatively very similar Saanich Inlet at Low Sea-State Figure 6 shows an example sonar image from Saanich Inlet, collected under calm conditions (wind speed < 3 knots). The target was the big (91 cm diameter) sphere at a depth of 4 m. The ship speed at this time was approximately 2.5 knots. The target is distinctly visible at 530 m range approximately 5 to the right of the sonar main axis. The sonar target appears elongated in the angular dimension due to the 1.8 beamwidth; the apparent target width decreases at shorter ranges. This target could be clearly tracked inwards over a period of 7 minutes to a Closest Point of Approach (CPA) near 40 m. The large intense scattering features on the right side of the image 10 DRDC-RDDC-2016-R053

21 are the remnant bubbly wake from a previous pass by the ship. A low-intensity ring corresponding to the direct seabed echo is visible at 200 m range. While in this particular image the target has high signal strength relative to the local background, there are other regions of higher sea-surface backscatter in the image against which the target would not be so easily detectable. There is also an acoustic shadow due to the ship s hull on the port (left) side of the image at ranges less than 200 m. Figure 6: Forward-looking HF sonar image to 600 m range, 1711Z April 9 in Saanich Inlet. Wind speed = 2.5 knots. Range rings are in 120 m increments. Arrow denotes location of big target sphere, suspended at 4 m depth. Results of the target detection and background extraction process are shown in Figure 7. There are 490 samples spanning ranges from 49 m to 572 m. The most obvious characteristic of the data is strong fluctuation in both the target and background estimates. There is also a very high SNR (expressed in db), varying between 10 db (at long ranges) to more than 30 db in some regions. The overall average SNR was 22.4 db. The minimum SNR was approximately 10 db, which confirms a commonly-held assumption of the active detection threshold of 10 db (more on this in Section 5). The simple target detection process used herein became very unstable when the SNR dropped below 10 db. The target intensity roughly follows the expected -20*log 10 [r] variation, however there are significant variations which are hypothesized to be due to acoustic convergence and shadow zones created by reflection and refraction near the sea-surface. This will be investigated later using acoustic models (Section 5). The background approaches the sonar self-noise level (which increases due to the TVG) at ranges beyond 500 m. The target echoes also reach the sonar analog-to-digital clipping limit near 66 db (sonar units) in a few places. DRDC-RDDC-2016-R053 11

22 Figure 7: Comparison of target echo and background intensity (db re sonar units) vs. range for approach run against the big sphere starting 1710Z April 9 in Saanich Inlet. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping and noise thresholds. Figure 8: Apparent Target Strength (db re m 2 ) vs. range for approach run against the big sphere starting 1710Z April 9 in Saanich Inlet. Dashed line shows sonar clipping threshold. Figure 8 shows the conversion of raw sonar amplitude into Apparent TS (ATS). In the absence of non-spherical propagation loss effects this should yield a constant value with range (ignoring fluctuations). However, the ATS clearly shows an increase with range in the nearest 270 m, with several peaks and drop-outs underlying the short-spatial scale fluctuations. The overall average ATS is -8.0 db. It is hypothesized that the true sphere TS (near -14 db) can be seen from m range (more on this in Section 3.1.4). 12 DRDC-RDDC-2016-R053

23 3.1.2 Strait of Georgia at Sea-State 3 Figure 9 shows a target detected under somewhat higher sea-state conditions. This example is taken from the Strait of Georgia on April 11, under winds of 12 knots. This environmental condition is just at the threshold for breaking wave activity [1], so sub-surface bubble layers are expected to be weak or negligible. In this example the big sphere target at 4 m depth is located at 275 m range; it was first detected at a maximum range of 390 m. Under these conditions the sea-surface reverberation was confined to a zone up to roughly 180 m from the sonar, allowing target detection beyond. As the ship approached the target it became more difficult to detect the target inside this shorter-range zone of higher surface reverberation. This existence of this surface reverberation maximum can be attributed to the upward-refracting acoustic propagation at this location. Note also a distinct acoustic shadow on the left (port) side of the image up to 160 m range created by the ship s hull. Figure 9: Forward-looking HF sonar image to 400 m range, 1735Z April 11 in Strait of Georgia. Wind speed = 12 knots. Range rings are in 80 m increments. Arrow denotes location of large target sphere, suspended at 4 m depth. Figure 10 shows the corresponding target and background intensities extracted as the ship approached from over 350 m range to a CPA near 25 m. Similar to the previous example the SNR are relatively high (overall average 20.1 db with standard deviation of 5.0 db). The target echo generally follows the -20log 10 [r] trend line except for a clear drop-out between 200 m and 250 m range. This drop-out in target echo strength beyond 200 m coincides with a clear decrease in the background reverberation outward of its peak at 200 m (see Figure 9). This drop-out in target echo and background is likely due to a near-surface shadow zone starting at 200 m range. The background shows a strong reverberation peak near 200 m, suggesting the sonar beam is strongly focused on the sea-surface at this range, but is otherwise of similar level to that seen in Saanich Inlet (Figure 7). The target echo also shows some clipping at ranges less than 50 m. DRDC-RDDC-2016-R053 13

24 Figure 10: Comparison of target echo and background intensity (db re sonar units) vs. range for run against the big sphere starting 1734Z April 11 in Strait of Georgia. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping and noise thresholds Strait of Georgia at Sea-State 4 Figure 11 shows the corresponding sonar image taken under the highest wind speeds in the sea-trials, near 16 knots, in the Strait of Georgia. On April 12 the target spheres were suspended at 1 m depth, expected to enhance the impact of near-surface bubbles on sonar performance. At this wind speed, bubble layers or clouds were expected to exhibit exponentially decreasing densities with depth scales of 0.5 to 0.8 m [1]. At wind speeds near 16 knots detection ranges were generally limited to less than 250 m. A key characteristic of this location was the patchiness of the observed background reverberation. These patches appear to have spatial dimensions approximately m. In this snapshot the target was readily detected in-between patches of relatively intense reverberation. If the target had been located inside one of the reverberation patches it would have been much more difficult to detect. The background reverberation reaches a consistent peak between 60 m and 100 m range; this can be attributed to the upward-refracting conditions focusing the sonar beam on the surface. Once again the acoustic shadow extending up to 60 m range on the port side of the ship is clear. 14 DRDC-RDDC-2016-R053

25 Figure 11: Forward-looking HF sonar image to 200 m range, 2108Z April 12 in Strait of Georgia. Wind speed = 16 knots. Range rings are in 40 m increments. Arrow denotes location of large target sphere suspended at 1 m depth. Figure 12 shows the corresponding target echo and background intensity measurements. A key feature here is a drop-out in target detection between 90 m and 110 m range due to a local maximum in the sea-surface reverberation. The target echo also shows strong ( ±10 db) variability with a spatial scale of approximately m. Visual inspection of the sequence of snapshots suggests that these are due to surface waves, which can be seen to move through the images. Finally, beyond 110 m range the target echo drops below the -20log 10 [r] trendline, likely due to absorption of the sonar signal by near-surface bubbles at shorter ranges. The target echoes showed signs of clipping at ranges less than 80 m. DRDC-RDDC-2016-R053 15

26 Figure 12: Comparison of target echo and background intensity (db re sonar units) vs. range for run against the big sphere starting 2108Z April 12 in the Strait of Georgia. Dashed lines show expected -20log 10 [r] variation in target echo and sonar clipping threshold Summary Results The maximum detection range under the varying conditions is one of the most obvious parameters to calculate. Based on previous investigations [1, 2] it was hypothesized that this would be strongly controlled by the wind speed. These maximum detection ranges were found as a by-product of the target echo vs. background investigations (i.e., as shown in Figures 7, 10, and 12) conducted for every forward-looking sonar run against the two sphere targets. These are summarized in Figure 13. While there is a clear trend of decreasing detection range under increasing winds, it is not a simple one. The overall maximum detection range is near 575 m, but some scatter is present at ranges m at low wind speeds (< 4 knots). A threshold for wave breaking above 6 knots was chosen to separate the wind-induced effect. A best-fit line (intercept = 17.0 knots, slope = knot/m, correlation coefficient r 2 = 0.455) is shown, however there is significant scatter in the data. This analysis suggests that these targets would not be detectable at wind speeds higher than 17 knots. Unfortunately the present data set does not cover the higher sea-state conditions required to confirm this assertion. At wind speeds between 8 and 10 knots the clear difference between the Saanich Inlet and Strait of Georgia locations also highlights the importance of local acoustic propagation effects. 16 DRDC-RDDC-2016-R053

27 Figure 13: Summary of maximum detection ranges from all forward-looking sonar tests with big and small spheres in Saanich Inlet and Strait of Georgia, April Best fit line to data where wind > 6 knots is shown. Using the calibration results it is desirable to estimate the true acoustic target strength (TS) of the sphere targets. However, it was clear in all the data sets that the Apparent TS (ATS) generally varied with range due to acoustic propagation effects (example shown in Figure 8 above, and to be discussed in Section 3.2 below). The ATS usually showed a roughly constant value, assumed to be an estimate of the true TS, at ranges less than approximately 100 m. However, at very close ranges (< 50 m) the sonar data frequently showed clipping (sonar raw intensity > 66 db). Thus, averaged ATS were computed over to range interval 50 to 100 m for each of the various approach runs. It was found that there was no significant difference in these true TS estimates between the two locations and at varying wind speeds. The overall averaged TS estimates were thus (-14.0 ±1.4) db and (-14.4 ±1.5) db for the big and small spheres, respectively. These estimates fall in-between the geometric TS value for the two spheres (see [3]) of -15 db and db. The fact that these two in situ TS estimates are similar in spite of the difference in sphere diameter (71 cm vs. 91 cm) suggests that the chain ballast made a contribution to the overall TS. The background reverberation showed variations up to 10 db between locations due to the combined effects of wind speed variations and local acoustic propagation effects. Figure 14 shows a comparison of averaged background Reverberation Level (RL) between the three example locations. These curves were averaged in angle over a 10 sector near the center of the sonar aperture and in time over 200 pings. Sonar self-noise was only a minor contributor at ranges greater than 500 m (i.e., only for the April 9 Saanich Inlet case). Short range (< 20 m) peaks in all data are due to the sonar beam-pattern and interference from ship s hull echoes. The low wind speed Saanich Inlet case shows the lowest overall RL. The two higher wind speed cases in the Strait of Georgia show peaks and drop-outs attributed to acoustic propagation effects. The fact that the higher wind speed cases (2108Z April 12) shows slightly lower RL (than the 12 knot case) beyond 100 m is attributable to bubble layer saturation effects, whereby (at higher wind speeds) bubbles at shorter ranges attenuate echoes from bubbles at longer ranges. DRDC-RDDC-2016-R053 17

28 Figure 14: Comparison of averaged Reverberation Level (db re 1 µpa) vs. range for runs against the big sphere in Saanich Inlet and Strait of Georgia on April 9, 11, and Transmission Loss Effects Quantitative analysis of the data uncovered plentiful evidence for non-spherical acoustic propagation effects. This was seen as large (up to 15 db) anomalies in the apparent TS, both positive and negative. Generally these were attributed to the combined effects of surface reflections, refraction by near-surface sound speed gradients, and acoustic extinction due to bubbles. Figure 15 shows a comparison between three separate instances of ATS for the big sphere. The key differences between these environments are wind speed (2.5, 12, and 16 knots) and sound speed profiles (upward vs. downward refracting, see Figure 2). The April 9 data show strong positive ATS anomalies, enhancing the target detectability both in terms of SNR and maximum detection range. The April 11 data show a middle case, with both positive and negative apparent TS anomalies. The highest wind speed case (April 12) shows much lower ATS, which is attributed to acoustic extinction by the bubble layer, limiting detection the range to 190 m. All three data sets show significant variability (even after adjacent-averaging) with peaks and troughs approximately ±3 db and spatial scale approximately m. It is speculated that these peaks and drop-outs in ATS are due to acoustic interference effects between the direct and surface-reflected acoustic paths (as sketched in Figure 3). 18 DRDC-RDDC-2016-R053

29 Figure 15: Comparison of apparent TS (db re m 2 ) vs. range for big sphere on April 9, 11, and 12 under wind speeds of 2.5, 12, and 16 knots, respectively. Data are 10-pt. adjacent averaged. Dashed line shows estimated true TS of db. Target depths were 4 m on April 9 and 11 and 1 m on April Variability of Target Echoes One of the most obvious characteristics of these sonar trials was the strong ping-to-ping variability in both the target echoes and background reverberation. Assuming that calculation of ATS removes the dominant geometric trends with range, it is useful to consider the statistical characteristics of the ATS over the entire approach run. Note that using the data in this manner includes both temporal variations, for example due to surface wave effects, and spatial variations due to the decreasing range to target. A probability distribution of the target ATS from the April 9 data (Figure 8) was calculated, as shown in Figure 16. In this case the ATS data appear to be Gaussian distributed, with a width (2 x standard-deviation) of 14.8 db. Analysis reported by Dahl and Plant [6] using a fixed geometry concluded that a Gaussian distribution is appropriate for logarithmic variables (such as ATS) scattered near the sea-surface. Their levels of variance were somewhat smaller (standard deviation of 5.6 db) because their analysis only included temporal variations in sea-surface scattering. The fact that the Gaussian-fit mean value (-8.1 db) is larger than the estimated true TS of the sphere (-14.3 db, see Section 3.1.4) again suggests that acoustic propagation effects are (on average) enhancing the target echoes, improving the sonar detection performance. DRDC-RDDC-2016-R053 19

30 Figure 16: Frequency distribution of apparent TS over all ranges for detection run against the big sphere starting 1710Z April 9 in Saanich Inlet. Best fit Gaussian distribution (mean = -8.1 db, width = 14.8 db, r 2 = 0.926) shown in red line. 3.4 Propeller Cavitation Effects While the ship was traveling at constant speed there appeared to be no interference from propeller cavitation. The maximum ship speed reached during the forward-looking target approach runs was less than 5 knots, so propeller cavitation effects were not expected. However, during periods when the ship accelerated, particularly at speeds above 6 knots, strong interference effects attributed to propeller cavitation were observed. Figure 17 shows an example. Compare the background level with similar sonar setup in Figure 6. In this example the ship was accelerating from 7.5 to 8 knots during a self-noise test in Saanich Inlet on April 9. The signature of propeller cavitation noise appears as a strong increase in background noise at longer ranges (due to amplification by the sonar TVG) with some modulation due to the propeller blades. This modulation appears as bands roughly 40 m wide with a peak to peak spacing of approximately 80 m. This spacing in range corresponds to a period of 0.11 s. This cavitation noise has the potential to completely mask target echoes at greater ranges. A quantitative comparison of background levels between low speed (3 to 5 knots) and this example found an approximate 15 db increase in background noise levels at ranges beyond 300 m. The CCGS Vector has a single, three-bladed, variable-pitch propeller with diameter of 1.8 m. At this speed the propeller RPM was near 180, corresponding to a blade rate of 9 Hz (period of 0.11 s). This supports the hypothesis that the cavitation noise interference has a blade rate modulation. It is believed that at higher ship speeds (> 8 knots) this cavitation interference will pose a strong limitation on HF sonar performance. Unfortunately during this trial it was impossible to assess performance at higher ship speeds due to safety limits on use of the sonar strut. 20 DRDC-RDDC-2016-R053

31 Figure 17: Forward-looking HF sonar image to 600 m range, 2118Z April 9 in Saanich Inlet during period of ship acceleration from 7.5 to 8 knots. Range rings are in 120 m increments. Arrows denote bands believed to be blade rate modulations. DRDC-RDDC-2016-R053 21

32 4 Aft-Looking Sonar HF sonar performance in the ship wake was specifically assessed in a series of tests on April 13 in Saanich Inlet. In all these tests the winds were light (< 3 knots) and seas were calm. The wake was the dominant acoustic feature in these tests. Tests were conducted at ship speeds from 6 to 7.5 knots, this being the upper limit of safe operation of the sonar strut. It should be noted that this speed generated only a relatively modest ship wake. Greater wake width, depth, and densities have been previously observed with this same ship at speeds up to 12 knots [7]. Based on a previous study on the wake of CCGS VECTOR [7], the expected maximum wake width at this speed was approx. 60 m, with the wake core deepening over the first 180 s up to a maximum of approximately 6 m. At the ship s transom the wake core depth can be assumed to be near 3 m. In that earlier study the wake was observed to persist as a distinct acoustic feature for periods greater than 6 minutes at low speeds (5 knots). This increased to over 10 minutes for higher ship speeds. This wake lifetime is much larger than the equivalent sonar ranges explored in these tests. Given that the Sharko tow depths were between 5 m and 24 m (greater depth at greater range), then the target was usually below the wake core depth. This geometry implies that wake-bubbles should not cause any acoustic extinction of the direct-path target echo, but may cause some extinction of the surface-reflected paths. The vertical beamwidth of the sonar (13.5 ) ensures that echoes from both the target and wake bubbles will be combined, so that wake backscatter sets a threshold for target detection. 4.1 Sonar Data Examples The Sharko towed target depth varied with the amount of tow-cable scope, as shown in Figure 18. The Sharko depth generally varied between 5 m and 24 m at ranges from m. The target depth was controlled by the combination of the tow-rope shape and drag and the Sharko weight relative to the water, and depends strongly on the towing speed relative to the water. The target pull-ins showed a slightly shallower approach due to the fact that the target speed relative to the water was 3.9 knots higher than the ship speed. Figure 18: Summary of Sharko depths during aft-looking HF sonar tests April 13 in Saanich Inlet. Black lines show approximate -3 db limits of sonar vertical beam. 22 DRDC-RDDC-2016-R053

33 4.1.1 Fixed-Range Towing Fixed-range towing tests were conducted at ranges from 100 m to 450 m. Beyond 450 m range the target was very difficult to detect visually. Sonar data was collected over a 5 minute period at each range to assess target echo variability. Figure 19 shows an example with Sharko towed at 150 m range. Recall that the sonar was mounted with a 9 counter-clockwise offset, so that the wake appears to have clockwise rotation. This presented no difficulties in processing, so corrections were not applied. Figure 19: Aft-looking HF sonar image to 200 m range, 1828Z April 13 in Saanich Inlet. Range rings are 40 m increments. Arrow denotes location of Sharko target, at 16 m depth. The general characteristic of the wakes observed in these tests was that they were not homogeneous, but rather consisted of isolated scattering patches of dimension up to 10 m. In the sonar frame of reference these patches appeared to be advected away from the sonar at the ship speed. The Sharko target echo was then identifiable as a stationary feature against this moving background. Note that the apparent intensity of the Sharko target is similar to that of the scattering patches, making it somewhat more difficult to detect the target in the near-field (ranges < 60 m) where the density of scattering patches is much higher. The fact that the wake apparently decreases in intensity at greater range is hypothesized to be a result of acoustic extinction within the wake, rather than any dissipation of the wake itself. The maximum wake age (time after wake generation at the ship stern) in this image is 65 s, which is considerably shorter than the previously measured wake persistence ( 360 s, see [7]). The maximum wake age assessed in these fixed-range towing trials (ranges up to 450 m) was near 120 s. Following a similar target following algorithm as employed in the forward-looking tests, estimates of the target echo and background intensity were extracted on a ping-by-ping basis from the sonar snapshots. A typical result spanning 220 s is shown in Figure 20. The figure shows that both target and background sonar intensities exhibited roughly constant average values overlain DRDC-RDDC-2016-R053 23

34 with ping-to-ping fluctuations. The level of short-period fluctuation in this example is approximately ±2 db. The mean SNR is 22 db, which is very good. Target echo fluctuations will be discussed further in Section 4.3. Figure 20: Temporal variation in target echo and background intensity for stationary tow at 220 m range in ship wake starting 1618Z April 13 in Saanich Inlet. Ship speed during this period was 7.2 knots. Figure 21: Apparent target strength (db re m 2, black squares) and SNR (db, red circles) vs. range from all fixed-range towing tests, April 13 in Saanich Inlet. Figure 21 shows a compilation of Apparent TS and SNR from all the fixed-range towing tests. The ATS results show the mean ±1 standard deviation averaged over a period of 3 to 5 minutes. The ATS results show a 5 db increase as range increases from 100 to 450 m, attributable to 24 DRDC-RDDC-2016-R053

35 acoustic propagation enhancements due to acoustic refraction and inclusion of surface-reflected multi-paths. Also, the level of fluctuation (size of error bars) generally increases with range. Generally the averaged SNR values were very good (> 20 db) at ranges up to 350 m, falling to 13 db by 450 m. The ATS estimate at 100 m range shows a somewhat lower mean value and a much larger variability due to variations in depth (12 16 m) during the run, which put it on the edge of the sonar main beam (recall that sonar beam-pattern effects are not compensated for in the ATS value). Disregarding the anomalous data point at 100 m range, the fixed-range results suggest a minimum ATS for the Sharko target near -20 db (re 1 m 2 ) Target Pull-Ins The second type of target detection test involved pulling in Sharko at 2 m/s (3.9 knots) relative to the ship from a start range near 210 m. Six separate pull-in runs were conducted at ship speeds between 6 and 7.5 knots. A 200 m sonar range setting was used for all the pull-in tests. Sharko depths during the pull-in runs are shown in Figure 18. Apart from ping-to-ping variability (similar to that observed in the fixed-range tests), the separate pull-in runs were all essentially the same in terms of target echo strength and SNR. Thus the results were averaged over all six runs. This was done by averaging the target and background intensities within 5 m range bins from m. Averaged pull-in results are shown in Figure 22. This shows relatively smooth variations in target echo and background intensity with range, approximately following -20 and -30*log[r] variations, respectively. The -20*log[r] variation in target echo suggests a constant ATS value. The -30*log[r] variation in background intensity is created by the linear increase in sonar sampling volume with range. Overall the SNR are good, increasing with range from 14 to 24 db. Figure 22: Comparison of averaged target, background, and SNR vs. range from six separate pull-in runs in arbitrary sonar units. Dashed lines show -20log 10 [r] and -30log 10 [r] variations. DRDC-RDDC-2016-R053 25

36 The averaged target echo intensity can be converted to the Apparent TS, shown in Figure 23. While removing the dominant geometric spreading effects, the ATS profile shows that there is still some target echo variability even after averaging over multiple runs. The best-fit line shows a small range-dependence. Recall that the ATS parameter is not corrected for acoustic propagation effects nor for sonar vertical beam-pattern. Acoustic propagation effects should diminish at short ranges. Based on the known target depths during the pull-in runs the sonar Beam-Deviation Loss (BDL, db, see Eq.(3)) can be calculated. The two-way BDL for the target depths encountered in these tests increases from 0 to 2 db over ranges from 20 m to 200 m. Thus, the minor decrease in ATS can be explained by sonar beam-pattern effects. Also, since the BDL at short ranges is effectively zero, then the intercept of the best-fit line (-19 db re m 2 ) can be taken as an estimate of the true TS of the Sharko target (at bow incidence). This is consistent (within experimental uncertainty) with the results from the fixed-range towing (near -20 db, see Figure 21). Figure 23: Apparent target strength (db re m 2 ) estimated from average of six separate pull-in runs. Red line shows best fit line (intercept = db, slope = db/m) Target Detection During S-Turns A separate fixed-range target detection test was conducted while the ship conducted a series of S-turns (through roughly ±45 relative to the mean heading) at speeds of 6.5 to 7 knots. Sharko was towed at a fixed distance of 210 m behind the ship at a depth near 16 m. The ship turns forced the Sharko target to tow outside of wake, generally on the inside of the turn, with frequent wake crossings. Figure 24 shows an example of the sonar data. Generally the signal to noise properties were at least as good as in previous straight-line tests, although the wake itself appears to be a stronger target than in previous constant-heading runs (contrast wake strength compared to Figure 17). It is expected that ship wakes are stronger in turns. In addition to forcing Sharko outside of the wake, the incidence angle of the sonar beam relative to Sharko was also forced away from directly on the bow (as it was in the previous tests). It was speculated that, due to its construction and presence of two acoustic target spheres within the main body, Sharko would show variation in TS with incidence angle, generally increasing in a 26 DRDC-RDDC-2016-R053

37 complicated way towards a broadside maximum. This effect is evident in Figure 25 (Apparent TS) by large positive excursions in ATS at various points through this series of ship turns. While there are similar ping-to-ping fluctuations as previously observed, these positive excursions are far larger than previous ATS estimates. Unfortunately, no simple relationship between Sharko position, wake locations, and ATS could be deduced from this test. Figure 24: Aft-looking HF sonar image to 300 m range during S-Turn, 1626Z April 13 in Saanich Inlet. Range rings are 60 m increments. Arrow denotes location of Sharko target. DRDC-RDDC-2016-R053 27

38 Figure 25: Apparent target strength (db re m 2 ) vs. time following Sharko towed target at 210 m range through series of S-Turns, 1624Z-1630Z April 13 in Saanich Inlet. 4.2 Wake Backscatter and Extinction It was hypothesized in earlier work that bubbles in the ship wake would both (a) generate increased backscattered signal relative to the ambient conditions and (b) absorb and scatter the sonar signal as it passes through (a process collectively known as extinction). The first effect can be quantified through a comparison between the averaged background reverberation levels vs. range both inside and outside of the wake, as shown in Figure 26. This shows an increased backscatter level within the wake of between 5 and 15 db, varying with range from a maximum near 40 m to a minimum near 200 m. The result is nearly identical at the two speeds investigated. 28 DRDC-RDDC-2016-R053

39 Figure 26: Comparison of in-wake and non-wake reverberation level vs. range profiles at two speeds, taken near 1800Z April 13 in Saanich Inlet. While there is clear evidence for the increasing background levels inside the wake, there is no solid evidence for wake extinction effects on the target echo strength. For example with the Apparent TS measurements in the fixed-range towing tests (Figure 21), the ATS was observed to generally increase with range. If acoustic extinction effects due to bubbles were present then a general decrease in ATS should have been observed. Bubble acoustic extinction losses can exceed 0.1 db/m (see analysis in [1, 2]). The relatively small decrease in ATS with increasing range observed during the target pull-in runs (Figure 23) was adequately explained by the sonar vertical beam-pattern effects. The reason for this lack of acoustic extinction effect in the target echo is simply that the target depth (between 6 and 25 m, see Figure 18) placed it below the wake core (expected to deepen with distance behind the ship from roughly 3 to 6 m see [7]). The direct-path from sonar to target was thus unaffected by wake bubbles. Furthermore, for deeper targets the surface-reflected acoustic paths have somewhat higher grazing angles, of the order 10, thus traveling shorter distances through the bubble layers and hence suffering only minor extinction losses. Note that this conclusion would not necessarily hold for much deeper wakes, generated for example by larger ships operating at higher speeds. 4.3 Target Echo Variability In both the fixed-range and pull-in target detection runs significant ping-to-ping variability was observed. It is hypothesized that this variability is due to acoustic interference between the direct and surface-reflected acoustic paths, inevitably randomized due to small scale fluctuations in acoustic path length caused by surface roughness and wake turbulence. In the fixed-range tests the target depth and bearing were approximately constant, thereby allowing an assessment of the statistical properties of these fluctuations on a temporal basis alone. Figure 27 shows a typical example. The distribution of ATS over this period is well-matched by a Gaussian distribution, but with a considerably smaller width (4 db) compared to the forward-looking tests (near 15 db, see Figure 16). The Gaussian-distributed character of this result is in agreement with finding in Dahl DRDC-RDDC-2016-R053 29

40 and Plant [6], but now with a much smaller width. A possible explanation for this smaller width is that the interfering surface-reflected path is slightly reduced in amplitude by bubble extinction effects, allowing only partial constructive or destructive interference with the direct path. Other fixed-range towing tests at different ranges showed Gaussian widths varying from 4 to 10 db, the larger values coming from tests showing some variation in target depth. Figure 27: Distributions of apparent target strength during fixed-range towing at 220 m range 1618Z April 13. Best fit Gaussian distribution (mean = db, width = 4.0 db, r 2 = 0.994) shown in red line. 30 DRDC-RDDC-2016-R053

41 5 Modeling HF Sonar Performance Sonar performance models potentially allow a more general prediction of the environmental limitations faced by a sonar; however these predictions need to validated against at-sea data. Modeling sonar performance must necessarily include the physics of acoustic propagation, including boundary scattering and internal refraction, combined with a variety of system and environmental parameters. In this work it was expected, both from previous experience and from the initial sea-trials results, that both sonar electronic (self-) noise and sea-surface reverberation will pose limitations to target detection with this sonar. This work will utilize the active sonar Signal Excess (SE, in db) for the target as a function of range. In this work SE is defined as the excess of the target echo SNR over a Detection Threshold (DT) based on signal-processing considerations. The target is said to be detectable at ranges where SE > 0 db. SNR is the ratio of target echo Sound Pressure Level (SPL) to the combination of the sonar Noise Level (NL) and Reverberation Level (RL) due to backscatter from the sea surface. The NL is assumed to be dominated by internal electronic noise in the sonar system. Although ambient noise can be included in the model (based on wind speed), this was found to be negligible compared to sonar self-noise at this frequency (90 khz). Volumetric backscatter, such as due to zooplankton, will also be ignored. The effects of volume scattering from sub-surface bubbles, however, cannot be ignored. The fundamental model for SE uses the active sonar equation (in db), i.e., SE = SL + TS - 2 TL - (RL NL) - DT, (5) where SL is the sonar Source Level, TS is the Target Strength, and TL is the Transmission Loss from the source to target. This CW sonar is assumed to have no active processing gain. The reverberation level is calculated based in part on the sonar beamwidths and pulse length. The denotes a power sum between the two terms. Using the SE or SNR values, the sonar Probability of Detection (Pd) can be calculated using models outlined in Trevorrow and Myers [2]. In active monostatic sonar it is usual to assume a DT = 10 db (which is in agreement with results from Sections 3 and 4). The calculation also requires a probability of false alarm, which is set at 0.1%. With these parameters a SE of 0 db corresponds to a Pd = 50%. The range where SE consistently drops below 0 db (or equivalently Pd drops below 50%), herein denoted the detection range, will be taken as a Measure of Performance (MoP) for comparison between environments. The effects of different oceanic environments enter through the acoustic transmission loss and surface and seabed reverberation terms. The NATO acoustic modeling suite ESPRESSO (Extensible Performance and Evaluation Suite for Sonar) [8], version 1.05 (November 2008) was used to perform the basic acoustic propagation, scattering and noise modeling needed to produce the results shown below. The tool was developed by the NATO Centre for Maritime Research and Experimentation in La Spezia, Italy. The software implements a number of standard models for sonar performance prediction. A modified version of BELLHOP, a public-domain ray-tracing model developed by Porter [9], was chosen as the propagation sub-model. This was implemented DRDC-RDDC-2016-R053 31

42 in ESPRESSO as described in [10] for bottom reflection and backscatter, surface reflection and backscatter, and loss due to near-surface bubbles were developed by the University of Washington Applied Physics Laboratory (APL-UW), described in [11]. Most of the APL-UW models have been validated for the khz range, and some have been validated up to 100 khz. Seawater absorption loss at the sonar operating frequency (following [5]) is also included. ESPRESSO does not model sea-surface height variations due to surface waves. In ESPRESSO the environmental inputs consist of a sound speed profile, water depth, seabed properties, and a surface sea-state controlled by wind speed. The water depth, sediment type, and sound speed profiles are all assumed to be range-independent. The various sonar input parameters for modeling the SM-2000 are summarized in Table 2. The assumed TS values were taken from in situ estimates for the target spheres (-14 db) and Sharko (-19 db) targets, as described in previous sections. The outputs include reverberation levels, SE vs. range and depth, and detection probabilities. Generally the predictions were generated at 1 m intervals out to 600 m range and 0.5 m intervals in target depth up to 20 m. Generally the ESPRESSO model showed that the sonar detection performance was limited by sea-surface reverberation at shorter ranges, switching to noise-limited at longer ranges. This is in general agreement with the experimental results discussion above. The sound speed profile was found to by a highly important factor, in some cases creating acoustic focusing and shadow zones that controlled target detection almost independent of wind speed. Parameter Table 2: ESPRESSO input parameters for the SM-2000 sonar models. Operating frequency Pulse length Source Level Sonar Depth Sonar elevation angle Value 90 khz 1.0 ms 205 db re µpa at 1 m 3.5 m Vertical beamwidth (both transmit and receive) 13.5 Receiver horizontal beamwidth 1.8 Receiver bandwidth Receiver spectral self-noise level Receiver Directivity Index (DI) Water depth 0 (horizontal) 800 Hz 50 db re µpa/hz 30 db 200 m Seabed sediment type silty-clay Target backscatter Target Strength (TS) -14, -19 db re m 2 The sound speed profiles were distinctly different in Saanich Inlet as compared to the Strait of Georgia, as shown in Figure 2. Generally the Saanich Inlet conditions were downward-refracting, whereas the Strait of Georgia was strongly upward-refracting. Note also day-to-day variability in the upper 20 m in Saanich Inlet; it is suspected that this also indicates some degree of spatial variation along the inlet on any given day. Unfortunately the spatial variation in near-surface conditions was not measured during the sea-trials. In this modeling, an average profile for 32 DRDC-RDDC-2016-R053

43 Saanich Inlet on April 9 and 10 was utilized for assessing the forward-looking tests. This was done because the target detection results (e.g., Figure 13) showed no significant difference between the two days. The aft-looking assessment used the April 13 profile (shown in Figure 2). Figures 28 and 29 compare the predicted SE between Saanich Inlet and the Strait of Georgia conditions. Note a difference in wind speed (5 vs. 10 knots) which is appropriate for the two locations. Recall that SE > 0 db defines the regions of target detectability, both of which are far from uniform in depth. The Saanich Inlet conditions feature a near-surface (4 to 5 m depth) sound channel that strongly enhances target detectability out to ranges of 600 m and beyond. Otherwise the Saanich Inlet prediction features a downward-refracted detection zone and a near-surface (1 3 m) shadow zone that prevents target detection beyond a range of approximately 300 m. Because the target detectability is controlled by the sound speed profile, these results are relatively robust with respect to small changes in wind speed. This result explains the observed sonar detection ranges up to 600 m range against target spheres at 4 m depth. Figure 28: Predicted sonar SE vs. target range and depth for the April 9/10 Saanich Inlet conditions, with assumed wind speed of 5 knots. Target TS = -14 db. Figure 29: Predicted sonar SE vs. target range and depth under April 11 Strait of Georgia conditions, with assumed wind speed of 10 knots. Target TS = -14 db. DRDC-RDDC-2016-R053 33

44 The Strait of Georgia SE prediction, shown in Figure 29, shows distinctly different behaviour due to the upward-refracting sound speed profile. The SE values are generally lower throughout due to the higher sea-surface reverberation level due to the higher wind speed. The effect of increasing wind speed is to reduce the SE values throughout. Note that in the uppermost 2 m the target detectability drops away with range; this is due to bubble-induced target echo extinction losses. For targets at 1 m and 4 m depth the region of target detectability (SE > 0) is broken into two zones (0 to 200 m and a narrow patch near 350 m) with a shadow zone in between. In this Strait of Georgia environment it is predicted that targets will not be detectable beyond 370 m range at any depth. The maximum observed sonar detection range in the Strait of Georgia was near 400 m under winds near 10 knots (see Figure 13). Figure 30: Predicted sonar Pd vs. range at 4 assumed wind speeds for a 1 m deep, -14 db target under Strait of Georgia conditions. An example of the conversion of SE to Pd is shown in Figure 30 for the Strait of Georgia conditions. The two zones of target detectability are clear. Using the Pd = 50% MoP, the effect of increasing wind speed is to dramatically reduce the detection range. The predicted target detection range in Figure 30 drops from 430 m to 50 m as the wind speed increases from 5 to 20 knots. Recall in the sea-trials that the maximum detection range dropped to below 200 m as the wind speed reached approximately 15 knots. The conditions and target depths were distinctly different for the aft-looking wake tests on April 13 in Saanich Inlet. In this case the winds were light and the Sharko target was towed at depths of 5 24 m. This analysis ignores the effects of the ship wake, which cannot be easily modeled in ESPRESSO. Figure 31 shows the predicted SE vs. range and depth under these conditions. For deeper targets the predicted detectability zone extends to greater horizontal range, which matches with the towing profile of the Sharko target which is deeper at greater range. The predicted detection range for the Sharko target at 20 m depth is near 440 m. This is in good agreement with the experimental result of maximum detection near 450 m (see Section 4.1.1). Furthermore in the inner 300 m the SE values along the target depth profile are all in excess of 10 db. This result is in agreement with the relatively high SNR observed in the sea-trials. Figure 31 predicts relatively good sonar performance in the upper 5 m at ranges up 300 m. This is considered erroneous due to bubble-induced acoustic extinction effects within the ship wake. 34 DRDC-RDDC-2016-R053

45 During the sea-trials it was not possible to explore this range-depth zone due to the towing characteristics of Sharko. It is hypothesized that target detection would not be possible at depths inside the wake core. Figure 31: Predicted sonar SE vs. target range and depth under April 13 Saanich Inlet conditions, with assumed wind speed of 2 knots. Target TS = -19 db. Black line shows approximate target depth vs range. Figure 32: Comparison between measured wake RL vs. range, taken 1758Z April 13 in Saanich Inlet, with ESPRESSO prediction at wind speeds of 10, 15, and 20 knots. Moreover, ESPRESSO does not provide bubble scattering models appropriate for ship wakes, i.e., it is not possible to fake the presence of a wake by using a larger wind speed. This is shown in Figure 32 through a comparison of Reverberation Level (RL). The ESPRESSO predictions, based on the APL-UW models for wind-driven seas [11], show the effects of bubble extinction in two ways: (i) a decrease in longer range RL to a saturation value as the wind speed increases to 15 knots and above, and (ii) an initial increase then a decrease in the peak RL value near 20 m range. In comparison with the sea-trial data it is clear that while ESPRESSO does predict the peak values (ranges m) reasonably well, it does not produce the same longer range dependence in the RL. This can be attributed to the use of a bubble model that is not appropriate for wakes. DRDC-RDDC-2016-R053 35

46 The result is that ESPRESSO predictions for wakes based on a fictional wind speed will, outside of a near-field zone, over-predict the RL and hence under-predict the sonar performance. Using the Pd = 50% MoP, the predicted detection range for various target depths in the different locations can be calculated, as summarized in Table 3. These predictions are all in reasonable agreement with the sea-trial results. The Saanich Inlet Apr 9/10 predictions show the most extreme difference between 1 and 4 m depth, due to the refractive sound channel centered at 4 5 m depth. Recall that the Saanich Inlet tests on Apr 9/10 were all conducted with the target spheres at 4 m depth, so there is no sea-trial data to verify the 1 m prediction. The sea-trial results in Saanich Inlet under light winds (under 6 knots, see Figure 13) produced an average detection range of 504 ±45 m, which is in reasonable agreement with this prediction. Table 3: ESPRESSO predictions of SM-2000 sonar detection ranges. Location/Date TS target depth wind = 5 knots wind = 10 knots wind = 15 knots wind = 20 knots Saanich Inlet, Apr. 09/10-14 db 1 m 4 m 309 m 626 m 213 m 626 m 98 m 619 m 50 m 556 m Strait of Georgia, Apr. 11/12-14 db 1 m 4 m 431 m 410 m 247 m 246 m 133 m 238 m 54 m 238 m Saanich Inlet, Apr db 10 m 15 m 20 m 322 m 422 m 441 m 305 m 421 m 440 m 316 m 421 m 440 m 305 m 417 m 440 m The ESPRESSO predictions for the Strait of Georgia conditions (wind speeds 10 to 16 knots) suggest detection ranges between 130 m and 250 m, with improved performance for a 4 m target. This is somewhat pessimistic as compared to the sea-trial results. In the Strait of Georgia sea-trials the target spheres were deployed at 4 m depth on April 11 and at 1 m depth on April 12. Detection ranges up to 400 m under wind speeds near 10 knots were observed on both days, and on April 12 the targets were detected up to 340 m range under wind speeds near 15 knots (see Figure 13). A contributing factor is that the target echo included some contribution from the chain ballast suspended approximately 2 m below the sphere. This would have improved target echoes in the 1 m depth cases (April 12). Another potential explanation for this discrepancy is that the Strait of Georgia sound speed measurement (reported in Figure 2) was conducted approximately 5 km to the north-east of the sonar test location; some spatial variation in the near-surface sound speed profile may have been present. This highlights the importance of conducting in situ environmental measurements at the same location and time as the sonar tests. Table 3 shows that the predicted detection ranges for the Sharko target in the aft-looking under-wake tests were in good agreement with the experimental result (maximum detection near 450 m at depths near 24 m). There is very little variation in detection range due to changing the model wind speed. 36 DRDC-RDDC-2016-R053

47 6 Summary Discussions The HF multibeam sonar operated well in both the forward- and aft-looking target detection tests from a ship, producing a valuable data set covering a variety of sea states and near-surface sound-speed conditions. In forward-looking tests hollow steel spheres suspended 1 4 m below the surface were used as targets. These sphere targets were routinely detected at ranges up to 580 m in low sea-states in Saanich Inlet, but detection range was reduced to roughly m under moderate sea states (winds up to 16 knots) encountered in the Strait of Georgia. While this suggests wind speed was a controlling factor in target detectability, the effect of near-surface sound speed gradients was also found to be important. Increasing wind speed was observed to produce both an increase in background reverberation and a reduction in apparent target strength. Both of these effects reduced the signal to noise properties of the target echoes, resulting in reduced target detection ranges. Observed SNR varied between 10 db (the lower limit of target detectability) and 30 db. Extrapolations from the observed sonar performance suggested that these targets would not be detectable at wind speeds exceeding 17 knots; however sea-trial data necessary to confirm this hypothesis was not collected. If this turns out to be a valid trend, this potentially places strong wind and sea-state limitations on future naval applications. However, some evidence for bubble saturation effects, whereby the observed reverberation level does not increase with increasing wind and sea-state, was observed at wind speeds from 12 to 15 knots. This has the potential to stop the trend of decreasing detection range with increasing wind speed. While no difference in detectability was observed between 1 and 4 m target depth (see Figure 13), modeling suggests improved detection performance against deeper targets. It was found that the relatively strong near-surface sound speed gradients in both locations created acoustic convergence and shadow zones which modified the target detectability. This was particularly clear in the Saanich Inlet test where a sub-surface refractive sound channel was observed to bias the apparent target strength by up to +15 db relative to simple spherical spreading predictions. In some cases the refraction effects produced local maxima in surface reverberation, inducing drop-outs in target detectability at some ranges. This observed sensitivity to near-surface sound speed gradients was not anticipated prior to the sea-trials, and thus only minimal (e.g., single location, daily) sound speed profiling was conducted. A key lesson-learned is that timely, local environmental measurements are necessary to understand sonar performance (at least in littoral regions). Considerable ping-to-ping variability in both target echo strength and background reverberation levels was observed. It is hypothesized that this target echo variability is due to acoustic interference between the direct and surface-reflected acoustic paths, inevitably randomized by the small scale fluctuations in acoustic path length caused by surface roughness and wave-induced turbulence. Observed distributions in apparent target strength were found to be Gaussian-distributed with variation (standard deviation) ±5 to ±8 db relative to the mean. This variability is believed to have both temporal and spatial components. The background reverberation also exhibited a combination of short period, small scale fluctuations and larger-scale scattering patches. It is recommended that further work be conducted to assess the causes and implications of this variability. DRDC-RDDC-2016-R053 37

48 During the aft-looking sonar tests a towed target was detectable up to 450 m range behind the ship at speeds from 5 to 7.5 knots. The towed target depth varied from 5 to 24 m, generally below the ship wake. The wake itself was observed to generate between 5 and 15 db increase in the background levels (dependent on range), while still allowing observed target SNR between 15 and 25 db. The wake was observed to have little influence on the apparent target strength. Similar to the forward-looking tests, significant levels of ping-to-ping variability were observed in both the target echo and background scattering from the ship wake. While these results are encouraging, it should be noted that the ship wake depths and densities at the maximum sea-trial speed (7.5 knots) were relatively modest. The maximum ship speed was limited by safety considerations in use of the sonar strut. Greater wake width, depth, and densities were previously reported with this same ship at speeds up to 12 knots [7]. Larger, higher-speed ships (e.g., HALIFAX-class frigates) have much deeper and denser wakes and so a much greater level of wake backscatter and acoustic extinction is to be expected. Therefore, sonar tests at higher ship speeds and possibly with larger ships should be undertaken. This will require some additional engineering changes to the sonar strut on CCGS Vector, for example through the use of streamlined fairings, or alternate arrangements on bigger ships. Relatively strong acoustic interference attributed to propeller cavitation was observed in both forward- and aft-looking tests when the ship accelerated at speeds above 6 knots. This cavitation generated approximately 15 db increase in background noise levels, with some evidence for blade-rate modulation. This measurement occurred near the cavitation inception speed; at higher ship speeds cavitation effects should increase. Thus, there is strong potential for ship-generated noise to render the sonar useless at higher speeds. However, it should be noted that the HF sonar used in these sea-trials was mounted relatively close to the propeller (5 m to starboard); other hull-mounted applications may provide greater physical separation and hence less interference. Other bigger ships will also have a higher propeller cavitation inception speed. A sonar performance model (NATO ESPRESSO) was useful in elucidating environmental influences on the results. Generally this model produced quantitative predictions in reasonable agreement with the sea-trials results. It was particularly useful in quantifying differences in sonar performance caused by the different sound speed profiles between Saanich Inlet and the Strait of Georgia. This limited validation now provides some confidence in the model, which may now be used to extrapolate sonar performance to different environments and greater target depths. These sonar tests were all conducted at relatively low target speeds relative to the sonar, which is appropriate for obstacle avoidance applications, but misses important acoustic effects occurring with higher speed ships and against high-speed targets such as torpedoes. For example, in moving ship applications some parts of the target echo, sea-surface backscatter reverberation, and wake scattering may be shifted out of the receiver band by own-ship Doppler effects, reducing their apparent amplitude. Similarly, the Doppler shift of a high speed target might push the echo completely out of the sonar receiver band. In the default settings for this SM-2000 sonar the receiver bandwidth was tightly coupled to the sonar pulse length, making it susceptible to these effects (see Annex A, Section A.6). Any future application against high speed targets would need to utilize a much wider receiver bandwidth. This can be done with the existing sonar. In future, adding a Doppler sensing capability to the sonar would potentially allow improved discrimination between background reverberation and high-speed targets. 38 DRDC-RDDC-2016-R053

49 This report makes use of acoustic calibrations of the multibeam sonar. While time-consuming, this effort had several benefits, namely (i) a verification of the manufacturer s technical specifications, and (ii) enabling quantitative sonar output which then allowed comparison with quantitative physical modeling. While most naval applications are concerned only with relative signal strength (i.e., target echo vs. reverberation), generating absolute acoustic quantities allows a better assessment of sonar performance. The calibration results generally confirmed sonar source levels, beam-patterns, time-varying gain functions, and clipping and self-noise thresholds. Importantly, the calibration results allowed estimates of the true target strengths of the target spheres (with ballast) and the Sharko towed target. These estimates would have been impossible to produce accurately from simple geometric formulae. The TS estimates were then used in sonar performance models to produce realistic predictions of target SNR and detection range. DRDC-RDDC-2016-R053 39

50 This page intentionally left blank. 40 DRDC-RDDC-2016-R053

51 References [1] Trevorrow, M., Measurements of near-surface bubble plumes in the open ocean with implications for high-frequency sonar performance, J. Acoustical Soc. Am. 114(5), [2] Trevorrow, M., and Myers, V., Modeling obstacle avoidance sonar performance in the presence of near-surface bubbles, presented at 11 th European Conference on Underwater Acoustics, Proc. of Meetings on Acoustics 17. [3] Trevorrow, M., and Vagle, S., Summary of HF sonar sea-trials on CCGS VECTOR, April 2015, DRDC-RDDC-2015-R245. [4] Medwin, H., and Clay, C., Fundamentals of Acoustical Oceanography (Academic Press, San Diego). [5] Francois, R., and Garrison, G., Sound absorption based on ocean measurements Part II: boric acid contribution and equation for total absorption, J. Acoust. Soc. Am. 72, [6] Dahl, P., and Plant, W., The variability of high-frequency acoustic backscatter from a region near the sea surface, J. Acoust. Soc. Am. 101(5), [7] Trevorrow, M., Vagle, S., and Farmer, D., Acoustic measurements of micro-bubbles within ship wakes, J. Acoustical Soc. Am. 95(4), [8] Davies, G., and Signell, E., Espresso Scientific User Guide, NATO SP [9] Porter, M., and Bucker, H., Gaussian beam tracing for computing ocean acoustic fields, J. Acoust. Soc. Am. 82(4), [10] Meyer, M., and Davies, G., Application of the method of geometric beam tracing for high frequency reverberation modelling, Acta Acustica 88(5), [11] APL-UW, High-frequency ocean environmental acoustic models handbook, Technical Report 9407, Applied Physics Laboratory, University of Washington. [12] Foote, K., Chu, D., Hammar, T., Baldwin, K., Mayer, L., Hufnagle, L., and Jech, J., Protocols for calibrating multibeam sonar, J. Acoust. Soc. Am. 117(4), [13] Cochrane, N., Li, Y., and Melvin, G., Quantification of a multibeam sonar for fisheries assessment applications, J. Acoust. Soc. Am. 114(2), [14] Trevorrow, M., and Crawford, A., Calibration of a high-frequency multibeam sonar for water volumetric surveys, DRDC Atlantic TM [15] Vagle, S., Foote, K., Trevorrow, M., and Farmer, D., Absolute calibrations of monostatic echosounder systems for bubble counting, IEEE J. Oceanic Eng. 21(3), DRDC-RDDC-2016-R053 41

52 This page intentionally left blank. 42 DRDC-RDDC-2016-R053

53 Annex A HF Sonar Calibrations An effort was made to acoustically calibrate the SM-2000 sonar. This was done in order to verify the manufacturer-provided technical specifications, understand the sonar performance limitations, and enable quantitative sonar outputs. This process involved multiple distinct measurements to quantify the acoustic output of the sonar. Calibration procedures for multibeam sonars have been previously described in References [12 15]. While some of these measurements could be made at ranges up to 3.5 m in the DRDC Atlantic acoustic test tank, far-field target sphere measurements had to be conducted at the DRDC Acoustic Calibration Barge. The following were specific areas of investigation: 1. Background noise levels measured sonar signal output while the sonar transmitter was turned off, at varying range and receiver gain settings. 2. Clipping levels measured using a separate transmitter to generate 90 khz signals on the main sonar axis, with varying receiver gain. 3. Transmitter source levels measured using a calibrated hydrophone on the main sonar axis, with varying transmitter power levels and pulse lengths. 4. Transmitter horizontal and vertical beamwidths measured with a single hydrophone while either rotating the sonar head (horizontal) or varying the hydrophone vertical location (vertical). 5. Echo strength from reference target spheres conducted by suspending a sphere in the sonar beam and rotating the sonar so the target can be seen across the sonar aperture with successive pings. The apparent angular width of the target is then a measure of the sonar beamforming resolution. 6. Receiver bandwidth (Doppler capability) measured receiver response vs. frequency using a separate transmitter to generate either continuous or pulsed-cw signals. These will be described in turn below. The sonar data were extracted from the raw sonar recordings using a two-stage process, as explained in Section 2.3. The SM-2000 Time-Varying Gain (TVG, in db) was set to the system default of 20 log 10 [r] r, (A.1) DRDC-RDDC-2016-R053 43

54 where r is the range in metres and 1 is the acoustic absorption, having the value db/m. A fixed gain offset is available, but this was set to 0 db for the field trials and most of the acoustic calibrations. This TVG is limited to a maximum value of 122 db, then remains constant. This limit only occurs when using an additional fixed gain offset. Note that the default value of 1 under-predicts the true value, which should be close to db/m under the sea-water conditions encountered in these measurements (using relations found in [5]). It was expected that some calibration parameters would be dependent on the sonar range settings, for example due to changes in pulse length and bandwidth settings (see Table 1, Section 2.2). Therefore all of the measurements were repeated using multiple sonar range settings. A.1 Sonar Self-Noise Levels The background self-noise levels in the sonar were measured while the sonar transmitter power was turned off, at varying range and receiver gain settings. In this mode the internal self-noise of the sonar was amplified by the TVG, so that these measurements also served as verification of the TVG variation. The raw sonar amplitude was averaged over all 128 beams. A minimum of 20 pings (snapshots) were averaged. Figure A.1: Comparison of sonar self-noise intensity (db re sonar units) at various sonar range settings. TVG given in Eq.(A.1) with -92 db and -82 db offsets. As shown in Figure A.1 the background self-noise was strongly controlled by the TVG, and had a mild dependence on the sonar range setting. In this figure a fixed offset to TVG curves was applied to match the raw sonar intensity. There appeared to be two components of the self-noise: (i) a small transducer noise level amplified by the TVG, and (ii) a fixed (in time or range) lower limit arising after the TVG. The pre-tvg noise appeared to have two distinct values with a distinct drop (approximately 5 db) near m range. It is hypothesized that the higher noise level is due to cross-talk from the digital data telemetry output of the sonar head. Beyond 250 m range the noise levels all asymptote to a single TVG curve. 44 DRDC-RDDC-2016-R053

55 The post-tvg noise dominated at short ranges, and was distinctly different for the various sonar range settings. It strongly increased for the longer range settings, varying from -14 db for the 60 m setting to +2 db for the 600 m setting. This increase can be partly explained by the fact that the different sonar settings have a varying digital A/D gain (not shown in Table 1, Section 2.2), which increases from 1 to 32 times as the sonar range increases from 100 m to 600 m. This digital gain effect on self-noise was partially counteracted by the decreased receiver bandwidth of the longer range settings. A measurement of self-noise levels at varying gain offsets was performed to verify linearity in the preamplifier and signal processing chain, with result shown in Figure A.2. The self-noise increased in almost exact correspondence to the fixed gain offset (0, +10, +20, and +30 db). The maximum noise levels near +32 db (re sonar units) were created by the TVG ceiling. As in the previous figure, there appeared to be a fixed (post-tvg) self-noise level of order 1 sonar unit. Figure A.2: Comparison of sonar self-noise intensity (db re sonar units) at various gain offsets for the 600 m sonar range setting. A.2 Clipping Levels It was not possible to drive the sonar to clipping using only the internal self-noise. Thus a separate measurement was conducted using an external transmitter positioned on the main sonar axis at a range of 2.5 m. The standard TVG setup (0 db offset) was used and the sonar transmitter was turned off. The clipping level was explored by varying the voltage applied to the external transmitter. A fixed clipping level of 66 db (re sonar units) was observed at a variety of external transmitter power levels. No evidence was seen for so-called soft-clipping due to pre-amplifier or signal-detector non-linearity at higher amplitudes. DRDC-RDDC-2016-R053 45