Measurement and Modelling of Atmospheric Acoustic Propagation Over Water

Transcription

1 Copy No. Defence Research and Development Canada Recherche et développement pour la défense Canada DEFENCE & DÉFENSE Measurement and Modelling of Atmospheric Acoustic Propagation Over Water Cristina Tollefsen E. C. Murowinski Sean Pecknold Defence R&D Canada Atlantic Technical Memorandum DRDC Atlantic TM December 2010

2 This page intentionally left blank.

3 Measurement and Modelling of Atmospheric Acoustic Propagation Over Water Cristina Tollefsen Defence R&D Canada Atlantic E. C. Murowinski Defence R&D Canada Atlantic Sean Pecknold Defence R&D Canada Atlantic Defence R&D Canada Atlantic Technical Memorandum DRDC Atlantic TM December 2010

4 Principal Author Cristina Tollefsen Approved by Daniel Hutt Head/Underwater Sensing Approved for release by Calvin Hyatt Head/Document Review Panel c Her Majesty the Queen in Right of Canada as represented by the Minister of National Defence, 2010 c Sa Majesté la Reine (en droit du Canada), telle que représentée par le ministre de la Défense nationale, 2010

5 Abstract General interest in understanding atmospheric acoustic propagation over water has increased in recent years, driven primarily by concerns about noise from offshore wind farms. In addition, there are specific naval interests such as evaluating the performance of directional acoustic hailing devices used at sea, determining the potential environmental impact of naval gunfire exercises, and understanding the in-air acoustic footprint of maritimebased military assets. Atmospheric acoustic propagation is strongly affected by the environment, including sea surface roughness and atmospheric parameter profiles (temperature, wind velocity, humidity, and turbulence). However, published measurements and realistic modelling of the effect of environmental parameters on over-water acoustic propagation are sparse. An experiment was designed to measure atmospheric acoustic transmission loss over water by using an acoustic source on a small boat and a receiver on a barge. Point measurements of atmospheric parameters and directional ocean wave spectra were made in the vicinity the experiment. Also, separate data from two types of atmospheric parameter profiles - measured radiosonde profiles and modelled profiles generated by Environment Canada s Global Environmental Multiscale (GEM) model - were made available for this investigation. The dependence of measured transmission loss on source-to-receiver range, atmospheric parameter profiles, receiver height, and wind speed was explored, and impulsive sources were used to observe separate multipath arrivals. Measured acoustic transmission loss was compared with the output of the Sensor Performance Evaluator for Battlefield Environments (SPEBE) atmospheric acoustic propagation model developed by the United States Army Research Laboratory (ARL). Résumé Depuis quelques années, on s intéresse de plus en plus à l étude de la propagation acoustique au-dessus de l eau, en raison surtout des préoccupations relatives au bruit provenant des parcs éoliens en mer. On souhaite en outre, dans le secteur de la Marine, évaluer la performance des dispositifs d appels acoustiques directionnels pour usage en mer, déterminer l impact environnemental possible des exercices de l artillerie navale et comprendre l empreinte acoustique dans l air des ressources militaires maritimes. La propagation acoustique atmosphérique est très sensible à l environnement, y compris les profils des paramètres atmosphériques (température, vitesse du vent, humidité et turbulence) et la rugosité de la surface. Par contre, peu de données ont été publiées sur les mesures et la modélisation réaliste des effets des paramètres environnementaux sur la propagation acoustique au-dessus de l eau. Une expérience a été conçue pour mesurer les pertes de transmission acoustique au-dessus de l eau au moyen d une source acoustique installée sur un petit bateau et de récepteurs installés sur un chaland. Des mesures par points des paramètres atmosphériques et spectres de vagues directionnelles océaniques ont été effectuées aux environs du site de l expérience. De plus, d autres données de deux types de profil des paramètres at- DRDC Atlantic TM i

6 mosphériques - des profils de radiosonde mesurés et des profils modélisés produits par le modèle global environnementale multi-échelle (GEM) d Environnement Canada - ont été utilisées pour cette recherche. On s est penché sur le rapport entre les pertes de transmission mesurées et la distance entre la source et le récepteur, les profils des paramètres atmosphériques, la hauteur du récepteur et la vitesse du vent, et on a employé des sources impulsives pour observer des réceptions distinctes par trajets multiples. Les pertes de transmission acoustique mesurées ont été comparées aux résultats de l évaluateur des performances de capteurs pour les environnements de combat (SPEBE), un modèle de propagation acoustique atmosphérique mis au point par l ARL (Army Research Laboratory) des États-Unis. ii DRDC Atlantic TM

7 Executive summary Measurement and Modelling of Atmospheric Acoustic Propagation Over Water Cristina Tollefsen, E. C. Murowinski, Sean Pecknold; DRDC Atlantic TM ; Defence R&D Canada Atlantic; December Background: General interest in understanding atmospheric acoustic propagation over water has increased in recent years, driven primarily by concerns about noise from offshore wind farms. In addition, there are specific naval interests such as evaluating the performance of directional acoustic hailing devices used at sea, determining the potential environmental impact of naval gunfire exercises, and understanding the in-air acoustic footprint of maritime-based military assets. Atmospheric acoustic propagation is strongly affected by the environment, including sea surface roughness and atmospheric parameter profiles (temperature, wind velocity, humidity, and turbulence). However, published measurements and realistic modelling of the effect of environmental parameters on over-water acoustic propagation are sparse. Principal results: An experiment was designed to measure atmospheric acoustic transmission loss over water by using an acoustic source on a small boat and a receiver on a barge. Point measurements of atmospheric parameters and directional ocean wave spectra were made in the vicinity the experiment. Additional environmental data available include two types of atmospheric parameter profiles: measured radiosonde profiles, and modelled profiles generated by Environment Canada s Global Environmental Multiscale (GEM) model. The dependence of measured transmission loss on source-to-receiver range, atmospheric parameter profiles, receiver height, and wind speed was explored, and impulsive sources were used to observe separate multipath arrivals. Measured acoustic transmission loss was compared with the output of the Sensor Performance Evaluator for Battlefield Environments (SPEBE) atmospheric acoustic propagation model developed by the United States Army Research Laboratory (ARL). Significance of results: Atmospheric acoustic transmission loss over water was measured in a variety of atmospheric conditions, and the technical challenges encountered in making the measurements will inform future field programs. Propagation model results from a model not optimized for use over water were generated using both measured and modelled inputs for atmospheric parameter profiles, and qualitatively compared to transmission loss measurements. Future work: The acoustic receivers need to be calibrated in order to make a quantitative comparison between modelled and measured transmission loss. The surface roughness DRDC Atlantic TM iii

8 data have not yet been analyzed, and a more realistic surface scattering and loss model should be explored. A simple ray tracing model should be used to determine the origins of the observed multipath arrivals. The source has strong harmonics extending as high as 10 khz; the analysis could be expanded to include the transmission loss of the harmonics. In preparation for anticipated future experiments, the data acquisition hardware is being redesigned, alternative sources are being considered, and new propagation models are being evaluated. iv DRDC Atlantic TM

9 Sommaire Measurement and Modelling of Atmospheric Acoustic Propagation Over Water Cristina Tollefsen, E. C. Murowinski, Sean Pecknold ; DRDC Atlantic TM ; R& D pour la défense Canada Atlantique ; décembre Contexte : Depuis quelques années, on s intéresse de plus en plus à l étude de la propagation acoustique au-dessus de l eau, en raison surtout des préoccupations relatives au bruit provenant des parcs éoliens en mer. On souhaite en outre, dans le secteur de la Marine, évaluer la performance des dispositifs d appels acoustiques directionnels pour usage en mer, déterminer l impact environnemental possible des exercices de l artillerie navale et comprendre l empreinte acoustique dans l air des ressources militaires maritimes. La propagation acoustique atmosphérique est très sensible à l environnement, y compris les profils des paramètres atmosphériques (température, vitesse du vent, humidité et turbulence) et la rugosité de la surface. Par contre, peu de données ont été publiées sur les mesures et la modélisation réaliste des effets des paramètres environnementaux sur la propagation acoustique au-dessus de l eau. Principaux résultats : Une expérience a été conçue pour mesurer les pertes de transmission acoustique au-dessus de l eau au moyen d une source acoustique installée sur un petit bateau et de récepteurs installés sur un chaland. Des mesures par points des paramètres atmosphériques et spectres de vagues directionnelles océaniques ont été effectuées aux environs du site de l expérience. De plus, d autres données de deux types de profil des paramètres atmosphériques - des profils de radiosonde mesurés et des profils modélisés produits par le modèle global environnementale multi-échelle (GEM) d Environnement Canada - ont été utilisées pour cette recherche. On s est penché sur le rapport entre les pertes de transmission mesurées et la distance entre la source et le récepteur, les profils des paramètres atmosphériques, la hauteur du récepteur et la vitesse du vent, et on a employé des sources impulsives pour observer des réceptions distinctes par trajets multiples. Les pertes de transmission acoustique mesurées ont été comparées aux résultats de l évaluateur des performances de capteurs pour les environnements de combat (SPEBE), un modèle de propagation acoustique atmosphérique mis au point par l ARL (Army Research Laboratory) des États-Unis. Importance des résultats : Les pertes de transmission acoustique atmosphérique au dessus de l eau ont été mesurées dans diverses conditions atmosphériques, et les difficultés rencontrées sur le plan technique au moment de prendre les mesures permettront d adapter les prochains programmes sur le terrain. Les résultats de la modélisation de la propagation issus d un modèle non optimisé pour l utilisation au dessus de l eau ont été générés au DRDC Atlantic TM v

10 moyen de données mesurées et modélisées sur les profils des paramètres atmosphériques, puis comparés, sur le plan quantitatif, aux mesures des pertes de transmission. Travaux à venir : Les récepteurs acoustiques doivent être calibrés pour réaliser des comparaisons quantitatives entre les pertes de transmission modélisées et mesurées. Les données sur la rugosité de la surface n ont pas encore été analysées, et un modèle plus réaliste de diffusion et de pertes en surface devrait être examiné. Un simple modèle de traçage de rayons devrait être utilisé pour déterminer les origines des réceptions à trajets multiples observées. La source présente une forte harmonique pouvant atteindre 10 khz ; la portée de l analyse pourrait être élargie afin d examiner la perte de transmission des harmoniques. En prévision d autres expériences futures, on revoit actuellement la conception du matériel d acquisition des données, on envisage d utiliser d autres sources et on évalue de nouveaux modèles de propagation. vi DRDC Atlantic TM

11 Acknowledgements The authors would like to thank P. Shouldice, D. Graham, P. Anstey, R. Johnson, M. Fotheringham, and LS D. Ratelle for technical support. MetOc Halifax kindly provided meteorological measurements and model outputs. DRDC Atlantic TM vii

12 This page intentionally left blank. viii DRDC Atlantic TM

13 Table of contents Abstract... i Résumé... i Executive summary... Sommaire... Acknowledgements... iii v vii Table of contents... ix List of figures... xi 1 Background Acoustic transmission loss Experiment Experimental geometry Data processing Modelling Results and discussion Overview of model results Moving-source recordings Range dependence of received SPL Microphone height Wind speed Stationary recordings Impulsive sources DRDC Atlantic TM ix

14 4 Challenges and future work Challenges and solutions Future work Conclusions Annex A: Equipment Annex B: Data Processing B.1 File synchronization B.2 Position data B.3 Blast start and end times B.4 Received SPL B.5 Noise calculation Annex C: Additional data C.1 Run summary C.2 Range dependence of SPL: Additional Runs References List of abbreviations x DRDC Atlantic TM

15 List of figures Figure 1: Figure 2: Figure 3: Experimental setup. Zodiac runs were done toward and away from the ACB (grey rectangle) in various directions. Coordinate system for wind velocity is indicated at top right by u x and u y. Water depths and contours (m, dm) are indicated in blue The receiver used in the experiments: (a) the TetraMic TM, (b) with foam windscreen in place, (c) with hairy wind screen in place over the foam Microphone directions. Channels are numbered as indicated on the diagram (not to scale) Figure 4: Locations for the Zodiac, relative to the ACB (green marker at N, W) for the first two stationary recordings (Point A) and the third stationary recording (Point B). Point C is the location of the impulsive source, described in Section 3.4. The ADCP location ( N, W) is indicated with a blue marker. Scale is at bottom left Figure 5: Sample Zodiac GPS track (green line). The ACB location at N, W is indicated with a green marker. Scale is at bottom left Figure 6: Figure 7: Figure 8: Effective sound speed c ef f as a function of height for (a) positive x-direction, (b) negative x-direction, (c) positive y-direction, and (d) negative y-direction, for measured (radiosonde) and modelled (GEM) atmospheric parameter profiles Effective sound speed c ef f as a function of height for (a) positive x-direction, (b) negative x-direction, (c) positive y-direction, and (d) negative y-direction, for measured (radiosonde) atmospheric parameter profiles. The resulting transmission loss calculated by SPEBE is plotted in Figure Sample SPEBE output: transmission loss (db) as a function of latitude and longitude, for a receiver at the centre of the image ( ) and a source at any point in the image, with topographic contours (m) shown as solid black lines. The atmospheric parameter profiles used are shown in Figure 7. The transmission loss along the black dashed line for a source at the black circle ( ) is interpolated and plotted as a function of range in Figure Figure 9: Interpolation of transmission loss from sample SPEBE output shown in Figure DRDC Atlantic TM xi

16 Figure 10: Modelled transmission loss centred on the ACB (white dot) for (a) 530 Hz, radiosonde data, (b) and 670 Hz, radiosonde data, (c) 530 Hz, GEM output, (d) 670 Hz, GEM output, with dates and times as indicated on the plots. The black line indicates the profile extracted and plotted in Figure Figure 11: Modelled transmission loss along the black line in Figure 10 (due east of the ACB) for 530 Hz (blue) and 670 Hz (red), for (a) radiosonde data and (b) GEM output Figure 12: Received SPL, noise levels, and modelled transmission loss in arbitrary db at 530 Hz as a function of range for a relative wind direction of 8 (downwind). The wind speed was 3.4 m/s, and the TetraMic TM was at 3.76 m height Figure 13: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 8 (downwind). The wind speed was 3.4 m/s, and the TetraMic TM was at 3.76 m height Figure 14: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 95 (crosswind). The wind speed was 2.7 m/s, and the TetraMic TM was at 3.76 m height Figure 15: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 177 (upwind). The wind speed was 3.3 m/s, and the TetraMic TM was at m height Figure 16: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 140. The wind speed was 4.9 m/s, and the TetraMic TM was at m height. 19 Figure 17: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 147. The wind speed was 5.0 m/s, and the TetraMic TM was at 3.61 m height. 20 Figure 18: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 175. The wind speed was 1.8 m/s, and the TetraMic TM was at m height. 21 Figure 19: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the first stationary recording, taken at Point A (Figure 4). The TetraMic TM was at m height and the wind was 6.1 m/s from 121 T xii DRDC Atlantic TM

17 Figure 20: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the second stationary recording, taken at Point A (Figure 4). The TetraMic TM was at m height and the wind was 3.5 m/s from 302 T Figure 21: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the third stationary recording, taken at Point B (Figure 4). The TetraMic TM was at m height and the wind was 5.8 m/s from 321 T Figure 22: Received SPL (arbitrary units) for each receiver channel for four impulsive noise events (a-d). The times of the direct and refracted arrivals are plotted as red and blue vertical lines, respectively Figure 23: The ACB, with the approximate location of the TetraMic TM indicated by a red circle Figure B.1: Diagram of signal and noise calculation windows. The blue window is the 3-s signal window, and the two red windows are the 3-s before and after noise windows. The buffer time used in the analysis is T B = 1 s. The actual signal is plotted in blue on the bottom half of the diagram Figure C.1: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 7. The wind speed was 4.2 m/s, and the TetraMic TM was at 3.76 m height.. 41 Figure C.2: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 31. The wind speed was 1.6 m/s, and the TetraMic TM was at m height. 42 Figure C.3: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 33. The wind speed was 1.9 m/s, and the TetraMic TM was at m height. 43 Figure C.4: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 73. The wind speed was 1.0 m/s, and the TetraMic TM was at m height. 44 Figure C.5: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 77. The wind speed was 1.8 m/s, and the TetraMic TM was at 3.76 m height. 45 DRDC Atlantic TM xiii

18 Figure C.6: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 90. The wind speed was 0.9 m/s, and the TetraMic TM was at m height. 46 Figure C.7: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 94. The wind speed was 1.2 m/s, and the TetraMic TM was at m height. 47 Figure C.8: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 114. The wind speed was 0.9 m/s, and the TetraMic TM was at m height. 48 Figure C.9: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 123. The wind speed was 1.6 m/s, and the TetraMic TM was at m height. 49 Figure C.10:Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 126. The wind speed was 0.7 m/s, and the TetraMic TM was at m height. 50 Figure C.11:Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 139. The wind speed was 4.6 m/s, and the TetraMic TM was at m height. 51 Figure C.12:Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 166. The wind speed was 1.1 m/s, and the TetraMic TM was at m height. 52 Figure C.13:Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 167. The wind speed was 3.0 m/s, and the TetraMic TM was at m height. 53 Figure C.14:Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 180. The wind speed was 1.6 m/s, and the TetraMic TM was at m height. 54 xiv DRDC Atlantic TM

19 1 Background General interest in understanding atmospheric acoustic propagation over water has increased in recent years, driven primarily by concerns about noise from offshore wind farms. In addition, there are specic naval interests such as evaluating the performance of directional acoustic hailing devices used at sea, determining potential environmental impact of naval gunfire exercises, and understanding the in-air acoustic footprint of maritime-based military assets. Atmospheric acoustic propagation is strongly affected by the environment, including atmospheric parameter proles (temperature, wind velocity, humidity, turbulence) and sea surface roughness. The maximum audible range in air for sound sources at sea will change with weather conditions and sea state; unfortunately, previous experimental work in the area of atmospheric acoustic propagation over water is sparse. In 1961, Wiener measured acoustic transmission loss in foggy conditions using a fog horn as the acoustic source [1]. Salomons demonstrated that water surface waves can strongly affect transmission loss in long-range over-water propagation [2]. Boué showed that spherical spreading (see Equation 1) is an appropriate model up to 700 m range [3]. Bolin and Boué showed that accurate predictions in shadow zones 1 rely on inclusion of atmospheric turbulence in transmission loss models [4]. 1.1 Acoustic transmission loss The simplest representation of acoustic transmission loss is modelled using geometric spreading. Spherical spreading is appropriate when a source is in an unbounded, homogeneous medium: ( ) R TL= 20log, (1) R 0 in which TLis the transmission loss in decibels (db) relative to a reference range R 0 (generally 1 m), and R is the range to the receiver. In practice, spherical spreading is used as a first approximation in many applications, and is valid as long as a source is far from any boundaries and the medium is reasonably homogenous. For a source in a waveguide, that is, a medium with two parallel boundaries where the distance between the boundaries is much less than the range of the measurement, cylindrical spreading is the appropriate model for the transmission loss ( ) R TL= 10log, (2) where again R is the range to the receiver and R 0 is the reference range. 1 Shadow zones are regions where no sound rays arrive when using ray-tracing propagation models. R 0 DRDC Atlantic TM

20 Atmospheric absorption by dissipative processes in the atmosphere causes additional losses beyond the geometric spreading expected based on Equations 1 and 2, and depends strongly on the frequency of the sound, as well as temperature and humidity of the atmosphere [5]. Beyond the two simple geometric spreading models are a wide array of computational models, including ray-based models, parabolic equation models, and fast field program models [5]. 2 Experiment An experiment was performed in July 2010 to measure the atmospheric acoustic transmission loss over water, and an atmospheric acoustic propagation model was used to compare measured and modelled transmission loss. The experimental geometry is described in Section 2.1, the data processing is described in Section 2.2 and Appendix B, and the propagation model is described in Section Experimental geometry In order to measure the acoustic transmission loss over water as a function of range under a variety of atmospheric conditions, an experiment was performed in the Bedford Basin, Halifax, Nova Scotia on board DRDC Atlantic s Acoustic Calibration Barge (ACB), located at N, E. An overview of the experimental setup will be given here; full equipment details, including serial numbers and instrument settings, can be found in Appendix A. A schematic diagram of the experimental setup is shown in Figure 1. The acoustic source was a dual-tone Nautilus 3500 motorcycle horn with nominal frequencies of 530 Hz and 670 Hz and a measured on-axis source level of 115 db re 20 μpa at 1 m. The source was mounted aft-facing at 1.25 m height above the water on a Zodiac inflatable boat. A Sony Linear PCM Recorder was used in the Zodiac to monitor the source. The receiver was a Core Audio TetraMic TM mounted above the barge structure at m above the water surface 2. The TetraMic TM is a directional array of four cardiod microphones (Figure 2); the raw audio acquired by each microphone capsule is stored as a separate.wav file. The raw audio can then be processed in conjunction with the manufacturerprovided microphone calibration data to create four channels: one omnidirectional channel and three dipoles 3. A windscreen provided by the manufacturer was mounted on the TetraMic TM to minimize wind noise. The windscreen was shaped like a cylinder with a hemisphere attached to one end, and consisted of two parts: a foam core (11.4 cm high and 7.9 cm in diameter) with a hairy cover. With both the foam core and the cover in place, the windscreen was 15.0 cm high and 11.4 cm in diameter. For some runs, the TetraMic TM 2 The positive x-direction in the TetraMic TM coordinate system was facing the stern of the ACB. 3 Dipoles are oriented along the x-, y-, and z-axes in the TetraMic TM coordinate system. 2 DRDC Atlantic TM

21 305 weather station & gps [ height 9.6m ] TetraMic TM & gps [ height 10.2m ] ADCP u y u x 268 ambient noise SPL meter [ height 1.9m ] sample run direction source [ height 1.25m ] m 38 Figure 1: Experimental setup. Zodiac runs were done toward and away from the ACB (grey rectangle) in various directions. Coordinate system for wind velocity is indicated at top right by u x and u y. Water depths and contours (m, dm) are indicated in blue. was moved to a lower position (3.61 m to 3.76 m above the water surface) because the wind noise was causing clipping 4 on the recording. The TetraMic TM microphone directions are shown schematically in Figure 3. Three types of recordings were made with the TetraMic TM : moving source, stationary source, and (improvised) impulsive sources. In all cases, a series of three timing blasts (short, quick blasts of the horn) were sounded when the Zodiac was near the ACB at the beginning and end of each run to synchronize the audio tracks with each other and with the GPS. The GPS time at the start of the third timing blast was noted in the log book, and the two audio tracks were synchronized by calculating their cross-correlation, as described in more detail in Appendix B.1. For the moving source recordings, the Zodiac was driven away from and towards the ACB at speeds of 4 to 8 knots to distances of up to 3 km. One person on the Zodiac was assigned to sound the horn for specified lengths of time at specified intervals that were timed using a watch, and the Zodiac was turned around for the inward-bound run when the horn was no longer audible on board the ACB. There were a total of 20 moving-source runs during which the horn was sounded every 20 s for 5 s; for three runs, the horn was sounded every 10 s for as short a blast as possible (in practice, about 0.3 s). 4 The signal exceeded the maximum analog-to-digital (A/D) range available. DRDC Atlantic TM

22 Figure 2: The receiver used in the experiments: (a) the TetraMic TM, (b) with foam windscreen in place, (c) with hairy wind screen in place over the foam. (a) (b) NW N 1 2 Barge (side view) 3 4 Barge (top view) Figure 3: Microphone directions. Channels are numbered as indicated on the diagram (not to scale). For the stationary source recordings, the horn was sounded every 20 s for 5 s for a total of five minutes, followed by as-short-as-possible blasts every 10 s for a total of five minutes. For the first two stationary recordings, the Zodiac was tied to a buoy at 286 m range from the ACB (Point A in Figure 4). For the third stationary recording, the Zodiac was tied to the ACB with a rope and drifted downwind until it came to a stop at a distance of 119 m from the ACB, and the horn blasts began once the Zodiac had stabilized in heading and range (Point B in Figure 4). For impulsive sources, recordings were made by popping balloons while the Zodiac drifted from a range of 95 m to 121 m (Point C in Figure 4). The ambient noise level at the ACB and horn source level were measured several times 4 DRDC Atlantic TM

23 Figure 4: Locations for the Zodiac, relative to the ACB (green marker at N, W) for the first two stationary recordings (Point A) and the third stationary recording (Point B). Point C is the location of the impulsive source, described in Section 3.4. The ADCP location ( N, W) is indicated with a blue marker. Scale is at bottom left. Figure 5: Sample Zodiac GPS track (green line). The ACB location at N, Wis indicated with a green marker. Scale is at bottom left. each day with a Brüel & Kjær (B & K) sound pressure level (SPL) meter. Directional ocean surface wave spectra were measured using a Teledyne RD Instruments Acoustic Doppler Current Profiler (ADCP) operating at 300 khz. GlobalSat global positioning system (GPS) antennas were mounted on the east and west corners of the ACB to monitor barge heading and position. A handheld GPS receiver (Garmin Oregon 55T) was used on the Zodiac to record its position during the runs (a sample track is shown in Figure 5). Point measurements of temperature, wind velocity, humidity, and air pressure were made at the ACB at 9.6 m height using a Vaisala WXT520 meteorological station. The meteorological station was aligned with its north direction parallel to the ACB bow-stern axis and the wind directions were corrected in the post-processing stage using the measured ACB heading. Vaisala Radiosondes were launched from Canadian Forces Base (CFB) Halifax (6 km SE of the ACB) on each day at 0930 and 1230 local time to record atmospheric DRDC Atlantic TM

24 parameter profiles. Parameter profiles from Environment Canada s Global Environmental Multiscale (GEM) model [6, 7, 8] were available at 0900, 1200, and 1500 local time each day. 2.2 Data processing A general outline of the data processing steps will be given in this section, while the details of the data processing can be found in Appendix B. Data processing consisted of several steps that were implemented using MATLAB: synchronizing the files, identifying horn blasts, and calculating the received SPL. The audio files fell into two groups: the stereo monitor tracks recorded using the Sony recorder on the Zodiac, and the four mono receiver tracks recorded using the TetraMic TM mounted on the ACB. Before the automated steps of the data processing could be performed, the data files were listened to and examined using the spectrogram function in Audacity (audio mixing software). The left channel of the stereo monitor track was used to determine blast start times and file synchronization. The start time of the third timing blast for each monitor and receiver file was noted to the nearest second, relative to the beginning of the file. In addition, the time of the first data blast and the time after which there were no more data blasts were noted for each monitor file. The times of the first and last data blasts and the third timing blast were used as inputs to the automatic blast identification algorithm described below and in Appendix B. The cross-correlation of the horn timing blasts between the monitor and receiver files was used to accurately determine the lag between files (Appendix B.1). The range to the Zodiac at the mid-point of each blast was calculated using the ACB and Zodiac GPS coordinates (Appendix B.2). Start and end times for each horn blast were detected using the monitor track recorded by the Sony recorder on the Zodiac (Appendix B.3). The received SPL was calculated from the receiver files using the detected blast time from the monitor track and taking into account the delay as the sound travelled to the receiver 5, along with a buffer to avoid horn start and end effects (Appendix B.4). Noise levels were estimated 2 s before and after each horn blast and interpolated to obtain the noise level during each blast (Appendix B.5). In order to calculate the received SPL, the frequency band chosen had to be wide enough to take into account the horn frequency instability and the Doppler shift resulting from the relative motion of the Zodiac and the ACB. The horn frequency varied between 525 Hz and 544 Hz for the nominal 530 Hz tone, and 666 Hz and 688 Hz for the nominal 670 Hz tone. The maximum Doppler shift at 8 knots and 688 Hz is 8 Hz. Therefore, the received SPL was averaged across a frequency band wide enough to account for Doppler shift, horn frequency instability, and a small additional buffer (525 ± 25 Hz and 677 ± 25 Hz). 5 The sound speed was calculated using the mean measured air temperature for the run. 6 DRDC Atlantic TM

25 The same two frequency bands were used to estimate the noise level during each blast (Appendix B.5). 2.3 Modelling The Sensor Performance Evaluator for Battlefield Environments (SPEBE) [9] is an inair acoustic propagation and sensor performance model that can compute acoustic transmission loss given inputs for terrain, ground type, meteorology, and source and receiver specficiations. A set of modelling runs was undertaken in order to compare the measured transmission loss with that obtained by using SPEBE in conjunction with both measured and modelled atmospheric parameter profiles. The terrain was input from a Digital Terrain Elevation Database [10], using a 2-km by 2- km square centred on the barge microphone location at N, W. The ground type in SPEBE is range-independent; therefore, an infinite impedance water surface was used, with a root-mean-square (RMS) roughness height of 0.2 m. Over a water surface, the roughness height implemented in SPEBE affects only the atmospheric parameter profiles and is not used for calculating reflection coefficients; preliminary tests suggest that the model results were not sensitive to the choice of roughness height. The source height was 1.25 m, and the receiver height was 10.4 m, 3.76 m, or 3.61 m, corresponding to the receiver height for the run being modelled. Transmission loss was modelled at 530 Hz for all the runs and 670 Hz for some of the runs, using the fast field program model included in SPEBE [11, 12]. MetOc Halifax provided profiles of temperature, wind velocity (i.e., speed and direction), relative humidity, and atmospheric pressure. Profiles were obtained from radiosondes launched at CFB Halifax and from Environment Canada s GEM weather model. The modelled profiles were adjusted using measured surface data from the North Magazine Jetty on the NE shore of the Bedford Basin ( N, W) except for the sea level pressure, which was obtained from the Shearwater Airport ( N, W). SPEBE can accept a variety of formats for meteorological inputs. For the model runs presented here, parameters were provided as a function of height in metres: temperature T in degrees Celsius; the wind speeds u x and u y in m/s in the east-west (x) and north-south (y) directions (see Figure 1 for convention); and the dimensionless specific humidity q. The specific humidity was computed from the relative humidity U, pressure p (in millibars), and temperature T using [13, 14] where q = r 1 + r = r wu 1 + r w U, (3) ( ) ( ) ew (T ) ew (T ) r w = ε = , (4) p e w (T ) p e w (T ) DRDC Atlantic TM

26 (a) 20 Jul x direction GEM (12:00) Sonde (12:29) (b) 20 Jul 2010 x direction GEM (12:00) Sonde (12:29) (c) 20 Jul y direction GEM (12:00) Sonde (12:29) (d) 20 Jul 2010 y direction GEM (12:00) Sonde (12:29) Height (m) Height (m) Height (m) Height (m) c eff (m/s) c eff (m/s) c eff (m/s) c eff (m/s) Figure 6: Effective sound speed c ef f as a function of height for (a) positive x-direction, (b) negative x-direction, (c) positive y-direction, and (d) negative y-direction, for measured (radiosonde) and modelled (GEM) atmospheric parameter profiles. and ( ) 17.67T e w (T )=6.112exp. (5) T Figure 6 is a plot of the effective sound speed as a function of height in the atmosphere c ef f (z) along the positive and negative x-directions for measured (radiosonde) and modelled (GEM) atmospheric parameter profiles. The sound speed as a function of height was first calculated from the temperature profile T (z): c(z)=c 0 T (z) T 0 (6) where T (z) is measured in Kelvin, c 0 = 331 m/s, and T 0 = 273K. (Equation 6 is strictly valid only for dry air, but the correction for a nonzero relative humidity will increase the sound speed by no more than 0.3% [5]). The effective sound speed as a function of height c ef f (z) plotted in Figure 6 is calculated to first order by summing the sound speed profile c(z) for still air and the component u(z) of the wind velocity in the direction of interest [5]: c ef f (z)=c(z)+u(z) (7) 8 DRDC Atlantic TM

27 The sign of the sound speed gradient c ef f / z determines whether the atmosphere is downward-refracting ( c ef f / z > 0) or upward-refracting ( c ef f / z < 0) [5]. In downwardrefracting conditions, there is a turning point for acoustic rays at some height in the atmosphere, so that in a ray model, rays interact repeatedly with the ground surface. The result is that at some ranges there are many possible ray paths and high received SPL, and at other intermediate ranges there are fewer or no rays and correspondingly lower received SPL. In upward-refracting conditions, rays are refracted away from the ground, and there is a region where no rays arrive called the shadow zone ; in practice, sound penetrates the shadow zone through diffraction and turbulent scattering [5]. For the profiles c ef f (z) along the positive and negative x-directions (Figures 6a and b), the radiosonde profile (red line) in Figure 6b has a generally positive gradient over the whole profile and is therefore downward-refracting; the other three profiles are upward-refracting. For the profiles c ef f (z) along the positive and negative y-directions (Figure 6c and d), the lower 150 m of the radiosonde profiles is downward-refracting, while the entire GEM profile is upwardrefracting. Atmospheric turbulence can also have an effect on acoustic transmission loss, and several options for modelling atmospheric turbulence are available in SPEBE [15]. The turbulence models selected use a combination of the Mann rapid distortion theory model [16] to generate velocity fluctuations caused by wind shear, the Hunt/Graham/Wilson modification [17] of the von Karman turbulence model to generate turbulence caused by buoyancy, and an isotropic von Karman model [18] to generate temperature fluctuations. The parameters used for model input were an inversion layer height z i = 1000 m, a friction velocity u = u 0 /20, and a surface-layer temperature scale T = u 0 /50, where u 0 is the surface wind speed. These correspond in general to SPEBE s sunny day profile case, and should therefore be reasonably representative of the conditions on July and July 23. The weather on July 22 was cloudy with some rain. Figure 8 is a sample of transmission loss calculated using SPEBE for a receiver at the centre of the image (indicated by ) and a source at any position within the image, using the measured atmospheric parameter profiles from 09:29 on 23 Jul The anisotropy in the modelled transmission loss to the west of the ACB are likely due to a combination of atmospheric refraction and the local topography (there is land to the west of the ACB). Transmission loss along the black line representing a Zodiac track in Figure 8 connecting the receiver with a source position ( ) is plotted as a function of range in Figure 9. 3 Results and discussion Representative model results for different atmospheric inputs and source frequencies are discussed in Section 3.1. The three types of recordings (moving source, stationary source, and impulsive source) were acquired to investigate different effects. The moving-source recordings (Section 3.2) were used to examine the effects on received SPL of source- DRDC Atlantic TM

28 400 (a) 23 Jul x direction 400 (b) 23 Jul 2010 x direction 400 (c) 23 Jul y direction 400 (d) 23 Jul 2010 y direction Height (m) Height (m) Height (m) Height (m) c eff (m/s) c eff (m/s) c eff (m/s) c eff (m/s) Figure 7: Effective sound speed c ef f as a function of height for (a) positive x-direction, (b) negative x-direction, (c) positive y-direction, and (d) negative y-direction, for measured (radiosonde) atmospheric parameter profiles. The resulting transmission loss calculated by SPEBE is plotted in Figure 8. receiver range, microphone height, and wind speed. The stationary recordings were used to attempt to quantify the time variability in received SPL by removing the range dependence (Section 3.3). A first attempt at acquiring impulsive source recordings in order to determine the effects, if any, of multipath propagation, is described in Section Overview of model results Model runs were performed at 530 Hz for all the available atmospheric parameter profiles. However, in order to determine whether the transmission loss differs between the two source frequencies of 530 Hz and 670 Hz, the SPEBE model was run at both frequencies for two sets of atmospheric conditions. Figure 10 is a plot of the modelled transmission loss as a function of latitude and longitude for 530 Hz and 670 Hz using a measured atmospheric profile (Figures 10a and b, respectively), and for 530 Hz and 670 Hz using a modelled atmospheric profile (Figures 10c and d, respectively). For a given atmospheric parameter profile, the plots at the two frequencies look qualitatively similar with downward refraction causing alternating rings of lower and higher transmission loss; however, the detailed shape of the transmission loss pattern differs between frequencies. The transmission loss along the black line in Figure 10 is extracted and plotted as a function of range from the ACB in Figure 11. There are significant differences in transmission loss for the two frequencies, especially for the case in Figure 11a. The downward-refracting 10 DRDC Atlantic TM

29 SPEBE output, 530 Hz, radiosonde launch, 09:29 on 23 Jul Latitude Longitude Transmission loss (db) Figure 8: Sample SPEBE output: transmission loss (db) as a function of latitude and longitude, for a receiver at the centre of the image ( ) and a source at any point in the image, with topographic contours (m) shown as solid black lines. The atmospheric parameter profiles used are shown in Figure 7. The transmission loss along the black dashed line for a source at the black circle ( ) is interpolated and plotted as a function of range in Figure Transmission loss (db re 1 m) Figure 9: Interpolation of transmission loss from sample SPEBE output shown in Figure 8. DRDC Atlantic TM

30 (a) 530 Hz, 12:32 on Jul 22, radiosonde 30 (b) 670 Hz, 12:32 on Jul 22, radiosonde 30 Latitude Transmission loss (db) Latitude Transmission loss (db) Longitude Longitude 100 (c) 530 Hz, 12:00 on Jul 23, GEM 30 (d) 670 Hz, 12:00 on Jul 23, GEM 30 Latitude Transmission loss (db) Latitude Transmission loss (db) Longitude Longitude 100 Figure 10: Modelled transmission loss centred on the ACB (white dot) for (a) 530 Hz, radiosonde data, (b) and 670 Hz, radiosonde data, (c) 530 Hz, GEM output, (d) 670 Hz, GEM output, with dates and times as indicated on the plots. The black line indicates the profile extracted and plotted in Figure 11. atmosphere results in local minima in transmission loss at different ranges for each frequency: 600 m, 1200 m, and 1800 m at 530 Hz, and 700 m and 1450 m at 670 Hz. The differences in transmission loss between the two frequencies in Figure 11b are generally less than 5 db. 3.2 Moving-source recordings Received SPL was calculated for the 20 moving-source runs for distances up to 2.2 km (the maximum audible range). Of the meteorological parameters that affect atmospheric acoustic propagation, the wind velocity (i.e., direction and speed) relative to the sourcereceiver direction most strongly affects the received SPL. In general, wind velocity varies with height in the atmosphere (e.g., Figures 6 and 7) and the SPEBE model takes the full vertical wind profile into account. In order to organize the datasets, a representative relative wind direction was calculated using the average wind velocity measured at the ACB during a given run. Using the source-to-receiver direction as a reference for the wind 12 DRDC Atlantic TM

31 Transmission loss (db) Transmission loss (db) (a) :32 on Jul 22, radiosonde 530 Hz 670 Hz (b) :00 on Jul 23, GEM 530 Hz 670 Hz Figure 11: Modelled transmission loss along the black line in Figure 10 (due east of the ACB) for 530 Hz (blue) and 670 Hz (red), for (a) radiosonde data and (b) GEM output. velocity vector, the downwind wind direction is 0, the crosswind direction is 90, and the upwind direction is 180. Table C.1 in Appendix C.1 is a summary of runs and relative wind velocity for all of the moving-source runs. Wind speed varied from 0.9 m/s to 5.0 m/s, and the relative wind direction varied from 7 (nearly downwind) to 180 (upwind). Figure 12 is a plot of received SPL as a function of range for all four raw TetraMic TM channels and the 530-Hz nominal horn tone. Generally, the highest received level is observed on the microphones most closely aligned with the run direction (microphones 3 and 4). For ease of interpretation, the remaining plots of received SPL as a function of range will only include the microphone measuring the highest received level. The moving-source recordings allowed for examination of the range dependence of the received SPL (Section 3.2.1), the dependence of received SPL on receiver height (Section 3.2.2), and the effect of changes in wind speed (Section 3.2.3) Range dependence of received SPL Figure 13 is a plot of received SPL as a function of range for a downwind run for the 530- Hz and 670-Hz nominal horn tones (Figures 13a and b, respectively). For the run in Figure 13, the receiver had been lowered to a height of 3.76 m above the water, and mounted on the east corner of the barge deck, because wind noise had been causing clipping in its original position. The maximum range at which the horn is detectable above the noise floor is 1800 m. The higher-spl spikes at 1200 m and 1450 m (Figure 13) may be due to acoustic convergence zones caused by a downward-refracting atmosphere. Since the DRDC Atlantic TM

32 Run 27, 530 Hz Mic1 Signal Mic1 Noise Mic2 Signal Mic2 Noise Mic3 Signal Mic3 Noise Mic4 Signal Mic4 Noise 12:00 GEM 12:32 Sonde 15:00 GEM Run Start/End: 13:08 to 13: Figure 12: Received SPL, noise levels, and modelled transmission loss in arbitrary db at 530 Hz as a function of range for a relative wind direction of 8 (downwind). The wind speed was 3.4 m/s, and the TetraMic TM was at 3.76 m height. absolute calibration of the receiver is not known, the modelled transmission loss can only be compared with measurements in a qualitative way. A similar pattern of maxima and minima can be seen in two of the SPEBE results (using 12:00 GEM and 12:32 radiosonde profiles), although the maxima and minima occur at different ranges than in the measurements. In contrast, the SPEBE output using the 15:00 GEM profile drops off smoothly with increasing range to a range of 1500 m, then suddnely drops by 10dB. Figure 14 is a plot of received SPL as a function of range for a crosswind run. The receiver was at a height of 3.76 m on the east corner of the ACB. At 700 m range, the received SPL drops suddenly by 15 db and remains low, but the horn is still detectable above the noise floor until the end of the run at 1300 m range (Figure 14). None of the modelled lines for transmission loss agree very well with the measurements: the GEM output does not show the sharp drop-off at 700 m range, and the radiosonde output shows maxima and minima that are not evident in the data. Figure 15 is a plot of received SPL as a function of range for an upwind run, with the TetraMic TM at a height of m. The received SPL first drops below the noise floor at 500 m range, is faintly audible until 600 m range, and then drops below the noise floor again. The modelled SPL using the radiosonde profile shows a steep drop-off between 100 m and 600 m range that is similar to the measurements, whereas the model results using the GEM profiles show an essentially linear decrease in received level with range. 14 DRDC Atlantic TM

33 (a) Run 27, 530 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:32 Sonde 15:00 GEM Run Start/End: 13:08 to 13: (b) Run 27, 670 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:32 Sonde 15:00 GEM Run Start/End: 13:08 to 13: Figure 13: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 8 (downwind). The wind speed was 3.4 m/s, and the TetraMic TM was at 3.76 m height. DRDC Atlantic TM

34 (a) Run 28, 530 Hz Mic4 Signal Mic4 Noise 12:00 GEM 12:32 Sonde 15:00 GEM Run Start/End: 13:38 to 13: (b) Run 28, 670 Hz Mic4 Signal Mic4 Noise 12:00 GEM 12:32 Sonde 15:00 GEM Run Start/End: 13:38 to 13: Figure 14: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 95 (crosswind). The wind speed was 2.7 m/s, and the TetraMic TM was at 3.76 m height. 16 DRDC Atlantic TM

35 (a) Run 10, 530 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:30 Sonde 15:00 GEM Run Start/End: 13:23 to 13: (b) Run 10, 670 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:30 Sonde 15:00 GEM Run Start/End: 13:23 to 13: Figure 15: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 177 (upwind). The wind speed was 3.3 m/s, and the TetraMic TM was at m height. DRDC Atlantic TM

36 3.2.2 Microphone height The effect of microphone height can be determined by comparing Figures 16 and 17, which were recorded with the receiver at m and 3.61 m height, respectively. The two recordings were made on the same day (23 July), with comparable wind speeds (4.9 m/s and 5.0 m/s, respectively) and in similar directions (source-to-receiver bearings of 274 and 316 true, respectively). The greatest effect of height is on the noise level, which is between -65 db and -70 db for the m receiver height, and between -70 db and -80 db for the 3.61 m height. Auditory inspection of the recorded tracks reveals that there are two main sources of noise on the recordings for m height: severe radio interference, and wind noise. For the recordings at 3.61 m height, the main source of noise is the sound of water lapping against the side of the ACB. As a result, the 3.61 m height recording is much cleaner Wind speed Most of the wind speeds experienced during the experiment were low compared to what might be experienced on the open ocean: the maximum wind speed during the experiment was 5.0 m/s (10 knots). Within the dataset there are some pairs of runs with low (< 2 m/s) and high (> 2 m/s) wind speeds and comparable relative wind directions. Figures 15 and 18 are plots of received SPL as a function of range with a wind speed of 3.3 m/s and 1.8 m/s (relative directions of 177 and 175 ), respectively. In both cases, the receiver was at m height. The maximum range for which the horn was detectable above the noise was 600 m for the 3.3 m/s wind speed, compared with 1000 m for the 1.8 m/s wind speed, suggesting that even at low overall wind speeds, a slight increase in wind velocity can result in a significant difference in maximum detectable range. 3.3 Stationary recordings In order to quantify the time variability in received SPL, three recordings were made with the source stationary; however, each recording was plagued by a different set of problems and none resulted in data that were entirely satisfactory. The first two stationary recordings took place while the Zodiac with the source was tied to a buoy 286 m away from the ACB (Point A in Figure 4), and the third recording took place with the Zodiac tied to the ACB with a rope at 119 m range (Point B in Figure 4). The specific details of the stationary measurements as described in the following paragraphs illustrate several of the difficulties of making acoustic transmission loss measurements over water from two moving platforms in the middle of a city. Figure 19 is a plot of the received SPL as a function of time for the first stationary recording, made on 22 July while the Zodiac was at Point A (Figure 4). Received SPL measurements that experienced clipping are plotted as hollow plot symbols, while unclipped data are 18 DRDC Atlantic TM

37 (a) Run 32, 530 Hz Mic4 Signal Mic4 Noise 09:00 GEM 09:29 Sonde 12:00 GEM Run Start/End: 11:35 to 11: (b) Run 32, 670 Hz Mic4 Signal Mic4 Noise 09:00 GEM 09:29 Sonde 12:00 GEM Run Start/End: 11:35 to 11: Figure 16: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 140. The wind speed was 4.9 m/s, and the TetraMic TM was at m height. DRDC Atlantic TM

38 (a) Run 34, 530 Hz Mic3 Signal Mic3 Noise 12:00 GEM 15:00 GEM Run Start/End: 14:11 to 14: (b) Run 34, 670 Hz Mic3 Signal Mic3 Noise 12:00 GEM 15:00 GEM Run Start/End: 14:11 to 14: Figure 17: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 147. The wind speed was 5.0 m/s, and the TetraMic TM was at 3.61 m height. 20 DRDC Atlantic TM

39 (a) Run 19, 530 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:39 Sonde 15:00 GEM Run Start/End: 12:45 to 12: (b) Run 19, 670 Hz Mic3 Signal Mic3 Noise 12:00 GEM 12:39 Sonde 15:00 GEM Run Start/End: 12:45 to 12: Figure 18: Received SPL, noise levels, and modelled transmission loss in arbitrary db as a function of range for (a) 530 Hz, and (b) 670 Hz, for a relative wind direction of 175. The wind speed was 1.8 m/s, and the TetraMic TM was at m height. DRDC Atlantic TM

40 plotted as filled symbols. The recording shown in Figure 19 suffered from two significant sources of noise: clipping, due to the wind speed of 6.1 m/s; and a resonant whistling noise that originated from a pipe attached to the structure of the ACB. The pipe had a series of holes in the southeast-facing direction, so that when the wind blew from the southeast (as it did during the dataset shown in Figure 19), the pipe acted like a flute and produced tones in the range of Hz that were comparable to or greater in intensity than the received level from the horn. The stationary recording was repeated on 23 July with a different wind direction, in an attempt to overcome the problems encountered in the first recording. Figure 20 is a plot of the received SPL as a function of time for the second stationary recording, made while the Zodiac was tied to the same buoy (Point A in Figure 4). However, the wind blew the Zodiac to the far side of the buoy (relative to the ACB) during the recording. The direct line-ofsight to the ACB was blocked by the buoy and the Zodiac was unable to maintain a constant heading. The horn was not omnidirectional; therefore, the combination of the moving Zodiac and the buoy blocking the direct line-of-sight resulted in unusable measurements. The third stationary recording was made on 23 July with the Zodiac and source at Point B. The recording was started once the Zodiac came to a stop with a stable heading. Figure 21 is a plot of received SPL as a function of time for the third recording. However, noise was again a problem: the wind speed was relatively high (5.8 m/s), resulting in substantial clipping as indicated by the hollow plot symbols in Figure 21; and there were audible commercial radio signals being picked up on the recording; therefore, it is not clear that the data are reliable. 3.4 Impulsive sources In order to attempt to separate multipath arrivals, latex balloons were used as impulsive sound sources. The Zodiac with the balloons was drifting at 128 m range from the ACB, at Point C in Figure 4, and the balloons were popped at intervals of s at a height of 2 m above the water surface. Figure 22 is a plot of received SPL for each TetraMic TM channel as a function of time for each of four impulsive source events. As shown schematically in Figure 3, Microphones 3 and 4 were approximately facing the Zodiac, while Microphones 1 and 2 were facing away from the Zodiac; Microphones 1 and 4 were facing upward while Microphones 2 and 3 were facing downward. As a result, the direct-path arrival is barely detectable on Microphone 1; however, Microphone 1 measures the highest level for a multipath arrival approximately 20 ms after the direct-path arrival. The time difference between the two arrivals for each impulsive source event is shown in Table 1, and is quite stable with an average value of 23.4 ms. 22 DRDC Atlantic TM

41 (a) Run 25, 530 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 11:29 to 11:34 Mic Time since the start of the first blast (s) (b) Run 25, 670 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 11:29 to 11:34 Mic Time since the start of the first blast (s) Figure 19: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the first stationary recording, taken at Point A (Figure 4). The TetraMic TM was at m height and the wind was 6.1 m/s from 121 T. DRDC Atlantic TM

42 (a) Run 30, 530 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 11:00 to 11:05 Mic Time since the start of the first blast (s) (b) Run 30, 670 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 11:00 to 11:05 Mic Time since the start of the first blast (s) Figure 20: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the second stationary recording, taken at Point A (Figure 4). The TetraMic TM was at m height and the wind was 3.5 m/s from 302 T. 24 DRDC Atlantic TM

43 (a) Run 33, 530 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 12:54 to 12:59 Mic Time since the start of the first blast (s) (b) Run 33, 670 Hz, received level ( ), clipped received level ( ), noise ( ) Mic1 Mic2 Mic3 Run Start/End: 12:54 to 12:59 Mic Time since the start of the first blast (s) Figure 21: Received SPL and noise levels in arbitrary db as a function of time at (a) 530 Hz, and (b) 670 Hz, for the third stationary recording, taken at Point B (Figure 4). The TetraMic TM was at m height and the wind was 5.8 m/s from 321 T. DRDC Atlantic TM

44 (a) Mic1 Mic2 Mic3 Mic4 (b) Time (s) Mic1 Mic2 Mic3 Mic4 (c) Time (s) Mic1 Mic2 Mic3 Mic Time (s) (d) Mic1 Mic2 Mic3 Mic Time (s) Figure 22: Received SPL (arbitrary units) for each receiver channel for four impulsive noise events (a-d). The times of the direct and refracted arrivals are plotted as red and blue vertical lines, respectively. 26 DRDC Atlantic TM

45 Table 1: Summary of balloons and time differences ΔT between direct-path and refractedpath arrivals. Balloon number ΔT (ms) ± ± ± ± 0.03 In order to determine whether the strong multipath arrival was due to a surface reflection, the path length difference between the once-reflected path and the direct path was calculated using the method of images. The resulting path length difference of 0.32 m coupled with the speed of sound of m/s (calculated using the measured temperature) would result in a delay between the direct-path and once-reflected path of 0.9 ms much smaller than the observed delay of 23.4 ms. In addition, the multipath arrival was seen most strongly on Microphone 1, one of the upward-facing microphones. Therefore, it is likely that the second strong arrival is either a downward-refracted path or a reflection from the structures above the barge roof (see Figure 23). The approximate synchronization of the two audio tracks with the GPS positions meant that it was not possible to calculate the total travel time with sufficient accuracy to obtain additional insights into the propagation. 4 Challenges and future work The experiments and modelling described in Section 3 were useful as a preliminary investigation into atmospheric acoustic propagation over water. Section 4.1 outlines some of the challenges faced when making atmospheric acoustic transmission loss measurements over water, and explores solutions that were either implemented in the present experiment or planned for future experiments. Future work is outlined in Section Challenges and solutions The challenges encountered in this experiment were valuable for identifying improvements when designing future atmospheric acoustic experiments. The major challenges encountered include: time synchronization, source and receiver directivity, analog-to-digital (A/D) conversion problems, source stability, radio interference, and wind noise. Each of these is described in detail in the following paragraphs. The speed of sound in air is nominally 340 m/s; therefore, at a range of 1 km sound takes 3 s to travel from the Zodiac to the ACB. In order to accurately detect faint horn blasts in DRDC Atlantic TM

46 Figure 23: The ACB, with the approximate location of the TetraMic TM indicated by a red circle. the receiver track, and correlate the blasts with source position and weather information, it was necessary to synchronize the two recorded tracks to the time recorded with the GPS receiver. The tracks were synchronized in post-processing using the three short, quick blasts sounded on the horn when the Zodiac was near the ACB (less than 20 m away) at the beginning and end of each run. The time at which the third timing blast began was noted in the logbook. In the post-processing phase, cross-correlations were performed on the monitor and receiver tracks (described in more detail in Appendix B.1) to find the lag between the tracks and then calculate the received level for the appropriate time window containing the horn blast; this procedure worked very well. In future experiments, the recording equipment should be outfitted with a time synchronization track that will be recorded along with the audio tracks. Neither the source nor the receivers were omnidirectional. Preliminary tests performed before the experiment revealed that the horn source level was approximately 5 db re 20 μpa lower in the aft direction than in the forward direction. The relative levels of the two fundamental tones also differed from fore to aft. Depending on winds and currents, the Zodiac was not necessarily facing directly toward or away from the ACB, and there was no way of measuring the heading of the Zodiac; therefore, there is some uncertainty in the source level along the line joining the Zodiac and the ACB for any given run. For future experiments, it will be necessary to either use an omnidirectional source, or have a way to monitor the source heading during each run. 28 DRDC Atlantic TM

47 The receiver consisted of four microphones with cardioid pickup patterns facing in different directions as described in Section 2.1 and shown schematically in Figure 3. If the heading of the ACB were accurately known, a correction could be applied to account for the microphone sensitivity in the direction of the Zodiac. The overall heading of the ACB was calculated using the GPS units at each corner, which were acquiring position data once per second. A GPS data quality check was performed by calculating the range between the two corner GPS units, fixed at a separation of m. The range calculated on each pair of GPS points varied from m to m, and the calculated instantaneous heading of the ACB varied between -67 and 2. In reality, it is unlikely that the ACB heading varied by more than ±10 from its mean value of 315, since it is moored at all four corners. By eliminating GPS points with ranges outside a fixed window around the true separation (e.g., ±2 m or ±7 m), the calculated instantaneous ACB heading varied less; however, eliminating data points resulted in time periods in which the ACB heading was unknown. The two-corner GPS solution for measuring barge heading was theoretically sound, but in practice it would have required GPS units with greater accuracy than the stated ±5 m, such as differential GPS units [19]. The TetraMic TM is one of a class of soundfield microphones intended to be used by beamforming the four raw tracks to provide four new tracks: one omnidirectional and three dipole tracks. Instead of correcting for the cardioid pickup patterns on the raw microphone tracks, the beamformed omnidirectional track could then be used for analysis, and the ACB heading would become irrelevant. However, the data acquisition hardware was not well synchronized between tracks, and appeared to drop samples on some tracks: in other words, the four raw tracks differed in length for many of the recordings. The dropped samples did not occur consistently on one particular track, and it was unknown whether the samples were dropped at the beginning, middle, or end of the recording (or some combination). The numbers of dropped samples ranged from 32 to 192, corresponding to 0.7 ms to 4.0 ms delay between tracks at 48 khz sampling frequency, affecting the beamforming. Therefore, the analysis of data beamformed in post-processing was not pursued. The acquisition hardware will be upgraded to avoid synchronization problems in future experiments. The source frequency was not stable and the instability appears to be temperature-related. The horn frequencies were lower when the ambient temperature was lower, and the source frequencies increased over the course of a run, sometimes by as much as 5 to 10 Hz, suggesting that heating of the horn caused by repeated blasts was changing the fundamental horn frequencies. There were also occasions when the source suffered from an audible vibrato effect during a blast. The source was monitored by using a Sony PCM recorder on board the Zodiac; however, for future experiments it would be better to obtain a more stable, omnidirectional source. The short (1 m) analog run of receiver wires from the microphones to the A/D conversion hardware acted as an antenna for radio signals, resulting in audio recordings with static, buzzing noises, and audible pickup of commercial radio transmissions. The problem was DRDC Atlantic TM

48 reduced to some extent by shielding the affected wires; however, it was not completely eliminated. Part of the problem was that the experiment took place in the middle of the city where there are many sources of radio frequency (RF) interference; it is expected that measurements in more remote locations will not suffer from the same effects. As mentioned above, the data acquisition hardware is also being redesigned, and shielding from RF interference will be among the many design considerations. The windscreen provided by the microphone manufacturer was used to reduce wind noise (Figure 2); however, the shape of the four-microphone array resulted in there being less foam between the microphones and the turbulent outside air than the 4- or 5-cm thickness of foam often used for in-air acoustic measurements. In addition, the analog gain knob on the A/D converter could only reproducibly be set to its maximum setting, and the maximum gain setting resulted in the wind noise causing clipping when the wind speeds were above about 4-5 m/s, or when there were significant gusts. A larger windscreen should be used for future measurements, and the data acquisition system redesign will allow for a variable gain setting with fixed and repeatable increments; however, wind noise is a universal challenge to air acoustics and no perfect solution has yet been found. 4.2 Future work The receivers need to be calibrated in order to make a quantitative comparison between modelled and measured transmission loss. The surface roughness data acquired with the ADCP have not yet been analyzed, and a more realistic surface scattering and loss model should be explored. The modelling effort should be expanded to complete the modelling for the 670-Hz source frequency. A simple ray tracing model should be developed to determine the origins of the multipath arrivals described in Section 3.4. The horn also has strong harmonics extending as high as 10 khz; the analysis could easily be expanded to include the transmission loss of the horn harmonics. In preparation for anticipated future experiments, the data acquisition hardware is being redesigned (Section 4.1), alternative sources are being considered, and new propagation models are being evaluated. 5 Conclusions Atmospheric acoustic transmission loss was measured as a function of range in a variety of atmospheric conditions. For a 115-dB re 20 μpa at 1 m source, maximum audible ranges at wind speeds of 3 m/s were 1800 m for the downwind propagation direction, 1300 m for the crosswind direction, and 600 m for the upwind direction. Two different types of atmospheric parameter profiles were used as inputs to an atmospheric acoustic propagation model: profiles measured with a radiosonde, and modelled 30 DRDC Atlantic TM

49 profiles from Environment Canada s GEM model. Comparisons between measured and modelled transmission loss were qualitatively good for the downwind and upwind runs, but not for the crosswind run; however, given the limited number of runs overall, it would be rash to generalize further. Differences were observed between the predicted transmission loss calculated using measured and modelled parameter profiles, but neither provided consistently better transmission loss estimates. The effect of microphone height and wind velocity on received SPL were examined. Noise increased by 5-15 db when the microphone height was increased from 3.61 m to m height. Maximum audible range for the source decreased from 1000 m to 500 m when the wind speed increased from 1.8 m/s to 3.3 m/s. Stationary measurements were made with the intention of quantifying the variability in received SPL when the source-receiver range was fixed. However, problems with source positioning and background noise resulted in data that are inconclusive. Latex balloons were used as impulsive sound sources, resulting in identification of multipath propagation to the receivers. The path for the strong multipath arrival is unknown, but it does not originate from the air-water boundary. Numerous technical challenges were overcome during the experiment and the lessons learned are guiding upgrades to existing equipment, driving future equipment purchases, and influencing experiment designs. DRDC Atlantic TM

50 This page intentionally left blank. 32 DRDC Atlantic TM

51 Annex A: Equipment Table A.1 is a summary of brands and model numbers of equipment used in the experiment, and instrument settings where relevant. Table A.1: Details of experimental equipment. Make and Model Serial Number Comments Brüel & Kjær db range: 2260 Observer 30.7 db db (ambient), 40.7 db db (source level) Bandwidth: 1 3 -octave Peaks over: 110 db Time: broadband status (fast), spectrum measurement (fast) Frequency: broadband measurement (A&L), broadband status (L), spectrum measurement (L) Measurement time: 1:00 (ambient), 0:10 (source level) Microphone S. I. Correction: Frontal Core Audio TetraMic TM 2122 Garmin Oregon 55T GlobalSat BU-353 USB BUG BUG Nautilus 3500 horn Part No Sony PCM D-50 MEA recorder #2 Linear PCM Recorder f samp = 44.1 khz Teledyne RD Instruments khz ADCP WHS300-1-UG22 Operating in WAVES mode Vaisala WXT520 D s averaging interval meteorological station DRDC Atlantic TM

52 This page intentionally left blank. 34 DRDC Atlantic TM

53 Annex B: Data Processing The following sections contain detailed descriptions of the data processing steps: calculations for the time synchronization of the recordings (Appendix B.1), calculation of the range from the ACB to the Zodiac (Appendix B.2), processing of the monitor tracks to find the horn blast start and end times (Appendix B.3), calculation of the received SPL (Appendix B.4), and calculation of the receiver noise level (Appendix B.5). B.1 File synchronization The monitor and receiver audio tracks were synchronized by calculating the cross-correlation between the two files for the three short, closely-spaced timing blasts. A section of each audio file was analyzed, with a buffer of 10 s before the third blast and 5 s after the third blast. The offset between the files τ of fset was already known to the nearest second from preliminary visual observation of each file. The envelope of the timing blasts was extracted using a low-pass filter on the squared pressure time series from each file. The cross-correlation between the enveloped time series was calculated, and the lag was taken to be the time τ max on the lag interval [ 1,1] at which the cross-correlation reached a maximum value. The lag between the files was therefore the sum of τ max and τ of fset. The lag between monitor and receiver track was calculated once for each TetraMic TM channel (i.e., four times for a given run), and the mean lag for each run was calculated by averaging the resulting four lag estimates. B.2 Position data The distance from the ACB to the Zodiac was calculated using the Haversine formula, in which the great-circle distance Δr between two locations with latitudes φ 1 and φ 2 and longitudes λ 1 and λ 2 is given by: ( ( ) ( ) ) Δφ Δλ Δr = 2Rsin 1 sin 2 + cosφ 1 cosφ 2 sin 2 (B.1) 2 2 where R is the radius of the earth, Δφ = φ 2 φ 1, and Δλ = λ 2 λ 1. All angles in Equation B.1 are in radians, and the units chosen for the radius of the earth R will determine the units of the great-circle distance Δr. The GPS at the east corner of the ACB was closest to the receiver (3.48 m away) and was used as the receiver location. There was a short time on 21 July when the east corner GPS was not logging data; during that time the west corner GPS was used as the receiver location. DRDC Atlantic TM

54 B.3 Blast start and end times The Sony recorder on the Zodiac was used as a monitor track in order to calculate the start and end times for each horn blast. The monitor track was examined manually in order to determine (to the nearest second) two times bounding the blast series: the time of the start of the first data blast (i.e., not the timing blasts), and a time after which there were no more data blasts. A MATLAB script was then used to automatically extract the sample number of the start and end of each horn blast on the monitor track by looking for rising and falling blast edges as follows: 1. Read in the section of the.wav file centred at the first blast start time ± 0.5 s, 2. Bandpass filter between 480 Hz and 720 Hz, 3. Calculate the spectrogram P( f,t) using a 4096-point FFT, 4. For each nominal horn frequency, determine which frequency bin f max had a maximum in P( f,t) within a 50-Hz window surrounding the nominal frequency, when averaging P( f,t) over time 5. Detect the time at which P( f max,t) rises above a threshold P thresh : First blast: P thresh = 0.5(max[P( f max,t)] + min[p( f max,t)]) Subsequent blasts: P thresh =< P below > +0.8(< P above > < P below >), where < P above > is the mean value of P( f max,t) above P thresh, and < P below > is the mean value of P( f max,t) below P thresh 6. Examine the series of P( f max,t) relative to P thresh : If P( f max,t) > P thresh for all t or P( f max,t) < P thresh for all t, then we have not detected an edge. Determine whether we are guessing too early (P( f max,t) < P thresh ) or too late (P( f max,t) > P thresh ). Adjust the guess accordingly by 0.5 s and repeat the above series of steps. Otherwise, we have found an edge at time t edge at which the threshold is first exceeded. 7. Use the measured blast start time to calculate a guess for the start time of the next blast, and repeat until the end of the data is reached The edge times t edge are measured relative to the start time of the.wav file. They can be converted into UTC time using the synchronization pulses. The same analysis is used to find the blast end times with logic adjusted accordingly. 36 DRDC Atlantic TM

55 (T 2 -T 1 )-2T B 3 s 3 s Noise #1 2T B T B Signal (~ 3 s) 2T B Noise #2 Detected horn start (T 1 ) Detected horn stop (T 2 ) Figure B.1: Diagram of signal and noise calculation windows. The blue window is the 3-s signal window, and the two red windows are the 3-s before and after noise windows. The buffer time used in the analysis is T B = 1 s. The actual signal is plotted in blue on the bottom half of the diagram. B.4 Received SPL Once each horn blast s start and end time were detected on the monitor track, the corresponding start and end times on the receiver track were calculated, taking into account the range to the ACB and the speed of sound in air at the temperature measured during the run. A buffer of 1 s was used on both ends of the time window for calculating received SPL so that any edge effects from the turning on or off of the horn could be avoided (see schematic diagram in Figure B.1). Therefore, for a blast of exactly 5 s, the middle 3 s of data were analyzed; in practice, the length of the blast varied somewhat because the blasts were timed by a human with a watch. For each blast, approximately 3 s of data were read in from the receiver file using MAT- LAB, and the power spectral density was calculated using Welch s method [20] with a 4096-point FFT. The 32-bit floating-point samples of the pressure time series P(t) in the.wav files ranged in value from +1 to 1. The SPL estimate was marked as clipped if more than of the pressure samples were clipped (that is, P(t) > 0.99). The power spectral density was then averaged across the frequency band of interest (525 ± 25 Hz and 677±25 Hz for each of the fundamental horn tones). The width of the frequency band was chosen to be wide enough to account for the Doppler shift (based on the maximum Zodiac speed) as well as the maximum frequency drift of the horn. DRDC Atlantic TM