CAPTURING ANCIENT THEATERS SOUND SIGNATURE USING BEAMFORMING

Transcription

1 CAPTURING ANCIENT THEATERS SOUND SIGNATURE USING BEAMFORMING Mojtaba Navvab College of Architecture and Urban Planning, University of Michigan, 2 Bonisteel Blvd. Ann Arbor, MI, USA, moji@umich.edu Fabio Bisegna and Franco Gugleirmetti Dept. Astronautical, Electrical and Energetic Engineering, University of Sapienza, Via Eudossiana, 8-84 Roma, Italy Ancient theatres are used as venues for a variety of live performances, and there is a demand for detailed information relating to sound fields within these remarkable historical buildings. The objective of this study is to demonstrate the ability to visualize sound fields that represent the characteristics of all architectural elements and their contributions to the room acoustics within historical places. This study presents application of beamforming combined with computer simulation as a close numerical examination to provide evidence for the relevant acoustical aspects of ancient theatres, based on modern room acoustic indicators. Beamforming technique is used to capture the sound signatures of internationally well-known Flavian Amphitheatre Rome Coliseum. Results of measured room acoustic characteristics are compared to the latest room acoustic indicators according to ISO standard. In addition to historical preservation, the simulation results, based on selected parametric studies, inform the enhancement of the current use of the spaces during public events. Real time measured data are used to visualize the audible impact of architectural elements, and data is used as inputs to simulation models representing these acoustic conditions within a virtual environment for the public to experience the past.. Introduction There is a strong demand from the public for access to the outdoor archaeological sites during daytime and even more during night time, both for archaeological visits and for the organization of several types of cultural events, ranging from concerts and symposia to sport or comedy shows. Most archaeological sites cannot sustain such activity along with their in-progress cultural heritage work. The building officials and historical societies that manage the use and operation of these ancient sites or theatres have outlined a series of guidelines for their use by the public that includes the role of acoustics and lighting techniques in the modern use of these ancient places. Many researchers in pursuit of accurate reconstruction through possible alternatives of material and design evolution have studied the acoustic properties of ancient performance spaces. This paper describes a new approach in measuring and mapping the contribution of sound reflected component in frequency and time domain for evaluation and virtual recreation of ancient theatres. 2. Methodology Architecturally speaking, not all ancient theatres are designed equally since the space geometry for each period varies. Most significantly, for our purposes, there are noticeable differences between

2 an open-air theatre and an amphitheatre []. The contributions from direct and reflected sound components using impulse (large scale balloon and sweepers) as a source was determined along with the maximum equivalent A-weighted sound pressure level (SPL) for various locations within the space. Additionally, a Head Related Transfer Function (HRTF) [2, 3] was utilized along with the use of Noise Image Software [4] in order to determine which portions or direction of the viewing areas were contributing to the majority of the sound pressure level reaching a particular receiver s location or source localization due to the sound source position. 2. Space Geometry The Anfiteatro Flavio (Roman Coliseum) better known as the Coliseum was built by between the 72 (Vespasiano), and the 8 D.C., with Tito. It is located in the valley among the hills of Palatino, Esquilino and Celio, previously part of the Domus Aurea. It was built in the shape of an ellipse, with the major axis 88m, the minor axis up to 56m, a circumference of about 527m, and a height of about 5m. The inner arena is about 86x54m. Different theories estimate its capacity as anywhere from 4 persons to more than 7. The inside is almost in ruins, without the surface of the arena, while the outside walls still exist, given damage by lighting, earthquakes, and fire. In ancient times, it was covered by an enormous curtain (velarium) to protect the visitors from the sun. It was used with its original function for the last time during the reign of Teodorico, in the VI sec., while in the medieval era it was used as a fortress before being despoiled from its exterior material (marble). Current condition of the Coliseum is shown in Figure. The partial sand covered arena is used as a stage for music performance. Figure : Surface characteristics within the current condition of the Coliseum. 2.2 Acoustical measurements on-site and model calibration Different acoustical measurements were made inside the Coliseum. Single calibrated sound level meters were used to measure the background noise levels during the course of the day from people or traffic. The background noise level is based on measurements on 6- microphone positions for the site. A Brüel & Kjær sound level meter was type 277 with dual-channel input (microphone, sound intensity probe, accelerometer or direct signal) with 4.2 Hz to 22.4 khz broadband linear frequency range and 6.6 to 4 db at A-weighted dynamic range with supplied microphone, type 489. Sound fields within the arena were mapped using a spherical beamforming array with the following specifications given this particular application: 2 microphones,.6 cm diameter, and Carbon fibre structure. Dynamic of the microphones is within 35 db -3 db. Recommended mapping frequencies are between khz - 2 khz. Typical measurement distance from source is.4-2 m. Data recorder is at 92 khz with sampling frequency of 48 to 44 channels, per -inch rack (24 channels per card), Integrated PC with Windows 7. (embedded) and operated with Noise Image 2 ICSV23, Athens (Greece), -4 July 26

3 (NI) software using mobile power supply/battery pack. The reverberation time and most indicators recommended by the ISO standard [5] were computed based on the data from the impulse response measurement, utilizing this 2-microphone spherical beamforming array using the Beamforming technique. The acoustic images consist of colour contours indicating where the most significant noise sources are located. Detailed review of this technique is explained in references [4]. Figure 2 shows the typical system components used by the Acoustics Camera and the actual experimental setting for measurement within the site. The system consists of a spherical microphone array with 2 channels and data recorders with a camera, paired with noise image software running on a laptop PC. Microphone Array Sphere48 Calibration Tester PC Video Camera Data Recorder mcdrec 72 Extra Sensors Figure 2: Acoustic Camera and its actual set up within Rome Coliseum. 2.3 Application of the Head Related Transfer Function Head Related Transfer Function (HRTF) describes how a given sound wave input is filtered by the diffraction and reflection properties of the body before the sound reaches the eardrum. Hearing with both ears is defined as binaural hearing. The sound that is received from both ears localizes the direction and true spatial perception of sounds. The geometrical parameters of the human body have a major impact on binaural hearing. The application and the analysis of the specifically collected room acoustic measurements by the Acoustic Camera allows examination of this concept and the creation of a possible new room acoustic index that captures the sound signature of a space. 2.4 Computer Simulations Based on historical records and on-site measured data, different CAD models of the interior space of the Coliseum were created to meet the requirements for acoustic modelling and simulations. The computer models were based on the condition of having the source on the stage, a scenario often required the simulation of the theatre fully occupied to account for total absorption of the space per past historical record. The measured results were used as input data to simulate the sound source in other computer software programs. An acoustic program was used for detailed calculation and parametric studies. Materials applied were estimated from architectural records and observation. The collected data was used as a database and input toward various simulations. EASE program is based on the hybrid method, a combined method of image source models (ISM) and a ray tracing method (RTM) [6]. This method allows using the ray tracing process to find a receiver hit by a ray. The objective with this analysis was to create a model that would allow exploring use of different surface materials while observing the differences in room acoustic conditions. The selected materials from the EASE library provided the acoustical absorption in the desired frequency range from 63 Hz to 8 khz in octave bands. To observe the room geometry and to find possible reflection sequences, the ray tracing method is applied. Figure 3 shows example of EASE output and ray distribution with the current surface of the Coliseum. The hybrid model is the method used in EASEaura, the main computer-simulation software in this research. Briefly, this model can be described as running a specular ray tracing process, which finds a receiver hit by a ray. As a result, the corresponding image source must be audible [6]. ICSV23, Athens (Greece), -4 July 26 3

4 Figure 3: Computer simulation of ray tracing and scale model methods for Rome Coliseum. 2.5 Room Acoustic Parameters One of the key parameters is clarity (C 5, C 8 ) which is the ratio of energy before and after 5, and 8 milliseconds (ms) in decibels respectively. C5 is used to predict intelligibility of the spoken word. The clarity corresponds to early-to-late arriving sound energy ratios at 5 ms or an 8 ms early time limit, depending on whether the results are intended to relate to conditions for speech or music, respectively. It is recognized that the Coliseum is an open-air theatre that requires measurements in real time. Calculating the reverberation times was made possible using the beamforming and NI software. Sound field mapping made it possible to compute the L p (pressure level) and L W (power) at any given point. Sound Strength, G, is defined as the gain from the sound pressure level, which is produced by the same omni-directional sound source, with the same power level, in a free field at a distance of m from the sound source. To comply with this criterion, the power level was measured outside of the site with the identical source. The measured L P (pressure level) and L W (power) at a selected point is used to estimate the gain (G) from the sound pressure level using the recommended procedure within ISO standards. Table I is a list of all parameters for description of the acoustical quality in rooms (ISO/DIS 3382) [5]. Table I: Parameters for description of the acoustical quality in rooms (ISO/DIS 3382) T3/s EDT/s D/% C8 Ts/ms G/dB LF/% LFC/% = Reverberation time, computed from -5 to -35 db of the decay curve = Early Decay time, computed from to - db of the decay curve = Definition, Early (-5ms) to total energy ratio in db = Clarity (C 7, C 5, C 8 ) the ratios in db of energy before and after 7, 5, 8 milliseconds = Centre Time, time of. moment of the energy impulse response = S Strength level related to Omni-directional free field radiation over m distance = Early lateral (5-8ms) energy ratio, cos 2 (lateral angle) = Early lateral (5-8ms) energy ratio, cos (lateral angle) 3. Results & Analysis The Coliseum was selected for its capacity seating area for the spectators. Specific measurements and simulations were made for its room acoustic performance as presented for selected variables in Figures 4 to 7. As part of the room acoustic analysis the sound conditions inside the space including the peak impulse measurements were captured similarly to crowd noise. Measured data using AC were converted from pressure units to relative sound intensity level distribution and presented equal to the HRTF distribution as described in section 4. The results for crowd noise in the Coliseum and impulse or use of sweeper as a source were measured and plotted for various slices of time and selected locations within the arena; related examples are shown in Figures 4. The peak noise levels are clearly identified in purple colours in 3D architectural representations. The sound intensity from the surrounding crowd noise and any other contribution from wall surface areas as viewed by 4 ICSV23, Athens (Greece), -4 July 26

5 Reverbration (Sec) The 23 rd International Congress on Sound and Vibration spherical microphone array are shown in Figure 4 (left and right). During the time allocated for our studies, the sound in the Coliseum was measured to predict as to what impact the planned renovations/restoration would have in restoring it to its original appearance and material. Various parametric studies were conducted. Acoustic Photo files were created for as many locations and time intervals required for the computer simulations, using AC data, which was post-processed to eliminate background noise. Using 3D acoustic views and sound wave behaviour, the simulations facilitate the visualization of the data. Figure 4: Measured crowd noise propagated in Coliseum, Purple = high, Blue = low sound pressure levels in dba, and sound generated by sweeper and surfaces exposed to first reflection. Tourist participation was controlled by the announcement made during the measurements and the public willingness and effort to create a loud noise when requested. If all individuals or 5, original occupancy spectators during ancient times had yelled at the same intensity, the measurement would have increased to dba, Leq, within the centre of the Coliseum, which is a significant sound increase. The reverberation times are estimated based on the simulation of the existing condition of the Coliseum and its original conditions; with and without the spectators. Parametric studies using computer simulation with roof top cover full open and closed are reported. The calculated and measured reverberation times for selected positions within the Coliseum at various band of frequency are shown in Figure 5. The measured results while utilizing the beamforming technique allow for comparison with simulation results since there were no other unique data available in literature for the Rome Coliseum. The measured RT values (Fig. 5 right side) throughout the arena were used to calibrate the simulated room acoustic models (Fig.5 left side) and selected materials used in parametric studies for wall surface characteristics are shown in Table I below. Colosium Material Choice SU-Layers Zones Existing option- Main body Surface Brick Unglazed Brick Unglazed Seats Broken a=% Concrete Rough Floor Horizontal a=% Concrete Rough Stage-vertical Brick Unglazed Brick Unglazed Vertical- Brick Unglazed Brick Unglazed Vertical-st Brick Unglazed Brick Unglazed Vertical Vertical-2nd Brick Unglazed Brick Unglazed Walls Vertical-2.5 Brick Unglazed Brick Unglazed Vertical-3rd Brick Unglazed Brick Unglazed Vertical-3.5 Brick Unglazed Brick Unglazed Stage Stage Soil Rough Soil Rough Stage-under Soil Rough Soil Rough Opening Opening Opening Opening Top Opening Opening Fabric Shade Occupancy Empty Empty Fully Occupied Below Arena Above Arena.5 Arena W/Sweeper.5 Above W/Sweeper Frequency (Hz) Figure 5: RTs time estimates based on simulation and the measured data within Rome Coliseum. Proposed procedures in ISO/DIS 3382 were used to calculate the reverberation, clarity and sound strength G for selected positions and various distances away from the centre stage made within the arena. Figure 6 shows the calculated clarity of C 5 and C 8 based on measured data. The estimated sound strength (G) values for selected points are identified on a grid point shown as a top view plan; and also an example of parametric study of fabric shades impact on sound are presented in sound mapped image format in Figure 7. The results show the contribution of reflected sound and the decrease in clarity and G levels as a function of distance from the source location. Measurement locations close to wall surface show higher delta or more sound strength as shown in plotted data in Figure 7. Such data including reverb times and early decay times are useful for positioning of loud speakers in an open-air theater during special musical events by sound engineers. The C 8 values ICSV23, Athens (Greece), -4 July 26 5

6 Sound Strngth (G) Clarity (C 5 ) Clarity (C 5, C 8 ) Clarity (C 8 ) The 23 rd International Congress on Sound and Vibration expressed in decibels in the 5, and 2 Hz bands are critical for music application. The C 5 and C 8 clarities estimations are presented on the right and left side of Figure 6 respectively. Their difference for upper and lower level seating area are shown in the middle of the Figure 6., These differences show the contributions from the remaining walls within the Coliseum to the sound level energy at critical frequencies for speech and music Frequency (Hz) Pos-2 Pos-3 Pos-4 Pos-5 Pos-6 Pos C5_AVG-UP C5_AVG-DN C8_AVG-UP C8_AVG-DN Frequency (Hz) Frequency (Hz) Figure 6: Sound clarity (C 5, C 8 ) estimates based on the measured data within Rome Coliseum Pos- Pos-2 Pos-3 Pos-4 Pos-5 Pos Distance from Centre stage (m) G Rome Poly. (G Rome) Figure 7: Sound strength (G) estimates based on the measured data within Rome Coliseum. 4. Capturing Space Sound Signatures This section describes the process of obtaining the similar data to HRTF and its conversion to Spatial Frequency Response Surfaces (SFRS) through the format used by MATLAB data files of compensated Head-Related Impulse Response (HRIR) measurements. The HRTF varies with azimuth and elevation [2]. To better understand its dynamic behaviour, an algorithm was developed for computing the variation in sound pressure received at the location by an acoustic camera from all the surfaces of the space as a function of direction and range to the sound source. The impulse response generated from the large balloon, clapper, yachting cannon, or yelling crowd were measured experimentally. The impulse response P is a function of azimuth, elevation, and time. In the data files, P is sampled in space and time. The values of azimuth, elevation, and time are specified by discrete indices; Naz, Nel, Nt and P (Naz, Nel, Nt) is a 37 x 73 x 8 ms (8 seconds) limit of fast recording time as a 3-dimensional array. Thus, HRIR values are given for 73 different azimuths, 37 different elevations, and at least 8 ms instances in time. The azimuth angle and the elevation angle are measured in a head-cantered inter-aural polar coordinate system (Figure 8). The azimuth is the angle between a vector to the sound source and the midsaggital or vertical median plane, and varies from -9 to +9. The elevation is the angle from the horizontal plane to the projection of the source into the midsaggital plane, and varies from -9 to +27. The following are the coordinates in which the data is reported to a conversion matrix for plotting by MATLAB. (, ) corresponds to a point directly ahead, (, 9 ) corresponds to a point directly overhead, (, 8 ) corresponds to a point directly behind, (, 27 ) corresponds to a point directly below, and (9, ) corresponds to a point directly to the right, (-9, ) corresponds to a point directly to the left. Figure 8 on the left shows the concept for conversion of this set of data through a matrix as plotted in a 3D form (right). The maximum, minimum, and mean plotted data are in colour dots. 6 ICSV23, Athens (Greece), -4 July 26

7 Figure 8: Schematic view of the coordinates for HRTF and plotted data by MATLAB. The acoustic camera microphone positions were converted to altitude, azimuth, and positions uniformly. This angular increment divides the full circle into equal parts in altitude and azimuth, and used in MATLAB, the elevation angle corresponding to N elevation is the N th element of the P vector elevations = The temporal sampling frequency is fs = 96 khz. Max, Min and Average of the sound intensity levels measured by each microphone in the array are presented in the same coordinate within the AC 2 array and translated to the coordinates used for application of HRTF. The interpolated results through a smooth function for total viewing are presented in Figure 9 (left) which shows the measured acoustic signature of the Coliseum in normalized unit (dba) including the incoming sound from the crowd and reflected sound from the upper level seating/standing areas. There is some contribution from the soft sand on the arena s ground or lower level cavity where animals used to be kept according to historical record. The sound from background noise is shown in the centre of the figure and the noise reaching the right side or right ear is shown on the right side of the Figure 9. HRTF provides experiencing the sounds envelop within total immersive environment [7-8]. Figure 9: Crowd noise level Leq (dba re 2 Pa) within the Coliseum (Left), Background Noise (Centre) and Noise from One side based on the HRTF. There are many other possibilities to isolate or highlight the data in terms of source directivity for its reflected components from various architectural elements or crowd noise. This could be clearly demonstrated using the time series data as well as filtering for specific spectral band of frequencies. Application of immersive virtual environment technology for sound perception is achieved through auditory stimuli based on the results of simulated, auralized, and reproduced sounds within computer-simulated space of the existing conditions. This immersion capability as shown in Figure within a 3D virtual reality laboratory that allows stimulation of all human sensory subsystems in a natural way within this immersive environment. The Coliseum sound conditions are simulated in a virtual laboratory where the subject uses special viewing glasses and headphones for best realistic sensation utilizing HRTF with a head-tracking device while listening to an auditory event of a simulated space in real-time and experiencing the sound of the past while viewing an accurate historical representation[9]. ICSV23, Athens (Greece), -4 July 26 7

8 S3 S2 S4 S S5 Surround 5. Speakers Acoustic Camera 5. Conclusion Figure : Views of virtual reality lab, virtual scene and virtual auditory space. The Coliseum sound conditions with open-air as compared to partially enclosed space are difficult to understand without the reflection contribution of each surface and their directionality. Change of surface angles and absorption characteristic within computer simulation allow observing the sound behaviour including reflection and direction. The proposed method in measuring the incoming sound from all directions using beamforming techniques allows capturing the unique acoustic characteristics of a space. This method includes ISO standard indicators as well as a 3D format of the sound signature that represents sound reflection contributions with characteristics of all architectural elements to the space. Viewing each space signature contributes to a better understanding of the specific contribution from each architectural element, through the reproduction of their unique room acoustic condition though aurulization and virtual representation. Acknowledgment Special thanks to Ministry of Cultural Heritage representative Rossella Rea, the superintendent Anna Maria Moretti. SAPIENZA - University of Rome s staff, Laura Monti., Andrea Carraro, Jonida Bundo, and the technical support from GFai Tech GmbH, Berlin, Germany. REFERENCES Haddad, N., Criteria for the Assessment of Modern use of Ancient Theatres and Odea, International Journal of Heritage Studies,3:3, , (27). 2 Algazi, V. R., Duda, R. O., Duraiswami, R., Gumerov, N. A., and Tang, Z. Approximating the Head- Related Transfer Function using Simple Geometric Models of the Head and Torso, The Journal of the Acoustical Society of America, 2, , November ( 22). 3 Hartmann, W., M., How We Localize Sound, Physics Today, (999). 4 Döbler, K., Heilmann, G., Perspectives of the Acoustic Camera, Proceedings of the 44 th Inter-Noise Congress & Exposition, 9-2 August, (25). 5 ISO 3382-: 29, Acoustics - Measurements of Room Acoustic Parameters, Part: Performance spaces. 6 EASE, Enhanced Acoustic Simulator for Engineers, (26), The EASE Software Suite - Acoustic Modelling and Simulation, [Online.] available: 7 Navvab, M., et al., Dynamic Variation of the Direct and Reflected Sound Pressure Levels using Beamforming, Proceedings of 4 th Berlin Beamforming Conference, February, (22). 8 Navvab, M., et al., Simulation, Visualization and Perception of Sound in a Virtual Environment using Beamforming, Proceedings of 4 th Berlin Beamforming Conference, February, (22). 9 Hall, T., Navvab, M., Maslowski, E., and Petty, S., Virtual Reality as a Surrogate Sensory Environment, in Advances in Robotics and Virtual Reality, Springer-Verlag GmbH, Heidelberg, Germany (2). 8 ICSV23, Athens (Greece), -4 July 26