A COMPARISON OF SOURCE TYPES AND THEIR IMPACTS ON ACOUSTICAL METRICS

Transcription

1 A COMPARISON OF SOURCE TYPES AND THEIR IMPACTS ON ACOUSTICAL METRICS By KEELY SIEBEIN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN ARCHITECTURAL STUDIES UNIVERSITY OF FLORIDA

2 2012 Keely Siebein 2

3 To my parents 3

4 ACKNOWLEDGMENTS I would like to thank my mother, father, sisters, brothers, nephews and the rest of my family for all your love and support throughout my life. I would like to thank Professors Gold and Siebein for your continuous support, guidance and encouragement throughout the process of my thesis. I d also like to thank my fellow acoustics students, Sang Bong, and Sang Bum, Cory, Adam B, Adam G., Jose, Jorge and Jenn and any others who helped me with the data collection and reviewing of the material and for keeping me on track with the process. I d also like to thank Dave for your continuous support throughout this entire process, for believing in me and doing all that you could to help me throughout the course of my studies. I d also like to thank my wonderful friends, for believing in me and encouraging me throughout my studies. Thanks to my Dad, for being my father, professor, boss, but most importantly, my mentor. It s an honor to have studied with you and learned from you in this capacity and I will treasure these years and the time I spent learning under your guidance forever. Thank you all for everything. 4

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS... 4 LIST OF TABLES... 7 LIST OF FIGURES... 8 ABSTRACT CHAPTER 1 INTRODUCTION LITERATURE REVIEW Architectural Acoustics Fine Structure of Reverberation Impulse Response Overview Comparison of the Early and Late Portions of the Impulse Response Acoustical Criteria Reverberation Early Decay Time Clarity/Intelligibility Diffuse Sound Field Soundscape Theory Worship Space Acoustics Performance Space Acoustics Balloon Pop Study Omnidirectional Speaker Study Directivity of Sources Reverberation Time: Traditional Equations and Acoustical Standards Traditional Equations Acoustical Standards ISO ASTM E Just Noticeable Differences in Acoustical Metrics Reverberation Time Early Decay Time Clarity (C 80 ) ACOUSTIC SOURCE COMPARISON STUDIES General Information Floor Plan

6 Reverberation Materials Pilot Study Method Electronic Signal Results Natural Acoustic Signals Anechoic voice and music as measurement sources Live music as a measurement source Balloon pop as a measurement source Results Pilot Study Conclusions Source Signal Comparison Method Study 1: Omnidirectional, Directional and Calculated Comparisons Electronic signals Reverberation time Clarity (C 80 ) Study 2: Anechoic Music and Speech and Balloon Pop Stimuli Comparison RT Comparison RT EDT Comparison Natural acoustic sources Scientifically calibrated sources Reverberation Time Comparison Results Early Decay Time Comparison Results C 80 Comparison Results Just Noticeable Differences Reverberation time just noticeable differences Early decay time just noticeable differences Clarity (C 80 ) just noticeable differences Discussion Conclusions Future Studies APPENDIX A IMPULSE RESPONSE PRINTOUTS FOR WIN MLS SOURCES B OMNIDIRECTIONAL AND DIRECTIONAL LOUDSPEAKER COMPARISON LIST OF REFERENCES BIOGRAPHICAL SKETCH

7 LIST OF TABLES Table page 3-1 Just noticeable difference in reverberation time ranges grouped by method Just noticeable difference in reverberation time ranges grouped by receiver position Just noticeable difference ranges for early decay time grouped by method Just noticeable difference ranges for early decay time grouped by receiver position C 80 just noticeable difference ranges grouped by method C 80 just noticeable difference ranges grouped by receiver position

8 LIST OF FIGURES Figure page 2-1 Polar plot of the directionality of human voice in 500 and 4,000 Hertz octave bands. Credit: M. D. Egan, Architectural Acoustics, p Polar plot of the directionality of trumpet in 220, 480, 920, 1840 and 4,000 Hertz octave bands. Credit H. F. Olson, Music, Physics and Engineering, p Directivity patterns of medium and large balloons. Credit: Pätynen et al., Investigations on the balloon as an impulse source, p Polar plot of JBL EON15 G2 in the 500, 1,000 and 4,000 Hertz frequencies showing and -9dB radii Polar plot of the Norsonic 276 loudspeaker, which was provided by the manufacturer as a comparable polar plot to the Norsonic 223 loudspeaker. Credit: Nor276 Dodecahedron Loudspeaker [cited 12 July 2012]. Available from Standard Deviation of Reverberation Times from all Scientifically Calibrated Signals Comparison of all scientifically calibrated Reverberation Time measurements Middle frequency Reverberation Times for average scientifically calibrated source signals with 99% Confidence Interval Floor plan of Baughman Center showing source and receiver locations Reverberation Times for the anechoic speech, music, live music and Balloon Pop sources in receiver position Reverberation Times in the middle frequencies for all source types Floor plan of Baughman Center indicating source types and positions and receiver positions Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 1 st receiver position Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 2 nd receiver position Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 3 rd receiver position

9 3-11 Average measured and calculated Reverberation Times for the Baughman Center Clarity (C 80 ) values measured at the first receiver position with omnidirectional and directional loudspeakers Clarity (C 80 ) values measured at the second receiver position with omnidirectional and directional loudspeakers Clarity (C 80 ) values measured at the third receiver position with omnidirectional and directional loudspeakers Placement of cursors to derive acoustic metrics for the Traditional Method using the end note of the Trumpet Song at 1,000 Hertz as an example Placements of cursors to derive acoustical metrics from the Expanded Method End note of Trumpet Song with 1,000 Hertz octave filter applied to signal. Note the times, as they are used in the next step Cursor placement after filtering and listening to Trumpet Song stop chord. Note placement of cursors corresponds to times identified in Figure New wav file made by clipping the top image between the Reflected Sound and Noise Floor cursors The new wav file from Figure 3-19 was run through WinMLS software program and acoustic metrics derived from this wav file Reverberation Time comparison for Pilot Study, Study 1 and Study 2 in the 500, 1000 and 4000 Hertz octave bands Female Talker RT EDT analyzed using thetraditional Method Female Talker RT EDT analyzed using the Expanded Method Female Talker RT EDT analyzed using the Method Female Talker RT EDT analyzed using the WinMLS Method Trumpet Song RT EDT analyzed using thetraditional Method Trumpet Song RT EDT analyzed using the Expanded Method Trumpet Song RT EDT analyzed using the Method Trumpet Song RT EDT analyzed using the WinMLS Method

10 3-30 Balloon Pop RT EDT analyzed using the Traditional Method Balloon Pop RT EDT analyzed using the Expanded Method Balloon Pop RT EDT analyzed using the Method Balloon Pop RT EDT analyzed using the WinMLS Method Averaged MLS measurements dervied from the directional loudspeaker subtracting EDT from RT at receiver locations R1, R2 and R Averaged MLS measurements dervied from the omnidirectional loudspeaker subtracting EDT from RT at receiver locations R1, R2 and R Female Talker and Trumpet Song Reverberation Times normalized to average MLS data Balloon Pop source graphs showing similar RT values across receiver positions and methods Graphs of Female Talker Normalized Reverberation Times using Expanded and Methods, which result in similar values Graphs of Trumpet Song Normalized Reverberation Times using Expanded and Methods, which result in similar values Spectrograph of the Female Talker last syllable with concentration of sound energy in the 6,500 to 9,200 Hertz octave bands Spectrograph of the Trumpet Song end note. The Trumpet Song has more concentrated sound energy in many frequencies than the Female Talker Spectrograph of the Balloon Pop. The solid decay of green, yellow, red and orange indicate very strong levels across all frequencies Female Talker source analyzed with Traditional, and WinMLS methods resulting in similar normalized EDT values Female Talker Normalized EDT values for Receiver Positions 1 and Trumpet Song Normalized EDT values for Traditional and WinMLS methods Trumpet Song Normalized EDT values for Expanded and methods Graphs showing large variation in left and right receiver normalized EDT values for Trumpet Song compared to fairly consistent Female Talker values

11 3-48 Graphs of Balloon Pop analyzed with Expanded and methods resulting in similar normalized EDT values Balloon Pop analyzed using the WinMLS method resulting in EDT values that are centered around average MLS data Balloon Pop source at receiver positions 1 and 3, showing differences in left and right receivers as the receiver position moves farther from the source Graphs of Female Talker Normalized C 80 values using Expanded and Methods, which result in similar values Graphs of Trumpet Song Normalized C 80 values using Expanded and Methods, which result in similar values Graphs of normalized C80 values for Balloon Pop source for the 3 receiver positions Graphs showing Female Talker having similar JND values in the 4,000 Hertz octave band for Expanded and methods Graphs showing Trumpet Song having similar JND values in the 4,000 Hertz octave band for Expanded and methods Balloon Pop Expanded and Methods showing similar Just Noticeable Differences in Early Decay Time Balloon Pop WinMLS method showing low JND values for Early Decay Time Graphs showing Female Talker having similar JND values for C 80 across the octave bands for Expanded and methods Graphs showing Trumpet Song having similar JND values for C 80 across the octave bands for Expanded and methods Impulse response graphs of various sources used in study Average RT EDT values for Female Talker source Average RT EDT values for Trumpet Song source Average RT EDT values for Trumpet Song source Average Normalized Reverberation Time for Female Talker source Average Normalized Reverberation Time for Trumpet Song source Average Normalized Reverberation Time for Balloon Pop source

12 3-67 Average Normalized Early Decay Time for Female Talker source Average Normalized Early Decay Time for Trumpet Song source Average Normalized Early Decay Time for Balloon Pop source Average Normalized Clarity (C 80 ) for Female Talker source Average Normalized Clarity (C 80 ) for Trumpet Song source Average Normalized Clarity (C 80 ) for Balloon Pop source Reverberation Time comparison of 3 studies showing higher RT s for scientifically calibrated sources and lower RT s for natural acoustic sources A-1 MLS Directional Front: Receiver 3_ A-2 MLS Directional Front: Receiver 3_ A-3 MLS Directional Front: Receiver 3_ A-4 MLS Directional Corner: Receiver 3_ A-5 MLS Directional Corner: Receiver 3_ A-6 MLS Directional Corner: Receiver 3_ A-7 Sine Sweep Directional Front: Receiver 3_ A-8 Sine Sweep Directional Front: Receiver 3_ A-9 Sine Sweep Directional Front: Receiver 3_ A-10 Sine Sweep Directional Corner: Receiver 3_ A-11 Sine Sweep Directional Corner: Receiver 3_ A-12 Sine Sweep Directional Corner: Receiver 3_ A-13 Multiple Sine Sweep Directional Front: Receiver 3_ A-14 Multiple Sine Sweep Directional Front: Receiver 3_ A-15 Multiple Sine Sweep Directional Front: Receiver 3_ A-16 Multiple Sine Sweep Directional Corner: Receiver 3_ A-17 Multiple Sine Sweep Directional Corner: Receiver 3_ A-18 Multiple Sine Sweep Directional Corner: Receiver 3_

13 A-19 Piano Middle C Note: 500 Hertz A-20 Piano Middle C Note: 1,000 Hertz A-21 Piano Middle C Chord: 500 Hertz A-22 Piano Middle C Chord: 1,000 Hertz A-23 Piano Song Stop Chord: 500 Hertz A-24 Piano Song Stop Chord: 1,000 Hertz A-25 Anechoic Trumpet Music: 500 Hertz A-26 Anechoic Trumpet Music: 1,000 Hertz A-27 Anechoic Female Talker: 500 Hertz A-28 Anechoic Female Talker: 1,000 Hertz B-1 MLS Directional Front: Receiver 1_ B-2 MLS Directional Front: Receiver 1_ B-3 MLS Directional Front: Receiver 1_ B-4 MLS Directional Front: Receiver 2_ B-5 MLS Directional Front: Receiver 2_ B-6 MLS Directional Front: Receiver 2_ B-7 MLS Directional Front: Receiver 3_ B-8 MLS Directional Front: Receiver 3_ B-9 MLS Directional Front: Receiver 3_ B-10 MLS Omnidirectional Front: Receiver 1_ B-11 MLS Omnidirectional Front: Receiver 1_ B-12 MLS Omnidirectional Front: Receiver 1_ B-13 MLS Omnidirectional Front: Receiver 1_ B-14 MLS Omnidirectional Front: Receiver 1_ B-15 MLS Omnidirectional Front: Receiver 2_

14 B-16 MLS Omnidirectional Front: Receiver 2_ B-17 Omnidirectional Front: Receiver 2_ B-18 MLS Omnidirectional Front: Receiver 2_ B-19 MLS Omnidirectional Front: Receiver 2_ B-20 MLS Omnidirectional Front: Receiver 3_ B-21 MLS Omnidirectional Front: Receiver 3_ B-22 MLS Omnidirectional Front: Receiver 3_ B-23 MLS Omnidirectional Front: Receiver 3_ B-24 MLS Omnidirectional Front: Receiver 3_

15 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science in Architectural Studies A COMPARISON OF SOURCE TYPES AND THEIR IMPACTS ON ACOUSTICAL METRICS Chair: Martin Gold Cochair: Gary Siebein Major: Architecture By Keely Siebein August 2012 The purpose of this study is to compare acoustical measurements made with different source types in a relatively reverberant room to determine if ISO 3382 monaural acoustic parameters such as Reverberation Time (RT), Early Decay Time (EDT), and Clarity Index (C 80 ), yield different results for natural acoustic versus calibrated source stimuli. The source stimuli used in the study included Maximum Length Sequences (MLS), a running train of speech, a running piece of music and a balloon pop. The scientifically calibrated method is then compared to acoustical measurements obtained from natural acoustic sources, which include anechoic recordings of voice and music played through a directional loudspeaker in the front of the room to simulate activities that would normally take place in the room, such as a person speaking and music being played during a worship service. This analysis is performed to determine if there are differences in acoustic room parameters using natural acoustic sources. This study essentially compares the effects of different source stimuli on measured acoustic parameters. 15

16 It was found that different source signals and receiver locations significantly affect the acoustic metrics derived from the acoustical measurements due to variations in frequency, level and directionality. 16

17 CHAPTER 1 INTRODUCTION The goal of this study is to determine if different sound sources have an effect on the acoustical metrics derived from acoustical measurements taken in a room. Acoustical metrics are typically given as a single number average value for the entire room. These measurements are usually performed in accordance with ISO 3382, with an omnidirectional loudspeaker playing a Maximum Length Sequence (MLS) or sine sweep signal from a loudspeaker from which impulse responses are derived. Are the acoustical metrics derived from these measurements applicable to all the sounds that are normally heard in the room? This question is pertinent because if one can understand how rooms affect the specific sounds produced within them, one can more accurately design for improved acoustics in these spaces. One cannot design a room solely based on numerical criteria. The numerical criteria do not address the room as a whole; they serve as an attempt to quantify the complex interactions of sound and environment. Recent research in soundscape theory emphasizes the importance of understanding how rooms are used by the people who inhabit them, and what kinds of sounds are made and listened to in these spaces. This soundscape method is at the heart of this study, in that by identifying sounds that might typically take place in the Baughman Center and deriving acoustical metrics from them, a deeper understanding of how the space affects these different sound sources is possible. Research has been conducted in the past ten years based on findings that there are multiple acoustic metric values in one space. For example, Taeko Akama et al. conducted research that entailed a matrix of between 511 and 1427 receiver position- 17

18 some in almost every seat of a concert hall- to determine how the different seat positions affect the acoustic metrics such as Reverberation Time, Early Decay Time and Clarity. 1 Their study revealed that each seat has different direct and reflected paths from the source to the receiver, and therefore sounds will be heard differently at each seat. Based upon these observations, this study was conducted to determine what effect different sources have on acoustical metrics and to relate acoustical metrics taken in general accordance with ISO 3382 to metrics derived from natural sources that might actually take place in the room. 1 T. Akama, H. Suzuki, and A. Omoto, Distribution of Selected Monaural Acoustical Parameters in Concert Halls, Applied Acoustics 71 (2010):

19 CHAPTER 2 LITERATURE REVIEW Architectural Acoustics Acoustics is a relatively young science, with the first scientific experiments performed in the early part of the twentieth century by Wallace Clement Sabine. While W. C. Sabine is most noted for his studies on Reverberation Time; he also recognized basic architectural acoustical concepts, principles of clarity, loudness, distinctness, low background noise, and uniform sound distribution. 1 He understood basic architectural concepts as they related to acoustics, such as elevating the sound source in an auditorium above the audience, and raising the rear of the audience seating area to allow for better comprehension of speech and music. He also understood that putting a wall behind the sound source would provide a surface that sound may strike and propagate back into the audience as early reflections making the sounds louder than they would otherwise be, and that adding a roof to an open auditorium would ensure that sound would not escape the main volume of the audience area. 2 Sabine understood basic acoustic concepts as they relate to the design of auditoria, and that he understood many of the intricacies that are found in the study of reverberation. Sabine s grasp of the complexity of reverberation is evident in his explanation of direct and reflected sounds. He refutes what he describes as being the traditional way of thinking; that direct and reflected sound reinforce each other all the time and therefore increase the loudness of the sound. While this may happen if the sound waves are properly aligned, this is not always the case. Sabine describes the 1 Wallace Clement Sabine, Collected Papers on Acoustics (New York: Dover Publications, 1964), 4. 2 Sabine, Collected Papers, 5. 19

20 physical characteristics of sound waves and the motion of compression and rarefaction. If one sound path is shorter than the other due to it striking a surface in the room, one wave s compression may meet at the same time as another wave s rarefaction and therefore result in silence. 3 His scientific approach to studying reverberation allowed for a deeper and quantifiable understanding of what was then considered the phenomenon of sound. Sabine defines architectural acoustics as it related to the design of a room by defining two room variables: shape and material. 4 His reverberation experiments started with identifying the sound absorption power of various materials. The Reverberation Time was measured in the room before the addition of any materials, and then in increments of amounts of materials. The first experiment involved bringing in seat cushions from the nearby Sanders Theater into the Fogg Lecture Hall and measuring how long the reverberation lasted in the room with different amounts of cushions. Other materials were tested as well including chenille curtains, oriental rugs, Herez (Persian rug), Deminjik, Hindoostanee, cretonne cloth, canvas and hair felt. Sabine even tested the absorption characteristics of a man and woman in a smaller room. 5 Sabine was able to draw several conclusions from the initial experiments that can still be applied today. The first conclusion was that the duration of audibility of the residual sound is nearly the same in all parts of the auditorium. 6 Testing to come to 3 Sabine, Collected Papers, Sabine, Collected Papers,10. 5 Sabine, Collected Papers, Sabine, Collected Papers,17. 20

21 this conclusion was performed in Stienert Hall, where Reverberation Times at eight receiver positions were tested. The Reverberation Times varied from 2.12 to 2.27 seconds (0.15 seconds difference). 7 While it is true that Reverberation Time may not vary greatly depending upon the receiver location, the paths the sound reflections take from source to receiver changes, and the impulse response will reveal a different pattern of reflections. However, this concept was not discovered for over 80 years after Sabine s initial experiments. Similarly, Sabine tested various source positions in a room to see if changing the source position had an effect on Reverberation Time. He found that the duration of audibility is nearly independent of the position of the source. 8 An experiment was performed in the Jefferson Physical Laboratory in which a fixed observer measured the Reverberation Time in six different source positions in the room. This test resulted in Reverberation Times of 3.90 to 4.00 seconds (0.10 seconds difference). 9 Lastly, an experiment was performed in the lecture room of the Fogg Museum to determine if the locations of the absorbent material had an effect on the Reverberation Time. Fifty meters of cretonne cloth were hung in four sections of the room, and the observer remained in a fixed position. The resulting Reverberation Times after the addition of the cloth were 4.83 to 4.92 seconds, a difference of 0.09 seconds. 10 He concluded that the efficiency of an absorbent in reducing the duration of the residual sound is, under ordinary circumstances, nearly independent of its 7 Sabine, Collected Papers, Sabine, Collected Papers, Sabine, Collected Papers, Sabine, Collected Papers,

22 position. 11 This is true, however, Sabine also noted that placement of the absorbent material is of importance when designing the room. Sabine s initial experiments paved the way for architectural acoustics to move out of what was considered a black art into the realm of science by performing replicable scientific experiments in many rooms with a variety of conditions. It was said that Sabine had a strong desire to ensure that his experiments were able to withstand scrutiny. In the Introduction to the Dover Edition of W. C. Sabine s Collected Papers on Acoustics, Frederick V. Hunt said Sabine had an uncommon reluctance to publish the results of an experiment until he was sure it was done so well that nobody would ever need to repeat it. 12 Fine Structure of Reverberation Reverberation Time was considered the only acoustical parameter necessary for describing room conditions for approximately 50 years after Sabine s experiments. From the Reverberation Time criteria, many other criteria were developed that break apart the reflection structure associated with the sound decay to better understand how the room affects the sounds within it. Criteria were developed that were based on Sabine s exponential decay. It was found that during the early part of the sound decay, a different slope of the level versus time curve is shown depending on whether the excitation in the room is steady-state or impulsive. 13 Atal et al. as summarized by Cremer recommended the Initial Reverberation Time as a room acoustics criterion instead of the traditional Reverberation Time. It is denoted 11 Sabine, Collected Papers, Sabine, Collected Papers, XV. 13 Lothar Cremer and Helmut A. M ller, Principles and Applications of Room Acoustics. (London: Applied Science, 1982),

23 as T I. It is defined in terms of the time required for a 60 db decay, but comes from a straight line fitted to the slope of the 1 st 160ms or the first 15dB of decay. This is related to the backwards integration method Schroeder developed to record Reverberation Times in a hall. It was found that there are extreme fluctuations that follow the cessation of steady-state excitation. These fluctuations do not allow for accurate assessment of the early part of the reverberation decay using classical methods. 14 Kurer and Kurze showed that different source receiver locations produce different initial slopes given by T I. These studies expand upon Sabine s original testing in which he tested the Reverberation Times in various source and receiver locations throughout the room and found small variations of up to.15 seconds using the full 60 db of decay. They also discovered that T I may exceed T when evaluated from -5 and -35 db, resulting in sagging and ballooned curves of level vs. time. Kurer and Kurze also found that using the region between 0 and -20dB to define the initial slope was helpful. It was denoted as T A, or Anfangsnachhallzeit, which means Initial Reverberation Time. 15 V. Jordan restricted the decay range to just 0 to -10dB, which he referred to as the early decay time or EDT. 16 It is represented by T E. Cremer comments that linguistically, the use of Early seems as though it should follow, not precede Initial. Cremer also comments that measuring with the backwards integration of an impulse allows the shortening of the range from 0 to -10dB, especially if the slope is evaluated with a computer and not by hand Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics,

24 Impulse Response Overview Building upon Sabine s original findings and using complex physical acoustical equations, Manfred Schroeder developed an alternative method of measuring Reverberation Time in a room and published his first paper on this new method in He called this the integrated tone-burst method, but it would later be referred to as the integrated impulse response method. 18 Schroeder recognized that with the traditional methods of obtaining Reverberation Times, the randomness of the source signal created difficulties. Differences in the decay curves are found with a bandpass-filtered source signal despite keeping the source and receiver positions the same due to the mutual beating of normal modes of different natural frequencies. 19 These differences are caused by the initial amplitudes and phase angles of the normal modes being different as the source signal shuts off. 20 Averaging the measurements together usually compensates for these differences. Schroeder states that averaging the measurements fails for two reasons: first, because it does not reveal the true nature of the decay curve, and second, because it cannot detect multiple decay rates and especially high initial decays. Schroeder asserts that the initial part of the decay curve contains valuable information, as recent research had indicated that may be as important as the later decay when analyzing the reverberance of a space. Schroeder states that the new method that he has developed will have the same effect as averaging an infinite number 18 M. R. Schroeder, New Method of Measuring Reverberation Time, The Journal of the Acoustical Society of America 37 (1965): Schroeder, New Method, Schroeder, New Method,

25 of bandpass-filtered noise decay curves. 21 Indeed, the impulse response method that was ahead of its time and has only recently been acknowledged as a scientifically calibrated method for obtaining the Reverberation Time in International and American standards. The standards will be discussed in a later portion of the text. An impulse response is composed of the direct sound and the discrete, separate reflections from various room surfaces that follow it. Several methods were identified to attempt to represent the best method of revealing the impulse response. There are several methods to provide the impulsive sound that is measured by the impulse response. A pistol such as one used for starting races was often used when testing the Reverberation Time in a room. W. Reichardt has shown that the spectral distribution of energy approximately corresponds to speech. The pistol also has a preferential directional response, as the axis of the barrel is directional. 22 It is for this reason a pistol is considered appropriate for evaluating speech qualities in a room. Another method is a loudspeaker that radiates a pre-filtered signal, such as a Gauss tone impulse. Gauss-tone impulses, which are considered especially distortionless, are short pure-tone impulses whose time enveloped is shaped like the bell-shaped Gaussian distribution. However, these curves may represent responses that are representative of our ears, yet experience has shown they are not well suited for detailed analysis with echograms. Lehmann demonstrated in an experiment that using Gauss-tones can lead to very different impulse responses when changing the source and receiver only several centimeters. Our ears do not detect these kinds of minute differences in distances in 21 Schroeder, New Method, Cremer and M ller, Room Acoustics,

26 large halls; therefore this method is not best suited for impulse response measurements. 23 Schodder performed experiments in which source and receiver locations were varied in the Royal Festival Hall, and impulse responses recorded. Upon review and comparison of the impulse responses, it was clearly evident that source and receiver positions had a great effect on the impulse response. Each source and receiver position had extremely varied impulse responses from one another. 24 Cremer suggests that based upon this evidence, it is not fitting to refer to the acoustical attributes of a hall without at least mentioning where the source and receiver were located for the measurement. 25 Comparison of the Early and Late Portions of the Impulse Response It has been established that the direct sound, first reflection and subsequent reflections up to 50 ms are regarded as useful, and considered the limit of perceptibility. Thiele in 1953 proposes a measure called Distinctness, which compares the useful sound with the total sound in ratio form. 26 The equation is: This equation is the integral of the useful sound from 0 to 50ms of the sound pressure squared (p 2 ) over the integral of sound pressure squared (p 2 ) of the total sound from 0 to. Cremer proposes to call this the distinctness coefficient, since it involves a ratio 23 Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics,

27 of energies and the term Distinctness implies a subjective evaluation. 27 Beranek and Schultz modified this equation to form the reverberant-to-early-sound ratio. This equation is: The new equation takes the integral of the sound pressure squared (p 2 ) from 50ms to divided by the integral of sound pressure squared (p 2 ) from 0 to 50ms. It was found that 10 * log of this ratio is the reverberation index of Hallmass. The equation is: ) db These metrics only consider sound after 50ms to be reverberant sound. However it was found that the limit of perceptibility is not a constant time, but rather depends on the character of the signal. 28 Based on this idea, Reichardt proposed a value of 80ms as the limit of perceptibility of music. He wanted to replace the Distinctness measure with Clearness in reference to music. 29 His equation is: This equation takes the ratio of the integral of sound pressure square from 0 to 80ms divided by the integral of the sound pressure squared from 80ms to. The previous 3 equations are problematic in that a very slight shift in arrival time of a strong reflection 27 Cremer and M ller, Room Acoustics, L. L. Beranek and T. J. Schultz, Some Recent Experiences in the Design and Testing of Concert Halls with Suspended Panel Array, Acustica 15 (1965): Cremer and M ller, Room Acoustics,

28 can significantly change the value of the criterion. 30 Lochner and Burger are able to avoid this problem when defining the Signal to Noise Ratio. They introduced a weighting factor ( ) that lessens the change from useful to non-useful sound that depends on the delay time and the relative level of the reflection. 31 This new equation reads: The useful energy is the integral from 0 to 95ms of sound pressure squared multiplied by the weighting factor ( ). The detrimental energy is the integral from 95ms to of sound pressure squared. Jordan goes on to propose another similar criterion called the rise time ( ). This is based on the fact that forward integration over the impulse response corresponds to the onset of steady-state excitation. It is defined as the moment in the impulse response where the early and late energy are equal and as such is also the time required to reach a level 3 db below the asymptotic final level. 32 The equation is This equation shows the integral from 0 to the rise time ( ) of sound pressure squared is equal to the integral from the to of the sound pressure squared. Comparing the 30 Cremer and M ller, Room Acoustics, J. P. A. Lochner and J. F. Burger, The Subjective Masking of Short Time Delayed Echoes, Their Primary Sounds, and Their Contributions to the Intelligibility of Speech, Acustica 8 (1958): V. L. Jordan, The Building-up Process of Sound Pulses in a Room and its Relation to Concert Hall Quality, Proceedings of the 3rd ICA, Stuttgart, 1959, Vol. II, (Amsterdam: Elsevier, 1961),

29 early to late energy in an impulse response allows for several methods of analysis that are useful when determining acoustical characteristics of a room. 33 Acoustical Criteria This study focuses on several acoustical criteria based on room geometry and materiality, including Reverberation Time, Early Decay Time, Clarity/Intelligibility and a diffuse sound field. These concepts are expanded upon below. Reverberation Reverberation is defined as the sound that persists in a room after a source is suddenly stopped. 34 It is measured by the Reverberation Time, which is the amount of time it takes a sound to decay 60dB from the level of the stimulus such as the stop chord. The Sabine Reverberation Time equation is V is the room volume in ft 3. S is the surface area in ft 2 and α is the average absorption coefficient in the room. This requires general information about the room, including the volume and absorption in the room. The reverberation is controlled by the ceiling height, the volume of the space and the materials in the space. A desired Reverberation Time can provide a general idea as to how tall the ceiling should be, but does not describe where absorption needs to be placed in the room. In performance spaces, it is necessary to have an adequate ceiling height and volume so that there is enough reverberation in the space. The audience is generally the largest sound absorbent area 33 Cremer and M ller, Room Acoustics, L. L. Beranek, Concert and Opera Halls: How They Sound (Woodbury, NY: Published for the Acoustical Society of America through the American Institute of Physics, 1996),

30 in a room that is designed for music. Reverberation Time is one of the most commonly used criteria for assessing the acoustical performance of a room. Early Decay Time Early Decay Time measures the first 10, 15 or 20 seconds of sound decay. It is related to clarity, envelopment and spaciousness. Early Decay Time is defined by Jordan as the RT corresponding to the slope measured over the first 10 ms of decay. 35 Jordan goes on to argue that EDT is one of the most relevant acoustical parameters in that it is extremely easy to measure, it has the best established correlation with subjective assessment, and EDT measurements in models and the actual halls result in comparable data. 36 Jordan examines two main interpretations of EDT measures. The first is that the EDT is used when calculating the Inversion Index, which is the ratio of the EDT in the audience divided by the EDT value of the stage area. Jordan also suggests that one can compare the RT and EDT at individual locations in the audience. He states the RT and EDT should be close, and in halls where the RT is low, having a slightly higher EDT is preferred. 37 Jordan s data in the Metropolitan Opera House present differences in RT and EDT between 0.01 and 0.12 seconds. 38 Clarity/Intelligibility Clarity is defined by Egan as the degree to which individual notes are distinct or stand apart. 39 A more technical definition is the ratio of early sound energy to late or 35 V. L. Jordan, Acoustical Criteria for Auditoriums and Their Relation to Model Techniques. The Journal of the Acoustical Society of America 47 (1970): Jordan Acoustical Criteria Jordan Acoustical Criteria Jordan Acoustical Criteria M. D. Egan, Architectural Acoustics. (New York: McGraw-Hill, 1988),

31 reverberant sound energy. 40 Clarity is measured by the Clarity Index or C 80. This is the ratio of the first 80 ms of an impulse sound arriving at a listener s position divided by the energy in the sound after 80 ms. 41 Clarity is the term associated with music, while intelligibility is the term associated with the clear perception of speech. Clarity is enhanced by coplanar reflections that reach the listener between 30 and 50 ms for speech signals, and 30 to 80 ms for music. 42 If these reflections occur within the first 50 to 80 ms, they reinforce the loudness of the sound as well. The shaping of the ceiling should provide these coplanar reflections so that the audience receives multiple reflections. There is also the concept of horizontal clarity, which is described as the ability to hear successive notes in a piece of music. 43 In a main theater space, the ceiling over the stage and the audience should be shaped so that these reflections will occur. Clarity, according to Beranek, is the degree to which the discrete sounds in a musical performance stand apart from one another. 44 This is quantitatively measured by the ratio of the early sound in the first 80ms to the reverberant sound from 80ms on. In this equation, 10 * log is multiplied by the ratio of the integral of the first 80ms of sound pressure squared over the integral of 80ms to of sound pressure squared. For 40 Beranek, Concert and Opera Halls, Beranek, Concert and Opera Halls, P. S. Veneklasen Design Considerations from the Viewpoint of the Professional Consultant In Auditorium Acoustics, ed. by Robin MacKenzie. (New York: John Wiley and Sons, 1974), Beranek, Concert and Opera Halls, Beranek, Concert and Opera Halls,

32 a concert hall, the 80ms time period is considered the optimal time period for early reflections to arrive to increase the clarity in the room. 45 One coplanar ceiling reflection in the first 40ms is desirable to create this effect. 46 Diffuse Sound Field A diffuse sound field occurs when sounds propagate equally over the space. It is measured by the Inter Aural Correlation Coefficient (IACC). It is the measure of the difference in sounds arriving at the two ears of a listener facing the performing entity in a hall. 47 It relates to the amount of diffusion present in a space. The IACC metric ranges from 0 to 1, and lower IACC values are desirable. To achieve low IACC values, it is necessary to provide diffusion in the low, middle and high frequencies by providing diffusion of different widths. Beranek reported measured IACC values of 31 concert halls and found that 74% had IACC values of less than 0.40 in both the 500 and 1,000 Hertz octave bands 48 While the IACC metric is not examined in this study, the concept of a diffuse sound field is important as it is the basis for acoustical measurements. The standards dictate the use of an omnidirectional loudspeaker so that it may create a diffuse sound field in the room. However, it is nearly impossible to have a truly diffuse sound field in any room, as the room is almost always never completely diffuse and there are very few if any truly omnidirectional sound sources that would excite the room in this manner. 45 W. Reichardt, Abel Alim and W. Schmidt, Clarity, Acustica 32 (1975): Veneklasen, Design Considerations, Beranek, Concert and Opera Halls, Beranek, Concert and Opera Halls, 535,

33 While understanding the technical definitions of acoustical criteria are necessary to provide a background for the study, a philosophical framework is needed to put the ideas into context and to develop a measurement method that is founded on solid scientific and perceptual principals. Soundscape Theory Murray Schafer is credited with coining the term soundscape that is often referenced in the field of acoustics. Schafer simply defines soundscape as any acoustic field of study. 49 While leaving the definition simple, it allows interpretation and does not place unnecessary restrictions on attempting to specifically define the term. He challenges humans to consider their views of sound and noise by asking, is the soundscape of the world an indeterminate composition over which we have no control, or are we its composers and performers, responsible for giving it form and beauty?. 50 Humans are responsible for a great number of sounds in the soundscape at any given period of time, from planes and automobiles, to the hum of a laptop in a coffee shop and the clicking of fingers on the keyboards to a ringtone heard as a cell phone rings, we introduce a multitude of sounds into our environment. Similarly, humans are not only responsible for making sounds, but they are responsible for creating spaces where sounds are altered. From the acoustic designer and architect who mold the buildings floors, walls and ceilings, to the interior designer who designs the interior of the space to 49 R. M. Schafer, The Soundscape: Our Sonic Environment and the Tuning of the World (Rochester, VT: Destiny Books, 1994), Schafer, The Soundscape, 5. 33

34 the end user, who changes the space even more once occupied, humans are responsible for creating the space in which sounds take place. 51 Soundscape theory offers the possibility for architects, interior designers, landscape architects and urban planners to integrate the conscious design of the sonic attributes of the environments they plan, design and construct as a part of their typical design process for interior and exterior spaces. 52 Soundscape theory can be used to assist in identifying the types of sounds that typically take place in a building so that the acoustical attributes can be designed with them in mind. By identifying the constituents that make up the soundscape in a worship space, typical source types and locations can be identified, as well as itineraries and acoustic calendars can be defined to provide a more holistic understanding of the space. With knowledge of the types of sounds that typically take place in the building, how do typical scientifically calibrated acoustic measurements compare to the sound sources that excite the room? Are the metrics derived from the scientifically calibrated sound sources comparable to metrics derived from natural acoustic sources that would typically be heard in the room? Schafer likens the idea of a soundscape to a composition, in which all the parts of the composition are carefully planned and executed: the notes, pauses, tempo, and rhythm. The notes in a song can be likened to the sound sources in a soundscape: some occur by themselves, some occur together. There are pauses in the composition, as there are pauses in the soundscape, where sounds cease and there are periods of 51 B. Blesser and L. Salter, Spaces Speak, Are You Listening?: Experiencing Aural Architecture. (Cambridge, Mass: MIT Press, 2007), G. Siebein, Essential Soundscape Concepts for Architects and Urban Planners (paper presented at Designing Soundscape for Sustainable Urban Development Conference, Stockholm, Sweden, September 30 October 1, 2010). 34

35 silence. The tempo of the song is likened to the tempo of the soundscape; some constituents have an intrinsic speed or pace at which the sounds occur, such as a bird s call or a cicadas hum. The rhythm of the music or the regular reoccurrence of refrains, choruses, or notes can be likened to the acoustic calendar of a soundscape. The acoustic calendar is documented by cataloging the specific sound sources and determining their hourly, daily, weekly, monthly and yearly cycles. A soundscape can be thought of similarly, and can be designed so that the acoustic attributes of indoor and outdoor spaces are consciously composed as the spaces are designed and built. 53 Similarly, Siebein discusses using soundscape analysis, sophisticated measuring and modeling of spaces to allow stakeholders to aurally preview the space before it is built so as to intelligently design and shape the space. Given this perspective, perhaps using acoustical measurements derived from sound sources that are typical of the room as the basis of the computations for the auralizations could provide even more robust modeling of the space. Using analysis methods that have been developed to understand how sound moves throughout a room, one can see how real sound sources behave in a room as well. The impulse response contains information about how the sound interacts with the room. By studying parts of the impulse response, one can understand more about sound and its interactions with the room surfaces. This study examines Reverberation Time, Early Decay Time and Clarity in relation to scientifically calibrated sound sources and natural acoustic sources, such as anechoic speech and music. 53 Siebein, Essential Soundscape. 35

36 Worship Space Acoustics Youngmin Kwon performed a study in which acoustical measurements were conducted using a single omnidirectional loudspeaker located in the middle of the stage as the source and compared it to measurements conducted using an array of multiple directional loudspeakers set up on the stage to simulate an orchestra. The directional loudspeakers were carefully placed in locations where instruments are typically placed, and each speaker was to represent an individual instrument section of a full orchestra. The loudspeaker was placed so that its directionality matched the directionality of the instrument it was to simulate. It was found that there were not significant differences in RT and EDT, however, there were differences in C 80 values depending upon receiver location. 54 Soeta et al. compared acoustical measurements made in 4 churches ranging in room volume from 1,000 and 12,600 m 3. Sources were arranged using locations proposed by Martelotta et al., as well as locations described in the Catholic liturgy. A directional loudspeaker at 1.5m above the finish floor was used to simulate sound propagating into the room from a person speaking in the room at 2 source locations. The directionality of the loudspeaker was measured and compared to the directionality of a human mouth. Soeta et al. found that even though the directivity pattern of the directional loudspeaker did not directly correspond to the human mouth, it provided a better approximation than an omnidirectional speaker would of the directivity pattern of a human mouth. It was observed that there was little variation between source and receiver position for the RT and EDT from 125 to 1,000 Hertz. It was found that using 54 Y. Kwon, Quantitative and Qualitative Analyses of Under-Balcony Acoustics with Real and Simulated Arrays of Multiple Sources. (PhD diss., University of Florida, 2006),

37 the sources that faced away from the congregation resulted in lower Strength (G), total amplitude of reflection (A) and C 80 than using the sources facing towards the congregation. 55 Performance Space Acoustics In 1991, John Bradley performed a study comparing three concert halls: the Amsterdam Concertgebouw, the Vienna Musikvereinsaal, and Boston Symphony Hall. Acoustic metrics were derived from measurements taken in multiple source and receiver locations. It was found that the Reverberation Times did not vary widely throughout the rooms (+ 0.1 s in the middle and high frequencies). 56 Reverberation Times and Early Decay Times had a tendency to have similar values despite being derived from measurements from three different source locations (+ 0.1 s in the middle frequencies). 57 When different receiver positions were used, the values for the Reverberation Times remained mostly constant, however EDT values varied in each hall. Interestingly, Bradley found that where EDT values increase with increasing source-receiver distance, C 80 values decrease with increasing distance. 58 He also found that the EDT values and RT values varied from each other by approximately seconds Y. Soeta, K. Ito, R. Shimokura, S. Sato, T. Ohsawa, and Y. Ando, Effects of Sound Source Location and Direction on Acoustic Parameters in Japanese Churches, The Journal of the Acoustical Society of America 131 (2012): J. S. Bradley, A Comparison of Three Classical Concert Halls, The Journal of the Acoustical Society of America 89 (1991): Bradley, A Comparison, Bradley, A Comparison, Bradley, A Comparison,

38 Balloon Pop Study Recent studies were performed by Pätynen et al. that examined the directivity pattern and power spectrum of balloons. Balloons of different color, size and inflation were tested in an anechoic chamber with a spherical microphone array. It was determined that larger balloons had more energy and balloons that were more inflated had more energy content in the high frequencies. It was also determined that while balloon directivity patterns were mostly stable, they do not satisfy stipulations in ISO 3382 regarding the source having omnidirectional propagation. 60 Therefore, if Balloon Pops are to be a sound source in acoustical measurements, it is best to use large balloons that have been inflated as much as possible. Omnidirectional Speaker Study A study was completed by Ricardo San Martin and Miguel Arana on differences in acoustic metrics derived from performing acoustical measurements with four different omnidirectional speakers in five halls using simulations using Odeon room acoustics software. A grid of over 3,000 receiver positions were measured as the source was rotated 5 o. It was found that below 1,000 Hertz, differences in acoustic metrics were not prominent. However, in the 1,000 and 2,000 Hertz octave bands, uncertainties based on the subjectively perceivable change was from 15% to 40% respectively. Octave bands higher than 2,000 Hertz were found to have deviations greater than half the Just Noticeable Difference of the acoustic metric in over 80% of the receiver positions J. Patynen, B. F. G. Katz, and T. Lokki, Investigations on the Balloon as an Impulse Source. The Journal of the Acoustical Society of America 129 ( 2011): R. S. Martin and M. Arana. Uncertainties Caused by Source Directivity in Room-Acoustic Investigations. The Journal of the Acoustical Society of America 123 (2008):

39 Therefore, even omnidirectional loudspeakers do not have entirely reproducible signals, despite the intent of the standards that prescribe the use of them. Directivity of Sources The way in which sound radiates from the source to the receiver in each octave band is referred to as the directivity of the source. The directivity of a source is often plotted on a polar coordinate graph to show the directional characteristics of the source. The directionality of a human voice is plotted in Figure 2-1. The high frequencies, which are associated with consonant sounds, have much more energy to the front of the speaker, as opposed to the rear, which can be reduced up to 20dB from the front. 62 Figure 2-1. Polar plot of the directionality of human voice in 500 and 4,000 Hertz octave bands. Credit: M. D. Egan, Architectural Acoustics, p. 83 The directionality of a trumpet is shown in Figure 2-2. The lower frequencies tend to be more omnidirectional, while the high frequencies tend to propagate directly from 62 Egan, Architectural Acoustics,

40 the horn of the instrument. 63 Figure 2-2. Polar plot of the directionality of trumpet in 220, 480, 920, 1840 and 4,000 Hertz octave bands. Credit H. F. Olson, Music, Physics and Engineering, p. 235 Figure 2-3 shows the directivity patterns of medium and large balloons. The balloons tend to have more omnidirectional directivity in the 1,000 and 2,000 Hertz octave bands than in the 125, 250, 500, and 4,000 Hertz octave bands. Figure 2-3. Directivity patterns of medium and large balloons. Credit: Pätynen et al., Investigations on the balloon as an impulse source, p. 30 The directivity pattern of the JBL EON15 G2 that was used in these studies is shown in Figure 2-4. The low and middle frequencies have directivity patterns that propagate omnidirectionally. However, the high frequencies including 8,000 Hertz and above tend to propagate towards the front of the speaker. 63 H.F. Olson, Music, Physics and Engineering. (New York: Dover Publications, Inc.,1967),

41 Figure 2-4. Polar plot of JBL EON15 G2 in the 500, 1,000 and 4,000 Hertz frequencies showing and -9dB radii Figure 2-5 shows the approximate directivity pattern of the Norsonic 223 dodecahedral loudspeaker. Exact polar plots of the Norsonic 223 dodecahedral loudspeaker were not available, as the loudspeaker is an older model. However, the manufacturer provided the polar plot in Figure 2-5 as an approximation of the Norsonic 223, as it is a polar plot of a similar updated dodecahedral loudspeaker. The low and middle frequencies of 100, 315 and 1,000 Hertz tend to radiate omnidirectionally. The higher frequencies of 3,150 Hertz and above radiate directionally from each of the twelve speakers, which are shown with the brown undulations in Figure 2-5. Reverberation Time: Traditional Equations and Acoustical Standards Traditional Equations Several methods for obtaining the Reverberation Time are discussed below, as they are used in this study. The Sabine Equation for Reverberation Time is generally used for rooms that are not highly absorbent and when there is a mostly diffuse sound field. 64 The Equation is as follows: 64 Sabine, Collected Papers,

42 ( ) T is the Reverberation Time in seconds. V is the room volume in ft 3. α is the total surface area of room absorption in sabins. Figure 2-5. Polar plot of the Norsonic 276 loudspeaker, which was provided by the manufacturer as a comparable polar plot to the Norsonic 223 loudspeaker. Credit: Nor276 Dodecahedron Loudspeaker [cited 12 July 2012]. Available from The Norris Eyring equation is another equation used to estimate the Reverberation Time in a room. It is generally used for rooms that are more absorbent. The Norris- Eyring equation is as follows: T is the Reverberation Time in seconds. V is the room volume in ft 3. S is the total surface area in ft 2. is the mean sound absorption coefficient Beranek, Concert and Opera Halls,

43 Acoustical Standards ISO 3382 The ISO 3382 Standard is the International standard for the Measurement of the Reverberation Time of rooms with reference to other acoustical parameters. The ISO 3382 standard details measurement procedures for the measurement of room acoustic parameters. It contains descriptions of measurement techniques using both the interrupted noise and the impulse response methods to derive acoustical parameters. The ISO 3382 standard defines Reverberation Time as the time required for the sound pressure level to decrease by 60 db, at a rate of decay given by the linear leastsquares regression of the measured decay curve from a level 5dB below the initial level to 35 db below. 66 The standard states that the sound source should be as close to omni-directional as possible. The microphone positions should be placed no closer than 2 meters, so they are at least half a wavelength apart. 67 For low and normal coverage measurements, a minimum of 2 source positions should be used and located in the typical source positions. For example, in a room such as a worship space or classroom, where a speaker is typically located in the front of the room facing the audience, the loudspeaker should be placed in this same position. Three to four microphone positions are recommended for the receivers. 68 It is noted that the measurements should include octave bands from 63 to 4,000 Hertz in concert halls and 66 ISO 3382:1997, Acoustics Measurement of the reverberation time of rooms with reference to other acoustical parameters, International Organization for Standardization, Geneva, Switzerland ISO 3382, ISO 3382, 5. 43

44 rooms for speech. 69 For impulse response measurements, a linear least-squares fit line is used to determine the slope of the reverberation decay. The impulse response method is used in this study as the scientifically calibrated acoustical measurement system. As such, the ISO standard is referenced and provides the basis for which the scientifically calibrated measurements are taken, as it is the only standard the specifically deals with the measurement of rooms using the impulse response method. ASTM E 2235 The ASTM E 2235 Standard is the only American standard the deals with Reverberation Time. However, this standard must be used in conjunction with acoustical testing of partitions. Unlike the ISO 3382 standard, it does not dictate measurement procedure to determine Reverberation Time on its own. The ASTM E2235 standard states specifically in section 5.5, This test method shall not be used when room sound absorption or decay rate is to be used directly to satisfy some criterion, for example in a room that must not be overly reverberant so speech will be intelligible. 70 It adds to this in a Note following section 5.5, stating The uncertainty of the room absorption will usually be too high and additional measurements are necessary. The ASTM standard requires an omnidirectional speaker, but makes an allowance stating In practice, using multiple driver elements to cover different frequency ranges and placing sources in trihedral corners of the room will be adequate. 71 This typically 69 ISO 3382, ASTM Standard E 2235, 2004, Determination of Decay Rates for Use in Sound Insulation Test Methods (West Conshohocken, PA: ASTM International, 2004), ASTM E 2235,

45 means a 2 or 3 way loudspeaker, which is inherently directional. This however, is not explicit in stating what the intent of the expression is. It may be allowing for directional loudspeakers, as long as they are placed in the corners of the room. The standard calls for at least one source position, but does not specify or recommend the location it should be placed. 72 At least three receiver locations are called for. 73 When determining the decay rate, the standard specifies that the first point to be included in the analysis should be as soon as practical after the sound has been switched off and no more than 5dB below the level when the sound was on. 74 Section 16.1 specifies that all points of analysis must be 10 db above the background noise level. Despite that this standard is not to be used as the sole method for determining the Reverberation Time in a room, it must be referenced for this study as it is the only American standard that addresses this measurement protocol. Just Noticeable Differences in Acoustical Metrics A Just Noticeable Difference (JND) is defined by Martellotta as the smallest perceivable change detected by 50% of the subjects. In statistical terms, this value is the average of JND s provided they are available. 75 JND s of the Reverberation Time and C 80 values for each sound source tested are discussed below. Reverberation Time In terms of Reverberation Time, Cremer cites an article written by H.P. Seraphim in 1958 in which approximately 500 test subjects were given listening tests to determine 72 ASTM E 2235, ASTM E 2235, ASTM E 2235, F. Martellotta, The Just Noticeable Difference of Center Time and Clarity Index in Large Reverberant Spaces. The Journal of the Acoustical Society of America 128 ( 2010):

46 Just Noticeable Differences in Reverberation Time. The relative difference limen for Reverberation Time was noted as δt/t which is denoted by the Greek letter delta, referring to the change in Reverberation Time divided by the Reverberation Time. It was found that in the region of greatest sensitivity (T between 0.6 and 4.0 s), the value of δt/t is between 3 and 4%. 76 Cremer continues that some subjects were musically trained, and that 4% should be used as the just noticeable difference term. 77 Several references to the subjective difference limen were also found, however they did not include the original source information where the number originated. Kleiner, Klepper and Torres mention in Worship Space Acoustics that changes of less than 5% in Reverberation Time can be noticed. 78 Ingolf Bork wrote a series of papers on acoustical measurement simulations in which he uses 5 ms as the subjective difference limen for Reverberation Time. 79 The original source from which he obtained the 5ms for the subjective difference limen is not indicated in the text of the article. Early Decay Time Martellotta performed a study on the Just Noticeable Difference in Center Time and Clarity in reverberant spaces in Martellotta includes a table that has Just Noticeable Difference values of 0.05% for Early Decay Time in his study 80. In a study 76 Cremer and M ller, Room Acoustics, Cremer and M ller, Room Acoustics, M. Kleiner, D. L. Klepper, and R. R. Torres, Worship Space Acoustics (Ft. Lauderdale: J. Ross Publishing, 2010), I. Bork, Report on the Second International Round Robin on Room Acoustical Computer Simulation. The Journal of the Acoustical Society of America 105 (1999): Martellotta, The Just Noticeable Difference,

47 performed by Skålevik that establishes that Reverberation Time, volume and sourcereceiver distance can predict sound level, reverberance, clarity, apparent source width and listener envelopment. The 0.05% value for JND in EDT is echoed in Skålevik s study, but does not list an original source as to how this value was determined 81. Clarity (C 80 ) John Bradley et al. performed a study in which the JND for the clarity index for speech, C 50, was tested. Subjects listened to simulated sound fields and made determinations between which were the same and which were different. 82 This experiment determined that there is a 1.1 db JND for C 50, and that the equation used to generate the relationship can also be applied to C 80, resulting in a JND of 0.9 db for C Ahearn et al. performed a study in the past several years in which 51 musically trained subjects were given listening tests to determine JNDs for music motifs in an anechoic chamber that was digitally configured to have 1.6 and 2.1 second Reverberation Times. Results from all participants yielded little information, however, when 17 selected subjects choices were analyzed, it was determined that 1.6dB was the JND M. Skålevik, Reverberation Time: The Mother of All Acoustic Parameters Proceedings of the 20th International Conference on Acoustics. Sydney, Australia (August 23-27, 2010): J. S. Bradley, R. Reich, and S. G. Norcross. A Just Noticeable Difference in C50 for Speech. Applied Acoustics 58 (1999): Bradley et al., A Just Noticeable Difference, M. J. Ahearn, Matthew J. Schaeffler, Robert D. Celmer, and Michelle C. Vigeant. Investigation of the Just Noticeable Difference of the Clarity Index for Music, C80. The Journal of the Acoustical Society of America 126 (2009):

48 Martellotta performed an experiment in which acoustic measurements for 3 large spaces with different long Reverberation Times were performed to determine if the Reverberation Time affected the JND for clarity metrics. 85 The Reverberation Times for the spaces were 2.1, 4.1 and 6.0 seconds. 86 It was found that in these larger spaces with longer Reverberation Times, the JND for C 80 was db. 87 This value will be used as it relates to spaces with longer Reverberation Times than the other literature. The technical concepts, background, relevant studies, standards and soundscape theory serve as the theoretical and philosophical framework that is at the heart of the study. Scientifically calibrated acoustical measurements are not representative of the real sounds that happen in a room. By analyzing a space with the soundscape method and understanding the acoustic events that actually take place in a space, one can begin to appreciate how those sounds are affected by the room. By deriving acoustic metrics from natural acoustic sources one can begin to have a deeper understanding of how a space affects sounds that take place within it. 85 Martellotta, The Just Noticeable Difference, Martellotta, The Just Noticeable Difference, Martellotta, The Just Noticeable Difference,

49 CHAPTER 3 ACOUSTIC SOURCE COMPARISON STUDIES General Information The Baughman Center is a non-denominational chapel located on the University of Florida campus. Constructed in 2000, it is made of low-e energy efficient glass, and the exterior is Florida cypress. The interior of the chapel is painted structural steel, southern yellow pine tongue in groove roof planking and gypsum board. Its area measures approximately 1500 feet 2, and its dimensions are 65 6 x It can accommodate up to 96 people seated. 1 Floor Plan The floor plan in the Baughman Center is considered a rectangular shape. The measurements are 21 6 wide by 72 long. The distance from the rear of the stage to the furthest pew is approximately 62. According to Kleiner, Klepper and Torres, 60 is the maximum distance an untrained speaker can naturally project into a space without the need for amplification. 2 Reverberation Average Reverberation Times in worship spaces vary from 1 second to 3 seconds. For a room with a volume similar to the Baughman Center, the desired Reverberation Time is.07 seconds for unassisted speech to 1.3 seconds in the 500 Hertz octave band for chamber music and choral music for a space that is approximately 1,152m 3. The measured Reverberation Time in the Baughman Center is approximately 2.6 seconds in 1 Baughman Center [Web site] (2010); available from Internet; accessed 10 August M. Kleiner, D. L. Klepper, and R. R. Torres, Worship Space Acoustics (Ft. Lauderdale: J. Ross Publishing, 2010),,

50 the 500 Hertz octave band using the Maximum Length Sequence (MLS) measurement procedure. At the highest peak, the ceiling height is 44 feet, the lowest point is 20 feet 6 inches, and on average 32 feet 4inches. Materials Rooms for worship are generally made of reflective surfaces with little sound absorption. The materials in the Baughman Center are hard reflecting surfaces. The roof is constructed of Southern yellow pine tongue-in-groove planking. The floor is made of travertine marble and the stage area is made of maple with inlaid cherry around the exterior. The walls are made of gypsum board and glass windows. There are two rows of seating in the Baughman Center, made of maple and cherry, with a light blue fabric upholstery on the seats and backs of the pews. Pilot Study The pilot study was inspired by a fellow University of Florida student wishing to obtain simple Reverberation Time measurements for the space. When in the space, it was clear that the building design did not affect sound in a desirable way. Because of its religious/spiritual context and the fact that the space is used by musical performers from the University of Florida, the building should be able to provide adequate acoustics for speech and music. However, this is not the case. Speech and music in the space sound muddy and individual sounds tend to run together. Due to the difficulties for both types of communication, the idea to perform multiple acoustical measurements with varying source signals was developed to determine if acoustic metrics derived from the various sounds in the room are affected by the signal type. Anechoic speech and music signals are compared to scientifically calibrated signals to determine if the different source signals result in different acoustic metric values. 50

51 Method Source signals were generated using WIN MLS software and a laptop computer. A sine sweep, multiple sine sweep and Maximum Length Signals (MLS) were generated by a JBL EON15 G2 loudspeaker. The JBL EON15 G2 loudspeaker is a two-way speaker that has a frequency response from 39 to 18,000 Hertz. 3 The loudspeaker was placed in the front center of the stage and the left corner of the stage and measured in three receiver locations throughout the audience area. The signal was received by an Earthworks microphone connected to the Digigram VXPocket 2 instrumentation quality sound card input in the computer. The signal is processed on a laptop computer using WinMLS software to derive the impulse response and calculate acoustical measurements such as reverberation time. The microphone was mounted on a stand approximately 42" (seated ear height)above the floor. The measurement performed at each receiver location was generated three times. Electronic Signal Results The standard deviation of each source type is shown in Figure 3-1. The Sine Sweep in the front of the room, the MLS signal in the corner and the MLS signal in the front had standard deviation values greater than 0.1 in the 63 Hertz octave band. The remaining octave bands four all source signals had standard deviations as low as 0.1. The measured Reverberation Times for each signal source are shown in Figure 3-2. The values in the 63 Hertz octave band varied from 1.19 to 2.46 seconds across all sources signals. However within each source type, the standard deviation is below 0.1 for all but 3 source signals. 3 Nor276 Dodecahedron Loudspeaker [website] available from Internet; accessed 12 July

52 Standard Deviation of Reverberation Times in Seconds Standard Deviation of Scientifically Calibrated Signals Receiver Position 3 Dir corner sine sweep Dir Front Sine Sweep Octave Band Center Frequency in Hertz Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Front Dir Front MLS Dir Corner MLS Figure 3-1. Standard Deviation of Reverberation Times from all Scientifically Calibrated Signals To focus the scope of this study on the qualities of the natural acoustic sources, the middle frequency 500 and 1,000 Hertz octave bands are compared to the additional sources, as these frequencies are the main speech frequencies. As shown in Figure 3-3, the average middle frequency octave bands for the electronic signals vary just 0.04 seconds in the 500 Hertz octave band and 0.2 seconds in the 1,000 Hertz octave band at the Receiver 3 position. The values for the 500 Hertz octave band are slightly higher than the 1,000 Hertz octave band for all the electronic signals. The Reverberation Time values in the 500 Hertz octave band are nominally the same based upon the 99% Confidence Interval for the average RT of all the sources. The directional source in the front of the room, and the sine sweep and multiple sine sweep sources with the directional loudspeaker placed in the corner of the room all fell outside the 99% Confidence Interval in the 1,000 Hertz octave band. Due to the nature of how the scientifically calibrated acoustical measurements are performed, and the type of signals used, the sound emitted in the room for these tests does not accurately reflect the 52

53 Reverberation Time in Seconds nature of sounds that are typically produced in the room. The scientifically calibrated acoustical measurements present a maximization of the room s acoustics to derive a Reverberation Time that will most likely never happen during the use of the room. 3.5 Reveberation Times for Scientifically Calibrated Sources Receiver Position 3 Dir corner sine sweep Dir corner sine sweep Dir corner sine sweep 3 Dir Front Sine Sweep Dir Front Sine Sweep Dir Front Sine Sweep Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Front Mult Sine Sweep Dir Front Mult Sine Sweep Dir Front Dir Front MLS Dir Front MLS Dir Front MLS Octave Band Center Frequency in Hertz Figure 3-2. Comparison of all scientifically calibrated Reverberation Time measurements Dir Corner MLS Dir Corner MLS Dir Corner MLS Sound sources in the room will be directional, such as a human voice and musical instruments, and will not create a diffuse sound field in the room. Therefore, additional measurements were taken that more closely resembled the activities that might normally happen in the room, to evaluate whether the scientifically calibrated test 53

54 Reverberation Time in Seconds signals provide a realistic indicator of the Reverberation Times associated with realistic sounds that take place in the room. 2.7 RT Comparison of Scientifically Calibrated Sources with 99% CI Average Scientifically Calibrated Sources Dir corner sine sweep Dir Front Sine Sweep Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Front Dir Front MLS Octave Band Center Frequency in Hertz Dir Corner MLS Figure 3-3. Middle frequency Reverberation Times for average scientifically calibrated source signals with 99% Confidence Interval Natural Acoustic Signals In a setting such as the Baughman Center, typical sound sources would include a pastor or speaker talking in the front of the room such as for a wedding or religious ceremony. Music from live sources such as a violin, trumpet, or trombone may also be heard for a musical recital, which sometimes takes place in the chapel. Anechoic source signals were chosen that resemble the acoustic events that actually take place in the room. These signals were generated and recorded at the three receiver positions so that they could be analyzed to derive acoustic metrics from them. 54

55 Anechoic voice and music as measurement sources An anechoic recording of a female voice and trumpet music from the CATT- Acoustic program was used so that the exact timing and level could be duplicated at each receiver position. These sources were played into the room using a directional loudspeaker placed at the front center of the room to simulate where a person might actually speak or where live music might be played. These signals were recorded with the M-Audio Microtrack II WAV recorder with an electret microphone at the three measurement locations. The recorded signals were processed using Acoustic Tools software. The stop chord at the end of the trumpet song and the last phrase of the female talker s sentence was analyzed for this study. Live music as a measurement source A middle-c note, middle-c Chord and a short phrase of Tales from the Vienna Woods from Concert Waltz Opera 325 by Johann Strauss Jr. were played on the electronic piano located in the rear of the room. These signals were recorded with the M-Audio Microtrack II WAV recorder at the three measurement locations. The recorded signals were processed using Acoustic Tools software. The stop chord at the end of the piano song phrase was analyzed for this study. The source was located in the opposite end of the room from the source location for the other measurements and is noted in Figure 3-4 as S3 Piano. Because of this, the receiver location is much closer to the source signal than the other source types. The path from the source to the receiver is much different for the piano sources than from the other measurements 55

56 Figure 3-4. Floor plan of Baughman Center showing source and receiver locations Balloon pop as a measurement source Balloons were used as a sound source as well, as they are known for having a strong broad-band spectrum burst of sound. A balloon was popped in the front center of the room, in the same location that the source playback loudspeaker occupied. The Mictrotrack II WAV recorder was used to record the balloon signal at the three receiver locations in the room. The balloon signal was processed using the Win MLS software and Acoustic Tools. The Win MLS software provides a Reverberation Time (T30) without any manual calculation or manipulation. Acoustic Tools requires user input to determine Reverberation Times. Results Figure 3-4 shows the average Reverberation Times in the middle frequencies for the speech, music and balloon signals. The values for Reverberation Times range from 1.64 to 3.08 in the 500 Hertz octave band and range from 1.96 to 2.91 in the 1,000 Hertz octave band. The values for the 500 Hertz octave band are greater than the 1,000 Hertz octave band for the Balloon Pop measurements, the Female Talker, and the middle C Chord. The values for the 500 Hertz octave band are less than the 1,000 56

57 Reverberation Time in Seconds Hertz octave band for the middle C note, the piano song and the anechoic music played through the loudspeaker. Appendix A provides the images from the analysis in Acoustic Tools for the Balloon Pop, anechoic Female Talker, Trumpet Song, and live piano note, chord and song phrase. When compared to the electronic sources, the anechoic speech, music, Balloon Pop analyzed with Acoustic Tools, and live music are outside of the blue cluster for at least one of the octave bands shown on Figure 3-6 that includes the electronic sources and the Balloon Pop analyzed with Win MLS. 500 and 1,000 Hertz T30 Values for Natural Acoustic Sources Receiver Position Balloon Pop MLS Balloon POP AT Female Talker Piano Middle C Note Piano Middle C Chord Piano Song Anechoic Music Octave Band Center Frequency in Hertz Figure 3-5. Reverberation Times for the anechoic speech, music, live music and Balloon Pop sources in receiver position 3 57

58 Reverberation Time in Seconds and 1,000 Hertz T30 Values Receiver Position Octave Band Center Frequency in Hertz Dir corner sine sweep Dir Front Sine Sweep Mult Sine Sweep Dir Corner Mult Sine Sweep Dir Front Dir Front MLS Dir Corner MLS Balloon Pop MLS Balloon POP AT Female Talker Piano Middle C Note Piano Middle C Chord Piano Song Figure 3-6. Reverberation Times in the middle frequencies for all source types Pilot Study Conclusions It was determined that while the Reverberation Times measured with a scientifically calibrated method may provide an idea of how the room responds, ultimately, the Reverberation Time will be affected by the types of sources. It was found the source type and source location provides variation in the Reverberation Times. The range of Reverberation Times for the sources tested varied 1.44 seconds in the 500 Hertz octave band and 0.95 seconds in the 1,000 Hertz octave band. It was determined that consistent methods for analyzing the natural acoustic sources were needed. The data varied greatly, and in order to determine if this was due to inconsistencies in the method or if the sources produced that much variation, new methods to analyze the data needed to be developed. Comparing other acoustic metrics from the acoustical measurements taken, such as EDT, C 80, STI or D 50 may provide additional insight into the study. 58

59 The impulse response based measurements are internally consistent, as each speaker position and source signal generated similar Reverberation Times in the room. As mentioned previously, when comparing the live music played on the piano, it would be helpful to test other signals at the same location so as to accurately compare the response of the room to that certain path between source and receiver. Source Signal Comparison Method Testing took place on August 24, 2010, September 14, 2010, October, 5, 2010 and on January 23, It was determined from the pilot study that all three receiver locations should be used. The apparent placement of the speakers in the electronic piano included a speaker that faces out from the rear of the piano and a speaker facing towards the piano player. It would be difficult to set up a speaker to run test signals through that matched this configuration, so it was decided to not concentrate on the data collected from the rear of the room. It was also decided to add the 4,000 Hertz octave band into the final analysis. The natural acoustic signals did not have as much energy in the low frequencies, so analyzing the middle and higher frequencies would allow more data that was collected to be used. Another study was performed that used the Pilot Study as the basis for the measurements, but provided alternative means of analyzing the data. A JBL EON15 G2 directional source playback loudspeaker was used as the source signal. The source playback loudspeaker was positioned in the front center of the room facing the audience, and in the front left corner of the room facing upward. Three receiver positions were used due to the small room size. Microphones were positioned at approximately 1.5m or seated ear-height. Source and Receiver locations are shown in Figure

60 Figure 3-7. Floor plan of Baughman Center indicating source types and positions and receiver positions Study 1: Omnidirectional, Directional and Calculated Comparisons Acoustical measurements were performed using an omnidirectional and directional loudspeaker with an MLS signal. ISO and ASTM dictate that an omnidirectional loudspeaker should be used for acoustical measurements, which will in theory create a diffuse sound field. However, most typical sound sources (human voices, instruments, etc.) are not omnidirectional and do not excite a room in such a way that they fully excite the reverberant field in the room. Therefore, it was proposed to examine if there are differences in using an omnidirectional and a directional loudspeaker on acoustic metrics. These impulse response measurements are compared to each other and to Reverberation Time values calculated using the Sabine Equation and the Norris-Eyring Equation. The Sabine Equation is a better indicator of the reverberation because the chapel is constructed of mostly reflective materials. The Norris-Eyring Equation is used for spaces that are sound absorbent. The calculated equations provide whole room averages, however each scientifically calibrated receiver position is analyzed independently and the results are shown in Figures 3-8 through

61 Electronic signals Maximum Length Sequence (MLS) signals were generated using WIN MLS software and a laptop computer. The MLS signal was generated from a JBL EON15 G2 directional loudspeaker placed in the front center of the stage and the left corner of the stage and measured in three receiver locations throughout the audience area. The JBL EON15 G2 is a two way loudspeaker with a maximum sound power level of 129dB. The MLS signals were also generated with a Norsonic 223 omnidirectional loudspeaker. The omnidirectional loudspeaker is a dodecahedral loudspeaker that consists of speakers with a maximum sound power outlet of 118dB 4. The measurement performed at each receiver location was repeated three to five times. Reverberation time The standard deviations for the Reverberation Times derived from acoustical measurements taken with the omnidirectional and directional loudspeaker at each receiver location are shown in Figures 3-8 through In the first receiver location, the directional loudspeaker has lower standard deviation values than the omnidirectional in all octave bands except the 63 Hertz band, which is less than one hundredth of a second in standard deviation from the omnidirectional. The omnidirectional speaker had a standard deviation of 0.23 in the 125 Hertz and 0.11 in the 1,000 Hertz octave band. Both the omnidirectional and directional loudspeakers had standard deviations of less than 0.05 in the 250, 2,000, 4,000 and 8,000 Hertz octave bands. 4 Norsonic 213 and 223 Dodecahedron Speaker Specifications [Web site] available from Internet; accessed 12 July

62 Standard Deviation of Reverberation Time in Seconds Standard Deviation of Reverberatio Time in Seconds 0.25 Standard Deviation of Omnidirectional and Directional Louspeakers T30 Measurements: Receiver Position Omnidirectional Receiver 1 Directional Receiver Octave Band Center Frequency in Hertz Figure 3-8. Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 1 st receiver position 0.25 Standard Deviation of Omnidirectional and Directional Louspeakers T30 Measurements: Receiver Position Omnidirectional Receiver 2 Directional Receiver Octave Band Center Frequency in Hertz Figure 3-9. Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 2 nd receiver position 62

63 Standard Deviation of Reverberation Time in Seconds For receiver position 2, the Reverberation Times measured with the directional loudspeaker had lower standard deviation values than the omnidirectional loudspeaker with the exception of the 4,000 Hertz octave band, which is less than 5 thousandths of a second standard deviation from the omnidirectional loudspeaker. The omnidirectional loudspeaker has standard deviation values of less than 0.1 seconds for all but the 63 and 125 Hertz octave band at this position. The directional loudspeaker has standard deviations of less than 0.03 in all octave bands Standard Deviation of Omnidirectional and Directional Louspeakers T30 Measurements: Receiver Position Omnidirectional Receiver 3 Directional Receiver Octave Band Center Frequency in Hertz Figure Standard deviation of Reverberation Time for the omnidirectional and directional loudspeaker in the 3 rd receiver position The standard deviation for the Reverberation Time measurements conducted with the directional speaker was higher than the omnidirectional in the 63, 250 and 2,000 Hertz octave bands for the third receiver location. The standard deviation for the Directional loudspeaker was below.016 for all octave bands and below 0.09 for all octave bands except the 63 Hertz. The omnidirectional loudspeaker had the highest standard deviation in the 500 Hertz octave band, with a value of 0.18 seconds. 63

64 Reverberation Time in Seconds In receiver positions 1 and 2, the Reverberation Time derived from the acoustical measurements made using the directional loudspeaker had lower standard deviations than the omnidirectional speaker in most octave bands. However, in the third receiver position, the highest standard deviation value for the Reverberation Time derived from acoustical measurements made using for the directional speaker at 63 Hz was The Reverberation Times derived using the omnidirectional speaker had higher standard deviations in receiver positions 1 and 2. The Reverberation Times derived from acoustical measurements using the omnidirectional and directional loudspeakers were also compared to calculated Reverberation Time values using the Sabine and Norris Eyring Equations. Figure 3-11 shows the averaged Reverberation Times derived from measurements made using the omnidirectional and directional loudspeakers, as well as calculated Reverberation Times derived from the Sabine and Norris Eyring equations. The calculated Reverberation Times are a good approximate for the whole room average measured Reverberation Times. 3.5 Comparison of Reverberation Time Measurements and Calculations Octave Band Center Frequency in Hertz Omni Average MLS Overall Directional Average MLS Overall Sabine Equation Norris Eyring Equation Figure Average measured and calculated Reverberation Times for the Baughman Center 64

65 Clarity (C 80 ) Value in db Clarity (C 80 ) The Reverberation Times measured with the omnidirectional and directional loudspeaker in each receiver position remained relatively consistent. The Clarity Index (C 80 ) or measure of the early to late energy ratio after 80 ms was examined as well. Graphs of the results for the C 80 values are shown in Figures 3-12 through Three to five measurements were made at each receiver position. These measurements are illustrated in the following graphs using the receiver number followed by the measurement number. For example, the third measurement taken at the 1 st receiver position is noted as 1-3. Clarity (C 80 ) Comparison between Omnidirectional and Directional Loudspeakers Receiver Octave Band Center Frequency in Hertz Omni MLS Receiver 1_1 Omni MLS Receiver 1_2 Omni MLS Receiver 1_3 Omni MLS Receiver 1_4 Omni MLS Receiver 1_5 Dir MLS Receiver 1_1 Dir MLS Receiver 1_2 Dir MLS Receiver 1_3 Figure Clarity (C 80 ) values measured at the first receiver position with omnidirectional and directional loudspeakers The C 80 values for the acoustical measurements made using the directional loudspeaker are higher than the values derived from acoustical measurements made using the omnidirectional loudspeaker in all octave bands except the 125 Hertz octave 65

66 Clarity (C 80 ) Value in db bands in the first receiver position. The only octave band that the directional loudspeaker had a negative value for was the 125 Hertz octave band. The omnidirectional loudspeaker achieved lower C 80 values in all but the 125 Hertz octave bands. C 80 values for the omnidirectional loudspeaker ranged from -1.6 to 2.3 db. The C 80 values in the 63 Hertz octave band are not noticeably different (0 Just Noticeable Differences or JND) between the loudspeaker types, the C 80 values in the 125 to 2,000 Hertz octave band range between 1 and 2 JND, and the C 80 values for the directional loudspeaker in the 8,000 Hertz octave band are 5 JND higher than the omnidirectional loudspeaker. 10 Clarity (C 80 ) Comparison between Omnidirectional and Directional Loudspeakers Receiver Omni MLS Rec 2_1 Omni MLS Rec 2_2 Omni MLS Rec 2_3 Omni MLS Rec 2_4 Omni MLS Rec 2_5 Dir MLS Rec 2_1 Dir MLS Rec 2_ Octave Band Center Frequency in Hertz Dir MLS Rec 2_3 Figure Clarity (C 80 ) values measured at the second receiver position with omnidirectional and directional loudspeakers The acoustical measurements derived from using the omnidirectional loudspeaker achieved slightly higher C 80 values in the 63 and 125 Hertz octave bands than acoustical measurements made with the directional loudspeaker in receiver position 2. 66

67 Clarity (C 80 ) Value in db However, the acoustical measurements derived from using the directional loudspeaker achieved higher C 80 values in the middle and high frequency octave bands. There was not noticeable difference in C80 values for the directional and omnidirectional loudspeaker in the 63 and 250 Hertz octave bands (0 JND). In the 125, 500, 2,000 and 4,000 Hertz octave bands, the JND ranged from 1-3 between loudspeaker types. The C 80 values for the directional loudspeaker in the 1,000 and 4,000 Hertz octave bands are 4 JND higher than the omnidirectional values Clarity (C 80 ) Comparison between Omnidirectional and Directional Loudspeakers Receiver Octave Band Center Frequency in Hertz Omni MLS Rec 3_1 Omni MLS Rec 3_2 Omni MLS Rec 3_3 Omni MLS Rec 3_4 Omni MLS Rec 3_5 Dir MLS Rec 3_1 Dir MLS Rec 3_2 Figure Clarity (C 80 ) values measured at the third receiver position with omnidirectional and directional loudspeakers The acoustical measurements derived from using the omnidirectional loudspeaker achieved negative C 80 values in all the octave bands in the third receiver position. The acoustical measurements derived from using directional loudspeaker had higher C 80 values than the omnidirectional in all octave bands. The values for C 80 derived from the directional loudspeaker were not noticeably different than values for C 80 derived from the omnidirectional loudspeaker (0 JND). The directional loudshpeaker had C 80 values 67

68 that were 1-3 JND higher than the omnidirectional loudspeaker in the 125 through 4,000 Hertz octave bands. The C 80 values for the directional loudspeaker in the 8,000 Hertz octave band are 4 JND higher than the omnidirectional loudspeaker. Study 2: Anechoic Music and Speech and Balloon Pop Stimuli Comparison This analysis looks at a modified version of the pilot study and compares the acoustical metrics derived from using anechoic speech and music and Balloon Pop stimuli by 4 methods. Anechoic music and speech were played through the JBL EON G12 directional loudspeaker and recorded with a binaural M-Audio WAV recorder to simulate how the room affects these signals with a reproducible means. The source location was at the front center of the room. The receiver locations were placed in the front, middle and rear of the room, in the same receiver locations used in the previous studies. Please refer to Figure 3-4 page 58 for the source and receiver locations. To analyze how the room would be affected by source located in the front center of the space, only the loudspeaker placed in the front of the room was analyzed. There were 4 binaural methods used to determine the acoustic metrics from the anechoic sound sources. Each method, Traditional, Expanded, and WinMLS, is described in detail below. Traditional method. The Traditional approach uses the decay curve of the source signal and the ISO 3382 protocol that dictates that the linear least squares regression start at a point 5 db below the initial sound level. 5 Acoustic Tools software was used to analyze the data. In the Intelligibility module of the software program, the direct sound cursor was placed at the top of the decay curve when the sound stopped. 5 ISO 3382:1997, Acoustics Measurement of the reverberation time of rooms with reference to other acoustical parameters, International Organization for Standardization, Geneva, Switzerland

69 The reflected sound cursor was placed at a point on the curve approximately 5 db down from the direct sound. The noise floor cursor was placed approximately 10dB above the ambient. By placing the cursors at these points, acoustical metrics are calculated within the Acoustic Tools software program, and results for Reverberation Time, Early Decay Time and C 80 are given, among others. An example of the placement of the cursors is given in Figure Expanded method. The Expanded approach includes more data than the Traditional method to plot the least squares fit line. The direct sound cursor is kept in the same place as the Traditional Method, however the reflected sound cursor is placed on the highest reflection just after the direct sound. The noise floor cursor is placed at the end of the decay, where the decay meets the noise floor, instead of being 10dB above it. These cursor positions are shown in Figure method. The observation method uses listening to determine when the direct sound stops and when the reverberant sound stops. The wav file used in the previous two methods is imported into the Audacity software program. The end note of the Trumpet Song and the last syllable of the Female Talker were analyzed. A 500, 1,000 and 4,000 Hertz octave filter was applied to the wav files independently, so that each octave band could be listened to separately. The wav file was played back via Sony MDR-V600 Dynamic Stereo Headphones. To determine where the direct sound stopped, the wav file was played back at normal speeds and slower speeds. To determine where the reverberant sound stopped, the wav files were played back at speeds generally between 0.2 to 0.4 times slower than the original sound. 69

70 Direct Sound Cursor Reflected Sound Cursor at -5dB from the Direct Sound Noise Floor Cursor at +10 db above Noise Floor Noise Floor Figure Placement of cursors to derive acoustic metrics for the Traditional Method using the end note of the Trumpet Song at 1,000 Hertz as an example 70

71 Direct Sound Cursor Reflected Sound Cursor at first reflection after Direct Sound Noise Floor Cursor at Noise Floor Noise Floor Figure Placements of cursors to derive acoustical metrics from the Expanded Method 71

72 The wav files were listened to several times at different speeds to adequately determine where the direct and reverberant sound stopped. The times when the direct sound stopped and the reverberant sound stopped were noted. The wav files were then opened in Acoustic Tools. In the Intelligibility mode, the Reflected Sound cursor was moved to the time the direct sound was heard to stop. The Noise Floor cursor was placed on the time where the reverberant sound was heard to stop. The Noise Floor cursor was adjusted so that it could provide a visual least squares fit line on the decay curve. The RT, C 80 and EDT were determined from positioning the cursors at these perceptual points. WinMLS method. The WinMLS method involves using the time of the direct sound from the previous methods, and the time of the perceived end of the direct sound and end of reverberation. This method essentially took the data from the Method and processed it using the WinMLS software program. The direct sound, end of direct sound and end of reverberation was noted for each octave band. A separate wav file was made that contained only the data between the times associated with the reflected sound cursor and noise floor cursor. This new wav file was imported and analyzed with the WinMLS software to determine acoustic measures. Certain frequencies did not have enough signal level for the program to recognize and therefore no data was recorded. Figures 3-19 through 3-20 show in detail each step of the process. 72

73 Direct Sound: Where Direct Sound was heard to stop Where Reflected Sound was heard to stop Figure End note of Trumpet Song with 1,000 Hertz octave filter applied to signal. Note the times, as they are used in the next step 73

74 Direct Sound Where Direct Sound was heard to stop Where Reflected Sound was heard to stop Noise Floor Figure Cursor placement after filtering and listening to Trumpet Song stop chord. Note placement of cursors corresponds to times identified in Figure

75 Figure New wav file made by clipping the top image between the Reflected Sound and Noise Floor cursors 75

76 Figure The new wav file from Figure 3-19 was run through WinMLS software program and acoustic metrics derived from this wav file RT Comparison The Reverberation Times derived from the natural acoustic sources were compared to the reverberation times derived from the scientifically calibrated sources. The Reverberation Times derived from the natural acoustic sources and scientifically calibrated sources were also compared to the Reverberation Times derived from the sources in the Pilot Study. The Reverberation Times derived from all three studies are shown in Figure Data from the scientifically calibrated sources in the 4,000 Hertz octave band is included in the figure even though it was not analyzed in the Pilot Study. This data is meant to provide a frame of reference for the data derived from the natural acoustic sources in the same frequency. There Reverberation Time data ranges from 0.71 to 1.80 seconds across the 500, 1,000 and 4,000 Hertz octave bands. 76

77 Reverberation Time in Seconds Octave Band Center Frequency in Hertz Pilot Study: Dir corner sine sweep Pilot Study: Dir Front Sine Sweep Pilot Study: Mult Sine Sweep Dir Corner Pilot Study: Mult Sine Sweep Dir Front Pilot Study: Dir Front MLS Pilot Study: Dir Corner MLS Pilot Study: Balloon Pop MLS Pilot Study: Balloon POP AT Pilot Study: Female Talker Pilot Study: Piano Middle C Note Pilot Study: Piano Middle C Chord Pilot Study: Piano Song Pilot Study: Trumpet Song Study 1: Dir Front MLS Study 1: Dodec Front MLS Study 2: Traditional Method: Balloon Pop Study 2: Traditional Method: Female Talker Study 2: Traditional Method: Trumpet Song Study 2: Expanded Method: Balloon Pop Study 2: Expanded Method: Female Talker Study 2: Expanded Method: Trumpet Song Study 2: Method: Balloon Pop Study 2: Method: Female Talker Study 2: Method: Trumpet Song Study 2: WinMLS Method: Balloon Pop Study 2: WinMLS Method: Female Talker Study 2: WinMLS Method: Trumpet Song Figure Reverberation Time comparison for Pilot Study, Study 1 and Study 2 in the 500, 1000 and 4000 Hertz octave bands 77

78 Difference of RT - EDT in Seconds RT EDT Comparison Natural acoustic sources By comparing the findings of this study to previous studies, one can attempt to show how the metrics derived from natural acoustic sources differ from metrics derived from scientifically calibrated sources. By subtracting the EDT from the RT derived from all 4 methods, it was found that the differences in the EDT and RT metrics are considerably larger than the differences of approximately 0.1 seconds found by Bradley. It should be noted that the metric described as the EDT for the natural acoustic sources is actually a relative value determined by obtaining the slope of the line that connects the direct sound cursor to the reflected sound cursor in Acoustic Tools. Using the Traditional method with the Female Talker stimulus, the differences between EDT and RT ranged from -.37 to 2.16 seconds. When averaging both the left and right receivers over all 3 receiver positions, it was found that the RT was.12 in the 4,000 Hertz octave band to.96 seconds in the 500 Hertz octave band longer than the EDT. Female Talker RT - EDT Traditional Method Octave Band Center Frequency in Hertz Traditional Right R1 Traditional Left R1 Traditional Right R2 Traditional Left R2 Traditional Right R3 Traditional Left R3 Figure Female Talker RT EDT analyzed using thetraditional Method 78

79 Difference in RT - EDT in Seconds The Female Talker source signal analyzed using the Extended Method resulted in differences in EDT and RT of -.55 to 2.53 seconds, which is shown in Figure When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 2.11 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 1.66 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.14 seconds. The Female Talker source signal analyzed with the Method resulted in differences in the RT and EDT of between and 2.2 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 1.17 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 1.32 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.09 seconds. Female Talker RT- EDT Expanded Method Octave Band Center Frequency in Hertz Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Figure Female Talker RT EDT analyzed using the Expanded Method 79

80 Difference in RT - EDT in Seconds Difference in RT - EDT in Seconds Female Talker RT - EDT Method Octave Band Center Frequency in Hertz Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Figure Female Talker RT EDT analyzed using the Method The Female Talker source analyzed with the WinMLS Method resulted in differences in the RT and EDT between and 0.60 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by.003 seconds. The averaged RT in the 1,000 Hertz octave band was less than the EDT by -.17 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.01 seconds. Female Talker RT - EDT WinMLS Method Octave Band Center Frequency in Hertz WinMLS Right R1 WinMLS Left R1 WinMLS Right R2 WinMLS Left R2 WinMLS Right R3 WinMLS Left R3 Figure Female Talker RT EDT analyzed using the WinMLS Method 80

81 The Female Talker source analyzed using the Expanded and methods resulted in differences in the RT EDT that were similar. The Female Talker source analyzed using the WinMLS method resulted in the lowest differences in RT EDT. When using the Traditional Method with the Trumpet Song as a stimulus, the difference between RT and EDT varied from to 2.59 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 1.33 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.21 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.41 seconds. The Trumpet Song analyzed with the Expanded Method resulted in differences between RT and EDT varying from to 2.5 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 1.19 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.32 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.35 seconds. The Trumpet Song analyzed with Method resulted in variances between the RT and EDT from to 2.19 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 0.36 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.41 seconds. The averaged RT in the 4,000 Hertz octave band was less than the EDT by seconds. 81

82 Difference in RT - EDT in Seconds Difference in RT - EDT in Seconds Trumpet Song RT - EDT Traditional Method Traditional Right R1 Traditional Left R1 Traditional Right R2 Traditional Left R2 Traditional Right R3 Traditional Left R Octave Band Center Frequency in Hertz Figure Trumpet Song RT EDT analyzed using thetraditional Method Trumpet Song RT - EDT Expanded Method Octave Band Center Frequency in Hertz Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Figure Trumpet Song RT EDT analyzed using the Expanded Method 82

83 Difference in RT - EDT in Seconds Trumpet Song RT - EDT Method Octave Band Center Frequency in Hertz Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Figure Trumpet Song RT EDT analyzed using the Method The Trumpet Song analyzed using the WinMLS Method resulted in differences in the RT and EDT of between and 0.52 seconds. When averaging the left and right side over all three receiver positions, the RT was less than the EDT in the 500 Hertz octave band by seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.17 seconds. The averaged RT in the 4,000 Hertz octave band was less than the EDT by seconds. Similar to the Female Talker, the Trumpet Song resulted in similar differences in the RT EDT using the Expanded and Method. The Trumpet Song analyzed using the WinMLS also resulted in the lowest differences in RT EDT. The Balloon Pops were also analyzed using the same methods that were used to analyze the natural acoustic stimuli. They were used as they provide an approximately broadband stimulus that excites the frequencies that are analyzed in this paper. Although balloons have been found to not meet the ISO 3382 standard for being an 83

84 Difference in RT - EDT in Seconds omnidirectional source, they were used in this study as a comparison against the natural acoustic sources. The Balloon Pop analyzed with the Traditional Method resulted in differences in RT and EDT ranging from to 1.55 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 0.61 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.87 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.79 seconds. The Balloon Pop analyzed with the Expanded Method resulted in differences between RT and EDT varying from to 2.1 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 1.17 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 1.59 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 1.14 seconds. Trumpet Song RT - EDT WinMLS Method Octave Band Center Frequency in Hertz WinMLS Right R1 WinMLS Left R1 WinMLS Right R2 WinMLS Left R2 WinMLS Right R3 WinMLS Left R3 Figure Trumpet Song RT EDT analyzed using the WinMLS Method J. Patynen, B. F. G. Katz, and T. Lokki, Investigations on the Balloon as an Impulse Source. The Journal of the Acoustical Society of America 129 ( 2011):

85 Difference in RT - EDT in Seconds Difference in RT - EDT in Seconds Balloon Pop RT - EDT Traditional Method Octave Band Center Frequency in Hertz Traditional Right R1 Traditional Left R1 Traditional Right R2 Traditional Left R2 Traditional Right R3 Traditional Left R3 Figure Balloon Pop RT EDT analyzed using the Traditional Method Balloon Pop RT - EDT Expanded Method Octave Band Center Frequency in Hertz Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Figure Balloon Pop RT EDT analyzed using the Expanded Method The Balloon Pop analyzed with the Method resulted in differences in the RT and EDT of between and 2.06 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 1.60 seconds. The averaged RT in the 1,000 Hertz octave band was 85

86 Difference in RT - EDT in Seconds greater than the EDT by 1.71 seconds. The averaged RT in the 4,000 Hertz octave band was greater than the EDT by 0.57 seconds. Balloon Pop RT - EDT Method Octave Band Center Frequency in Hertz Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Figure Balloon Pop RT EDT analyzed using the Method The Balloon Pop source analyzed with the WinMLS Method resulted in differences in the RT and EDT of between and 0.36 seconds. When averaging the left and right side over all three receiver positions, the RT was greater than the EDT in the 500 Hertz octave band by 0.06 seconds. The averaged RT in the 1,000 Hertz octave band was greater than the EDT by 0.08 seconds. The averaged RT in the 4,000 Hertz octave band was less than the EDT by 0.01 seconds. For the Traditional, Expanded and Methods, the EDT is regulated by where the Direct Sound cursor is placed on the curve. For the natural acoustic sources, the entire sound file had to be used when analyzing the decay curve. Because of this, even slight variations in the placement of the cursors could result in large differences in the EDT. 86

87 Difference in RT - EDT in Seconds Balloon Pop RT - EDT WinMLS Method Octave Band Center Frequency in Hertz WinMLS Right R1 WinMLS Left R1 WinMLS Right R2 WinMLS Left R2 WinMLS Right R3 WinMLS Left R3 Figure Balloon Pop RT EDT analyzed using the WinMLS Method A clipped file of just the end note could not be analyzed using Acoustic Tools because the signal energy was spread out so much that the decay slope no longer had a 45 o slope (which is preferred for calculated Reverberation Time). The slope for the clipped file was approximately 165 o, and therefore not ideal to analyze. The averaged Balloon Pop data analyzed using the WinMLS Method resulted in the lowest RT-EDT differences in the 500, 1,000 and 4,000 hertz octave bands. These low differences are most likely due to the Balloon Pop having a steady, easily discernible decay curve, and broadband spectrum. The low differences for the Balloon Pop analyzed using the WinMLS method are also most likely due to the fact that the Balloon Pop signal was run through the WinMLS computer software program whose analysis techniques are internally consistent and not based on human input. Overall, the differences in RT EDT ranged from -2.5 to 2.59 seconds using the natural acoustic sources. The differences in RT EDT for the scientifically calibrated sources ranged from -.19 to 0.32 seconds. The differences in RT- EDT for scientifically calibrated are approximately 8 times shorter than the differences in RT-EDT in the natural acoustic 87

88 Difference in RT - EDT in Seconds sources. These findings did not correspond to Youngmin Kwon s results, which did not find significant differences in the RT and EDT values 5. Scientifically calibrated sources Values for EDT and RT using the scientifically calibrated methods were also compared. It was found that with EDT and RT measurements derived from acoustical measurements using the omnidirectional loudspeaker in the first receiver position, the EDT values were generally higher than RT values in the 500 Hertz octave band. However, in the 1,000 and 4,000 Hertz octave bands, the EDT values were slightly lower than the RT values. Overall, the differences from subtracting EDT from RT with both loudspeaker types are similar, and ranged from +.01 to 0.3 seconds. This is a much smaller variation than the alternative source signals, however still a larger variation than what Kwon and Jordan have found. Directional RT - EDT Average Dir R1 Avg Dir R2 Avg Dir R3 Avg Octave Band Center Frequency in Hertz Figure Averaged MLS measurements dervied from the directional loudspeaker subtracting EDT from RT at receiver locations R1, R2 and R3 5 Youngmin Kwon, Quantitative and Qualitative Analyses of Under-Balcony Acoustics with Real and Simulated Arrays of Multiple Sources. (PhD diss., University of Florida, 2006),

89 Difference in RT - EDT in Seconds Omnidirectional Loudspeaker RT - EDT Average Octave Band Center Frequency in Hertz Dodec R1 Avg Dodec R2 Avg Dodec R3 Avg Figure Averaged MLS measurements dervied from the omnidirectional loudspeaker subtracting EDT from RT at receiver locations R1, R2 and R3 Reverberation Time Comparison Results The Reverberation Time data from methods 1-4 were normalized to the average Reverberation Time values from acoustical measurements made using the MLS signal at each receiver position. The average MLS values became the 0 and the values derived for the natural acoustic sources were plotted on the graphs with the average MLS values as the 0. Negative numbers indicated the values derived from the natural acoustic sources were below the average MLS data, while positive values indicated values higher than the average MLS data. This allows us to determine how different the natural acoustic sources are compared to the traditional scientifically calibrated sources. The normalized Reverberation Times were then compared by source, method and receiver position. It was found that the Trumpet Song had longer reverberation times than the Female Talker. Figure 3-36 shows graphs of the Female Talker and Trumpet Song Reverberation Time data normalized to the Average MLS data for the 1 st receiver position. The RT data for the Trumpet Song are closer to the normalized value of 0, and therefore have higher values than the Female Talker data. The Reverberation 89

90 Normalized Reverberation Time in Seconds Normalized Reverberation Time in Seconds Times derived from the Balloon Pop impulses tended to fall between the Female Talker and Trumpet Song Female Talker RT Normalized to Average MLS Receiver 1 Traditional Right R1 Traditional Left R1 Figure Female Talker and Trumpet Song Reverberation Times normalized to average MLS data The Reverberation Times derived using the Balloon Pop as the source signal resulted in all 4 methods having similar values. The similar values in RT are attributed to the Balloon Pop having high sound energy, having nearly equal energy per octave band and that it has nearly an omnidirectional dispersion pattern which uniformly excites the reverberant field in the room. The Reverberation Times derived from the natural acoustic source signals were different between methods and receiver position. The differences in RT between the Female Talker and Trumpet Song are most likely due to the limited sound energy, the sources exciting only limited frequencies and having directional characteristics from the directional loudspeaker. Figure 3-37 shows the normalized RT data for the Balloon Pop, revealing similar values for each method at each receiver position. Expanded Right R1 Expanded Left R1 Right R1 Left R1-3 WinMLS Right R WinMLS Left R1 Octave Band Center Frequency in Hertz Trumpet Song RT Normalized to Average MLS Receiver Traditional Right R1 Traditional Left R1 Expanded Right R1 Expanded Left R1 Right -1 R1 Left -2 R1-3 WinMLS Right R WinMLS Left R1 Octave Band Center Frequency in Hertz 90

91 Normalized Reverberation Time in Seconds Normalized Reverberatioin Time in Seconds Normalized Reverberation Time in Seconds For the natural acoustic sources, the Expanded and Methods resulted in similar values. Figure 3-38 shows graphs of the Female Talker and Trumpet Song analyzed using the Expanded and Methods Balloon Pop RT Normalized to Average MLS Receiver 1 Traditional Right R1 Traditional Left R1 Expanded Right R1 Expanded Left R1 Right R1 Left R1 WinMLS Right R WinMLS Left R1 Octave Band Center Frequency in Hertz Balloon Pop RT Normalized to Average MLS Receiver 3 Traditional Right R3 Traditional Left R3 Expanded Right R3 Expanded Left R3 Right R3 Left R3 WinMLS Right R3 WinMLS Left R Octave Band Center Frequency in Hertz Figure Balloon Pop source graphs showing similar RT values across receiver positions and methods Balloon Pop RT Normalized to Average MLS Receiver 2 Traditional Right R2 Traditional Left R2 Expanded Right R2 Expanded Left R2 Right R2 Left R2 WinMLS Right R WinMLS Left R2 Octave Band Center Frequency in Hertz 91

92 Normalized Reverberation Time in Seconds Normalized Reverberation Time in Seconds Normalized Reverberation Time in Seconds Normalized Reverberation Time in Seconds Female Talker RT Normalized to Average MLS Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R Female Talker RT Normalized to Average MLS Method Right R1 Left R1 Right R2 Left R Expanded Right R Expanded Left Octave Band Center Frequency R3 in Hertz -2-3 Right R3 Left R3 Octave Band Center Frequency in Hertz Figure Graphs of Female Talker Normalized Reverberation Times using Expanded and Methods, which result in similar values Trumpet Song RT Normalized to Average MLS Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R Expanded Left R3 Octave Band Center Frequency in Hertz Trumpet Song RT Normalized to Average MLS Method Figure Graphs of Trumpet Song Normalized Reverberation Times using Expanded and Methods, which result in similar values Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Octave Band Center Frequency in Hertz 92

93 After analyzing the Trumpet Song spectrograph using Acoustic Tools software, it was found that it contained high energy levels from 500, 630, 1,250, 1,600, 2,000, 2,500, and 3,150 Hertz. The Female Talker was analyzed in the same way, and it was determined that the signal had much less energy in all octave bands, and only had higher energy levels between 6,300 and 10,000 Hertz. This is due to the stop chord being an S sound on the end of the word letters. Figure 3-40 shows the spectrograph of the last syllable of the Female Talker audio file. Figure Spectrograph of the Female Talker last syllable with concentration of sound energy in the 6,500 to 9,200 Hertz octave bands The S sound made analysis more difficult because there were not high energy levels in the octave bands analyzed. The Trumpet Song had more energy in the octave bands that were analyzed than the Female Talker did, which resulted in higher sound 93

94 levels and higher measured Reverberation Times. The spectrograph for the Trumpet Song is shown in Figure Figure Spectrograph of the Trumpet Song end note. The Trumpet Song has more concentrated sound energy in many frequencies than the Female Talker The Balloon Pop had a more broadband signal and loud sound level, and was a more omnidirectional source which results in small differences in the Reverberation Times between the right and left receiver. Figure 3-42 shows the spectrograph of the Balloon Pop source signal. The natural acoustic sources have strong signals in limited frequencies, have weaker source levels, and have more variation in the right and left receiver positions. Overall, the Reverberation Times for the Balloon Pops show less variation between 94

95 methods, receiver positions, and left and right receivers than the natural acoustic sources due to consistency and strength of the source. Figure Spectrograph of the Balloon Pop. The solid decay of green, yellow, red and orange indicate very strong levels across all frequencies Much of the Reverberation Time data fell below the average MLS data. In fact, of the 216 receivers, 86% were below the average. Of the data that were above the average MLS, 63% were less than or equal to 0.2 seconds greater than the average. The Trumpet Song had the most data above the average MLS data, with 60% above the average. The Female Talker accounted for 27% of the data above the average MLS and the Balloon Pop accounted for just 13%. 95

96 Early Decay Time Comparison Results The Early Decay Times for all 4 methods were normalized to the average EDT values derived from acoustical measurements using the MLS signal. Early Decay Time (EDT) is typically defined as the time it takes a source level to decay by 10, 15 or 20 db. However, for this study, the EDT was calculated with the Acoustic Tools software by extrapolating from the slope between the direct sound cursor and the reflected sound cursor. Each method had different placements of these cursors, and therefore varying EDT values could have been the result of the method that was used. It was found that the Traditional, and WinMLS methods resulted in similar EDT values for the Female Talker source. Figure 3-43 shows these three methods. Receiver positions 1 and 3 had similar EDT values in the 500 and 4,000 Hertz octave bands, which are shown in Figure Receiver position 2 had slightly higher EDT values than the receiver positions 1 and 3. When using the Trumpet Song as the sound source, it was found that the Traditional and WinMLS methods had similar results. The graphs for these methods are shown in Figure Interestingly, the Expanded and methods had differences in the left and right signals at receiver positions 1-3. There are variations between left and right receiver signals with the Trumpet Song signal shown in Figure These variations are most likely due to the fact that the source signal is slightly different between right and left, and the source is listened to to determine where the direct and reflected sounds start and stop. The left and right signals are different and the method highlights these differences. The graphs for these are shown in Figure

97 Normalzed EDT in Seconds Normalzed EDT in Seconds Normalzed EDT in Seconds Female Talker EDT Normalized to Average MLS Traditional Method Traditional Right R1 Traditional Left R1 Traditional Right R2 Traditional Left R2 Traditional Right R3 Traditional Left R3 Octave Band Center Frequency in Hertz Female Talker EDT Normalized to Average MLS Method Right R1 Left R1 Right R2 Left R2 Right R3 Left R Octave Band Center Frequency in Hertz Female Talker EDT Normalized to Average MLS WinMLS Method WinMLS Right R1 WinMLS Left R1 WinMLS Right R WinMLS Left R2 WinMLS Right R3 WinMLS Left R3 Octave Band Center Frequency in Hertz Figure Female Talker source analyzed with Traditional, and WinMLS methods resulting in similar normalized EDT values 97

98 Normalzed EDT in Seconds Normalzed EDT in Seconds Normalzed EDT in Seconds Normalzed EDT in Seconds Female Talker EDT Normalized to Average MLS Receiver 1 3 Traditional Right R1 Female Talker EDT Normalized to Average MLS Receiver 3 3 Traditional Right R Traditional Left R1 Expanded Right R1 Expanded Left R Traditional Left R3 Expanded Right R3 Expanded Left R Right R Left R1 Octave Band Center Frequency in Hertz Figure Female Talker Normalized EDT values for Receiver Positions 1 and Right R Left R3 Octave Band Center Frequency in Hertz Trumpet Song EDT Normalized to Average MLS Traditional Method Octave Band Center Frequency in Hertz Traditional 3 Right R1 Traditional Left 2 R1 Traditional 1 Right R2 Traditional Left 0 R2-1 Traditional Right R3-2 Traditional Left R3-3 Trumpet Song EDT Normalized to Average MLS WinMLS Method WinMLS Right R1 WinMLS Left R1 WinMLS Right R2 WinMLS Left R2 WinMLS Right R WinMLS Left R3 Octave Band Center Frequency in Hertz Figure Trumpet Song Normalized EDT values for Traditional and WinMLS methods 98

99 Normalzed EDT in Seconds Normalzed EDT in Seconds Trumpet Song EDT Normalized to Average MLS Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Octave Band Center Frequency in Hertz Trumpet Song EDT Normalized to Average MLS Method Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Octave Band Center Frequency in Hertz Figure Trumpet Song Normalized EDT values for Expanded and methods The Trumpet Song had more variability in EDT values than the Female Talker. Figure 3-47 shows graphs of the Female Talker and Trumpet Song receiver position 3. The Trumpet Song has more variations, especially between the right and left receivers than the Female Talker. These variations between left and right receivers with the Trumpet Song source could be caused because the distance from source to receiver is increased and therefore the sound is being affected by the room surfaces more than the receivers that obtain more direct sound because they are closer to the source. The left receiver has higher EDT values than the right, perhaps due to reflections that focus more sound energy to the left ear due to its proximity to the wall surface. Interestingly, only the Balloon Pop resulted in similar normalized EDT values using the Expanded and approach (which had a tendency to be similar when analyizing the RT and C 80 ). Figure 3-48 shows graphs of these 2 methods. 99

100 Normalzed EDT in Seconds Normalzed EDT in Seconds When analyzing the Balloon Pop, the WinMLS method resulted in normalized EDT values that were similar to the average MLS EDT values. A graph of this data is shown in Figure Female Talker EDT Normalized to Average MLS Receiver Traditional Right R3 Traditional Left R3 Expanded Right R3 Expanded Left R3 Right R3 Left R3 WinMLS Right R WinMLS Left R3 Octave Band Center Frequency in Hertz Trumpet Song EDT Normalized to Average MLS Receiver Traditional Right R3 Traditional Left R3 Expanded Right R3 Expanded Left R3 Right R3 Left R3 WinMLS Right R WinMLS Left R3 Octave Band Center Frequency in Hertz Figure Graphs showing large variation in left and right receiver normalized EDT values for Trumpet Song compared to fairly consistent Female Talker values For the Balloon Pop source, receiver position 1 had the most uniform normalized EDT values across the 4 methods. As the receiver moved farther from the source, variations in EDT values were present between the left and right receivers. Figure 3-50 shows the normalized EDT values in receiver positions 1 and 3, where the differences between left and right are more prominent. 100

101 Normalzed EDT in Seconds Normalzed EDT in Seconds Normalzed EDT in Seconds Balloon Pop EDT Normalized to Average MLS Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Octave Band Center Frequency in Hertz Balloon Pop EDT Normalized to Average MLS Method Right R1 Left R1 Right R2 Left R2 Right R3 Left R Octave Band Center Frequency in Hertz Figure Graphs of Balloon Pop analyzed with Expanded and methods resulting in similar normalized EDT values Balloon Pop EDT Normalized to Average MLS WinMLS Method WinMLS Right R1 WinMLS Left R1 WinMLS Right R2 WinMLS Left R2 WinMLS Right R3 WinMLS Left R3 Octave Band Center Frequency in Hertz Figure Balloon Pop analyzed using the WinMLS method resulting in EDT values that are centered around average MLS data 101

102 Much of the EDT data fell below the average MLS data. Interestingly, as with the Reverberation Time, of the 216 receivers, 86% were below the average. However, of the 31 receivers that were above the average MLS data, only 32% were less than or equal to 0.2 seconds above the average. The Trumpet Song had the most data points above the average, with 58% of the data above the average MLS data. The Balloon Pop had 32% and the Female Talker had only 10% of the data with values above the average MLS data. C 80 Comparison Results The C 80 values for methods 1-4 were normalized to the average C 80 values derived from acoustical measurements using an MLS signal. These normalized values were compared by method and receiver position, similar to the RT comparison described above. It was found that the Trumpet Song resulted in lower C 80 values than the Female Talker in general. The Balloon Pop had similar left and right receiver values in all receiver positions and across all methods; however the Trumpet Song and the Female Talker did not. The balloon is a more broadband and louder source that excites the room more evenly than the speech or music. Like the Reverberation Time analysis, the Balloon Pop shows little variation in C 80 values between method, receiver position, and left and right receiver. Again, for both the Female Talker and Trumpet Song, C 80 values were similar when analyzed using the Expanded and methods. Figures 3-51 through 3-52 show graphs of the two natural acoustic sources analyzed using the Expanded and methods. Similar to the normalized Reverberation Time results, the Ballon Pop again had similar C 80 values across all 4 methods and receiver positions, which is shown in Figure

103 Normalized C 80 in decibels Normalized C 80 in decibels Normalzed EDT in Seconds Normalzed EDT in Seconds Balloon Pop EDT Normalized to Average MLS Receiver Traditional Right R1 Traditional Left R1 Expanded Right R1 Expanded Left R1 Right R1 Left R1 Octave Band Center Frequency in Hertz -2 WinMLS Right R1 WinMLS Left R1-3 Balloon Pop EDT Normalized to Average MLS Receiver 3 Figure Balloon Pop source at receiver positions 1 and 3, showing differences in left and right receivers as the receiver position moves farther from the source Traditional Right R3 Traditional Left R3 Expanded Right R3 Expanded Left R3 Right R3 Left R3 WinMLS Right R3 WinMLS Left R3 Octave Band Center Frequency in Hertz Female Talker C 80 Normalized to Average MLS Expanded Method Female Talker C 80 Normalized to Average MLS Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R Octave Band Center Frequency in Hertz Figure Graphs of Female Talker Normalized C 80 values using Expanded and Methods, which result in similar values Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Octave Band Center Frequency in Hertz 103

104 Normalized C 80 in decibels Normalized C 80 in decibels Trumpet Song C 80 Normalized to Average MLS Expanded Method 12 Trumpet Song C 80 Normalized to Average MLS Method 12 8 Expanded Right R1 8 Right R1 4 Expanded Left R1 Expanded Right R2 4 Left R1 Right R2 0 Expanded Left R2 0 Left R2-4 Expanded Right R3 Expanded Left R3-4 Right R3 Left R Octave Band Center Frequency in Hertz Figure Graphs of Trumpet Song Normalized C 80 values using Expanded and Methods, which result in similar values The natural acoustic sources do not excite the room uniformly across all octave bands, and as such some frequency bands have lower levels than others. The series of reflections in the first 80 ms varies greatly depending upon the source type and receiver position as well. David Prince had similar findings in his Master s thesis study of MLS Octave Band Center Frequency in Hertz and musical sources played through a violin structure. He found large variations in C 80 values (from -4.6 to 5.9 at the 1000 Hertz octave bands) when source positions varied. 6 6 D. Prince, Variation of Room Acoustic Parameters as a Function of Source Location and Directivity (Master s Thesis, University of Florida, 1994),

105 Normalized C80 in decibels Normalized C80 in decibels Normalized C80 in decibels Balloon Pop C80 Normalized to Average MLS Receiver 1 Traditional Right R1 Traditional Left R1 Expanded Right R1 Expanded Left R1 Right R1 Left R WinMLS Right R1 Octave Band Center Frequency in Hertz Balloon Pop C80 Normalized to Average MLS Receiver Traditional Right R2 Traditional Left R2 Expanded Right R2 Expanded Left R2 Right R2 Left R2 WinMLS Right R2 WinMLS Left R2 Octave Band Center Frequency in Hertz 12 Balloon Pop C80 Normalized to Average MLS Receiver 3 Traditional Right R Traditional Left R3 Expanded Right R3 Expanded Left R3 Right R Left R3 WinMLS Right R3 WinMLS Left R3 Octave Band Center Frequency in Hertz Figure Graphs of normalized C80 values for Balloon Pop source for the 3 receiver positions 105

106 The Trumpet Song source has longer Reverberation Times than the Female Talker and lower C 80 values. These findings are in accord with Beranek s findings in his concert hall research. He states, obviously, if there is no reverberation that is to say, the room is very dead the music will be very clear and C 80 will have a large positive value (in decibels). If the reverberation is large such as exists in a huge cathedral the music will be unclear and C 80 will take on a large negative value. 7 The C 80 graphs also indicate a large range in C 80 values, especially among the natural acoustic source signals. Youngmin Kwon s study on church acoustics agreed with these findings, indicating that there was much variation in the C 80 values when considering different receiver positions 8. As with the RT and EDT, much of the data was below the average MLS data. Of the 216 receiver, 73% were below the average. Of the 58 data points that were above the average, 41% were less than or equal to 1.38dB above the average, which is the Just Noticeable Difference value used in the study. The Female Talker had the most data points above the average, accounting for 43% of the all data above the average. The Trumpet Song accounted for 36% of the data above the average. The Balloon Pop had the least, with only 21% of data above the average. Just Noticeable Differences The Just Noticeable Differences for the RT and C 80 derived from acoustical measurements made using the alternative source signals were compared. It was 7 L. L. Beranek, Concert and Opera Halls: How They Sound (Woodbury, NY: Published for the Acoustical Society of America through the American Institute of Physics, 1996), Kwon, Under Balcony Acoustics,

107 determined that the JND s for Early Decay Time and Reverberation Time had greater variation than C 80. Reverberation time just noticeable differences Table 3-1. Just noticeable difference in reverberation time ranges grouped by method Just Noticeable Difference Ranges for Reverberation Time by Method Omnidirectional Female Trumpet Balloon Directional Talker Song Pop Speaker Speaker Traditional Expanded WinMLS The JND ranges for the Reverberation Times derived from the scientifically calibrated sources were between 0 and 1 for the directional loudspeaker and 0 and 2 for the omnidirectional loudspeaker. The lower JND values and smaller range suggest more consistency in the data derived from the scientifically calibrated sources. Compared to the scientifically calibrated sources, the Female Talker and Trumpet Song had JND values that were up to 11 times the JND values derived from the scientifically calibrated sources. The Female Talker had a higher Reverberation Time JND range than the Trumpet Song, and the Balloon Pop having the lowest JND range (from 0-8 JND s). The traditional method resulted in the largest range of JND values (1-8) for the Balloon Pop source. For the Female Talker source, the Traditional and 107

108 Just Noticeable Difference in Reverberation Time Just Noticeable Difference in Reverberation Time Just Noticeable Difference in Reverberation Time Just Noticeable Difference in Reverberation Time WinMLS methods resulted in the highest JND values. For both the Female Talker and the Trumpet Song, the Expanded and methods had similar JND values in the 4,000 Hertz octave band, shown in Figures 3-54 through Female Talker JND for RT Expanded Method Octave Band Center Frequency in Hertz Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Female Talker JND for RT Method Figure Graphs showing Female Talker having similar JND values in the 4,000 Hertz octave band for Expanded and methods Trumpet Song JND for RT Expanded Method Octave Band Center Frequency in Hertz Figure Graphs showing Trumpet Song having similar JND values in the 4,000 Hertz octave band for Expanded and methods Expanded Right R1 20 Expanded Left R1 Expanded 15 Right R2 Expanded 10 Left R2 Expanded Right R3 5 Expanded Left R Octave Band Center Frequency in Hertz Trumpet Song JND for RT Method Octave Band Center Frequency in Hertz Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 Right R1 Left R1 Right R2 Left R2 Right R3 Left R3 108

109 The Balloon Pop resulted in JND values that were similar using the Expanded and methods, and the WinMLS method resulted in the lowest JND values. Table 3-2. Just noticeable difference in reverberation time ranges grouped by receiver position Just Noticeable Difference Ranges for Reverberation Time by Receiver Omnidirectional Female Trumpet Balloon Directional Talker Song Pop Speaker Speaker Receiver Receiver Receiver The natural acoustic sources of the Female Talker and Trumpet Song had similar JND for RT values at each receiver position, while the Balloon Pop had significantly lower JND values. However, when compared to the scientifically calibrated sources using the directional and omnidirectional (of 0-2 JND), the three additional sources have much larger RT JND s. Early decay time just noticeable differences It was found that with both the Female Talker and the Trumpet Song that the JND s in EDT increased as the receiver position was farther away from the source. However, the converse was true with the Balloon Pop source: the JND s increased as the receiver position was closer to the source. 109

110 When using the Female Talker source, the Traditional and methods had similar results. Table 3-3. Just noticeable difference ranges for early decay time grouped by method Just Noticeable Difference Ranges for Early Decay Time by Method Omnidirectional Female Trumpet Balloon Directional Talker Song Pop Speaker Speaker Traditional Expanded WinMLS For all 4 methods, the JND values were higher in the 500 and 1,000 Hertz octave bands than the 4,000 Hertz octave band. It is assumed this is due to the fact that the Female Talker has more energy in the 4,000 Hertz octave band and therefore produces a more measureable signal. For the Trumpet Song source signal, the Expanded and methods resulted in similar values, with the exception of the Right Receiver position 3, which resulted in a JND of 27, which is shown in Table 3-4. The WinMLS method resulted in the lowest EDT JND values for the Trumpet Song source. Similar to the Trumpet Song source, the Balloon Pop source processed with the Expanded and methods resulted in similar values, shown in Figure

111 Just Noticeable Differences for EDT Just Noticeable Differences for EDT Table 3-4. Just noticeable difference ranges for early decay time grouped by receiver position Just Noticeable Difference Ranges for Early Decay Times by Receiver Omnidirectional Female Trumpet Balloon Directional Talker Song Pop Speaker Speaker Receiver Receiver Receiver Balloon Pop JND for EDT Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R Expanded Left R3 Octave Band Center Frequency in Hertz Figure Balloon Pop Expanded and Methods showing similar Just Noticeable Differences in Early Decay Time The WinMLS resulted in comparatively low JND values for the Balloon Pop, shown in Figure This is most likely due to the high sound energy across the octave bands. As the frequency increases, the JND values increased as well. JND Values in the 500 Hertz octave band ranged from 0-2, 1-2 in the 1,000 Hertz octave band, and 2-3 in the 4,000 Hertz octave bands Balloon Pop JND for EDT Method Right R1 Left R1 Right R2 Left R2 Right R Left R3 Octave Band Center Frequency in Hertz 111

112 Just Noticeable Differences for EDT Balloon Pop JND for EDT WinMLS Method WinMLS Right R1 WinMLS Left R WinMLS Right R2 WinMLS Left R2 Octave Band Center Frequency in Hertz Figure Balloon Pop WinMLS method showing low JND values for Early Decay Time The EDT metric resulted in the highest JND values of all the metrics analyzed in this study. The Female Talker resulted in JND s ranging from The Trumpet Song resulted in JND s ranging from The JND values ranged from 0-21 JND with the Balloon Pop source, which typically resulted in the lowest JND s for the RT and C 80. For the Trumpet Song and the Balloon Pop, the Expanded and methods resulted in similar values and the WinMLS method resulted in lower JND s than the other methods. For the two natural acoustic sources, the JND values increased as the receiver positions were farther from the source. However, with the Balloon Pop source, the opposite occurred, and the JND values increased as the receiver positions were closer to the source. Clarity (C 80 ) just noticeable differences The Trumpet Song has a slightly higher C 80 JND range than the Female Talker (0-8 as opposed to 0-6 JND s), however the Balloon Pop has the lowest JND range (from 0 5 JND s in all 4 methods). The Female Talker and Trumpet Song had the highest JND ranges in C 80 when analyzed using the WinMLS method. 112

113 Table 3-5. C 80 just noticeable difference ranges grouped by method Just Noticeable Difference Ranges for C 80 by Method Female Talker Trumpet Song Balloon Pop Directional Speaker Traditional Expanded Omnidirectional Speaker WinMLS The Female Talker had similar JND values when analyzed using the Expanded and methods, as did the Trumpet Song, which are shown in Figures 3-58 through All 4 methods resulted in similar JND values when the Balloon Pop source was used. Just Noticeable Difference in C Female Talker JND for C 80 Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Octave Band Center Frequency in Hertz Just Noticeable Difference in C Female Talker JND for C 80 Method Right R1 Left R1 Right R2 Left R2 Right R Left R3 Octave Band Center Frequency in Hertz Figure Graphs showing Female Talker having similar JND values for C 80 across the octave bands for Expanded and methods 113

114 Just Noticeable Difference in C80 Just Noticeable Difference in C Trumpet Song JND for C80 Expanded Method Expanded Right R1 Expanded Left R1 Expanded Right R2 Expanded Left R2 Expanded Right R3 Expanded Left R3 Octave Band Center Frequency in Hertz Trumpet Song JND for C80 Method Right R1 Left R1 Right R2 Left R2 Right R Left R3 Octave Band Center Frequency in Hertz Figure Graphs showing Trumpet Song having similar JND values for C 80 across the octave bands for Expanded and methods Table 3-6. C 80 just noticeable difference ranges grouped by receiver position Just Noticeable Difference Ranges for C 80 by Receiver Omnidirectional Female Trumpet Balloon Directional Talker Song Pop Speaker Speaker Receiver Receiver Receiver Both the Female Talker and Trumpet Song JND values increased as the receiver position moved closer to the source. The Balloon Pop had similar JND values across the frequencies in all 3 receiver positions. The C 80 JND values for the Female Talker tended to decrease as the receiver position was farther away from the source. The converse was true for the Trumpet Song, as the C 80 JND tended to increase as the receiver position was farther from the source. Interestingly, the Female Talker and 114

115 Trumpet Song had the same range in C 80 JND s in receiver positions 1 and 2, as shown in Table 6. Discussion People gather in spaces for different purposes, whether it be to worship, learn, live, listen, or to be entertained. The transfer of information takes places via communication, be it verbal, visual, tactile, or other sensory perceptions. Aural communication is the realm that acoustics deals with. When people communicate aurally with each other, the sounds they make are carried by direct and reflected paths throughout the room to reach the listener. The way in which these sounds interact with the room and the paths they take to reach the listener are the primary concern of architectural acoustics. When a room creates desirable reflections that increase the direct and reflected sounds, has uniform decay and is free from acoustic defects such as echoes, the sounds heard in the room will be judged by most to enhance the acoustics of the space. When the sounds from the source to the receiver are judged to not have the correct sonic characteristics, they are said to have acoustical defects. Acoustical measurements are typically made in rooms as a tool to aid in designing the rooms to have preferred acoustical qualities for the specific sound sources and to avoid the acoustical defects. There are typically specific issues that are associated with the room that should be studied, i.e.: one cannot clearly hear speech in the room, or there are echoes present that muddy the music. Acousticians are brought in to determine how to remedy these issues. Kirkegaard and Gulsrud suggest that before any acoustical measurements are made, the room be listened to, ideally using the sources that are known to produce the effects, so that the room attributes can be noted. They also suggest several other sources, including steady state noise and a metronome be 115

116 used to reveal time domain issues. Only after the room has been listened to carefully and problematic receiver positions are identified are acoustical measurements taken. 9 This is not necessarily the typical method of diagnosing a room, especially in modern research. Acousticians take acoustical measurements in accordance with the applicable standards, such as ISO 3382, to perform diagnostics of the room. The measurements consist of omnidirectional sound sources that excite the room with a broadband signal at relatively high levels. Measurements are derived from the backwards integration of the impulse response associated with each source. The problem therein lies in the fact that most sound sources found in a room are not omnidirectional, only excite a limited range of frequencies and are of varying sound levels. Conclusions There is the tendency for acoustical metrics to be given in the literature as whole room averages. There is typically one value for each acoustic metric for a given room. This has been shown to be quite limited and that in reality, there are actually a range of values for each metric that vary with source and receiver type and position. Kirkegaard and Gulsrud argue that parameters that depend strongly on direct sound and early reflection structure can be especially misleading if averaging over seating locations, since the most interesting and significant qualities of these parameters are often how they vary within the room. 10 Recent research that has involved matrices of multiple receiver locations has shown that the acoustic metrics vary across receiver positions. 9 L. Kirkegaard and T. Gulsrud, In Search of a New Paradigm: How Do Our Parameters and Measurement Techniques Constrain Approaches to Concert Hall Design? Acoustics Today 7 (2011): Kirkegaard and Gulsrud, In Search,

117 This is in accordance with the idea that different seat positions will have different paths from source to receiver, with different reflection patterns at each seat, and therefore sounds will be heard differently in different seating positions. Researchers have found that data vary across the room and acoustic parameters change based on the location of the source and receiver. It can be argued that this variance is not due to error in the measurements but rather it is a function of the different paths from source to receiver, varying frequency content and levels, and directionality that causes these differences. While scientifically calibrated acoustical metrics provide reproducible results, the ways in which the scientifically calibrated sound sources excite a room are significantly different than the sounds that are typically heard in the room. The level, frequency content and directionality of the sound source affect the acoustic metrics derived from the measurements. This relates to the idea that people sitting in different seats in the room hear sounds differently: it may not be a function of the individual s perception, but rather due to the way that sounds with limited bandwidth, level and directionality interact with the room surfaces. This study suggests that because there is such large variation in acoustical metrics when using sound sources that are closer to those that would typically take place in a room, other methods for taking acoustical measurements should be developed that more closely relate the metrics to the sounds that are normally heard in the room. The Female Talker resulted in the highest JND of 0 to 22 in Reverberation Time, while the Trumpet Song had the highest JND s of 0 to 8 in C 80. The Trumpet Song also had the highest JND values for EDT, ranging from The acoustic metrics derived from the scientifically calibrated methods had JND less than 1 for the C 80 metric, 117

118 between 0 and 4 JND in EDT and between 0 and 2 JND for the RT. The directional speaker had lower JND s than all other sound sources, with most of the data having a JND of less than 1. It should be noted that for this study the 2,000 Hertz octave band was not analyzed due to there not being enough source signal from the Female Talker. An important example of how real sounds in rooms do not have the adequate level across the octave bands to derive acoustical metrics is the s syllable at the end of letters which was analyzed had limited signal in the octave bands. When using a broadband source, there is adequate sound level, especially in the middle frequencies to derive acoustical metrics. However, the absence of signal with this source should not be considered a defect in the source but rather a prime example of how real sounds have limited bandwidth and level, and yet these are the sounds that are actually heard in the rooms. A comparison of the broadband impulse response from each of the source signals is shown in Figure Based on the impulse response graphs, one can see that the directional and omnidirectional loudspeakers have similar (but not identical) impulse responses. Each of the other sources has vastly different impulse responses, with different sound reflection patterns. It is the physical reality that these different sound sources, with different bandwidth, level, transient response and directionality will sound different at each receiver location and have different acoustical metrics associated with each condition. 118

119 Trumpet Song Female Talker Balloon Pop Directional MLS Omnidirectional MLS Figure Impulse response graphs of various sources used in study The data varies with source type and receiver position throughout the room. This variance is indicative of the different sources, the ways in which the data are processed and the receiver position in the room. The variances are not a result of error; the Talker and Trumpet Song have certain attributes that create these differences in acoustical metrics. The four methods of analyzing the recorded speech and music signals had similar results in many cases. The Traditional, Expanded and methods produced similar results, while the WinMLS method had more varied results. Interestingly, the Expanded and methods had extremely similar results, as the cursors for the reflected sound and noise floor were ultimately placed in almost the same position. This may indicate that while it is generally accepted that the reflected sound should be calculated from -5dB from the direct sound and 10 db above the ambient, our ears are able to hear more of the signal than what is included in the traditional slope. The WinMLS method, while obtaining similar results to the other methods with the Balloon Pop as the sound source, produced different results from the natural acoustic sources. This is most likely due to how the WinMLS software processes the data. It produces similar data with the sound source that is relatively 119

120 Difference in RT - EDT in Seconds broadband, has high levels and a straight, steady decay. The other sources have limited bandwidth and level and uneven decay patterns, which may not be analyzed properly by the analyzation software. To draw conclusions from the large amount of data that was collected, the left and right receivers in all 3 receiver positions were averaged to obtain single numbers per octave band for each source and method. From the RT EDT comparison, the WinMLS data not only had data that ranged from 0 to 0.2 seconds, which is comparable to previous studies by Bradley, but also had values that were within 0.2 seconds from the 0 value, which means they had almost the same values as the average RT EDT values derived using the MLS source. 2.5 Average RT - EDT: Female Talker Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Octave Band Center Frequency in Hertz Figure Average RT EDT values for Female Talker source 120

121 Difference in RT - EDT in Seconds Difference in RT - EDT in Seconds Average RT - EDT: Trumpet Song Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Octave Band Center Frequency in Hertz Figure Average RT EDT values for Trumpet Song source 2.5 Average RT - EDT: Balloon Pop Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Octave Band Center Frequency in Hertz Figure Average RT EDT values for Trumpet Song source 121

122 Normalized Reverberation Time in Seconds Reverberation time is typically considered a consistent metric when measured throughout a room. Wallace Sabine performed experiments in which he varied the source and receiver locations while measuring RT. He found little variance between the positions. However, performing measurements using the natural acoustic sources to derive RT s was found to have extremely high Just Noticeable Differences (up to 22 JND s) from the averaged RT data taken in general accordance with ISO The Expanded and Methods were within 0.01 and 0.06 seconds in the 1,000 and 4,000 Hertz octave bands for all 3 sources. The Traditional, Expanded and Methods were within.02 to 0.12 seconds of each other in all octave bands for Balloon Pop source, shown in Figure The WinMLS Method resulted in RT s that were very close to the average MLS data, ranging from just 0.03 to 0.18 seconds from the average. Average Normalized Reverberation Time: Female Talker Average Traditional Method Average Expanded Method Average Method Average WinMLS Octave Band Center Frequency in Hertz Figure Average Normalized Reverberation Time for Female Talker source 122

123 Normalized Reverberation Time in Seconds Normalized Reverberation Time in Seconds Average Normalized Reverberation Time: Trumpet Song Average Traditional Method Average Expanded Method Average Method Average Method Octave Band Center Frequency in Hertz Figure Average Normalized Reverberation Time for Trumpet Song source 0.5 Average Normalized Reverberation Time: Balloon Pop Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Reverberation Time for Balloon Pop source The EDT metrics produced the most variation between methods with up to 27 JND s Measurements made with the female talker produced EDT values that were 0.03 to 0.38 seconds from each other using the Traditional, and 123

124 Normalized EDT in Seconds Normalized EDT in Seconds WinMLS Methods, shown in Figure The Trumpet Song resulted in similar Expanded and methods, varying from 0.08 to.16 seconds from each other, shown in Figure The WinMLS EDT values were within.03 to 0.1 seconds of the average MLS data for the Balloon Pop source, shown in Figure Average Normalized Early Decay Time: Female Talker Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Octave Band Center Frequency in Hertz Figure Average Normalized Early Decay Time for Female Talker source 0.5 Average Normalized Early Decay Time: Trumpet Song Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Early Decay Time for Trumpet Song source 124

125 Normalized C 80 in decibels Normalized EDT in Seconds Average Normalized Early Decay Time: Balloon Pop Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Early Decay Time for Balloon Pop source The C 80 values had more clearly defined similarities between methods than the EDT. The C 80 values also resulted in the lowest JND s, ranging from 0-8. The Expanded and Methods resulted in similar C 80 values, differing between 0 and 0.6 db for the Female Talker source, shown in Figure All 4 methods had similar values for the Trumpet Song, differing by 0.17 to 1.22 db. The Traditional, Expanded and methods when analyzing the Balloon Pop shown in Figure 3-72 resulted in almost exactly the same values, differing by just 0 to 0.06 db Average Normalized C 80 : Female Talker Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Clarity (C 80 ) for Female Talker source 125

126 Normalized C 80 in decibels Normalized C 80 in decibels Average Normalized C 80 : Trumpet Song Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Clarity (C 80 ) for Trumpet Song source 3 Average Normalized C 80 : Balloon Pop Octave Band Center Frequency in Hertz Average Traditional Method Average Expanded Method Average Method Average WinMLS Method Figure Average Normalized Clarity (C 80 ) for Balloon Pop source As Figure 3-73 shows, much of the data derived from the natural acoustic sources was below that of the average MLS data. There are only 4 data points that are above the scientifically calibrated sources. These data are from natural acoustic sources analyzed during the Pilot Study, before consistent methods for analyzing natural acoustic sources were adopted. 126

127 Reverberation Time in Seconds Scientifically Calibrated Sources Natural Acoustic Sources Data from natural acoustic sources from Pilot Study before consistent methods for analyzing natural acoustic sources were developed Octave Band Center Frequency in Hertz Pilot Study: Dir corner sine sweep Pilot Study: Dir Front Sine Sweep Pilot Study: Mult Sine Sweep Dir Corner Pilot Study: Mult Sine Sweep Dir Front Pilot Study: Dir Front MLS Pilot Study: Dir Corner MLS Pilot Study: Balloon Pop MLS Pilot Study: Balloon POP AT Pilot Study: Female Talker Pilot Study: Piano Middle C Note Pilot Study: Piano Middle C Chord Pilot Study: Piano Song Pilot Study: Trumpet Song Study 1: Dir Front MLS Study 1: Dodec Front MLS Study 2: Traditional Method: Balloon Pop Study 2: Traditional Method: Female Talker Study 2: Traditional Method: Trumpet Song Study 2: Expanded Method: Balloon Pop Study 2: Expanded Method: Female Talker Study 2: Expanded Method: Trumpet Song Study 2: Method: Balloon Pop Study 2: Method: Female Talker Study 2: Method: Trumpet Song Study 2: WinMLS Method: Balloon Pop Study 2: WinMLS Method: Female Talker Figure Reverberation Time comparison of 3 studies showing higher RT s for scientifically calibrated sources and lower RT s for natural acoustic sources 127

128 Data from the same sources were reanalyzed using methods 1-4 and yielded results that were lower than data derived from the scientifically calibrated sources. On average, the natural acoustic sources had Reverberation Times that were between 0.28 and 0.44 seconds shorter than the scientifically calibrated sources because the natural acoustic sources had lower sound levels, limited bandwidth and did not fully excite the reverberant field of the room. For the Reverberation Time metric, 86% of the natural acoustic data was below the average MLS data. Of the 30 data points that were above the average MLS, 60% were from the Trumpet Song. The EDT metric also resulted in 86% of the data derived from the natural acoustic sources falling below the average MLS data. This is most likely due to the Trumpet Song having high sound levels in the 500, 1,000 and 4,000 Hertz octave bands that were analyzed. Again, the Trumpet Song accounted for most of the data points above the average MLS, with 58% above the average. The higher than average MLS values for RT and EDT are most likely due to the Trumpet Song having high sound levels in the 500, 1,000 and 4,000 Hertz octave bands that were analyzed. The values for the Clarity Index derived from the natural acoustic sources were generally below the average MLS values, with 73% of the natural acoustic data falling below the average MLS data. The Female Talker had the most data above the average MLS values, with 43% of the data above the average. The Female Talker had lower RT values, and as such, makes sense that it has higher Clarity values. The Clarity Index is dependent on the early and late reverberation. If there is less late reverberation, the C 80 values will be higher than if there were large amounts of late reverberation. 128

129 Future Studies While this study does not intend to prove a new methodology for taking acoustic measurements, it intends to highlight the degree of differentiation between certain acoustic metrics derived from more realistic or natural sound sources to provide an impetus for further study. For all 4 methods of analysis for alternative sound sources, the C80 values were between 0 and 8 JND s from the average obtained using the scientifically calibrated method. The Reverberation Time values for all 4 methods were from 0 to 22 JND s from the average Reverberation Time obtained in general accordance with ISO The large differences suggest that acoustical conditions vary in the room based on source type and listener location. It is suggested that similar studies be conducted in which more sources are used; including different words and sentences so that metrics may be derived from different words and syllables. Other instruments should also be used, which have different directivity patterns, bandwidth, content and level. Other larger rooms should be used, that would possibly have larger differences in the acoustical metrics derived using ISO It would be beneficial to compare a room whose metrics are more similar to a larger room that has larger variances from testing in accordance with ISO Using more receiver positions would also be helpful, so a grid could be made of the acoustic metrics. Testing of more than one room would also be beneficial, so as to compare the results across multiple room conditions. Subjective listening tests may also be conducted in the receiver positions to determine if the metrics derived from the natural acoustic source measurement methods relate to individual s perception of the natural acoustic source at that position. 129

130 APPENDIX A IMPULSE RESPONSE PRINTOUTS FOR WIN MLS SOURCES Figure A-1. MLS Directional Front: Receiver 3_1 Figure A-2. MLS Directional Front: Receiver 3_2 Figure A-3. MLS Directional Front: Receiver 3_3 Figure A-4. MLS Directional Corner: Receiver 3_1 130

131 Figure A-5. MLS Directional Corner: Receiver 3_2 Figure A-6. MLS Directional Corner: Receiver 3_3 Figure A-7. Sine Sweep Directional Front: Receiver 3_1 Figure A-8. Sine Sweep Directional Front: Receiver 3_2 131

132 Figure A-9. Sine Sweep Directional Front: Receiver 3_3 Figure A-10. Sine Sweep Directional Corner: Receiver 3_1 Figure A-11. Sine Sweep Directional Corner: Receiver 3_2 Figure A-12. Sine Sweep Directional Corner: Receiver 3_3 132

133 Figure A-13. Multiple Sine Sweep Directional Front: Receiver 3_1 Figure A-14. Multiple Sine Sweep Directional Front: Receiver 3_2 Figure A-15. Multiple Sine Sweep Directional Front: Receiver 3_3 Figure A-16. Multiple Sine Sweep Directional Corner: Receiver 3_1 133

134 Figure A-17. Multiple Sine Sweep Directional Corner: Receiver 3_2 Figure A-18. Multiple Sine Sweep Directional Corner: Receiver 3_3 DECAY CURVES OF ANECHOIC, PIANO AND BALLOON SIGNALS IN THE 500 and 1,000 HERTZ OCTAVE BANDS ANALYZED USING ACOUSTIC TOOLS Figure A-19. Piano Middle C Note: 500 Hertz 134

135 Figure A-20. Piano Middle C Note: 1,000 Hertz Figure A-21. Piano Middle C Chord: 500 Hertz 135

136 Figure A-22. Piano Middle C Chord: 1,000 Hertz Figure A-23. Piano Song Stop Chord: 500 Hertz 136

137 Figure A-24. Piano Song Stop Chord: 1,000 Hertz Figure A-25. Anechoic Trumpet Music: 500 Hertz 137

138 Figure A-26. Anechoic Trumpet Music: 1,000 Hertz Figure A-27. Anechoic Female Talker: 500 Hertz 138

139 Figure A-28. Anechoic Female Talker: 1,000 Hertz 139

140 APPENDIX B OMNIDIRECTIONAL AND DIRECTIONAL LOUDSPEAKER COMPARISON Figure B-1. MLS Directional Front: Receiver 1_1 Figure B-2. MLS Directional Front: Receiver 1_2 Figure B-3. MLS Directional Front: Receiver 1_3 Figure B-4. MLS Directional Front: Receiver 2_1 140

141 Figure B-5. MLS Directional Front: Receiver 2_2 Figure B-6. MLS Directional Front: Receiver 2_3 Figure B-7. MLS Directional Front: Receiver 3_1 Figure B-8. MLS Directional Front: Receiver 3_2 141

142 Figure B-9. MLS Directional Front: Receiver 3_3 Figure B-10. MLS Omnidirectional Front: Receiver 1_1 Figure B-11. MLS Omnidirectional Front: Receiver 1_2 Figure B-12. MLS Omnidirectional Front: Receiver 1_3 142

143 Figure B-13. MLS Omnidirectional Front: Receiver 1_4 Figure B-14. MLS Omnidirectional Front: Receiver 1_5 Figure B-15. MLS Omnidirectional Front: Receiver 2_1 Figure B-16. MLS Omnidirectional Front: Receiver 2_2. MLS 143