Sie sind auf Seite 1von 8

PAPERS

Room Acoustical Parameters in Small Rooms1


HASSAN EL-BANNA ZIDAN AND U. PETER SVENSSON, AES Member
(Zidan@iet.ntnu.no) (Svensson@iet.ntnu.no)

Acoustics Group, Department of Electronics and Telecommunications, Norwegian University of Science and Technology, NO 7491, Trondheim, Norway.

This paper presents a study of the room acoustics conditions in small shoebox shaped rooms. Barrons model is used to predict room acoustical parameters such as reflected energy, Grefl , and early reflected energy, G50refl . Measurements were carried out in three small rooms and detailed computer simulations using CATT Acoustic were done for different rooms. Comparisons between these measurements and detailed simulations, and the predictions based on Barrons model, are typically within 1 dB and +1 dB when averaged across a room, from 250 Hz to 2 kHz.

0 INTRODUCTION There are many different approaches for predicting the properties of the sound field in a room. Three classes of prediction methods are wave-theoretical methods, geometrical acoustics based methods, and statistical methods. The two former can give highly detailed information about a sound field: the sound pressure at any point in space inside a room. At the same time, these methods require, or can handle, geometry, boundary conditions, and source properties to the smallest detail. At the opposite end of the scale is the third approachstatistical methods. These use very few input parameters about a room and ignore most details, so the results that are achieved are not as accurate and detailed as the first two methods. Even with the tremendous progression of computation power and complex algorithm sophistication, the simple statistical methods are still very prominent. This is probably because of the usefulness of the models with few parameters, such that the influence of those basic parameters can be studied readily. One example is presented in [1] where the authors proposed an objective scheme for ranking halls based on a small set of room acoustical parameters. A common statistical model used in room acoustics is Barrons revised theory that views the sound field as consisting of direct sound and diffuse exponential reverberation only. It ignores the early reflections (i.e., the early part of the impulse response is viewed as part of the exponential diffuse field), it ignores possible non-diffuseness of the reverberation, non-exponential decays, and the room modes.
1 Some parts of this paper were presented (convention paper 8223) at the 129th Convention of the Audio Engineering Society, San Francisco, CA, 2010 November 47.

Some of these aspects could possibly be included in extensions to the model. Barrons model has been shown to predict the sound field in large auditoria quite well [2,3] and the largest inaccuracy seems to appear when space is subdivided into less diffuse sub-volumes [4]. Furthermore, better agreement is often found for the late part of the impulse response than for its early part. The same model has been evaluated in smaller auditoria [5]. The results verified the model with slight but consistent differences between the measured and the predicted strength levels, and a correction was proposed for possibly improved prediction accuracy. For classrooms, the relationship between early energy, in the form of the C50 -parameter, and the reverberation time has been analyzed in [6]. Measured values of C50 tended to be 13 dB higher, on average, than the diffuse-field relationship. The direct sound was ignored in the diffuse-field relationship, which to some degree explains why the measured values were higher. The purpose of this paper is to study how well Barrons model does in smaller rooms, similar to the study in [6]. Smaller rooms might be more demanding for such simplified methods but we aim at quantifying the deviations. As a part of this study, an impulse response model will be introduced that facilitates extensions to Barrons model and that can be used in further work where electroacoustically coupled rooms are studied. The paper is organized as follows. Section 1 briefly reviews Barrons model and its relationship to a simplified impulse response model. Section 2 presents the measurements and simulations done in different rooms to verify the model used. The achieved results are discussed in Section 3. Finally, the fourth section concludes this study.
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

62

PAPERS

ROOM ACOUSTICAL PARAMETERS IN SMALL ROOMS

1 THEORY 1.1 Impulse Response Models In any closed space or room, the acoustical properties of that room can be characterized and assessed by a room impulse response. Typically, impulse responses are viewed as being composed of three main components: direct sound; early reflections, which are a set of discrete reflections and can be chosen to be those that arrive within the first 50 80 ms of the first wave front; and late reflections that arrive after 5080 ms. While the direct sound is a well-defined component that is usually standing out in an impulse response, and which is typically analyzed for, e.g., detecting the location of a source [7], the transition from early reflections to late reflections is not possible to distinguish. Recent research has been done to determine the mixing time for this transition in room impulse responses [8]. A simple discrete-time room impulse response model that was introduced in [9], and expanded in [10], is used in the following way. It is composed of a direct sound unit pulse followed by an exponential-decay weighted noise sequence. Such a model can be used to find estimates of the direct sound energy, early energy, and late energy. These energy estimates are subsequently used in various room acoustical parameters such as Strength, Clarity, etc. Energy estimates can also be derived from Barrons model without introducing an equivalent impulse response model per se, but the impulse response model gives further flexibility, e.g., for the study of properties of convolved impulse responses [11]. Here, these three energy components will be denoted d, e, and l, respectively. In addition, an additional quantity for the reflected sound energy will be used: rev = e + l. They will be normalized relative to the direct sound energy at 10 m distance for an omnidirectional sound source with the same output power as the sound source used in a measurement. This normalization leads to the various Strength parameters, [12] and can be calculated directly in terms of the d, e, lenergies as follows G = 10 log10 (d + e + l ) dB (1)

via the reverberation time T60 , l50 = r ev exp 6 ln(10)0.050 T60 0.69 r ev exp T60 6 ln(10)0.050 T60 0.69 r ev 1 exp T60

(5)

e50 = r ev 1 exp

(6)

Furthermore, the direct sound energy is readily available as d= 100 r2 (7)

where r is the source-receiver distance in meters and is the directivity factor. For the reverberation energy, or reflected sound energy, several values have been proposed in the literature. A first is the classical expression for a constant reverberation energy in a room, r evclassical = 1600 A (8)

where the absorption area, A, can be written as A= 24 ln(10) V c T60 (9)

V is the room volume and c is the speed of sound. Beranek suggested a modification of the reverberation energy with ), being the the so-called room constant, R = A/(1 average absorption coefficient [13]. Thus, the reverberation energy is r ev Beranek = ) 1600(1 A (10)

The strength of the reverberation, i.e., all reflected sound, Grefl , is then given by G refl = 10 log10 (e + l ) dB (2)

The strength of the direct and early arriving sound over the first 50 ms, G50 , is given by introducing a time-limit specifier for the early (and late) energies, G 50 = 10 log10 (d + e50 ) dB (3)

Vorl ander [14] proved that the room constant, R, is just a special case of his correction factor that is eA / S , where (A = AEyring = S ln(1 ) + 4mV) and m is the air attenuation coefficient. Finally, Barrons revised formula [15] modified the reverberation term by a factor e2r /c , where = 3ln(10)/T60 is the decay constant, making the reverberation distance dependent, so that r ev Barr on = 1600 2r exp A c 31200T60 0.040r exp V T60

where e50 is the sound energy arriving within the first 50 ms (and l50 is the sound energy arriving after 50 ms). Finally, the direct sound, d, in Eq. (3) can be excluded in the final early reflected sound strength as G 50refl = 10 log10 (e50 ) dB (4)

(11)

Now, under the assumption that all reflected sound can be described as an exponential decay, e, l, and rev are related
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

By inserting the proper expressions from Eqs. (5)(11) into Eqs. (1)(4), estimates of the various G-parameters are found. These estimates can be based on the classical reverberation model, Beraneks or Barrons model, and they will be compared with measurements and computer simulations in Section 2.
63

ZIDAN AND SVENSSON

PAPERS

Table 1. Data for the three rooms measured.


1m

Room
1m

Size 5.7 5.7 2.7 = 88m3 5.8 5.9 3 = 104m3 10.3 6 2.9 = 180m3

T60,mid 0.75 s 0.77 s 0.5 s

fSchroeder 185 Hz 176 Hz 107 Hz

R1 R2 R3

Fig. 1. Top view of microphone and loudspeaker configuration setup in the three measured rooms.

1.2 Limitations and Possibilities with the IR Model There is no doubt that early reflections are significant in many respects but the simple IR model(s) ignore(s) these. Only comparisons with measurements and/or detailed simulations will indicate how severe this neglect is for the energy relations. The IR models could, in principle, be expanded if desired. The direct sound pulse can, e.g., be replaced by a measured impulse response in order to get a detailed direct sound impulse response for auralization purposes. Furthermore, very early reflections could be included in the direct sound as done in [10], where the influence of a reflecting table on a microphones response was studied. As mentioned above, the IR model offers the possibility to study convolutions of IRs. Such a convolution represents, e.g., an audio or telephone conference situation, where a microphone and a loudspeaker in one room (represented by a single room impulse response) are connected to a loudspeaker and a microphone in another room (also represented by a single room impulse response). The convolution can be assumed to smear out early reflection patterns and thereby reduce the problem caused by ignoring early reflections.

2 EXPERIMENTS A series of measurements was carried out as earlier reported in [10] and summarized in Section 2.1. Due to the limited scope of the measurement series, computer simulations were carried out as well as described in Section 2.2. Computer simulations have been shown to be quite accurate for strength parameters [16] in medium-sized and large auditoria. 2.1 Measurements A series of impulse response measurements was done in three small rooms using a small two-way loudspeaker (Genelec 1029A) as sound source. Its position was 1 m from one of the end walls and centered between the sidewalls as shown in Fig. 1. The height of the loudspeaker was 1.5 m from the floor to the center of the woofer. An omnidirectional 1/2 microphone (Norsonic 1220) was positioned
64

in a sequence of receiver positions, in steps of 1 m, starting at 1 m in front of the loudspeaker. The height of the microphone and of the loudspeaker was 1.5 m. Only measured positions on-axis of the loudspeaker were used in order to get the same directivity index for all the measured positions. The rooms had no furniture during the measurements, except for room R3, where tables were standing next to the side walls. Using symmetric measurement positions in a room could be viewed as a worst-case test because of accentuated interference effects caused by the two side-wall reflections. To further enhance deviations between measurements and predictions, the direct sound energy was excluded from the analysis. For the measurements, this was done simply by subtracting the direct sound energy from the impulse response energy values. The data of the rooms are shown in Table 1. For the calibration of the Strength parameter, a series of measurements was done in an anechoic room at distances 1, 2, 3, and 4 m, respectively. All these responses were rescaled to a distance of 10 m and the average value was chosen as representing a 10 m distance. All the measured impulse responses were filtered in octave bands and the total energy was summed for each band. Then a scaling factor was derived for each band such that the energy in each octave band for a 10 m free-field impulse response would be 0 dB. For the subsequent measurements, the gain settings were kept identical to the ones used for the anechoic calibration measurements. An important aspect is the directivity of loudspeaker, and, to a very minor degree, of the microphone. The standard ISO 3382 for room acoustical measurements indicates that both the loudspeaker and the microphone should be omnidirectional, which is practically unattainable for the loudspeaker. Allowable deviations for practical loudspeakers are given in ISO 3382, but here it is not critical to use a sound source that is omnidirectional. It is rather more important to use good estimates of the actual directivity factors. Simulations were done by an edge diffraction method in order to get the values of the directivity index in each octave band. This approach has been shown to be accurate [17]. Table 2 shows the on-axis directivity index of the loudspeaker averaged in octave bands. All IR measurements were done using the WinMLS software [18], and the reverberation times were calculated with this software as well. Strength parameters were, however, computed separately by Matlab scripts. As one example, Fig. 2 shows the reflected energy values, Grefl , as a function of source-receiver distance in the octave band range 125 Hz to 4 kHz. The corresponding error between the measured and the predicted reflected
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

PAPERS

ROOM ACOUSTICAL PARAMETERS IN SMALL ROOMS

Table 2. On-axis directivity index of the two-way Genelec 1029A loudspeaker, based on simulations. Octave band 125 Hz 250 Hz 500 Hz 1 kHz 2 kHz 4 kHz Directivity index 0.2 dB 0.6 dB 2.3 dB 5.2 dB 7.5 dB 4.6 dB

Grefl in room 2 - prediction error


125 250 500 1k 2k 4k

Grefl pred. error [dB]


125 250 500 1k 2k 4k 125 250 500 1k 2k 4k

26

Grefl in room 2 * = Measu rement - = Prediction

-1

24

1.5

2.5

3.5

Distance [m]
Fig. 3. Prediction error of the reflected energy, Grefl,Barron , for room 2.

22

Grefl [dB]

20

18

25

Mean error in Grefl. across positions Room 1 Room 2 + 5 dB Room 3 +10dB

16

20

14 1 1.5 2 2.5 3 3.5 4 4.5 5

Error in G refl [dB]

15

Distance [m]
Fig. 2. Measured and predicted reflected energy, Grefl, for room 2. Predictions were based on Barrons model.

10

energy, Grefl , is shown in Fig. 3, where predictions are based on Barrons model. Naturally, for such small rooms, a very weak distance dependence can be observed for the reverberation level. It can be seen in Fig. 3 that the prediction error is mostly within 1 and +2 dB for the octave bands 125 Hz to 2 kHz. Quite large errors occur for the 4 kHz octave band, the reasons for which will be discussed in Section 3. Figs. 2 and 3 show examples from one room, while results for all the three rooms are compiled in Fig. 4. Mean values, as well as the minimum and maximum values are shown. It can be noticed that the mean prediction errors are within 1, +2 dB from 250 Hz to 2 kHz. At 4 kHz the prediction is high on average, and the error varies between 2.4 dB and 3.5 dB. Fig. 5 shows a similar compiled view of the mean error of the early reflected energy, G50refl, Barron , for the three rooms. The mean prediction errors of the early reflected energy, G50refl,Barron , are within 1,+3 dB from 250 Hz to 2 kHz. These values are little bit higher than the mean prediction errors of the reflected energy, Grefl . Similarly to the situation for the reflected energy, the error is higher for the 4 kHz band, ranging from 3 dB to 4 dB.
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

-5 2 10

10

Frequ ency [Hz]


Fig. 4. Mean (and minimum and maximum) prediction error of reflected energy, Grefl,Barron , in the three rooms. Two curves were shifted for readability.

2.2 Simulations A series of computer simulations was done using the CATT acoustics software [19] for more investigations about the validity of the prediction model in small rooms. Three rooms were designed according to the data given in Table 3. Computer simulation software will usually require more input parameters about the room (such as the absorption coefficient, diffusion, etc.) than the simple impulse response models in Section 1. Consequently, the absorption coefficients of the rooms were chosen based on a simplified assumption that the rooms have one absorption coefficient for the floor and another one for all walls and the ceiling. The absorption coefficient of the floor was fixed in all the rooms to a value 0.05, while the absorption coefficient of walls and ceiling was chosen to give the designated reverberation times in Table 3 based on Sabines formula.
65

ZIDAN AND SVENSSON

PAPERS

Table 3. Data of the three rooms simulated Room R1 R2 R3 Size 5.7 5.6 3 = 95.76 m3 10 6 3 = 180 m3 8 7 3 = 168 m3 Absorption coefficient (walls/ceil) 0.18 0.35 0.29 T60,mid 0.5 s 0.5 s 0.6 s fSchroeder 145 Hz 105 Hz 120 Hz

25

Mean error in G50refl. across positions Room 1 Room 2 + 5 dB Room 3 +10dB


Grefl [ dB ]

Room 1, T60 = 0.5s, 10% diff.


24 23.5 23 22.5 22 21.5 21

20

Error in G50refl. [dB]

15

10

20.5 20

CATT results, 2S*99R Classical Beranek Line fit to CATT resu lts Barron's model
0 1 2 3 4 5 6

-5 2 10

10

Frequ ency [Hz]


Fig. 5. Mean prediction error of early reflected energy, G50refl,Barron , in the three rooms. Two curves were shifted for readability.

Distance [ m ]
Fig. 7. Simulated reflected energy, Grefl, CATT , compared to classical theory and Barrons model for room 1 with reverberation time 0.5 s and 10% diffusion.

Room 1, T60 = 0.5s, 60% diff.


24 23.5 23 22.5 22 21.5 21

Grefl [ dB ]

Fig. 6. Top view of two sources and a sample of the 99 receivers in a simulated room.

20.5 20

CATT results, 2S*99R Classical Beranek Line fit to CATT resu lts Barron's model
0 1 2 3 4 5 6

To cover a wide range of cases, each room was calculated with many different levels of diffusion, from which 10%, 60%, and 100% have been chosen to represent low levels, medium levels, and high levels of diffusion, respectively. Each room was simulated with the following source/receiver configuration setup. Two omnidirectional source positions were used. A first position was in the center of the room, 0.5 m from the front wall and a second position 1 m from the front wall and 0.5 m from the sidewall, as illustrated in Fig. 6. The 99 receivers were chosen to be located randomly with a constraint to be at least 1 m away from the first source and 0.5 m away from the second source and 0.5 m away from all the sidewalls. The height of the sources and the receivers was 1.5 m. Fig. 6 shows one example of a source/receiver configuration setup.
66

Distance [ m ]
Fig. 8. Simulated reflected energy, Grefl,CATT , compared to classical theory and Barrons model for room 1 with reverberation time 0.5 s and 60% diffusion.

The idea behind the random distribution was to get a range of variations as large as possible. In addition, the random distribution avoids a bias due to possibly systematic distributions. Fig. 7 and Fig. 8 demonstrate examples of the simulated reflected energy, G refl,CATT , compared with the classical theory, Barrons model, and Beraneks model. Fig. 7 gives an example with mostly specular reflections (10% diffusion) while Fig. 8 with more diffuse reflection (60% diffusion).
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

PAPERS

ROOM ACOUSTICAL PARAMETERS IN SMALL ROOMS

Table 4. Average absolute error for Grefl with Grefl, CATT as reference. Mean error for classical theory 0.69 dB 0.71 dB 0.70 dB 1.91 dB 2.13 dB 2.46 dB 1.26 dB 1.29 dB 1.41 dB Mean error for Beraneks model 0.26 dB 0.37 dB 0.46 dB 0.95 dB 1.31dB 1.67 dB 0.51 dB 0.73 dB 0.90 dB Mean error for Barrons model 0.10 dB 0.18 dB 0.26 dB 0.27 dB 0.47 dB 0.81 dB 0.18 dB 0.37 dB 0.50 dB

Room R1 T60 = 0.5s

Diffusion % 10% 60% 100% 10% 60% 100% 10% 60% 100%

R2 T60 = 0.5s

R3 T60 = 0.6s

Room 1, T60 = 0.5s, 10% diff.


24 23.5 23 24

Room 1, T60 = 0.5s, 60% diff.


23.5 23

CATT results, 2S*99R Line fit to CATT resu lts Barron's model

CATT results, 2S*99R Line fit to CATT resu lts Barron's model

G50refl [ dB ]

G50refl [ dB ]

22.5 22 21.5 21 20.5 20

22.5 22 21.5 21 20.5 20

Distance [ m ]
Fig. 9. Simulated early reflected energy, G50refl , compared to Barrons model for room 1 with reverberation time 0.5 s and 10% diffusion.

Distance [ m ]
Fig. 10. Simulated early reflected energy, G50refl , compared to Barrons model for room 1 with reverberation time 0.5 s and 60% diffusion. Table 5. Average absolute error for G50refl

All the rooms were simulated with a larger set of diffusion percentages. It was observed that for diffusion levels of 10%, 20%, and 40%, the values of the reflected energy, G refl,CATT , were all very similar. Only when the diffusion factor was 60% or more (80% and 100%) did the values differ markedly from those in Fig. 7. It can be noticed from Figs. 7 and 8 that the maximum difference in reflected energy, Grefl , between the CATT results and Barrons model does not exceed +/ 0.5 dB, which is to a high extent acceptable. Furthermore, it can be observed that Beraneks model corresponds closely to the average value through the room as explained in [14]. Table 4 summarizes the mean absolute error of the reflected energy, Grefl , according to classical theory, Barrons model, and Beraneks model across all the rooms and all the cases. The simulation results, Grefl, CATT , were considered as the reference. Figs. 9 and 10 show the simulated early reflected energy, G50refl , compared to Barrons model for a diffusion percentage of 10% and 60% respectively. Table 5 summarizes the mean error of the early reflected energy, G50refl , according to Barrons model across all the rooms and all the cases.
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

Room R1 T60 = 0.5s

Diffusion % 10% 60% 100% 10% 60% 100% 10% 60% 100%

Mean error for Barrons model 0.36 dB 0.38 dB 0.43 dB 0.49 dB 0.57 dB 1.01 dB 0.63 dB 0.63 dB 0.65 dB

R2 T60 = 0.5s

R3 T60 = 0.6s

It can be noticed that the mean error for early reflections, G50refl , was higher than for the reflected energy, Grefl , in all the cases. On average, for the nine configurations, the prediction error for G50refl is 0.2 dB larger than for Grefl . 3 DISCUSSION OF RESULTS The comparison of the reflected energy, Grefl , for the measurements done in the three rooms, and the predictions
67

ZIDAN AND SVENSSON

PAPERS

according to Barrons model, showed quite close agreement for the frequency bands 250 Hz to 2 kHz. The prediction error varied between 1 dB and +2 dB. For the 4 kHz band the error increased to between 2.5 dB and 3.5 dB. A possible cause for this noticeable error is the difficulty of accurately estimating the directivity index of the loudspeaker at higher frequencies. The simulation method modeled the drivers as a collection of point sources on the flat frontal baffle, while the loudspeaker does have a very shallow recession around the tweeter. Such a recession would act as a mild horn loading, which would be expected to lead to an increased directivity. Based on the detailed simulations with the CATT Acoustics software, Barrons model worked very well, too, for the prediction of average trends. The predictions are even closer to the detailed simulation values than to the measurements. This is not surprising since, among other factors, the detailed calculations used an omnidirectional source. As a consequence, the possible source of prediction error caused by an uncertain directivity was eliminated. Moreover, the trend lines of the results (strength values as function of distance) were the most interesting quantities to compare, and predicted trend lines did indeed agree well with the trend lines of the detailed simulations. Certainly, when looking at individual receiver points, the prediction error could be as large as observed for the measurements. When the diffusion was as low as 10%, as shown in Fig. 7, the mean error was below a few tenths of a dB. When the diffusion was increased to 60% or more, the mean error increased but was still bounded within 0.5 dB. For the early reflected energy, G50refl , the measurements show, as in Fig. 5, that the model also worked reasonably well although the prediction error was a little bit higher than the total reflected energy, Grefl . The prediction error varied between 2 dB and +2 dB. It is a bit surprising that the prediction error for Barrons model was smaller when distinct early reflections occurred (that is, when the diffusion was low), but Barrons model has been explained in terms of the image source model, which corresponds to a diffusion coefficient of zero [15]. It can be noticed that the error was only somewhat higher for the early reflected energy, G50refl , than the error of Grefl , for the same rooms and the same cases. These errors illustrate that on average in rooms, the simple model predicts fairly well the strength values in small shoebox-sized rooms with little or no furnishing. There is no doubt that the error in one specific receiving point might be substantially higher. However, the average behavior of the model might be very useful for studying the influence of parameters such as room volume, source-receiver distance, microphone, and loudspeaker directivity. Furthermore, as the impulse response model used in this paper is composed of two components only, direct sound and an exponential reverberation tail, it can be used to study the properties of two rooms or more through the convolution of room impulse responses as has been done in [10] and [11].
68

4 CONCLUSIONS A simple room impulse response model, corresponding to Barrons model, has been used to investigate the room conditions (early and late energy parameters) in small size shoebox-shaped rooms. Measurements and computer simulations were done to evaluate the simple model. Measurements show that the error was limited between 1 dB and + 2 dB for reflected energy, Grefl , across all the octave bands except at 4 kHz. For early reflected energy, G50refl , the prediction error did not exceed 2 dB. Detailed simulations showed that the mean error did not exceed + 0.8 dB for reflected energy, Grefl , and + 2.3 dB for the early reflected energy G50refl . In all the cases, for both measurements and simulations, the error for the early reflected sound, G50refl , was a bit higher than the error for reflected energy, Grefl , but still within acceptable limits. Therefore, it is suggested that the impulse response model can be useful for studying two rooms connected as in a video conference setup. 5 REFERENCES [1] S. Cerd a, A. Gim enez, and R. M. Cibri an, An Objective Scheme for Ranking Halls and Obtaining Criteria for Improvements and Design, J. Audio Eng. Soc., vol. 60, pp. 419430 (2012 June). [2] M. Barron, Using the Standard on Objective Measures for Concert Auditoria, ISO 3382, to Give Reliable Results, Acoust. Sci. & Tech. vol. 26, pp. 162169 (2005). [3] J. S. Bradley, Using ISO 3382 Measures, and Their Extensions, to Evaluate Acoustical Conditions in Concert Halls, Acoust. Sci. & Tech. vol. 26, pp. 170178 (2005). [4] J. S. Bradley, Experience with New Auditorium Acoustic Measurements, J. Acoust. Soc. Am., vol. 73, no. 6, pp. 20512058 (1983). [5] M. Aretz, and R. Orlowski, Sound Strength and Reverberation Time in Small Concert Halls, Applied Acoustics vol. 70, pp. 1099-1110 (2009). [6] E. Nilsson, Room Acoustic Measures for Classrooms, Proceedings of Internoise, Lisbon, Portugal (2010). [7] S. Tervo, T. Lokki, and L. Savioja, Maximum Likelihood Estimation of Loudspeaker Locations from Room Impulse Responses, J. Audio Eng. Soc., vol. 59, pp. 845 857 (2011 Nov.). [8] A. Lindau, L. Kosanke, and S. Weinzierl, Perceptual Evaluation of Model- and Signal-Based Predictors of the Mixing Time in Binaural Room Impulse Responses, J. Audio Eng. Soc., vol. 60, pp. 887898 (2012 Nov.). [9] U. P. Svensson, Energy-Time Relations in an Electroacoustic System in a Room, J. Acoust. Soc. Am., vol. 104, pp. 14831490 (1998). [10] U. P. Svensson and H. EL-Banna Zidan, Early Energy Conditions in Small Rooms and in Convolutions of Small-Room Impulse Responses, presented at the 129th Convention of the Audio Engineering Society (2010 Nov.), convention paper 8223. [11] U. P. Svensson, H. EL-Banna Zidan, and J. L. Nielsen, Properties of Convolved Room Impulse Responses, Proc. of the 2011 IEEE Workshop on Applications
J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

PAPERS

ROOM ACOUSTICAL PARAMETERS IN SMALL ROOMS

of Signal Processing to Audio and Acoustics (WASPAA), New Paltz, NY (2011). [12] ISO 3382:1997, AcousticsMeasurement of the Reverberation Time of Rooms with Reference to Other Acoustic Parameters, International Organization for Standardization (1997). [13] L. Beranek, Acoustics (McGraw-Hill, New York, 1954). [14] M. Vorl ander, Revised Relation between the Sound Power and the Average Sound Pressure Level in Rooms and Consequence for Acoustics Measurements, Acustica, vol. 81, pp. 332343 (2005).

[15] M. Barron and L. J. Lee, Energy Relations in Concert Auditoriums. I, J. Acoust. Soc. Am., vol. 84, pp. 618628 (1988). [16] M. Vorl ander, International Round Robin on Room Acoustical Computer Simulations, Proceedings 15th ICA Trondheim, Norway, pp. 689692 (1995). [17] U. Peter Svensson and K. Wendlandt, The Influence of a Loudspeaker Cabinets Shape on the Radiated Power, Proc. of Baltic Acoustic, Sept. 1721; J. of Vibroeng., No. 3(4), pp. 189192 (2000). [18] Winmls Software, http://www.winmls.com. [19] CATT Acoustics software, http://www.catt.se.

THE AUTHORS

Hassan Zidan Hassan EL-Banna Zidan was born in Cairo, Egypt in 1975. He received a B.Sc. and a M.Sc. in electronics and telecommunications engineering in 1998 and 2007 both from the Arab Academy for Science, Technology, and Maritime Transport (Alexandria/Cairo). Since 2008 he has been a Ph.D. candidate in the acoustics group at the Norwegian University of Science and Technology (NTNU), Trondheim, Norway. His research interests include audio signal processing, speech enhancement, and room acoustics.

Peter Svensson Since 1999 he has been a professor in electroacoustics at the Norwegian University of Science and Technology, Trondheim, Norway. His main research interests are auralization, especially computational room acoustics and sound reproduction techniques, measurement techniques, and perceived room acoustical quality. Prof. Svensson has published 33 journal papers and more than 100 conference papers. He is currently vice president of the European Acoustics Association and a past president of the Norwegian Acoustical Society. He is an associate editor in the field of electroacoustics for Acta Acustica united with Acustica. In 2001 he received a best paper award, together with Johan L. Nielsen, for authors 35 years or younger from the Journal of the Audio Engineering Society and in 2009 a best paper award from the IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), together with Haohai Sun and Shefeng Yan.

Peter Svensson received a M.Sc. degree in engineering physics in 1987 and a Ph.D. degree in 1994, both from Chalmers University of Technology, Gothenburg, Sweden. He has held postdoctoral positions at Chalmers University, University of Waterloo, Ont., Canada, and Kobe University, Japan.

J. Audio Eng. Soc., Vol. 61, No. 1/2, 2013 January/February

69

Das könnte Ihnen auch gefallen