Investigating listeners’ preferences in Detmold Concert Hall by comparing sensory evaluation and objective measurements
Banu Sahin, Sebasti` a V. Amengual, Malte Kob
Erich Thienhaus Institute, Detmold University of Music, Email: bsahinsound@gmail.com,{amengual, kob}@hfm-detmold.de
Introduction
For the halls which are used for speech and music, acous- tical design is as important as the architectural design.
Poor acoustical conditions may lead to expensive and time consuming changes after construction as in the case of the Philharmonic Hall in New York which was built in 1962 and reopened in 1976 as Avery Fisher Hall [1, p.336]. A large number of parameters have been derived from measurements to characterise the acoustic proper- ties of such halls [2] [3] [4]. On the other hand, subjective aspects of the acoustical measures also need to be consid- ered since the whole concern is to meet the expectations of the audience.
The standard ISO 3382-1 [5] presents a list of subjective listener aspects and its relations to acoustical properties of the room (see Table 1). However, the evaluation of subjective assessments is a challenge due to individual preferences of assessors. Each listener in a concert hall auditorium has an individual taste, thus the same acous- tical conditions may result in different subjective evalua- tions of listeners. For instance, a powerful sound may be preferred by some group of listeners while a clear sound quality may be preferred by the others. There are also other factors such as the preference of instrument, music genre and composer and as well as the psychological and the physical conditions of the subject. Therefore, a com- prehensive understanding of the concert hall acoustics requires a multidimensional investigation method [6].
In this paper, a set of measurements is presented for the purpose of investigating both physical and perceptual as- pects of the Detmold Concert Hall auditorium. To this end, several Room Impulse Response (RIR) and Binau- ral Room Impulse Response (BRIR) measurements were conducted in different seat positions throughout the au- ditorium area. Measured BRIRs were then convolved with anechoic music to generate auralizations which al- lowed the investigation of preference ratings in the lis- tening tests. All the acoustic quantities estimated from the RIR measurements were mapped and compared to the subjective listener aspects described in [5].
Measurements
For the RIR measurements an omnidirectional and a figure-of-eight microphone were used. BRIRs were mea- sured using an artificial head with pinna and ear canals.
The measurements were conducted simultaneously by at- taching the omnidirectional and the figure-of-eight micro- phones to the artificial head (see Fig. 1).
The state of the room was unoccupied also without per- former’s chairs, music stands and instruments except a grand piano on the stage.
A total of thirty-three seat positions were measured throughout the auditorium area (Figure 2). The mini- mum distance between two measurement positions was 2 m, as recommended in ISO 3382-1 [5]. The distance be- tween any microphone position and the nearest reflecting surface was not less then 1 m. The ears of the artificial head, the omnidirectional microphone and the figure-of- eight microphone were placed at approximately 112 cm, 135 cm and 139 cm of height from the floor, respectively.
The microphones were oriented perpendicularly to the seats in each measurement.
Figure 1: The impulse response measurements setup.
A modified version of the loudspeaker orchestra [7] was used as sound source to create a more similar source to the real situation. The modification was based on a previ- ous study [8] in order to allow comparability of results in the future. Eight loudspeakers were placed on the stage approximately 1.4 and 1.8 m of height from the floor as two rows on the stage. The set-up and relative distances between loudspeakers are shown in Fig. 3.
A logarithmic sweep signal was used for the excitation (20-20000 Hz) with a 24-bit quantization and 48 kHz sam- pling rate. The length of the sweep signal was approxi- mately 2,7 seconds. Measurements were performed with an average of 3 for each loudspeaker, in order to decrease the noise level. All the levels on the loudspeakers were set to the same level and flat equalization.
The equipment used for the measurements was as follows:
DAGA 2017 Kiel
249
Table 1: Acoustic quantities grouped according to listener aspects [5].
Subjective listener aspect Acoustic quantity Freq. avg.
(Hz) JND Typical
range Subjective level of sound G - Sound strength (dB) 500 to 1000 1 dB -2 dB; +10 dB Perceived reverberance EDT - Early decay time (s) 500 to 1000 Rel. 5% 1,0 s; 3,0 s
Perceived clarity C80- Clarity (dB) 500 to 1000 1 dB -5 dB; +5 dB
Apparent source width (ASW) JLF/LFC - Early lateral energy fraction 125 to 1000 0,05 0,05; 0,35 Listener envelopment (LEV) LJ - Late lateral sound level (dB) 125 to 1000 Unknown -14 dB; +1 dB
STAGE
2 4 6 8 10 12 14 16 18
14 16 18 20 22 24 26 28 30 32
SELECTED SEATS
67 52 57
131 121
163 168
206 211
216 280
290
296 372
377 448
454
459
470 546
557
361
551
62 126 158 274
535
541
367
464
383
285
Figure 2: Measured seats in the Detmold Concert Hall audi- torium. Colored seats are selected to use for the auralization and so for the listening tests.
• NTi M2010 omnidirectional microphone,
• Schoeps CCM8 figure-of-eight microphone,
• Neumann KU 100 artificial head,
• 8 x Neuman KH 120 A loudspeakers,
• RME: Micstasy preamp and AD/DA converter,
• RME: Madiface USB,
• RME: ADI-8 AD Sound Card,
To be able to calculate the absolute value of G, a set of sound pressure exposure level (LpE) measurements were held in the reverberation chamber of the Erich Thienhaus Institute. The measurements were conducted according to the standard ISO 3741 [9].
Listening tests
Subjective preference ratings were derived from the lis- tening tests performed using Sennheiser HD 650 head- phones. Subjects were asked to compare the aural- ized sounds of different listening positions in the con- cert hall auditorium. For the auralization purpose, 10 seats were selected (see Figure 2) by taking into ac- count the just noticeable differences (JNDs) of the mea-
Figure 3: The loudspeaker orchestra used as source.
sured room acoustics parameters (see Table 1). These seats were then auralized by convolving obtained BRIRs with an anechoic symphonic music recording [10]. The piece is a recorded excerpt of the Romantic-era composer A. Bruckner’s (1824-1896) Symphony no. 8 (II. move- ment) and its duration is approximately 42 seconds. This piece was selected due to its large and varying dynamics as well as the large orchestra size.
A two-alternative forced choice (2AFC) or in other words, a paired comparison test was used to derive preference ratings. Sixteen expert subjects were participated in the tests with ages ranged between 18-47 years old. The listening test was designed using Max/MSP software in a way that it enables the subject to listen and compare the two sound files by switching between stimuli in real time.
Each subject compared a total of 45 pairs and the test duration changed dramatically depending on the subject as between 11 and 45 minutes. The average duration was around 25 minutes.
DAGA 2017 Kiel
250
Figure 4: Standard acoustic parameters interpolated over the audience area. Black circular markers represent actual measure- ment positions.
Results
Monaural and binaural acoustic parameters were derived and mapped over the auditorium after interpolating the values. The parameters showing greater variations are presented in Fig. 4. The reverberation time (RT30) is ap- proximately 1.6 s and does not show significant variations over the space. On the contrary, EDT varies between 1.4 s at farthest positions and 1.6 s at central positions.
C80 ranges between 0 and 2.5 dB, with greater values at the back positions. In addition, frontal positions located closer to the organ present the lowest clarity values. This could be due to a scattering effect of the organ on the early lateral reflections. Strength (G) decreases progres- sively from 9 to 6 dB when moving further from the stage.
Late lateral sound level (LJ) follows a similar trend, with values ranging from 0.2 to -0.8 dB. Early and late inter- aural cross-correlation values tend to be higher at frontal and central positions, and they are centered at around 0.6 and 0.5, respectively.
The preference ratings from the listening test were ana- lyzed using a BTL model [11]. The predicted values show a clear preference for seat 126, with a value statistically significantly greater than any other seat. Seat 62 was the second most preferred seat, although the differences with seats 158, 274 and 367 are not significant. Seats 464, 285 and 383 conform a third group. The least pre- ferred seats are 535 and 541. Correlation analysis was
62 126 158 274 285 367 383 464 535 541
Seat number 0
0.1 0.2 0.3 0.4 0.5 0.6
BTL preference
Figure 5: BTL estimation.
performed relating subjective preference and standard acoustic parameters. However, due to the high differ- ence between the most preferred seat and the rest of the groups, this particular seat acted as an outlier, thus cre- ating a skewed distribution of the data. For this reason, the BTL analysis was performed again without including the most preferred seat. Finally, correlation analysis was performed using the filtered data. The analysis shows a clear relationship between the acoustic parameters G and LJ and subjective preference. The detailed correla- tion values are included in Table 2.
DAGA 2017 Kiel
251
Table 2: Correlation values between acoustical parameters and subjective preference. Significant correlations (p <0.05 and
p <0.01) are marked with underlined and bold fonts, respectively.
G EDT C80 JLF LJ IACCearly IACClate IACCtot
0.88 0.17 -0.29 0.59 0.83 0.67 0.4 0.71
Discussion
The standard ISO 3382-1 [5] was followed for the calcu- lations of acoustic parameters. However, the acoustic source was not omnidirectional, and instead it was a dis- tributed source. Although the directivity of the source was not investigated, it is assumed that reproducing sym- phonic music using a wider source supposes a better rep- resentation of a symphonic orchestra than a single loud- speaker. Additionally, the microphones were not oriented towards the center of the stage. Instead, seat positions were taken as the reference and so they were placed per- pendicular to the seats. This means that, since head tracking was not used for the binaural resynthesis, the direction of incidence of the direct sound was different depending on the listening position. However, and in spite of its probable impact on the listeners, it is not clear how to establish a relation between the incidence of the direct sound and any standard room parameter.
The preference results seem to be partially explained by the acoustic parameters and dominated by the sound strength (G), although other correlations are found. It is not known if the differences in late lateral sound level (LJ) are perceptually relevant, since the biggest differ- ence in the compared seats is around 1 dB. Finally, the two most preferred seats (126 and 62) are fairly close (in front of the stage) and present similar acoustic values, but their preference ratings are very different. This sug- gests that although standard parameters provide some insight, they could be insufficient to explain the subjec- tive aspects involved in rating the preference and goes in line with previous findings [12], [6].
Future work may comprise repeating listening tests by including head tracking on the auralization. Given that the preference is dominated by the sound strength (G), tests could be repeated performing loudness equalization in order to investigate other relationships between the room acoustic parameters and the subjective preferences.
Additionally, including more subjects and different mu- sical pieces in the listening tests would contribute to a generalization of the results.
Acknowledgements
Special thanks go to Timo Grothe, Martin Schneider, An- dreas Meyer, Tapio Lokki and Stephan M. A. Ernst. Also to the participants in the listening tests, who have will- ingly shared their precious time and effort. Part of the research work presented in this article has been funded by the European Commission within the ITN Marie Curie Action project BATWOMAN under the 7th Framework Programme (EC grant agreement no. 605867).
References
[1] R. T. Beyer, Sounds of our times: Two hundred years of acoustics, illustrated. New York: Springer Science and Business Media, 1999.
[2] L. L. Beranek, Music, acoustics and architecture, 1st Edition. New York: Wiley, Jun. 1962.
[3] M. Barron,Auditorium acoustics and architectural design. USA and Canada: E and FN Spon, 1993.
[4] R. Lacatis, A. Gim´eenez, A. B. Sevillano, S.
Cerd´aa, J. Romero, and R. Cibri´aan, “Histori- cal and chronological evolution of the concert hall acoustics parameters”, Acoustics’08, Paris, Jun.
2008.
[5] Acoustics – measurement of room acoustic param- eters – part 1: Performance spaces, 1st Edition, International Standards Organization, ISO 3382-1, Geneva, Switzerland, 2009.
[6] T. Lokki, J. P¨atynen, A. Kuusinen, and S. Tervo,
“Disentangling preference ratings of concert hall acoustics using subjective sensory profiles”, Jour- nal Ac. Soc. America, vol. 132, pp. 3148–3161, 5 2012.
[7] J. P¨atynen, S. Tervo, and T. Lokki, “A loudspeaker orchestra for concert hall studies”, The Seventh In- ternational Conference on Auditorium Acoustics, Oslo, Norway, Oct. 2008, pp. 45–52.
[8] S. V. Amengual, J. P¨atynen, and T. Lokki, “Phys- ical and perceptual comparison of real and focused sound sources in a concert hall”,Journal of the Au- dio Engineering Society, vol. 64, no. 12, pp. 1014–
1025, 2016.
[9] Acoustics – determination of sound power lev- els of noise sources using sound pressure – preci- sion methods for reverberation rooms, 4th Edition, International Standards Organization, ISO 3741, Geneva, Switzerland, Oct. 2010.
[10] J. P¨atynen, V. Pulkki, and T. Lokki, “Anechoic recording system for symphony orchestra”, Acta Acustica united with Acustica, vol. 94, pp. 856–865, 2008.
[11] F. Wickelmaier, “A matlab function to estimate choice model parameters from paired-comparison data”, Behavior Research Methods, Instruments, and Computers, vol. 36, no. 1, pp. 29–40, 2004.
[12] E. Kahle, “Room acoustical quality of concert halls:
Perceptual factors and acoustic criteria -– return from experience”, in Proceedings of the Interna- tional Symposium on Music and Room Acoustics 2013, ICA, Toronto, Canada, Aug. 2013.
DAGA 2017 Kiel
252