Barron - Using the standard on objective measures for concert auditoria, ISO 3382, to give reliable results.pdf - Akustyka(1) - dzidzia2603

Acoust. Sci. & Tech. 26, 2 (2005)

PAPER

Using the standard on objective measures for concert auditoria,

ISO 3382, to give reliable results

Mike Barron

Department of Architecture and Civil Engineering, University of Bath,

BATH BA2 7AY, UK

( Received 17 July 2004, Accepted for publication 4 December 2004 )

Abstract: The current version of the standard ISO 3382 has now been in existence for seven years,

yet for many the contents of Annexes A and B on newer measures remain confusing. A major issue is

the use to which these measures are put. Where the ‘new’ measures for auditoria diﬀer from other

acoustic parameters is that they refer to a range of subjective eﬀects, which are perceived

simultaneously. Using the newer measures requires a good understanding of the multi-dimensional

nature of music perception. Measurement data requires interpretation. When measurements are made

in unoccupied auditoria, the data requires correction to the situation with full audience. Another issue

is how to condense data measured across audience areas. The simplest approach is to present mean

values of the diﬀerent quantities, but this ignores the fact that many quantities vary signiﬁcantly with

location; the disappointment of sitting in a poor seat in an auditorium is no less for the knowledge that

the overall mean is good. Several of these issues are discussed here with the aim of promoting more

uniformity in the way the objective measures proposed in the Standard are applied by diﬀerent

research groups and companies.

Keywords: Auditorium acoustics, Concert halls, Reverberation time

PACS number: 43.55.Gx [DOI: 10.1250/ast.26.162]

1. INTRODUCTION

inter-aural cross correlation coe cient (IACC)—

Annex B

This paper will restrict itself to concert hall measure-

ments and will be concerned principally with the newer

measures from Annex A.

The 1997 revision of ISO 3382 was titled ‘‘Measure-

ment of the reverberation time of rooms with reference to

other acoustic parameters’’ [1]. The previous version from

1975 concerned itself exclusively with reverberation time.

The 1997 version is currently being revised into a Part 1

(the 1997 standard intended principally for performance

spaces) and Part 2 for reverberation time measurements in

ordinary rooms. The principle applied to Part 2 is that the

accuracy of measurement in ordinary rooms can be less

than in auditoria (mainly allowing for fewer source and

receiver positions).

The following measures are deﬁned in the 1997

standard:

reverberation time (RT) — main body of standard

sound strength (G) — Annex A

early decay time (EDT) — Annex A

balance between early and late arriving energy (C 80

and others) — Annex A

early lateral energy measures (LF and LFC)—

Annex A

2. SUBJECTIVE CRITERIA

There are many verbal expressions used for subjective

response to live music performance. It is certain that

further subtleties remain to be resolved. At least eight

subjective qualities are currently mentioned regularly, as

listed in Table 1 together with recommended objective

measures. Listeners with some experience of completing

questionnaires can usually comment on each of these

subjective qualities. There is substantial evidence however

that listeners vary in their preferences, so that they select

diﬀerent criteria when making an overall judgement.

Subjective studies to date conclude that listeners subdivide

into at least three groups: those that prefer either clarity or

reverberance or intimacy above other concerns [2, p. 188].

It is clear that a simplistic interpretation of the signiﬁcance

of the measures found in ISO 3382 is unwise.

As shown in Table 1, what was often called spatial

impression is now understood to comprise two separate

e-mail: m.barron@bath.ac.uk

162

M. BARRON: USING ISO 3382 TO GIVE RELIABLE RESULTS

Table 1 Subjective qualities in concert halls and their

possible objective correlates.

values should be corrected for these sensitivity diﬀerences.

Measurement of the total relative sound level (G)

depends on knowing the source sound power, or the

magnitude of the direct sound component. This should be

checked regularly, preferably on site before auditorium

measurements using a technique which enables the direct

sound to be isolated.

Subjective quality Objective measure

Clarity Clarity Index (C 80 )

Reverberance Early decay time (EDT)

Intimacy Total relative sound level (G)

Source broadening Early lateral energy fraction and sound

level

Listener envelopment Late lateral level

Loudness

Total sound level and source-receiver

distance

3.2. Audience Occupancy

In several respects, the usual measurement conditions

in halls diﬀer from the performance situation. A frequent

diﬀerence concerns occupancy, both in the audience

seating and on the stage. The ideal is either to make

occupied measurements or to include absorbers which

simulate people, such as that proposed by Hidaka,

Nishihara and Beranek [7].

In the case of measurements without audience, most

concert halls fortunately have well-upholstered seating

which, though never as absorbent as occupied seating, is

almost as absorbent. It is likely that in most concert halls

the correction of objective measures for the change of

reverberation time is suciently accurate. Corrections

should be applied to all measures except the spatial ones, as

outlined in Sect. 6.1.

Typical magnitudes of corrections are four diﬀerence

limen for RT and EDT and one and a half diﬀerence limen

for C 80 and G. These have been derived as follows from

reverberation time data and diﬀerence limen for the various

objective measures, as in Table 2. Occupied and unoccu-

pied reverberation times of 17 concert halls are given by

Hidaka et al. [7], Table 2. If the Vienna Musikvereinssaal

data is omitted because some of its seating is hard, then the

mean unoccupied and occupied RTs are 2.32 and 1.85 s,

with a ratio of 0.80. With a diﬀerence limen of 5% for RT,

this corresponds to four diﬀerence limen. A similar number

of diﬀerence limen will apply to EDT; though the ratio of

EDT to RT varies slightly [8], the ratio is independent of

actual RT value. The author’s revised theory for sound

level in rooms (Sect. 7 below) allows typical values of C 80

and G to be predicted: for the reverberation time change

mentioned, the maximum change of C 80 is 1.5 dB (for a

source-receiver distance of 10m) and for G is 1.6 dB (for a

Brilliance

Warmth

Bass level balance?

subjective eﬀects: source broadening and listener envelop-

ment. Bradley and Soulodre [3] have proposed the late

lateral level as a measure of envelopment. This is not

included in the 1997 version of the standard but is under

discussion for its revision. Some of this author’s views on

spatial issues are found in [4]. Total relative sound level is

often referred to as ‘Strength.’

An understanding of the signiﬁcance of each of these

proposed objective measures is enhanced by knowledge of

their history [5]. The history is important because sub-

jective experiments relating to concert hall listening are not

straightforward and have generally been conducted by

individuals working in diﬀerent labs around the world on

their own initiative (there being little economic imperative

in this area). The measures listed in the Standard are the

best currently available but could in most cases be

improved. One major di culty with the proposed objective

measures is their interdependence. For instance, reverber-

ation time inﬂuences C 80 , EDT and G.

A mystery at present in concert hall acoustics concerns

the subjective eﬀects of substantial diﬀusing surfaces on

the walls and ceiling of halls. There is some evidence that

listeners prefer diﬀuse conditions [6] but this is not

conclusive. The state of diﬀusion remains to be satisfac-

torily quantiﬁed and no suggestions have been oﬀered for

how we perceive diﬀusion.

3. OBJECTIVE MEASUREMENT

PROCEDURES

3.1. Calibration

For two of the objective measures (LF and G),

calibration is important for accurate results. For the

measurement of the early lateral energy fraction (LF),

measurement of the lateral portion is made with a ﬁgure-of-

eight microphone, with the lateral energy being compared

with that measured with an omni-directional microphone.

The maximum sensitivity of the ﬁgure-of-eight microphone

at the measuring frequencies should be measured relative

to that of the omni-directional microphone; an anechoic

chamber is likely to be the best location for this. Measured

Table 2 Possible frequency ranges for octave band

measurements in concert halls and subjective diﬀer-

ence limen after Bork [23].

Measure

Frequency range

(Hz)

Diﬀerence

limen

Reverberation time

125–4,000

Early decay time

125–2,000

Clarity Index (C 80 )

500–2,000

1 dB

Early lateral energy fraction, LF 125–1,000

0.05

Total relative sound level, G

125–2,000

1 dB

163

Acoust. Sci. & Tech. 26, 2 (2005)

source receiver distance of 40m). In both cases, this

corresponds to changes between one and two diﬀerence

limen.

complex directivity, which also changes depending on the

note being played. The standard measurement technique is

to use a single omni-directional source, usually a dodec-

ahedron loudspeaker. To appreciate the artiﬁciality of a

single source, one needs to listen to anechoically recorded

music played through an omni-directional source on a

concert hall stage; it is a lifeless listening experience. No

research into the signiﬁcance of this issue appears to have

been done.

3.3. Stage Occupancy

When one goes to make a measurement in a hall, one

can either ﬁnd the stage empty or occupied with chairs and

music stands. Of these, the latter is deﬁnitely to be

preferred. Ideally to match conditions with an orchestra,

chairs on stage would be occupied; however unoccupied

chairs and stands will partly obscure stage ﬂoor reﬂections,

which are the exception rather than the rule for symphony

orchestra performance. Measurement on an empty stage,

with in many cases ﬂoor reﬂections, will of course be

relevant to performances with small numbers of musicians.

The presence of chairs on stage also inﬂuences the

measured sound strength in the auditorium and may easily

reduce the sound level by a decibel or more; this is an

example of the eﬀect of absorption close to the source [9].

This has been observed in more than one location, most

recently in a large concert hall that was measured on

diﬀerent days, on one occasion with 50 chairs on stage and

the other with a bare stage (the source position was the

same for both measurements, namely 2m from the stage

front). The average auditorium level diﬀerence was 1.0 dB

[10]; that is about one diﬀerence limen. The ISO standard

rightly speciﬁes that the stage conditions should be

carefully recorded.

3.6. Receiver Positions

The ISO standard is speciﬁc about the minimum

number of microphone positions, depending on auditorium

size. These should be distributed uniformly about the

seating area. When measuring a symmetrical hall, if the

decision has been made to measure with a source (or

sources) only on the centre line, then microphone positions

only in one half of the hall may be used. In this case

microphone positions should not be within 1m of the line

of symmetry to avoid degenerate situations.

For audience conditions, there is no merit in measuring

too close to the source where the direct sound dominates. In

large concert halls a minimum source-receiver distance of

around 10m seems appropriate; this dimension is perhaps

best expressed in terms of the reverberation radius (where

the direct and reﬂected sound components are equal in

level). The reverberation radius is a function of the total

acoustic absorption; for concert halls with 10m 3 /seat and

an RT of 2 s, the reverberation radius varies between 4 and

7m for 1,000 to 3,000 seats. The suggested minimum

source-receiver distance is thus between 2.5 and 1.4 times

the reverberation radius and therefore in the region

dominated by reﬂected sound.

3.4. Source Locations

The possible inﬂuence of stage ﬂoor reﬂections should

also be taken into account when choosing sound source

locations. For this author’s measurements, the tendency has

been to use a single source position on the hall centre line

3m from the stage front. This location was chosen to

minimise the chance of stage ﬂoor reﬂections to the

audience occurring. With a full orchestra on stage, these

reﬂections will be obscured for most musicians. A stage

reﬂection can be expected to increase level by more than a

dB; that is in excess of one diﬀerence limen.

Using just a single source position can of course be

criticised since conditions are likely to vary with source

position on the platform. However this seems a lesser risk

than the inclusion of ﬂoor reﬂections for some source

positions and not others. In other words, in the absence of

chairs on stage, a single forward source position seems the

best compromise, when one wants to measure conditions

appropriate to an orchestra performing on stage.

4. MEASUREMENT FREQUENCIES

The ISO standard avoids being prescriptive about the

appropriate frequencies for measurement. Nor does the

standard say how results should be averaged to establish

the overall clarity, or whatever, in a concert hall. It is

recommended that measurements be taken in the six octave

bands from 125–4,000Hz. The standard suggests quoting

results by averaging over pairs of octaves to give low, mid-

and high frequency values. Bradley [11] uses this approach

for timbre-related parameters.

There are however two complications that occur at the

4,000Hz octave. Firstly the reverberation time etc. are

sensitive to air absorption, determined by temperature and

relative humidity. The main part of the standard states that

temperature and humidity should be measured, which

allows for correction of the reverberation time if measured

with non-standard temperatures or relative humidities. The

second di culty at 4,000 Hz is that a typical dodecahedron

loudspeaker (with a diameter in the order of 400mm)

3.5. Source Directivity

A much less manageable diculty with objective

measurements concerns the source. An orchestra occupies

around 200m 2 of stage with instruments which each have a

164

M. BARRON: USING ISO 3382 TO GIVE RELIABLE RESULTS

becomes directional at this frequency. One can compensate

for this diculty by making several measurements with

diﬀerent orientations of the loudspeaker, but this is time-

consuming. Behler and M ¨ ller [12] have solved this

problem by using a separate 100mm diameter dodecahe-

dron for high frequency measurements.

The author [9,13,14] has tended to measure over ﬁve

octaves 125–2,000Hz and divide results into a bass region,

125–250 Hz, and a mid-frequency region, 500–2,000Hz.

The major diﬀerences between the bass and mid-frequency

are diﬀerent amounts and type of absorption (usually panel

vs. porous absorption) and that the bass frequencies are

aﬀected by the seat-dip eﬀect [2, pp. 19–21], for which the

frequency of maximum absorption lies within the two

octaves 125 and 250 Hz. Since individual octave measure-

ments are inﬂuenced by interference, either constructive or

destructive, except in the case of reverberation time it is

preferable to use averages of several octaves where

appropriate, as elaborated below.

This last point, about averaging octave results to reduce

interference eﬀects, does not apply in the case of most

computer simulation programmes, since they usually

ignore phase. To gain a result equivalent say to the

average of 500–2,000Hz, a computation at only 1,000Hz

may be suitable, as long as absorption coe cients etc. for

1,000Hz are the average of those for the frequency range in

question.

Some recommended frequency ranges for the diﬀerent

measures are included in Table 2 [15]. In the case of C 80 ,

the literature is limited, though Beranek and Schultz [16]

suggest that low frequencies do not contribute to clarity.

Since low-frequency early sound levels are strongly

inﬂuenced by the seat-dip eﬀect, whose magnitude varies

depending on seat location [13], whereas clarity is aﬀected

by other concerns, measuring C 80 over the range 500–

2,000Hz looks appropriate. On the other hand, for the early

lateral fraction there is signiﬁcant evidence that low

frequencies are important whereas high frequencies are

less so [17–20].

Regarding quoting measured values, individual mean

octave values can be used for RT and EDT. Individual EDT

values at diﬀerent positions can be quoted as mid-

frequency and bass frequency values. For C 80 the three-

octave mean of 500–2,000Hz can be used, while for LF the

four-octave mean of 125–1,000 Hz is appropriate. Whether

G should be calculated as a full-frequency average or split

between bass and mid-frequencies depends on the situation

in hand.

With ﬁve or more measures at ﬁve or six octaves, a lot of

data is generated. To make sense of this plethora of

numbers, some averaging is appropriate.

5.1. Measurement Scatter

One issue relevant to measurement accuracy is the

variation of objective quantities for small movements of

the microphone [21]. To our knowledge, theoretical values

of scatter only exist for reverberation time and total sound

level (G). The measured scatter of reﬂected sound level in a

model diﬀuse space is illustrated in Fig. 1. This topic is

discussed in [22], which quotes the approximate theoretical

standard deviation proposed by Lubman and Schroeder:

s G r ¼

4:34

ð1Þ

B T

6:9

1 þ

where B is the bandwidth and T the reverberation time.

This relationship provides a justiﬁcation for averaging

objective measures over several octaves, as mentioned

above.

To experience subjective changes within concert halls

it is usually necessary to move to a seat position several

metres away, whereas objective data often changes

between one seat and its neighbour. The relevant compar-

ison is with subjective diﬀerence limen; limen listed by

Bork [23] were included in Table 2. It is thus tempting to

average over a few or many measurement positions.

Averaging over blocks of audience seating has its place,

classical theory

best-fit line

revised theory

Source-receiver distance, m

Fig. 1 Measured values of reﬂected sound level for 240

source-receiver pairs in a scale model diﬀuse space.

Reﬂected sound level values are relative to the direct

sound level at 10m from the source. The notional

model scale factor is 1:25; source-receiver distances

are full-size equivalents as is the 500 Hz octave at

which the measurements were made. Included on the

ﬁgure are predicted values according to classical and

revised theory [9].

5. AVERAGING OF RESULTS OF

DIFFERENT MEASUREMENT POSITIONS

The standard speciﬁes that for halls with more than

2,000 seats at least 10 seat positions should be measured.

165

Acoust. Sci. & Tech. 26, 2 (2005)

but the extreme of presenting hall average values needs

careful assessment.

measures are usually conducted with the hall unoccupied,

whereas we are generally interested in concert conditions

with a full audience. A measurement of occupied rever-

beration time is generally made soon after a hall opens. It is

therefore appropriate to make corrections to other objective

measures for the change in reverberation time between the

unoccupied and occupied state. In the case of measure-

ments in scale models, the model reverberation time is

often slightly diﬀerent to that expected in the real hall,

probably because of small inaccuracies in the absorption

coecients of model materials. Again corrections are

appropriate for RT change. The measures aﬀected are EDT,

C 80 and relative level. Spatial measures such as LF and

IACC are little aﬀected by RT change.

The following authors have suggested techniques for

correcting for reverberation time: Hidaka et al. [7], Bradley

[11] and Barron [2, p. 419]. Though the methods have

diﬀerent origins, they are likely to give similar results. The

accuracy of the correction will decrease for larger

reverberation time diﬀerences. When seating is well-

upholstered, the RT change is modest and corrections are

likely to be reasonably accurate. The discussion of criteria

below will relate to ﬁgures following a reverberation time

correction.

5.2. Whole-Hall Averages

Beranek in his extensive survey of world concert halls

[24] presents mean values for objective quantities and uses

these to establish guidelines for concert hall design. While

this limits the quantity of data one has to process, it tends to

make extracting signiﬁcant results dicult because the

means do not diﬀer much. For example, a comparison of

two British halls, one much liked and the other with

disappointing acoustics, is barely predicted by their mean

objective behaviour [5].

When an objective quantity varies only little through-

out a hall (or more precisely little more than the subjective

diﬀerence limen), then it is appropriate to talk about the

value of the quantity for the hall and work with the mean of

the objective quantity. This is generally the case with

reverberation time. If however the quantity varies signiﬁ-

cantly throughout the hall (relative to the diﬀerence limen),

then the mean value is only representative of a small

number of audience locations. The mean value says

nothing about the spread of values, nothing about the best

and worst seats. Most of the newer measures vary

signiﬁcantly within halls and there is usually a lot of

overlap between measured values of quantities such as C 80

between two halls. A satisfactory mean value only

indicates a tendency for the quantity/quality to be

satisfactory.

To give a simple example, the total relative sound level

(G) typically varies about 4.0 dB between seats 10 and

40m from the source in a large concert hall. This

corresponds to about four diﬀerence limen. The mean

value may apply to seat locations towards the rear of the

Stalls but says little about the level in the highest balcony.

It says little about the acoustic designer’s ability to provide

good acoustics throughout the auditorium. Including in

presented data the variation within halls is more dicult

but it indicates the full variety to be experienced within

individual concert halls.

One way of deriving a single ﬁgure of merit for halls

for a quantity such as C 80 is to quote the fraction of values

measured at diﬀerent positions in a hall which fall within

the preferred range for that quantity.

Where mean values are used for EDT and C 80 , they are

probably best calculated without including seats under

overhangs (Sect. 6.8). Mean values of LF need not exclude

these seats.

6.2. Simple Range Criteria

Many authors have provided recommended ranges for

objective measures for concert hall listening. This author’s

recommendations based on objective and subjective sur-

veys of concert halls are given in Table 3 [2, p. 61].

Under balcony overhangs, measured values tend to be

lower (for EDT and G) and higher (for C 80 ), as discussed in

Sect. 6.8 below. It may be argued that slightly less stringent

criteria are applied for EDT and C 80 in these locations.

When objective data for a hall is displayed, there is a strong

case for treating values from overhung seats separately. For

the same reason, mean values are probably better taken

omitting these locations — though the value of whole hall

mean objective measures has been questioned in Sect. 5.2

above.

The following discusses more elaborate ways in which

objective data can be analysed. In all cases apart from

reverberation time, values vary throughout auditoria. By

Table 3 Recommended ranges for objective measures

for concert halls.

Measure Acceptable range

Reverberation time (RT) 1:8 RT 2:2 s

Early decay time (EDT) 1:8 EDT 2:2 s

Early-to-late sound index (C 80 ) 2 C 80 þ2 dB

Early lateral energy fraction (LF)

6. INTERPRETATION OF OBJECTIVE

MEASURES

6.1. Correction for Reverberation Time Change

For tests in full-size halls, measurements of the newer

0:1 LF 0:35

Total relative sound level (G)

G > 0 dB (see text)

166

Barron - Using the standard on objective measures for concert auditoria, ISO 3382, to give reliable results.pdf

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: