Uniform derivation of optimum conditions for speech in rooms

(1)

Publisher’s version / Version de l'éditeur:

Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n’arrivez pas à les repérer, communiquez avec nous à PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca.

Questions? Contact the NRC Publications Archive team at

PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. If you wish to email the authors directly, please see the first page of the publication for their contact information.

https://publications-cnrc.canada.ca/fra/droits

L’accès à ce site Web et l’utilisation de son contenu sont assujettis aux conditions présentées dans le site

LISEZ CES CONDITIONS ATTENTIVEMENT AVANT D’UTILISER CE SITE WEB.

Building Research Note, 1985-11

READ THESE TERMS AND CONDITIONS CAREFULLY BEFORE USING THIS WEBSITE.

https://nrc-publications.canada.ca/eng/copyright

NRC Publications Archive Record / Notice des Archives des publications du CNRC :

https://nrc-publications.canada.ca/eng/view/object/?id=8afe9439-61b0-46bb-b283-389ad4403405 https://publications-cnrc.canada.ca/fra/voir/objet/?id=8afe9439-61b0-46bb-b283-389ad4403405

NRC Publications Archive

Archives des publications du CNRC

This publication could be one of several versions: author’s original, accepted manuscript or the publisher’s version. / La version de cette publication peut être l’une des suivantes : la version prépublication de l’auteur, la version acceptée du manuscrit ou la version de l’éditeur.

For the publisher’s version, please access the DOI link below./ Pour consulter la version de l’éditeur, utilisez le lien DOI ci-dessous.

https://doi.org/10.4224/40000478

Access and use of this website and the material on it are subject to the Terms and Conditions set forth at Uniform derivation of optimum conditions for speech in rooms

(2)

Ser TH1 w2 no. 239 c. 2 BISG

National Research Conseil national

19

Council Canada de recherches Canada Division of Division des

Building Research recherches

en

Gtirnent

Building

Research

Note

Uniform Derivation

of

Optimum

Conditions

for Speech

in

Rooms

by

J.S. Brad

ley

BRN

239

(3)

UNIFORM DERIVATION OF OPTIMUM CONDITIONS FOR SPEECH

I N

ROOMS

ANALYZED

by

J.S.

Bradley

Noise and Vibration Section Division of Building Research

BRN

239

E S N 0701-5232

Ottawa, November 1985

(4)

TABLE

OF CONTENTS

ABSTiUCT,k.E

'sdM,

~&

. . .

1 INTRODUCTION

. . .

2

PROCEDURE

. . . * . . .

3 RESULTS

. . .

4

Criteria from Single Measures

. . .

4

Combinations of S/N(A) and

RT

. . .

7

Optimum combinations of background level and

RT

. . .

11

CONCLUSIONS

. . .

17

(5)

ABSTRACT

The data from two previous studies of speech intelligibility in rooms are used to derive values of acoustical measures that would correspond to optimum conditions for speech in rooms. Optimum values of the A-weighted s ignal-to-noise ratio, the articulation index, the speech transmission index, and the 0.080-second useful-to-detrimental ratio were first determined as the point beyond which no increase in mean speech intelligibility score would be expected. Combinations of optimum reverberation time and background noise level were

calculated from equal useful- to-detrimental ratio contours. These optimum reverberation time and background noise levels are presented as simple design contours.

L'auteur utilise les donnkes de

deux

etudes antbrieures sur I'intelligibilitd de la parole dans

les salles pour determiner des valeurs de mesures acoustiques qui correspondent aux con-

ditions optimums de transmission de la parole en salle. Les valeurs optimums du rapport signal-bruit de valeur pondCrke A, de l'indice d'articulation, de l'indice de transmission de la parole et du rapport utilenuisible de 0,080 seconde ont d'abord Ct4 dbtermin6es

cornme constituant le point au-delk duquel on ne peut esp6rer augmenter l'indice moyen

d'intelligibilitk de la parole. Les cornbinaisons du temps

de

rCverbQation et du niveau

de bruit de fond optirnums ont &ti! calculkes k partir des 6quicontours du rapport utile- nufsible, Ces temps

de

rCverbkration et niveaux de bruit de fond optimums sont prhentk

(6)

INTRODUCTION

A large number of rooms are intended

for

speech communication between one talker and

a group of listeners. Such rooms vary in size from clssroons and small meeting room to large auditoria. It is essential that such rooms be designed so that the listeners experience

the highest possible level of speech intelligibility. Poorly designed classrooms may lead

to decreased levels of academic achievement, while unsatisfactrrry speech intelligibility in other rooms may lead to a Iass of revenue to the owners. In

all

cases poor conditions for speech

will lead

to less precise and more strained communication between talker

and

listener,

A number of acoustical quantities may

be

used to define ideal conditions for speech

in

rooms. Traditionally, the signal-to-noise ratio and room =oust ics aspects of the problem

have

been considered separately

in

terms of a maximum background noise level

and

an optimum reverberation time

(RT)

.

More recently, two types of measures have

been

devised that combine both aspects of the problem into one quantity. One of these, useful-to-

detrimental sound ratios, was first developed by Lochner and Burger [I1 after considerable experimental study of the response of the hearing system for conditions with a variety of reflection sequences, as are

found in

rooms- Recently useful-bdetrimentd ratios have

been evaluated in both the original form and in an even more successful simpIified form [2,3].

The

other type of measure, the speech transmission jnda (STI) is more arbitrary

in

its derivation and is calculated from modulation t r a d e r functions, and speech and noise levels 141.

There is presently little information concerning optimum conditions for speech in terms of the two new types of measures,

A

range of valum for optimum

RT

and maximum

back-

ground levels can be found in various textbooks. Unfortunately it is usnally difficult to trace the origin of these values, so that their validity cannot always be thoroughly evalu- ated, and the interpretation of such values is not always clear.

For

example, background noise level criteria, can be 'maximum tolerable levels', one step frum disaster, or 'ideal maxima', low enough t o provide truly optimum conditions. The difference could easily correspond to a range of 10 to 15 dB(A)

What is now needed is a unified procedure for deriving optimum conditions for speech in rooms, including a complete description of how the final values are obtained.

This

report includes such a derivation based

on

the combined data of two- previous studies of

speech

intelligibility in rooms [P,3]. Optimum values were determined for both of the newer types of quantities, as

well

as for the articulation index (AI) and A-weighted signal-twnoise ratios (SIN (A)). Optimum combmatiam of

RT

and A-weighted background level (BG(A)) were derived from usefu 1-to-detrimental ratios, as these most successfully combine

both

room acoustics and signal-to-noise ratio aspects of

the

overall problem. Optimum d u e s for

ST1

and Uso can be presented as single numbers, Ideal maximum background levels are presented as functions of both the speech source

level

and room volume, while optimum

(7)

PROCEDURE

The analyses in this paper were performed by combining the data from two previous studies in which acoustical ~rleasurements were obtained from the computer processing of pistol shot recordings. A large number of acoustical quantities were calculat-ed for comparison with speech intelligibility test _scores_obtainedat _the_same_locationsin the same rooms. Using the combined data provided a broader range of acoustical conditions on which to

base

the

derivation of optimum conditions. The data included measurements at 80 source

-

receiver combinations in 15 room with volumes from 253 to 20,000 m3 and

RT

values

from 0.39 to 3.8 seconds. Although the data were taken from two different studies, the two sets of data exhibited similar overall trends 131.

The

speech inteIligibility scores were produced using a Fairbanks rhyme test with the

test words embedded in a carrier phrase. The tests were performed using tape recorded material reproduced at known source levels through a loudspeaker with directional prop- erties similar to a human speaker. For each source - receiver combination, four different speech source Ievels were

used. Thus with

four source levels and 80 measurement positions, acousticaf measurements and group-averaged speech intelligibility scores were obtained for 320 conditions.

Acoustical measures were calculated in the standard octave bands from 125 to 8000 Hz. They included the reverberation time

(RT)

and Cso (the early-to late-arriving sound ratio with a 0.080-second early time limit). Background noise levels were measured from one-

minute recordings at each receiver position and

the

speech levels at each position were calculated from the source level, the reverberation time, the source - receiver distance,

and

the source directionality. 'Uao values, the useful-t* detrimental sound ratios with a

0.08&second early sound limit, were calculated from the corresponding early-to-Iate ratio (Cso) and

the

speech and noise levels at each position.

--

10 10g{C~0/(1+ (Cso

+

1)

r , / I a ) )

where C80,

the

linear early-twlate ratio, is given

by the

following:

where I, / I 8 is the ratio of A-weighted steady state background noise and speech intensities. Modulation transfer functions were also calculated from the recorded pulse responses using a technique described by Schroeder 151.

By

combining these values with speech and back- ground noise levels according to the procedure proposed by Houtgast et al. [4], ST1 values were produced. Further details of the measurement techniques can be obtained from the published accounts of the original studies [2,3,6].

(8)

Criteria from Single Measures-

Four independent acoustical quantities were first considered. Optimum conditions were defined as the point at which _100%_speechintelligibility _{was reached by}the _mean_trend

of the rhyme test results or where

the

mean trend showed no furkher increase in speech intelligibility score. Others have chosen lower criteria and there is necessarily some abi- trariness to such a limit. However, with a simple test such as the

Fairbanks

rhyme test,

the

mean trend

should

approach 100% under optimum conditions.

Figures 1 and 2 plot speech intelligibility scores

(SI)

vmus the articulation index (AI) and the A-weighted signal-to-noise ratio (S/

N(A)

)

respectively.

Both

are signal-&noise ratio type quantities and do not

include

the

influence of room acoustics factors. Figure

1

shows no substantial improvement

in

SI above an

AT

value of approximately 0.9.

Similarly,

Figure 2 indicates

that above

15 dB(A) S/N(A) there is no

further

improvement

in

SI

scares. Ignoring room acoustics factors, one can conclude that an

A1

of 0.9 or m SIN( A) of 15 dB(A) or greater represent optimum conditions for speech-

The A1

or S/N(A) values could

be

combined

with

RT

values to produce a compound predictor

that

would relate

SI

scores

to

_a_combined_{signal-twnoise ratio}_and_{room acoustics}

measure.

This

was done in two previous studies [2,3] using multiple regression analyses.

The

disadvantage of this approach is that without some other knowledge,

the

form of the combination of terms

is

arbitrary, and the resulting prediction equation is influenced by

the

range

and

distribution of

each

variable

in

the

available data. It is much more satisfactory to use a measure that combiies

both

aspects of the problem

in

_a_form_{based on}

independent previous research. Us&-to-detrimentd sound ratios, based on consider able laboratory research by Lochner and Burger, are certainly such a measure. The speech transmission index is also a m u r e

that

has received extensive evaluation as a predictor of speech intelligibility by its developers in the Netherlands. Thus ST1 and Uso represent

two _cambinations_of_{both a}_signaI-t_o-noise_{ratio and}_a_room_{acoustics measure From}

_which

ideal conditions for speech in rooms can be examined more completely.

Figures 3 and 4 plot measured

SI

scores against

ST1

and 1

kHz

Uso values, respective& From Figure 3 maximum speech intelligibility corresponds t o an ST1 of app~oximately 0.55,

while

Figure 4 indicates that optbum conditions occur for a 1 kHz UBo of

about

+4-0 dB. Table 1 s u r m d m s the obtained optimum values of these m u s t i d measures, as well as those for the 2

kHz Uso

and

the

1

kHz

Um

values.

The

2

kHz Uso

is

included

because it was faund to

be

more strongly related to speech intelligibility scores than

the

other measures.

U50

is the usehl-tt~ detrimental ratio derived from an early-%*late ratio

with

a 0.050-second early sound limit, and is included because the 0.05O-second early time limit

has

been used for speech by a number of other researchers.

(9)

20

0 0. 2 0.4 0 . 6 0.8 1.0

A R T I C U L A T I O M I N D E X

Figure 1 Measured speech intelligibility scores versus articulation index and best fit mean trend.

-15 - 1 0 -5 0 5 I 0 15 20

S I M I A

1,

d B ( A 1

Figure 2 Measured speech intelligibility scores versus A-weighted signal-to-noise ratio and

(10)

Figure 3 Measured speech intelligibility scores versus ST1 values and

best fit

mean trend.

- 2 0 - 1 5 - 1 0 - 5

a

op

d 6

Figure 4 Measured speech intelligibility scores versus 1 kHz USo values and best fit mean

(11)

TABLE f Optimum values for independent acoustical measures 0 _p_t_imum_Value 0.9 +15 dB(A) 0.55 +4.0 dB t2.0 dB t1.0 dB

An optimum S/N(A) of 15 dB(A) has been found in other recent studies 171, and an A1 value of 0.9 is generally considered to represent excellent conditions for speech. Houtgast 171 has labelled an ST1 of 0.6 as 'good'; this is only slightly larger than the optimum of 0.55 from this study-

Combinations of S/N(A) and

RT

Clearly the best way to evaluate a room for speech is to measure UHO or ST1 values directly. Until equipment to do

this

is readily available and in widespread use, it is still of interest to rdate an optimum

Uso

or ST1 value to related combinations of S /N(A) and

RT

values. Also an understanding of how S/N(A) and

RT

values combine to produce optimum conditions is valuable to better understand the inherent trade-offs in the design process. Finally, the values of S/N(A) and

RT

that are found to lead to optimum conditions c a n be compared to established values in the literature.

Houtgast et al. havedemonstrated how

ST1

can

be

calculated from S/N(A) and RT values and have published equal ST1 contours on a plot of S/N(A) versus

RT

[4]. These are reproduced as the dashed lines on Figure 5 for ST1 values of 0.4, 0.6, and 0.8. Thus all

combinations of S/N(AJ and

RT

that lie on one of these contours should correspond to equal speech intelligibility and the contours illustrate the trade-off between S/N(A) and

RT

that can be used in designing to achieve a particular level of speech intelligibility. As given in

Eq.

(I), Us0 c a n

be

calculated from Ggo and S/N(A) values.

To

associate Uso values with combinations of S/N(AJ and

RT

values it is necessary to relate Cso values to

RT

values. This can be done approximately by assuming ran ideal continuous exponential decay. Then the linear Cso can be calculated as foiIows:

With this approximate relationship one can calculate equal Uso contours from combina- tions of S/N(A) and 1

k H z

RT

values. These are also shown on Figure 5 for Uso values

(12)

of -2, 0. _2. _{4, and}₆_dB._Again_all_points_ala~ig_each_{contour are expected}_to_represent conditions of equal speech intelligibility. From the results in Figure 4, it was concluded that a

Us0

of +4 dB corresponded to optimum conditions for speech and therefore the

+

4 dB

Use

contour on Figure 5 represents optimum combinations of

RT

and S/N(A) val-

ues. It is clear that the two sets of contours in Figure 5 are different

and

cannot both represent equd speech intelligibility contours. Comparison of the two sets of contours in

this figure shows that the STT-based contours are more restrictive on

RT

values but

less

restrictive on S/N(A) value. compared to the

Use-based

contours. Of the three equd ST1 contours, the 0.6 ST1 contour comes closest to reprsenting just optimum conditions. It

suggests that a maximum

RT

of 6.9 seconds is tolerable for large signal-t-noise ratios or at the other extreme, similarly optimum conditions should exist for an S/N[A) value of 5 dB(A) and a O.Zsecond

RT.

This

5

dB(A)

S/N(A) value seems unreasonably small

for truly optimum conditions, particularly since overall a 15 dB (A) SIN( A) was found to be

ideal.

The

+4.0

dB

iJBo contour approaches a maximum

RT

of 0.85 seconds for large

S/N

(A) d u e s ,

which

k

similar

to

the

0.6

ST1

contour. However at very short

RT

values (0.2 seconds),

the

miniurn

S/N(A)

value is 10

-(A),

which & much larger than that of the

ST1

contour.

0 0.5 1 . 0 1 . 5 2.0 2- 5

R T ( A T 1 k H z ) ,

s

(13)

For a number of reasons it was judged better to derive optimum cornbinatior~s of

RT

and S/N(A) values from equal Uso contours. As discussed above, the equal Uao contours of Figure 5 seem intuitively more reasonable than the equal ST1 contours. The ns~ful-to- detrimental sound ratio concept is Founded on extensive laboratory research by Lochner and Burger that

established

a thorough understanding of

how the

hearing system reponds to speech

in

rooms. From analyses with the present combined data, UgO was found to be

as

good as or better

than all

other acoustical measures tested as a predictor of speech intelligibility scura. Finally ITso values are conceptually more satisfying in that one can readily comprehend the beneficial results

of

_stronger_early_{reflections and}_speech_levels_or

the detrimental effects of increased late reflections or background noise.

Both sets

of

contours on Figure 5 were derived assuming

ideal

continuous exponential decays. Figure

6

plots measured Cso value versus

RT

values dong with the prediction curve of Eq. (3) for an ideal exponentid decay. This figure illustrates

the

approxhak nature

of

_this_prediction,_showing_that_{errors of up}_to₅_{dB between measured}_{and predicted}

Cso values secured. Of course, the fact that one cannot accuratdy predict Cso values from

RT

value done further d m n s t r a h

the

need for newer measures such as UHO. Barron 181 has derived a more complete method for p~edicting Cso values from the

RT,

the room volume, and the source - receiver distance, which predicted the Cso values in the present study with an RMS error

of

1-20 dB, compared to the RMS error of 1.75

dB fur

the exponential theory.

If

one uses a

fixed

typical source - receiver distance, the estimation of CEO values can be

based

on

both

RT

and

the

morn volume.

Figure 7 presents equal USo contours on a plane of

S

/N(

A) versus 1

kHz

RT

values similar to Figure 5 but with

RT

values estimated from Cso values using Barroa's

Eq.

(7) for a

morn volume of 1 0m3. A source - receiver distance of 10 m was used. The contours of Figure 7 for a

rmm

volume of 1000 m3 are very similar ta those of Figure 5 arid for practical applications either

wmld be

satisfactory.

Figure 8 illustrates

the

influence of room volume on calculated +4 dB Uso contours. In this case the CgO

valus

were estimated using Barron b Eq. (7). Ruom volume has a Iarge effect on the calculated contours only for quite large moms (greater than 3000 m3 in volume]. Figure 8

implies

that

for larger rooms, longer reverberation times can

be

accepted as giving optimum resulk- for speech. This would agree qualitatively

with

current practice such as Knudsen

and

Harris's

[Q]

plot

of

_optimum_{reverberation}_{time for speech}_as_a_function_of room volume.

The

increased optimum reverberation time relative to a 1000 m3 room indicated by Figure 8 can

be

calcuIated quite accurately

by

an equation derived

from

this figure.

This

equation was obtained

by

fitting a regrasion line to the

ET

values versus room volume

for

an S/N(A) of 15 dB(A) on Figure 8.

The

resulting equation is:

where

V

is the room volume in m3, and

RT,

is the optimum

RT

for a 1000 m3 room. This. equation thus provides a simple procedure for determining opfimum

RT

values for speech

(14)

0 0.5 1.0

1.5

2.0 2.5 3 . 0 3.5 4 . 0 R E V E R B E R A T I O N TIME. s

Figure 6 Measured 1

kHz

Cso values versus

RT

md predict ion curve for ideal exponential

decays.

1 . 0 1.5 R T ( A T 1 kHz),

s

(15)

0 . 5 1.0 R T ( A T 1 kHz),

s

Figure 8 Uso equal to +4 dB contours based on Barron's equation for various room vol- umes in m3

as a function of room volume. Its validity depends on the concept of USo as an optimum combination of both a room acoustics and a signal-to-noise ratio measure, and on Barron's procedure for relating Cs0 values to

RT

values and room volumes. Both of these points were, of course, verified with respect to the present combined data.

Optimum combinations of badground level and

RT

As

a final result, it is desirable to derive both optimum

RT

and optimum A-weighted background noise levels (BG(A)J as a function of both the speech source level and the room volume. Equation (4) permits one to calculate optimum RT's once an optimum

RT

for a 1000 rn3 room has been obtained. One can calculate optimum BG(A) values in a manner similar to

that

used to obtain Figures 5 and 7, after

f i ~ t

calculating expected

speech Ievels in the room.

To

do this, representative speech source Ievels are required. Data by Pearsons et al.

[lo]

represent an extensive modern evaluation of this problem. They determined long term average speech levels for talkers at various levels of vocal

effort, dong with the standard deviation of the groups of talkers. They produced such data separately for males, females, and children. Table 2 gives their group mean values and standard deviations for males and females. It was judged that in small rooms (300 m3), subjects should

be

able to communicate with a 'normal' voice level, but in larger rooms subjects would reasonably expect to use a 'raised' voice level. It was also decided that

(16)

optimum canditims should be optimum for the large majority of talkers. Thus it

is

not satisfactory t o design only for the mean talker source level. Accordingly, in subsequent calcutatiom of optimum conditions,

the

mean female speech source levels less 1.0 standard deviation, as obtained from Pearsons et al., were used.

TABLE

2

Mean

speech source

levels

Males

Females

Mean Level Standard Mean Level Stmdard

Vocal Effort dB(A)

,

1 m Deviation, dB(A) dB(A), 1 m Deviation,

dB(A)

Casual 52

Normal 58

Raised 65

Loud 76

Shout 89

Knowing the speech levels at a distance

of

1.0 m f h m

the

source given in Table 2, one can

calculate the reverberant field speech levels in a room of given volume and reverberation time. Combining this with the calculations used to

obtain the

equal

Usa

contours of Figure 7, one can calculate equal

Uso

contours on a plane of BG(A) versus

RT

for various combinations of room volume

and

speech source level, Figures 9 and 10 repmsent two of many possible plots

from

this

procedure, Figure 9

_{is fm}

a 300 m3

room

and a 'normal' speech source level (55-4 dB(A)). Figure 10 is for s 1000 m3 room with a 'raised' voice s m e

level

(63-4

dB(A)).

From these plots one can arrive at combinations of BG(A) and

RT

that would

be

expetted to produce optimum conditions for speech from the +4 dB Us0 contour.

To avoid presenting a large number of such plots it is desirable t o derive a shnplified scheme for predicting optimum conditions from these plots.

From

Figure 10 one can select as a starting point

the

condition of 0.7 seconds

RT

and 37.5 dB(A) background level on the +4 dB

UaO

contour. Other points

on

this contour would also

be

expected to correspond to optimum

conditions,

but for practical reason6 this paint approximates a, good design

goal.

For much

higher

RT

values

on this contour considerably lower background levels are

necessary; fur much

lower

RT

values little is

gained in terms

of

less stringent backgmund

levels, and considerable

expense

would be

required f ~ r

the

absorptive treatmmt

to obtain

these shorter

RT

values.

In

addition, it

w w l d

be

difficult

to

add

enough absorptive t

At-

ment to greatly reduce the reverberation time without destroying necessary strong early rdections.

Thus

the combined values of _0.7_seconds

RT

_and_37.5

dB(A)

_represent₈_rea- sonable

initial

design point for this case of a _{1CMM m3}

room

_and_a'raised' voice _source

level.

Such a design point repmwnts the largest

RT

d u e that does not significantly increase

the

restriction

on

acceptable

background

noise leveb.

The corresponding design points for other room v o l and ~ speech ~ source levels can

readily be determined. The optimum RT

values

for other room volume c a n

be obtained

from Eq. (4). These are plotted versus the logarithm of the room volume on Figure 11.

(17)

0 0 - 5 1 . 0 1 . 5 2 . 0

R T ,

s

Figure

9 Equal Uso contours for a 300 m3 room and a 'normal' speech source level.

0 0.5 i . 0 1.5 2 . 0

R T .

s

(18)

Far comparison Knudsea and Harris's optimum Iine for speech is also given. Although tbe new results are not 1inea.r with the logarithm of the volume as are Knudsen and Harris's, for most room volumes the two approaches agree within about 0.1 second.

Once the optimum

RT

values far various room volumes had been determined using Eq. (41, the optimum BG(A) values were then read oKthe plots of BG(A) versus

RT

from the +4 dB Uso contours. These optimum BG(A) values are plotted versus room volume in Figure 12. The upper

T i

is for a 'raised' voice source level (63-4 dB(A)), while the lowest

line

is for

a 'normal' voice source

level

(55-4

dB(A)j. The

intermediate lines corrwpond t o 2

dg(A)

increments in speech source level.

Thus

Figure

12

summarizes

the

calculated optimum background noise

levels

for opt h u m speech conditions for combinations

of room volume

and

speech source level between 'normal' and 'raised' voice source

levels.

Combiming Figures I1 and 12 one can thus arrive at optimum

RT

and BG(AJ values for speech for a wide range of room volumes and speech source levels.

The results of Figure 12 can be reduced t o one design contour.

In

smaUa rooms, such _as classrmms and small meeting rooms, one expects t o communicate with a 'n0rm~1' voice

level, but

in

larger rooms one would expect to use a stronger vocal effort.

This

gradually increasing vocal effort with room siae can

be

appmxXhnated by drawing a horizontal line on the contours of Figure 12, as has been done on Figure 13. A maximum BG(A) value of 35

dB

(A) has been selected because it approximates the value calculated for a 300 m3 room and a 'normal' vocal &rt. It was

judged

that no more than a 'raised' vocd effort

was acceptable in any room volume; thus the design contour wen tually follows

the

'raised' vocal effort lime for larger room volumes. For very large roam volumes the design contour requires v q low background levels, or alternatively, electronically amplified speech. Figures 11 and 13 provide an easy-to-use procedure for determining optimum conditions for speech

in

_twms_of

_{the well}

_{known acoustical}_measures_of

RT

_{and backpund}_noise level.

Their

derivation has been presented in detail so that possible Zitations may be filly appreciated, axid so

that

where judgements have been

made,

others may make

smdl

adjustments to suit their own particular situation.

The optimum

RT

values of Figure 11 have b m compand with the well

known

optbum values of-Knudsen

and

Harris, and while the new results suggest smaller d a m for in- termediate room volumes, the

actual

difference~ are quite s d l .

Although

the results of some research studies have suggested that speech intelIiiiElity continues to improve as

RT

is

decreased to sera (113, such laboratory studies ignore the beneficial effects of

RT.

Thus the

contours of Figures 9 and 10 peak at

RT

mlu-

greater

than zero,

because some reverberation is

desirable

to provide increased speech levels.

Also

arguments based on reverberation ignore

the

value of reflecting surfaces to provide strong early reflections.

It

is more difficult t o compare

the

optimum BG[A) values of _this_study

_with

_published results because previous

data do

not explicitly specify BG(A) values as a function of room volume.

The

present resalts are derived from speech inteligibility tests involving subjects repraentative of am average audience.

It

is known

that

certain gmups

of

liiteners

(19)

1 i I I I

-

_-

-

P R E S E N T S T U D Y

-

--

-

K N U D S E N A N D H A R R I S [9]

-

_-

-

_-

-

_-

-

4 * / - # C / C

-

/ C

-

/ / d / #

-

_-

-

I I I I ₁+

Figure 11 Optimum reverberation time versus room volume.

3 00 1 0 0 0 3 000 1 0 000 3 0 000

V O L U M E , m3

(20)

Figure timum background noise level design contour.

require improved conditions for similarly optimum inteliigibility [11,12]. In particular younger children (under about 13 years old), older adults, and hearing- impaired subjects require more restricted conditiom, including Eowm

BG

(A) value and more controlled reverberation times.

To

accommodate the younger and older liiteners with

normal

hearing one s hodd probably consider a further 5

dB

[A) reduction in background levels. For special spaces such as large theatres, it

is

_{not always}_{adequate to}_have_satisfactory_{intelligibility} with a 'raised' voice level; it may be

desirable

for lower voice levels t o be intelligible sa that desired dramatic effects c a n be preserved. Recommended levels of 29 to 46 dB(A) were found in

the

literature for clas~rooms 131. The Swedish National Building Code requim that sf eady-state noise levels

in

classrooms from continuous n o d should not

exceed 30

dB(A)-

Berenek

(13) recommends 2130

dB(A)

for large theatres, up to 42

dB(A)

for _{small auditaria, and}₃₈_to₄₇

_dB(A)

_for_{meeting rooms}

_and

_classrooms._Thus_many

design criteria

in

the literature, according to the present study, would not provide optimum conditions for speech. The present results ~ e e m to differ most for smaller rooms, where much lower background levels are required. It is not possible to trace the source of

all

such criteria, or t o verify whetber these criteria are intended to _{produce truly}_optimum

conditions.

In

many cam^ they may represent no more than a good g u m at an upper bound, rather than an ideal goal,

(21)

(22)

9. Knudsen, V.O. and Harris, C.M. Acoustical Designing in .4rchitecture. John 12-iley

& Sons, New York, 194 (1965).

10. Pearsons, K.S., Bennett,

R.L.,

and

Fidell,

5. Speech Levels in Various Noise Envi- ronments. Bolt Beranek and Sewman h c . Report to USEPA, Pg-270-053, Canoga Park, CA (May 19771,

11. Nabelek,