Guidance on computer prediction models to calculate
the Speech Transmission Index for BB93
Version 1.0
1. Introduction
Section 1 of Building Bulletin 93 contains performance standards for speech intelligibility in open-plan
spaces. The guidance in Section 1.1.7 of Building Bulletin 93 states that a computer prediction model
should be used to calculate the Speech Transmission Index (STI) in the open-plan space. This
document contains additional guidance on the use of these computer models and suitable parameters
that should be used in these models.
2. Computer modelling
This Section contains guidance on the calculations that are required and suitable input data for speech
levels and speech directivity patterns. It also contains guidance on the calculation of background
noise, barriers, absorption and scattering.
The computer model should use octave bands from 125 Hz to 8k Hz, separate absorption and
scattering coefficients at these frequencies, separate background noise levels at these frequencies,
and calculate STI in accordance with BS EN 60268-16:1998 Sound system equipment – Part 16:
Objective rating of speech intelligibility by speech transmission index.
2.1 Calculations required
Section 1.1.7 of Building Bulletin 93 (BB93) states that for each open-plan space:
“The expected open-plan layout and activity plan should be agreed as the basis on which
compliance with BB93 can be demonstrated to the Building Control Body.”
The calculations described below are required for each class in this agreed open-plan layout and
activity plan; the full details that make up this layout and plan are described in BB93 Section 1.1.7.
The agreed layout should include areas where students are sitting. Guidance on defining these as
surfaces in the computer model is given in Section 2.4. Groups of students and their desks/tables
should be represented by single ‘boxed-in’ planes. Other ‘small’ objects (<2m 2 surface area) such as
the teacher and their desk should not be included as surfaces in the model.
Once the basic geometry of an open-plan space has been modelled, single point locations and
characteristics of sound sources and receivers (i.e. teachers and students) need to be defined.
Locations of teachers and students should be taken from the agreed open-plan layout.
The single points used for the source or receiver in the computer model should use the heights
prescribed in Section 1.1.7 of BB93. Guidance on human voice levels and directivities to be used in
the model is given in Section 2.2. The orientation of sources is discussed below.
For the calculation of STI, receivers are omni-directional.
Before calculating STI values, the “background noise level” should be established using the guidance
in Section 2.3.
1
Section 1.1.7 of BB93 states
“[the STI] performance standard applies to speech transmitted from teacher to student,
student to teacher and student to student.”
The following calculations are required:
(1) Teacher to student
This is to ensure that oral presentations by the teacher, with a raised voice, are intelligible to
the whole class (or, depending on the activity plan, a group). In this situation, the teacher is
the source and every student is a receiver – these locations are defined in the expected openplan layout. The teacher should be orientated to face the centre of the class or group. The
results should be expressed as the minimum and maximum STI values predicted for all
student locations.
(2) Student to teacher
This is to ensure that responses made by students to the teacher, in a raised voice, are
intelligible to the teacher. At least three separate calculations are required, one for each of
three different student source locations. The three student source locations should be chosen
to represent students at the rear of the class and one of them should be the furthest student
from the teacher. The teacher location and orientation remains the same as for (1). The
students are orientated so that they are facing the teacher. The results should be expressed
as the minimum and maximum STI values from the three (or more) student source positions to
the teacher.
(3) Student to student
This is to ensure that responses made by students to the teacher, in a raised voice, are
intelligible to other students in the class. At least three separate calculations are required one
for each of three different student source locations. All other student locations are receiver
positions. The source locations should be distributed in the student area with at least one at
the front and another at the back of the class. The speaking student should be orientated so
that they are facing the teacher. The results should be expressed as the minimum and
maximum STI values from the three (or more) student source positions to all the other student
positions in the class.
(4) Student to student
This is to ensure that conversation within groups of students, at normal voice levels, is
intelligible for the students in that group. Communication at very short distances (i.e. 1 m)
should be intelligible and should not need checking by calculation. The required calculations
for this situation are to assess speech intelligibility over slightly greater distances, typical of the
furthest distance between students in a group. Three calculations are required for each group:
one for each of three pairs of students selected to represent the furthest distances across
groups. The source and receiver in each pair should be a minimum of 3 m apart. If the openplan layout does not show groups or only has very small groups, then arbitrary locations in the
student area should be used that are 3 m apart. The results should be expressed as the
minimum and maximum STI values for the three (or more) pairs of students.
For each calculation run, the computer model should generate an impulse response at each receiver
for each octave band from 125 Hz to 8k Hz. From all of these impulse responses the model should
then calculate the Speech Transmission Index, STI, using 14 modulation frequencies in accordance
with BS EN 60268-16:1998. The STI predictions should be rounded to the nearest 0.05 step.
A summary of the computer model input for the different calculations is given in Table 1.
2
Minimum
number
of source
positions
Source(s)
Receiver(s)
Speech level
(see 2.2.1)
Height (m)
Facing centre
of student
group
ANSI raised voice
effort
0.8 (n)
1.0 (p)
1.2 (s)
Height (m)
Location
Orientation
Teacher to student
1
1.65
As defined in the
agreed open-plan
layout
Student to teacher
3
0.8 (n)
1.0 (p)
1.2 (s)
The furthest
student locations
from the teacher
Facing teacher
ANSI raised voice
effort
1.65
Student to student
3
0.8 (n)
1.0 (p)
1.2 (s)
At typical student
locations
Facing teacher
ANSI raised voice
effort
0.8 (n)
1.0 (p)
1.2 (s)
Facing other
student
ANSI normal voice
effort
0.8 (n)
1.0 (p)
1.2 (s)
0.8 (n)
At typical student
1.0 (p)
locations
1.2 (s)
n – nursery school, p – primary school, s – secondary school
Student to student
3
Table 1: Summary of calculations and calculation parameters required in the four situations
3
Location
One at every student
location in the group
as defined in the
agreed open-plan
layout
As defined in the
agreed open-plan
layout
One at every other
student location in
the group
as defined in the
agreed open-plan
layout
At least 3m from the
student speaking
2.2
Sound levels and directivity of speech sources
This section gives guidance on two aspects of modelling the human voice: sound levels and
source directivity. Both aspects should be applied to all speech sources being used to
calculate the STI.
2.2.1
Speech levels
Speech levels are typically expressed as sound pressure levels at a distance of 1m in front of
the speaker’s lips in the free field. For the purpose of STI modelling for BB93, suitable values
for the seven octave bands 125 Hz to 8k Hz are defined in ANSI 3.5:1997 – Methods for
calculation of the speech intelligibility index.
Table 2 contains the sound pressure levels at 1m in free field that should be used for normal
and raised voice effort. These values have been averaged from data for male and female
speakers, and hence should be used in the model to represent either male or female
speakers. (N.B. The 125Hz octave band value is not defined in ANSI 3.5-1997, and therefore
its value has been chosen to be 6dB lower than the 250Hz octave band.)
SPL
(dB re 2.10-5 Pa)
Normal voice effort
Raised voice effort
125
51.2
55.5
Octave band centre frequency (Hz)
250
500
1k
2k
4k
57.2
59.8
53.5
48.8
43.8
61.5
65.6
62.3
56.8
51.3
8k
38.6
42.6
dB(A)
59.5
66.5
Table 2: Speech spectra in terms of the sound pressure level at a distance of 1m in front of
the speaker’s lips in the free-field
2.2.2
Speech directivity
The human voice does not radiate sound equally in all directions. Typically, most of the
energy is radiated to the front of a speaker, particularly at higher frequencies. The source
directivity should be modelled by means of a three-dimensional directivity pattern. Suitable
directivity data is contained in the appendix of IRC Research Report 104 "Detailed Directivity
of Sound Fields around Human Talkers" (http://www.nrc.ca/irc/ircpubs). Horizontal and
vertical polar plots should be derived from these data, with a resolution in azimuth and
elevation of 30˚ or better. Any missing elevation data can be derived by linear interpolation
from the IRC report. From these two-dimensional polar plots a suitable three-dimensional
directivity pattern can be obtained using elliptical interpolation. For every octave band, a
separate directivity pattern should be created.
(N.B. The 125Hz octave band directivity data are not described in the IRC report, and
therefore data for the 160Hz third octave band have been used.)
Table 3a lists the directivity levels in dB, relative to the zero axis, in the horizontal plane.
Table 3b lists the directivity levels in dB, relative to the zero axis, in the vertical plane.
Directivity files that can be used in commercial software will be made available on the DfES
website (www.teachernet.gov.uk/acoustics).
4
Azimuth (º)
0 (front)
30
60
90 (left/right)
120
150
180 (back)
125
0.0
-0.2
-0.7
-1.5
-2.3
-2.8
-3.0
250
0.0
-0.3
-1.2
-2.8
-4.2
-5.3
-5.9
Octave band centre frequency (Hz)
500
1k
2k
0.0
0.0
0.0
0.3
-0.7
0.3
0.2
-0.5
-1.8
-1.8
-0.3
-5.6
-3.8
-3.0
-6.3
-5.3
-7.1
-10.4
-6.1
-5.0
-10.9
4k
0.0
-0.6
-2.0
-5.5
-10.0
-14.4
-16.6
8k
0.0
-1.1
-4.4
-8.9
-13.3
-16.8
-18.9
Table 3a. Directivity levels for the human voice, relative to the zero axis, in the horizontal
plane.
Elevation (º)
0 (front)
30
60
90 (up)
120
150
180 (back)
210
240
270 (down)
300
330
125
0.0
0.3
-0.3
-1.7
-2.2
-2.5
-3.0
-3.3
-3.0
-2.2
-1.3
-0.4
250
0.0
-1.0
-2.3
-3.3
-3.8
-4.8
-5.9
-6.6
-4.7
-4.0
-3.2
-1.6
Octave band centre frequency (Hz)
500
1k
2k
0.0
0.0
0.0
-0.2
2.4
0.3
-1.7
1.6
-0.1
-3.7
-0.2
-3.4
-3.4
-3.6
-4.7
-3.9
-1.2
-7.8
-6.1
-5.0
-10.9
-7.6
-8.9
-13.0
-6.3
-7.5
-15.4
-2.8
-3.7
-10.7
0.7
0.1
-6.0
3.2
-2.6
-1.2
4k
0.0
-2.1
-2.3
-5.8
-8.1
-13.3
-16.6
-18.6
-17.6
-11.0
-4.4
-0.5
8k
0.0
0.5
-1.1
-4.6
-9.4
-16.6
-18.9
-18.5
-16.9
-9.5
-2.1
2.2
Table 3b. Directivity levels for the human voice, relative to the zero axis, in the vertical plane.
An example of horizontal and vertical polar plots of the human voice directivity can be seen in
Figure 1. The 3D directivity pattern derived from these two polar plots is shown in Figure 2.
Figure 1: Typical horizontal and vertical polar plots of the directivity of the human voice at
250 Hz.
5
Figure 2: 3D voice directivity pattern, obtained from the horizontal and vertical polar plots.
2.3 Calculation of overall noise level (background noise)
The prediction of STI relies on accurate and realistic estimation of the overall noise level
(referred to as the background noise) in the open-plan space.
The activity plan should be used to establish the overall noise level due to the combination of
the indoor ambient noise level, all activities in the open-plan space (including teaching and
study), and transmitted noise from adjacent spaces.
Noise from “all activities in the open-plan space” includes teaching and studying but excludes
speech from the “source” teacher or and student(s) for which the STI is being calculated.
Noise sources that must be included are:
 teachers’ speech from surrounding areas (raised voice)
 students working and talking in surrounding class areas
 noise produced by equipment used in the space (e.g. machine tools, CNC machines,
dust and fume extract equipment, compressors, computers, overhead projectors,
fume cupboards)
 students working quietly in surrounding areas
 students listening to the speaking teacher or student
The activity plan described in Section 1.1.7, Section 1, BB93 will define the sources (e.g.
number of students talking) to be considered in the open-plan area.
The background noise is required in octave bands from 125 Hz to 8k Hz.
For open-plan classrooms the background noise levels should be calculated using one of the
following methods:
1. Computer prediction model.
2. Classical statistical room acoustics.
Sound power levels that should be used to calculate the background noise are contained in
Table 4.
6
Sound power levels (dB re 10-12 W)
Open-plan space – general working
(per 15 students)
Dining space
(per 60 students)
Speech at normal level (per person)
Quiet student being addressed by the
teacher (per student)*
125
Octave band centre frequency (Hz)
250
500
1k
2k
4k
8k
62
62
62
62
57
52
47
61
65
69
69
61
51
40
60
66
69
62
57
53
50
30
32
32
30
28
26
20
Table 4: Sound power levels to be used to calculate the overall noise level in the open-plan
space
* This sound power level should be used to account for noise from ‘quiet’ working and bodily functions
(i.e. breathing).
Where there is spatial variation of the background noise level predicted across the receiver
locations but the STI calculation in the computer model only allows for constant background
noise levels, then:
 the highest background noise level should be used at all receiver positions; or
 the model can be run repeatedly so that each receiver position has its corresponding
noise level.
When predicting background noise it is important to consider separately each activity zone or
classroom, as each area in an open-plan design can be exposed to a different level of noise.
2.4 Calculation of barrier diffraction
In open-plan spaces, screens or barriers are often used to reduce the sound transmitted
along the direct sound path and to provide absorption in the space. Although a barrier may
block the line-of-sight, some sound will diffract around the edges of the barrier. Diffraction
occurs when a sound wave travels within close proximity of an edge and in general it is more
significant at lower frequencies than at higher frequencies.
When modelling barriers, it is essential that the position and orientation be detailed precisely;
the height of the barrier is of critical importance and should be included in the model to at
least an accuracy of  5 cm.
The absorption and scattering coefficients of the barrier should be included for the octave
bands from 125 Hz to 8k Hz. For lightweight barriers it may be important to include
transmission loss in these octave bands
The computer prediction model should include a diffraction option. The diffraction option
should account for the wavelength of the sound being modelled. Diffraction should be
possible to predict from at least the edges prescribed by the model as diffracting, whether the
edges are formed from screens, acoustic barrier or room surfaces.
The computer prediction model should be capable of predicting at least one diffraction effect
per sound path. Available methods of modelling diffraction currently include path length
difference, the secondary source method, and the geometric theory of diffraction, although
this topic should be investigated further.
Note that some computer models use the term "diffraction" where actually an approximation is
being made that simply introduces scattering. In such cases, this is not appropriate for
calculating sound transmission over barriers.
7
2.5 Absorption and scattering of seated students
In general it is not practical or necessary to model each person and item of furniture
separately. This can result in models with a large number of surfaces, which may take a long
time to calculate and can be less accurate than simpler models.
An area of floor occupied by people should be modelled as an acoustically absorbent plane,
typically about 1 m above the floor, with appropriate absorption and scattering coefficients.
This audience plane must be “boxed in” with vertical planes forming, effectively, the sides,
front and back of the audience area. These vertical planes should have the same absorption
and scattering coefficients as the audience plane.
There is a great deal of published data for absorption coefficients of the audience in auditoria
and elsewhere. These are generally quoted as the amount of acoustic absorption (Sabines)
per square metre of floorspace occupied by the people, although some older data is quoted in
Sabines per person. The coefficients are a function of the seating density, furniture and
configuration. For example, an adult audience sitting in rows of upholstered chairs with a
density of two people per square metre will be more absorbent than students sitting on
wooden chairs at desks in a classroom (typically with a density of about one student per
square metre). Indicative data for absorption coefficients is on the website
www.teachernet.gov.uk/acoustics.
Research is currently being carried out on scattering coefficients for areas of people but in the
absence of data it is appropriate to model scattering coefficients for people using a value of at
least 0.7 in all frequency bands.
2.6 General note on absorption coefficients
Historically, most computer modelling has been limited to the octave bands from 125 Hz to
4k Hz. This is because measurements of acoustic absorption have, for technical reasons,
been limited to this range and there is little published data for acoustic absorption coefficients
above 4k Hz. For calculation of STI, the range has to be extended to 8k Hz and in the
absence of any other data, it is reasonable to use the same coefficients at 8k Hz as at 4k Hz.
At these high frequencies, the acoustic absorption of the air becomes significant, especially in
larger rooms, and this can have a real effect on speech intelligibility over large distances. The
air absorption is a function of ambient temperature, humidity and air density (barometric
pressure) and the model should take this into account. Except in special circumstances, it is
appropriate to assume air at 20 C, 50 % relative humidity and a density of 1.2 kg/m3.
2.7 General note on scattering coefficients
Different modelling systems use different methods for calculating the effect of scattering. In
general, scattering coefficients of at least 0.3 should be used except for large flat, acoustically
reflective areas. Research is currently being carried out on scattering coefficients for areas of
people.
3. Presentation of results
The following information should be submitted with the results of the STI calculations.
1.
2.
3.
4.
5.
6.
7.
8.
Software and version
Acoustic consultant and Operator
Number of rays and length of impulse
Absorption and scattering coefficients used in model
Diagram of the wire frame model with source and receiver positions
Background noise levels in octave bands used for each scenario
Assumptions made in calculating overall noise levels
Assumptions made in choosing source and receiver positions for calculation
8