Handbook for Sound Engineers

1392 Chapter 36

systems in low but noisy spaces. The AI method is not
able to take account of room reverberation or reflections.

In many situations, the background noise may not be
steady but vary over time. This is particularly the case in
many industrial complexes or transportation concourses.
Spectator sports also can exhibit highly variable crowd
noise levels dependent on the action at any given time.
Fig. 36-15 shows a typical noise profile for an under-
ground train station. Peaks of 90 dBA plus were
recorded as the trains moved in and out of the platforms.
A PA system would therefore need to generate at least
96–100 dBA in order to achieve an appropriate SNR at
these times.

Noise sensing and automatic level control are essen-
tial under such conditions, otherwise during the rela-
tively quiet periods when ambient levels drop down to

around only 66 dBA, significant startle may be caused

by such high-level announcements. (A better solution is

to store announcements and wait for the regularly occur-

ring quieter periods rather than trying to compete with

the background noise all the time.)

Spectator sports can also create wildly fluctuating

noise levels. Again if possible, announcements should

be made during the quieter periods, the levels of which

can be best determined by a statistical analysis of the

crowd behavior at the particular venue in question. Fig.

36-16 shows part of the time history for a soccer match.

Note that peak values in excess of 110 dBA can occur.

Figure 36-12. Comparison of speech intelligibility test for-
mats as a function of Articulation Index (AI) and Speech
Transmission Index (STI).

100

12 

9 

6 3 0 +3 +6 +9 +12 +15 +18

90

80

70

60

50

40

30

20

10

0

Sentences

(known to

listeners)

Percent of syllables, words, or sentences

understood correctly

Speech level minus noise level dBA

Sentences (first

presentation to

listeners)

PB words - single syllable

(1000 difference words)

(Write down answers)

Nonsense syllables

(100 difference syllables)

Rhyme tests

Note: These relations are

approximate. They depend upon

type of material and skill of

talkers and listeners.

B. Speech Transmission Index (STI)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0























Intelligibility score (%)

0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.0ST

Bad Poor Fair Good Excellent

A. Articulation Index (AI)

Test vocabulary

limited to 3z

PB words

Test vocabulary limited

to 256 PB words

Logatoms (CVC)

Diagnostic rhyme test

Short sentences

Numbers andspel alphabet

Figure 36-13. Effect of signal to noise ratio on %Alcons.

Figure 36-14. Spectral analysis of speech-to-noise ratio (one

octave band resolution).

100

50

20

10

5

2

10.0

9.0

8.0

7.0

6.0

5.0

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.75

0.50

RT60—

s

10 5 0 5 10 15 20 25 30

% ALcons

SNR—dB

20

40

60

80

125 250 500 1k 2k 4k 8k

Frequency–Hz

dB

Speech

Noise

125 250 500 1k 2k 4k 8k

Frequency–Hz

Speech

2040 Noise

60

80

dB

A. Good SNR. Good intelligibility—

noise does not mask speech.

B. Poor SNR. Poor intelligibility—note complete

masking of high frequency speech sounds by

the background noise.