Handbook for Sound Engineers

Designing for Speech Intelligibility 1395

same TEF data (see Section 36-13 on measuring intelli-

gibility) are:

•%Alcons 4.2%.

• Equivalent rasti 0.68.


• C50 9.9 dB.

An opportunity to exchange the high Q device for an
almost omnidirectional, low Q loudspeaker was taken
and found to have a profound effect on the perceived
intelligibility and the resulting ETC. This is shown in
Fig. 36-20, which presents an obviously very different
curve and pattern of sound arrivals. Clearly there is far
more excitation of the reflected and reverberant sound
fields. The D/R ratio is now 4 dB (a degradation of
some 12 dB) and other computed data is:

•%Alcons is now only 13%.

• C50 has been reduced to 3.6 dB.


• Equivalent rasti to 0.48.

All the indicators and even a visual inspection of the

graphs show there to be a significant reduction in the

potential intelligibility.

While visual inspection of an ETC can be very

enlightening, it can also at times be misleading. Take for

example the curve shown in Fig. 36-21. At first glance

this resembles the ETC for the low Q device shown

above and might suggest low intelligibility since no

clear direct sound component is visible. However,

densely distributed ceiling loudspeaker systems in a

controlled environment do not work in the same way as

point source systems in large spaces. In the former, the

object is to provide a dense, short path length sound

arrival sequence, from multiple nearby sources. The

early reflection density will be high and in well-con-

trolled rooms, the later arriving reflections and reverber-

ant field will be attenuated. This results in smooth

coverage and high intelligibility. In the case shown in

Fig. 36-21, the RT 60 was 1.2 s and the resulting C50 was

+2.6 dB and the Rasti was 0.68, results both indicating

high intelligibility, which indeed was the case.

It should not be forgotten that whereas it may well be

possible to produce high intelligibility in a localized

area, even in a highly reverberant space, extending the

coverage to a greater area will always result in reduced

intelligibility at this point, as the number of required

sources (additional loudspeakers) to accomplish the task

increases. This is primarily due to the resulting increase

in acoustic power fed into the reverberant field (i.e.,

increase in reverberant sound level) often referred to as

the loudspeaker system n factor.

36.7.1 Intelligibility Prediction—Statistical Methods

While it is relatively trivial to accurately calculate the

direct and reverberant sound field components by

Figure 36-19. ETC of a high Q (highly directional) loud-
speaker in reverberant church.

Figure 36-20. ETC of low Q (omnidirectional) loudspeaker
in a reverberant church.

Figure 36-21. ETC of distributed ceiling loudspeaker system

in an acoustically well-controlled room.

5.0

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

45.0

auto

0.0 200 400 600 800 1000

Time–ms

Filtered energy-time curve–dB (2000 Hz, 1.00 oct)