Handbook for Sound Engineers

Designing for Speech Intelligibility 1401

8. Comparison for high-frequency losses caused by
   long cable runs or excess atmospheric absorption.

To these may be added abnormal or deficient room
acoustics particularly if exhibiting strong reflections or
focusing.
Equalization is a thorny subject, with many different
views being expressed as to how it should be carried out
and what it can and cannot achieve. Suffice it to say that
equalization can make a significant improvement to
both the intelligibility and clarity of a sound system.
In some cases the improvements are dramatic—par-
ticularly when considering not so much the intelligibil-
ity per se but associated factors such as ease of listening
and fatigue. The essential point is that there is no one
universal curve or equalization technique that suits all
systems all of the time.
Two examples of this are given below. Fig. 36-29
shows the curves before and after equalization of a dis-
tributed loudspeaker system in a highly reverberant
church. The anechoic response of the loudspeakers in
question is reasonably flat and well extended at high fre-
quencies. Because the measurement (listening) position
is beyond the critical distance, the reverberant field
dominates and it is the total acoustic power radiated into
the space that determines the overall response.

The power response of the loudspeaker in question is
not flat but falls off with increasing frequency. (This is
the normal trend for cone-based devices but some
exhibit sharper roll-offs than others.) This, coupled with
the longer reverberation time at lower frequencies due to
the heavy stone construction of the building, results in
an overemphasis at low and lower midfrequencies. The
peak at around 400 Hz is due to a combination of power
response, mutual coupling of loudspeakers, and bound-
ary interaction effects. The resultant response causes
considerable loss of potential intelligibility as high-fre-
quency consonants are lost. Equalizing the system as
shown by the solid curve improved the clarity and intel-

ligibility significantly resulting in an improvement of

some 15%.

Fig. 36-30 shows a widely quoted equalization curve

for speech systems. This has been found to work well

for distributed systems in reverberant spaces, but it is

only a guideline and should not be regarded too rigor-

ously. Loudspeakers that have a better balanced power

response that more closely follows the on-axis fre-

quency response will exhibit less high-frequency

roll-off and will generally allow a more extended

high-frequency equalization curve.

An example of this is shown in Fig. 36-31. This is the

response of a distributed loudspeaker system employing

two-way enclosures in a reflective but well-controlled

acoustic environment. In this case, rolling off the

high-frequency response would be wholly inappropriate

and would degrade the clarity of the system.

Adding bass to a sound system may make it sound

impressive but will do nothing for the clarity and intelli-

gibility. Indeed, in general, such an approach will actu-

ally reduce the intelligibility and clarity particularly in

reverberant spaces. Where music as well as speech need

to be played through a system, different paths with differ-

ent equalization settings should be employed so that the

different requirements of each signal can be optimized.

36.12 Talker Articulation and Rate of Delivery

Whereas the sound system designer has some control or

at least influence over many of the physical parameters

that affect the potential intelligibility of a sound system,

an area where no such control exists is that of the person

using the microphone. Some talkers naturally articulate

Figure 36-29. Frequency response of a sound system in
reverberant church before and after equalization.

Loudspeaker system response in reverberant

space with and without equalization

125160200250351400500630800 1k1.25k1.6k2k2.5k3.1k4k5k6.3k8k10k12.5k

Frequency–Hz

dB No EQ

With EQ

5

0



5



10



15



20



25



30

Figure 36-30. Typical response guideline curve for speech

reinforcement systems.

2 dB 200 Hz to 1 kHz Rolloff 3 dB/octave

above 1 kHz

31.5 63 125 250 500 1k 2k 4k 8k 16k

Third octave center frequency–Hz

Select rolloff appropriate to

SPL–dB loudspeakers used