1402 Chapter 36
better than others and so the resultant broadcast
announcements are also inherently clearer.
However, it must not be forgotten that even good
talkers cause some loss of potential intelligibility. Peutz,
for example, found that good talkers produced 2–3%
additional Alcons loss over and above that caused by the
system and local environment. Poor talkers can produce
additional losses of up to 12.5%. It is therefore important
to design in some element of safety margin into a sound
system in order to compensate for such potential losses.
The rate at which a person speaks over a sound sys-
tem is also an important factor—particularly in rever-
berant spaces. Considerable improvement in intel-
ligibility can be achieved by making announcements at a
slightly slower than normal rate in acoustically difficult
environments such as large churches, empty arenas,
gymnasiums, or other untreated venues.
Training announcers or users on how to use the sys-
tem and how to speak into a microphone can make a sig-
nificant improvement. The need for proper training can
not be overstated but is frequently an area that is often
ignored. Prerecorded messages loaded into high-quality,
wide bandwidth digital stores can overcome certain
aspects of the problem.
For highly reverberant spaces, the speech rate needs
to be slowed down from the normal rate of speak-
ing—e.g., from around five syllables per second down
to about three syllables per second. This can be very dif-
ficult to do under normal operating conditions but care-
fully rehearsed, slower recordings can be very effective.
Equally, the author has found that feeding back a
slightly delayed or reverberated signal of the person
speaking (e.g., via headphones or an earpiece) can be a
very effective way of slowing down the rate of
speech—though this has to be carefully controlled and
set up, as too much delay can become off putting and
counterproductive.
Research has shown that intelligibility is improved
when the lips of the talker can be seen. At low levels of
intelligibility (e.g., 0.3 to 0.4 AI [Articulation Index])
visual contact can produce improvements of up to 50%.
Even with reasonably good intelligibility (e.g., 0.7 to 0.8
AI) improvements of up to 10% have been observed.
This suggests that paging and emergency voice alarm
systems may have a more difficult task than speech rein-
forcement systems where additional visual cues are gen-
erally also present.36.13 Summary of Intelligibility Optimization
TechniquesThe following tips should hopefully prove useful in
optimizing sound system intelligibility or act as a cata-
lyst for other ideas and design strategies. Although
some would appear very basic, it is remarkable how
many systems could be improved with just a minor
adjustment or simple redesign.- Aim the loudspeakers at the listeners and keep as
much sound off the walls and ceiling—particularly in
reverberant spaces or where long path echoes can be
created. - Provide a direct line of sight between the loudspeaker
and listener. - Minimize the distance between the loudspeaker(s)
and listener. - Insure adequate system bandwidth, extending from a
minimum of 250 Hz to 6 kHz and preferably
8–10 kHz.
- Avoid frequency response anomalies and correct
unavoidable peaks with appropriate equalization. - Try to avoid mounting loudspeakers in corners.
- Avoid long path delays (>45 ms). Use electronic
signal delays to overcome such problems where loud-
speaker spacing >20 ft/6 m (30 ft/9 m max). - Use directional loudspeakers in reverberant spaces to
optimize potential D/R ratios. (Use models exhibiting
smoothly controlled and reasonably flat or a gently
sloping power response if possible.) - Minimize direct field coverage variations. Remember
that variations of as little as 3 dB can be detrimental
in highly reverberant spaces. - Insure speech SNR is at least 6 dBA and preferably
10 dBA.
- Use automatic noise level sensing and gain adjustment
to optimize SNR where background noise is variable.
Figure 36-31. Frequency response curve of distributed
two-way loudspeaker system in reflective but well-con-
trolled acoustic space.
30 100 1000 10,000ML
SS
ATransfer Function Mag - dB volts/volts (0.30 oct)log frequency–Hz
15
20
25
30
35
40
45
50
55
60
65
autoLATERAL LOUDSPEAKER SYSTEM WITH POSITIVE D/A (AFTER EQ)