1400 Chapter 36
Where listeners are free to move about—e.g., on a
concourse or in a shopping mall—it may be possible to
have a greater variation in coverage and hence intelligi-
bility. However, with a seated audience or spectators in
an enclosed space, it is essential to minimize seat to seat
variations. In critical applications, variations in cover-
age may need to be held within 3 dB in the 2 kHz and
4 kHz octave bands. This is a stringent and often costly
requirement. To put this into perspective consider the
following example: assume a given space has an RT 60 of
2.5 s. Calculation shows that on-axis to the loudspeaker
at a given distance gives a value of 10%Alcons—an
acceptable value. However going off-axis or to a posi-
tion where the direct sound reduces by just 3 dB will
result in a predicted %Alcons of 20%—an unacceptable
value, see Fig. 36-22. This shows that it is vital to
remember off-axis positions as well as the on-axis ones
when carrying intelligibility predictions and system
designs. Particularly when it is considered that in many
applications, the potential intelligibility will be further
degraded by the presence of background noise— even
when it is not the primary factor.
36.10 Computer Modeling and Intelligibility
Prediction
Computer modeling and the current state of the art are
discussed in depth in Chapters 9 and 35 and so will only
be briefly mentioned here. The ability to accurately
predict the direct and reverberant sound fields and
compute the complex reflection sequences that occur at
any given point are truly remarkable advances in sound
system design. As we have seen, calculation of intelligi-
bility from the statistical sound fields alone is not suffi-
ciently accurate for today’s needs—particularly with
respect to distributed sound systems. The computation
of the reflection sequence and hence the impulse
response at a point allows far more complex analyses to
be carried out including predictions of the early-to-late
sound field ratios and the direct calculation of STI. (It
should be noted that some of the current simpler
programs and many of the earlier prediction programs,
although purportedly providing a prediction of STI, in
fact base this on a statistical %Alcons calculation and
convert the resulting value to Rasti. The accuracy of the
result value is therefore highly questionable.)
Some program, however, are capable of highly accu-
rate prediction, particularly as the precision of the loud-
speaker data increases to -octave bandwidths and 10
degrees or better angular resolution. Also as the comput-
ing power continually increases, greater reflection
sequence lengths and orders can be more practically
accommodated and hence more accurate reflection field
data can be calculated. The main restriction currently is
not the mathematical accuracy of the model itself, but
the time and effort required to build it in the first place.
For many schemes this is simply not economically via-
ble so some form of simple prediction routine, to at least
insure that the proposed system will achieve roughly the
right order of magnitude of intelligibility, is still
required.36.11 EqualizationIt is surprising how many sound systems are still
installed either with no or totally inadequate equaliza-
tion facilities. Yet the major variations in frequency
response (both perceived and measured) that systems
exhibit when normally installed can have significant
effect on the resultant intelligibility and clarity. Equally
many systems after they have been equalized often
sound worse than they did before.
This is primarily due to a lack of understanding on
behalf of the person carrying out the task. There would
appear to have been very little research carried out on
the effects of equalization on intelligibility. The author
has noted improvements of up to 15–20% on some sys-
tems, but otherwise the improvements that can be
gained are not well publicized.
There are probably about eight main causes of the
frequency response anomalies generally observed prior
to equalizing a sound system. Assuming that the loud-
speaker(s) has a reasonably flat and well-controlled
response to begin with these are:- Local boundary interactions, Fig. 36-7.
- Mutual coupling or interference between loud-
speakers. - Missynchronization of units in a cluster.
- Incorrectly acoustically loaded loudspeaker, (e.g., a
ceiling loudspeaker in too small a back box and/or
a coupled cavity). - Irregular (poorly balanced) sound power character-
istic interacting with reverberation and reflection
characteristics of the space. - Inadequate coverage, resulting in dominant rever-
berant sound off-axis. - Excitation of dominant room modes (Eigen tones).
(These may not appear as large irregularities in the
frequency response but subjectively can be very
audible and intrusive.)
(^1) » 3