GOLDBERG AND MÄKIVIRTA
AUTOMATED IN-SITU EQUALISATION
AES 114TH CONVENTION, AMSTERDAM, THE NETHERLANDS, 2003 MARCH 22-25
9
well damped having a T
30
of about 0.2 s across the
whole frequency range. The Schroeder curves indicate
a smooth, linear and very short decay in all octave
bands except at 62.5 Hz where low frequency room
effects frequently appear even in good rooms. There is
a high level (–6 dB) reflection about 3 ms after the
direct sound due to the floor reflection. Another high
level reflection (–9 dB) is seen about 5 ms after the
direct sound, and this is from the sidewall.
The frequency response before equalisation (Figure 6)
shows relatively flat mid and high frequencies above
1 kHz. There is a gain increase around 35 Hz.
Between 100 Hz and 1 kHz large notches and gain
increases affect the magnitude response to produce a
very non-flat response even after third octave
smoothing. Wide notches can be seen in the smoothed
response around 100 Hz, 200 Hz and 600 Hz. Gain
increases can be seen between these notches. The
rapid roll-off in measurements above 23 kHz is due to
the measurement system’s anti-aliasing filter, not the
loudspeaker’s natural response.
To determine room equalizer settings the default
frequency range was considered (–3 dB low frequency
cut-off to 15 kHz). Settings recommended by the
optimisation algorithm are shown in Table 12 and the
resulting room equaliser filter response in Figure 7.
Table 12. Case study 2, room response control
settings.
Control Setting
Treble Level
–3 dB
Midrange Level
–4 dB
Bass Level
0 dB
Bass Tilt
–6 dB
Bass Roll-off
0 dB
After equalisation, the response (Figure 8) is closer to
the target of a flat response. The broadband rms
deviation between the original responses (2.8 dB) and
the optimised responses (2.6 dB) shows a reduction of
7% (0.2 dB).
The gain increase at 35 Hz has been flattened due to
the bass tilt setting. The midrange/treble balance is
flatter due to the 1 dB decrease in the midrange level
relative to the treble level. The equalisation was
unable to improve the bass and midrange notches and
gain increases because the room response controls are
not designed to compensate for this type of acoustic
problem.
After equalisation the midrange gain increase around
300-400 Hz has become more prominent because the
bass response has been flattened. The subjective
impact of this change should be tested to see if the
resulting balance is disturbing.
In the 100 Hz to 1 kHz region there are cancellations
that should be removed by damping reflection
sources. The gain increase around 500 Hz is caused by
the floor reflection and should also be damped or the
loudspeaker cabinet raised up to reduced the level of
the floor reflection. The sidewall reflections should
also be damped to reduce their effects. Such narrow-
band problems remain because they cannot be fixed
using the room response controls.
4.3. Other Case Studies
Further detailed examples of the optimisation
algorithm performance in different acoustic condi-
tions, with smaller and larger loudspeaker systems and
with differing target responses can be found in [42].
g
q
y
p
g
g
Frequency [Hz]
10
100
1,000
10,000
P
re
s
. [P
a
] /
[v
o
lts
],
d
B
15
10
5
0
-5
-10
Figure 6. Case study 2, original frequency response, (light curve) unsmoothed magnitude response, (dark curve) the
third octave smoothed response.
Level, [dB]
