1620 Chapter 46
46.3.5.7 Estimate the Critical Distance DC
Critical distance is easy to estimate. A quick method
with adequate accuracy requires a sound level meter and
noise source. Ideally, the noise source should be band
limited, as critical distance is frequency dependent. The
2 kHz octave band is a good place to start when measur-
ing critical distance. Proceed as follows:
- Energize the room with pink noise in the desired
octave band from the sound source being
measured. The level should be at least 25 dB higher
than the background noise in the same octave band. - Using the sound level meter, take a reading near the
loudspeaker (about 1 m) and on-axis. At this
distance, the direct sound field will dominate the
measurement. - Move away from the loudspeaker while observing
the sound level meter. The sound level will fall off
as you move farther away. If you are in a room
with a reverberant sound field, at some distance the
meter reading will quit dropping. You have now
moved beyond critical distance. Measurements of
the direct field beyond this point will be a chal-
lenge for some types of analysis. Move back
toward the loudspeaker until the meter begins to
rise again. You are now entering a good region to
perform acoustic measurements on loudspeakers in
this environment. The above process provides an
estimate that is adequate for positioning a measure-
ment microphone for loudspeaker testing. With a
mic placement inside of critical distance, the direct
field is a more dominant feature on the impulse
response and a time window will be more effective
in removing room reflections.
At this point it is interesting to wander around the
room with the sound level meter and evaluate the
uniformity of the reverberant field. Rooms that are
reverberant by the classical definition will vary little in
sound level beyond critical distance when energized
with a continuous noise spectrum. Such spaces have
low internal sound absorption relative to their volume.
46.3.5.8 Common Factors to All Measurement Systems
Let’s assume that we wish to measure the impulse
response of a loudspeaker/room combination. While it
would not be practical to measure the response at every
seat, it is good measurement practice to measure at as
many seats as are required to prove the performance of
the system. Once the impulse response is properly
acquired, any number of postprocesses can be per-
formed on the data to extract information from it. Most
modern measurement systems make use of digital sam-
pling in acquiring the response of the system. The fun-
damentals and prerequisites are not unlike the
techniques used to make any digital recording, where
one must be concerned with the level of an event and its
time length. Some setup is required and some funda-
mentals are as follows:- The sampling rate must be fast enough to capture
the highest frequency component of interest. This
requires at least two samples of the highest
frequency component. If one wished to measure to
20 kHz, the required sample rate would need to be
at least 40 kHz. Most measurement systems sample
at 44.1 kHz or 48 kHz, more than sufficient for
acoustic measurements. - The time length of the measurement must be long
enough to allow the decaying energy curve to
flatten out into the room noise floor. Care must be
taken to not cut off the decaying energy, as this will
result in artifacts in the data, like a scratch on a
phonograph record. If the sampling rate is
44.1 kHz, then 44,100 samples must be collected
for each second of room decay. A 3-second room
would therefore require 44.1 × 1000 × 3 or 128,000
samples. A hand clap test is a good way to estimate
the decay time of the room and therefore the
required number of samples to fully capture it. The
time span of the measurement also determines the
lowest frequency that can be resolved from the
measured data, which is approximately the inverse
of the measurement length. The sampling rate can
be reduced to increase the sampling time to yield
better low-frequency information. The trade-off is
a reduction in the highest frequency that can be
measured, since the condition outlined in step one
may have been violated. - The measurement must hav a sufficient signal-to-
noise ratio to allow the decaying tail to be fully
observed. This often requires that the measure-
ment be repeated a number of times and the results
averaged. Using a dual-channel FFT or MLS, the
improvement in SNR will be 3 dB for each
doubling of the number of averages. Ten averages
is a good place to start, and this number can be
increased or decreased depending on the environ-
ment. The level of the test stimulus is also impor-
tant. Higher levels produce improved SNR, but can
also stress the loudspeaker. - Perform the test and observe the data. It should fill
the screen from top left to bottom right and be fully