The design of Allison Loudspeakers follows research by Roy Allison on the
room-loudspeaker interaction, which shows how reflected impedance from the
room boundaries increases a conventional loudspeaker’s power output at very
low frequencies, but decreases it significantly in the middle-bass range.
The mechanism by which this effect occurs can be understood as follows:
Consider a typical box loudspeaker system positioned in a room so that its
woofer cone is about two feet from each of the three nearest room surfaces,
say, the floor and two intersecting walls. When the speaker is radiating a
very low frequency the cone moves relatively slowly and over a relatively
long distance. If the radiated frequency is 40 Hz., for example, it takes
1/40th second (25 milliseconds) for the cone to execute one complete forward
and backward cycle. Each half of the cycle takes 12.5 milliseconds.
As the cone begins a forward movement it generates the start of a
compression wave. This impulse travels at the speed of sound (1,130 feet
per second) to each of the three room boundaries and is reflected back
toward the woofer cone, arriving there some 3.4 milliseconds after it left,
while the woofer is still generating the compression half of the sound
cycle. The reflected waves increase the instantaneous pressure seen by the
woofer and enable it to radiate more power than it could in free space---a
maximum of 9 dB more power at extremely low frequencies, for which the
reflected pressure is virtually in perfect phase coincidence with the
woofer’s motion.
But as the woofer tries to radiate higher frequencies, it must reverse its
motion more quickly. At 140 Hz., for example, the cone reverses direction
every 3.5 milliseconds. It begins its inward half cycle of motion
(attempting to create a rarefaction) just as the three compression wave
reflections begin to arrive back from the room boundaries two feet away. In
this case the reflected pressure is completely out of phase with the cone
motion, decreasing its radiation efficiency some 11 dB below the anechoic
output. That is the worse case: a 20 dB variation in power output (from
+9dB to -11dB) when the woofer is equidistant from the three nearest room
surfaces, from a loudspeaker system which measures flat in an anechoic
chamber.
Usually the boundaries are not equally distant from the woofer and the
effect is not as intense. Typically, the variation in power delivered by
the speaker to a listening room is 6 to 12 dB within the woofer range.
These effects simply do not exist in anechoic chambers, where loudspeakers
are commonly tested because there are no reflections from the chamber walls.
Measurements made in "live" rooms are complicated by the standing wave
resonances. Consequently a room’s influence on the actual power output of a
loudspeaker system, as a factor separate from other room effects, has never
been well understood.
An uncontrolled variation in system response of this magnitude would be
considered intolerable if it originated in an amplifier or in the laser of a
CD player. But it is just as audible when it originates in a loudspeaker.
If it could be eliminated or if its severity could at least be reduced
appreciably, an improvement should be expected in the accuracy of the
reproduced sound field.
Distortion and Clarity
Smooth, flat acoustic power output vs. frequency is necessary for resolution
of fine detail in reproduced music. Resolution is a property often called
transparency or clarity and without it no illusion of reality is possible.
But a flat power output curve alone is not sufficient to achieve this
quality. At least two other performance factors are just as important: low
distortion and wide dynamic range.
A loudspeaker with an appreciable amount of nonlinear distortion generates
audible spurious tones in addition to those it is supposed to reproduce.
Some of the "extra" sounds alter the original balance of harmonic overtones,
thus modifying the music’s timbre. Others are harmonically unrelated to the
original tones; their effect is to add new frequency components which amount
to a dissonant noise, masking subtle details of the desired program
material. Very low distortion at normal listening level is, therefore,
essential for clarity.
A good recording contains occasional peak sound levels that are 10 to 15 dB
higher than the average. If this dynamic range is to be reproduced
faithfully by the loudspeaker system, it must be able to respond linearly to
such peak inputs without any obvious strain or effort. Consequently, the
speaker’s ability to handle very high power input levels is of importance.