Inside Our Speaker
Tests
The initial step in evaluating a speaker is a battery of
measurements made in the National Research Council of Canada's anechoic chamber in Ottawa.
The speaker is rotated through many angles in several planes, and separate frequency
response curves recorded for a large number of positions. These are stored in computer
memory, and can be averaged or otherwise combined in a variety of ways. Our test reports
include a selection of these measurements. To view the measurements simply click the
graphic you see below that will be placed near the beginning of each product review.
If you click the graphic above you will see real
measurements
of Axiom Audio's Millennia M3Ti loudspeakers
The first three sets of curves should be read in
conjunction with one another, as their similarities or differences say a lot about a
speaker's performance.
The top set of measurements in Figure 1 contains three
separate curves to represent three characteristics of a speaker. The top curve is the
on-axis response, recorded directly in front of the speaker. If you actually listened to
the speaker in the anechoic chamber, that's what you'd hear. In real rooms, it's unlikely
that you would always be perfectly on axis, so a more realistic figure would take into
account some variation. The middle curve shows the response some 15 degrees off-axis. This
rather narrow beam is called the "listening window" and the on-axis and
15-degree off-axis measurements are averaged into a single curve in Figure 2. It
represents the direct sound you will hear from the particular speaker, seated more or less
in front of it.
Back in the top set of curves in Figure 1, the bottom curve
is taken at 30 degrees off axis, and with the 15-degree curves more or less represents the
spectral balance radiated generally into the listening room's reverberant field.
That is only part of the picture, however, as the acoustic
characteristics of a listening room play a large part in the overall sonic impression a
speaker gives, and to a considerable extent these are dependent on the sound radiated to
the sides.
The bottom set of curves in Figure 1 combines measurements
made 45, 60 and 75 degrees off axis, and these generally bounce off the walls and floor
close to the speaker. These near-field reflections combine with the direct sound in most
cases, and can have a profound effect on tonal balance; as it is harder to maintain
linearity far off axis, the radiation in this direction is often the downfall of a speaker
even if it has good response in the listening window.
All the curves contribute to a speaker's character, and
they help us to predict how a particular example will behave in a given acoustic space --
anomalies that show up in all curves will almost certainly be audible; ones that only
occur in certain directions may average out.
Another aspect of speaker performance is included in the
caption for Figure 1, although it is a separate test. A speaker's sensitivity shows the
acoustic output it achieves for a given input. Using frequencies from 300Hz to 3kHz -- the
most sensitive part of our hearing and where the vast majority of music is located -- an
average figure is calculated for an input signal of 2.83 volts. That rather odd number is
used because if the speaker was a perfectly resistive 8-ohm load, that would be one watt
of power being fed to the voice coil. The measurement is made 1 meter in front of the
speaker, which is why that is sometimes specified as 1 watt at 1 meter.
The sensitivity is stated in "dB SPL" -- decibels
referred to a standard sound pressure level. Low to mid 80s are considered insensitive;
high 80s to low 90s are quite sensitive; anything above about 95 dB SPL is very sensitive.
This all has to do with quantity of sound and has no effect on the sound quality.
Many speakers behave themselves when playing at low levels
but not when they are pushed. In Figure 3, the total harmonic distortion of a speaker is
shown, usually for two levels unless the higher level would damage the speaker. The first
curve shows distortion when the speaker is fed enough power to produce 90dB SPL at mid
frequencies; the second curve shows distortion for 95dB SPL. If the curves are similar,
the speaker should handle high power with relative ease; if the 95dB curve is
significantly higher than the 90dB, it suggests that the speaker will generally be happier
with lower levels (and if there is no dashed curve, it means that the speaker could not be
pushed to 95dB without severe distress).
Figure 4, upper curve, shows a speaker's impedance across
the frequency spectrum. For the most part, this will not affect sound quality, but if the
curve drops to a very low value (4 ohms or less, say) across much of the spectrum, the
speaker may present too difficult a load for some amplifiers to cope with; speakers with
low impedances should not be connected in parallel, as that lowers the total impedance
even further. Finally, the lower curve in Figure 4 represents the speaker's electrical
phase -- the relationship between parts of the spectrum. This should be as smooth as
possible.
This group of measurements has been chosen because it has
proved to be consistent with what we hear in our speaker listening tests. These are
performed in a specially designed listening room, which exhibits characteristics typical
of a normal domestic living room. The panel auditions speakers four at a time in a number
of rounds that assures each speaker has been in every position, each listener has occupied
every seat, and every speaker has been up against every other speaker in the batch. An
extensive selection of music is used, selected to show up individual characteristics of
speaker performance.
These double-blind listening tests are performed before the
panel has seen the results of the formal measurements, and include at least one speaker
not in the batch, to help prevent a tendency to "predict" what we are listening
to. Only when all tests have been completed are the various elements of the program
brought together. Then, if there are any inconsistencies between measurements and
auditioning, we go back to find out what's the problem. Only when everything fits is a
review ready to be written up.
...Ian G. Masters
ian@mastersonaudio.com
|
