THE ANECHOIC frequency response of a speaker tells a great deal about what
that speaker is capable of doing. However, none of us listens to a speaker
under anechoic conditions. Walls, floors, ceiling, and household furnishings
all form part of our acoustic environment. Many speaker systems are also designed
to use portions of a room to augment reproduction.
AUDIO provides a measurement specifically designed to indicate how the speaker will sound to you when you listen to it in a room. This is where the difficulty begins because two questions immediately present them selves: 1. what room shall we use, and 2. how should the measurement be made? Both of these considerations may be dealt with by considering some of the psychoacoustics of listening in a room. It has long been recognized that the first sound arrivals you hear substantially determine the quality of what we identify as the source of that sound. This is the "early sound" of architectural acoustics and amounts to the first few tens of milliseconds of sound arrival for a very large room.
Our ability to locate a source of sound based on the precedence effect is similarly associated with the first five to 50 milliseconds of arrival. Similarly, on the average, the pitch of a suddenly applied tone may be determined within 13 milliseconds independent of the value of that pitch. What we are referring to here are, of course, considerations based on dynamic pro gram material associated with real or apparent sources of sound for that material.
The later sound arrivals, which in a conventional room are due to sound scattering off various objects, tend to establish the ambience we normally think of as the "sound" of that room.
The earliest of these later arrivals yields the impression of reverberance, while later and later arrivals no longer fuse into a continuous sound pattern but dissociate into what we perceive as echoes. In many cases the net amount of sound can be greater for the reverberant sound than for the direct sound from a discrete source.
In that case, we are in what is known as the reverberant field. This still does not prevent our assignment of localization and quality for the apparent direct source of sound. You can quickly verify this by observing that you have no difficulty talking to a friend in a highly reverberant hall way-that is, until you block one ear and disable some of the psycho-acoustic mechanisms.
Our microphone is a single receptor analogous to that unblocked ear.
To measure the properties of room sound which may be best associated with quality, localization, and timbre, we therefore measure the spectrum of the early sound.
Our job is not complete, however, because we have to worry about the room and how it may differ from the room you might use for listening. The next time you are in an audio show room or a friend's home listening to reproduced music, notice where you tend to position yourself relative to the speakers if you have any freedom of choice. AUDIO's measuring location is based on such observations.
We place our microphone three meters (about 9 ft., 10 in.) from the front of the speaker. We assume you will want to sit down while listening, so the microphone is placed about at average ear height above the floor of one meter (about 3 ft., 3 in.). The speaker under test is placed exactly as recommended by the manufacturer, if such a recommendation is in the instructions you receive with the speaker. If the maker gives no recommendation or is unclear about speaker placement, we place the speaker according to our judgment.
Now we must concern ourselves with the room. AUDIO not only tries to use a reasonably conventional room geometry but has designed a measurement that can be duplicated by any well-equipped acoustic facility anywhere in the world, as it is import ant to produce repeatable, credible objective measurements. All we re quire of a room is that it have a floor, walls, and ceiling. The walls and ceiling are assumed to be hard reflecting surfaces, such as plaster or wood panel. The floor is assumed to be carpeted. The ceiling is assumed to be slightly over 8 ft. (251 cm) above the floor. The direct sound from the speaker to microphone takes about nine milliseconds for our three-meter measurement. No substantial article of furniture is allowed in the area where the first speaker reflection would have a path about twice the direct path. We do this not because we don't feel you should have lamps, chairs and the like that close to the line of sight from yourself to speaker, but because we want a reproducible measurement that doesn't depend upon detail layout of furnishings adjacent to the speaker.
For this test we energize the speaker with a special coherent signal which has equal energy density within 1/10 decibel for all frequency components from 20 Hz to 20 kHz.
Since the sound will take about nine milliseconds to travel to the micro phone we set up an acceptance window for the microphone signal which is centered at this nine milliseconds.
We allow the first 10 milliseconds of this early sound to pass to our processor through what is known as a Cauchy window for the amplitude of the components. This is, of course, an electrical gating of the microphone signal.
The transform of this time-gated signal is the frequency response of the first 10 milliseconds of sound you hear from that speaker, this is what is plotted as the three-meter frequency response. Depending upon physical placement of the speakers, this measurement contains the contributions of floor, wall, and ceiling reflections.
Because we take only a 10 millisecond "chunk" of sound for this measurement, the frequency response is accurate down to about 200 Hz. For that reason the plot only extends from 200 Hz to 20 kHz. The phase response is not plotted because it would be extremely difficult to interpret without additional measurements. Two amplitude measurements are made. One is made directly on-axis and the other 30-degrees off-axis to simulate stereo listening.
Is all this effort worth it? AUDIO believes it is because this measurement apparently correlates more nearly with the subjective sound impressions related to timbre than does the anechoic response. We compute the data for every 15th-octave interval throughout the useful audio range to cover every possible musical tone.
This gives a response that is far less smooth than good advertising copy might dictate, but it is the way you hear it.
The three-meter room response may be used in the same manner as the amplitude of the anechoic response. Quite often near crossover points you will see a great many interference nulls and peaks. This is due to differences in angular dispersion between crossed-over drivers causing a substantial interference due to floor or ceiling reflection. Compare the anechoic and three-meter test to verify this condition. If the anechoic response is uniform but the room response is jagged, then the implications are that first-order reflections are coloring the sound. This is an indicator also that stereo or quadraphonic sound images will be dispersed when they have substantial energy in that part of the spectrum.
The reason for this is that a left and right channel will never be exactly balanced for such absorption dips and the relative energy of sound in that range will rapidly shift from left to right with small changes in timbre or seating position.
Some speakers will measure better in the three-meter test than under anechoic conditions. Invariably these are the speakers which have been de signed more by how they sound than how they measure, and this is a very desirable property to look for.
Occasionally some drivers will have such a multiplicity of interference nulls that they are artificially smoothed by even a 15th-octave sampling. We will identify such conditions as they occur. It doesn't necessarily mean that it is a bad sounding speaker, but it does mean the measurement is not as smooth as it appears.
Periodic patterns on a linear frequency basis are indicators of sound coloration due to physical structure.
Many times a wall-mounted enclosure will create such patterns because of the acoustic discontinuity of its physical extension from the wall.
If this is the case, it will be more apparent in the 30-degree off-axis measurement than in the on-axis measurement.
A uniform roll-off of higher frequency energy as a trend in the measurement is no sign of acoustic problems, particularly off-axis. If bother some it can be touched up with conventional tone controls. Beware of severe or gross peaks and dips as a trend, however, if you demand ac curate sound. As a general rule, because this is a multipath acoustic interference situation, dips in response will be less objectionable than peaks more than 5 dB above the average through a given range.
by Richard C. Heyser
(Audio magazine, Jan 1975)
Also see:
The Acoustic Feedback Loudspeaker System (Jan. 1972)
Speaker Tests--Phase Response (by Richard C. Heyser) (Dec. 1974)
Doppler Distortion in Loudspeakers (Aug. 1970)
= = = =