Figure 18.4 illustrates three distinct sound fields in small listening rooms:
• The direct sound from the source to the listener, normally represented by the on-axis behavior of the loudspeaker.
• The early reflections, sounds that have been reflected only once on their way to the listener. These would be represented by measurements made at the appropriate off-axis angles, taking into account the positions of the room boundaries and the arrangement of loudspeaker and listener within them.
• The late reflections, arriving after multiple reflections from all directions. This would be called the reverberant sound in large performance spaces. This is the sum of all other sounds radiated by the loudspeaker in all possible directions and is described by the sound power.
FIGURE 18.4 The three sound fields in small rooms: (a) Direct sound. (b) Early reflections. (c) Late reflections.
It is believed that some combination of these is sufficient to describe much of what listeners hear from loudspeakers in small rooms. This much was evident in the early 1980s when the author set up a semiautomated data-gathering and -processing system at the National Research Council, in Ottawa.
A computer-controlled oscillator stepped through a logarithmically spaced set of 200 frequencies from 20 Hz to 20 kHz (20 measurement frequencies per octave). The anechoic chamber was calibrated at low frequencies by fixing the locations of the rotational center of the loudspeaker (the center of the front baffle) and the microphone, and deriving a correction curve based on true free-field measurements made on a 10 m tower outdoors (an alternative method of generating a reference curve is to employ the ground-plane technique; Gander, 1982).
It is believed that the resulting frequency response curves were accurate within ±1 dB down to 30 Hz. Of course, the correction applied only to monopole woofers. Even then, some had sufficient interference from port radiation that the sound power measurement became the definitive measure at low frequencies.
Measurements were made at 2 m, on both vertical and horizontal orbits, at angular increments of 15° in the frontal hemisphere and 30° in the rear hemisphere. The data were postprocessed by the computer to yield several individual and spatially averaged frequency responses, sound power, directivity index, and phase (Toole, Part 2, 1986).
Figure 18.5a shows some of the basic data on a loudspeaker designed to have a flat and smooth on-axis response, but clearly off-axis performance was not given comparable attention (from Figure 8.10, loudspeaker "E"). The raw data were then manipulated using dimensions from the prototype IEC listening room (Figure 4.10a) to generate a picture of the sequence of sounds that would arrive at the listener's ears in the room. Specular (mirror-like) reflections were assumed. Figure 18.5b shows those predictions.
FIGURE 18.5 (a) Anechoic data for a three-way loudspeaker with uneven directivity versus frequency. (b) Sounds arriving at the listening position predicted from anechoic measurements: direct sound; the early reflected sounds from floor, ceiling, and two side walls; and late reflected sounds. These are summed to show a predicted room curve, elevated by 10 dB for clarity. (c) The same loudspeaker measured at three locations in a real room, compared with the prediction in (b). From Toole, Part 2, 1986.
Let us pause at this point and ask the question: If we want to measure a loudspeaker, and from those measurements try to anticipate how it might sound in a room, what should we measure? The answer at low frequencies is sound power; it is the highest curve. However, we know that if the woofer radiation is omnidirectional, the shape of the curve at frequencies up to about 100 Hz will be the same in all of the curves, which it appears to be for this loudspeaker (for loudspeakers with collections of woofers, or a physically separated port, the sound power will be the true measure).
At the highest frequencies, it is the on-axis frequency response that is dominant; it is the highest curve. Over the rest of the frequency range, which includes voices and most musical instruments, the curves weave among each other and are never far apart. So the global answer to the question is that we must measure everything. No single curve tells the complete story. Performing an energy summation of the data, we get a curve that is a prediction of what might be measured in a room.
Taking the loudspeaker into the real room and measuring its performance in three locations, averaged over the listening area, yields the three curves shown in Figure 18.5c. As pointed out in Chapter 4, where these data are also shown, the standing waves in the listening room cause huge fluctuations at low frequencies.
However, at frequencies above about 300–400 Hz, the curves become quite similar to each other and also to the predicted room curve, which is superimposed. Obviously, it is possible to use anechoic chamber measurements to anticipate how a loudspeaker might sound in a room at frequencies above the transition frequency.
However, this observation is much more important than may appear. Some have argued that it is the shape of the room curve that determines how a loudspeaker sounds. If so, then one could ignore a loudspeaker's anechoic behavior and equalize the room curve to have the desired shape. Then the peculiarities of both the loudspeaker and the room are accommodated in the one action.
What is missing from this perspective is that two ears and a brain are far more analytical than an omnidirectional microphone and an analyzer. The measurement system simply adds up all sounds, from all directions, at all times, and renders a single curve. A loudspeaker turned to face the wall, after equalization, should sound like that same loudspeaker facing the listeners. It doesn't. There is insufficient data to describe the source and thereby how it interacts with the room boundaries.
Humans have a remarkable ability to separate a source from the room it is in and offer up detailed descriptions of how it sounds, even when the room is changed (see Chapter 11). We need measurements that describe the nature of the source and that provide insight into what happens in a room.
The loudspeaker in Figure 18.5 is an example of a common problem. Engineering convention dictates that the on-axis frequency response be as flat and smooth as possible. This, after all, is likely to appear, perhaps artistically enhanced, in the brochure.
However, this loudspeaker exhibits different directivity at different frequencies, and it is this pattern of directivity variation that shows up in the room curve (see Figure 18.5a and the early reflection curve in (b)). If equalization is used to flatten the room curve, the pristine on-axis curve will be lost - the only thing that was correct about the loudspeaker. An equalizer changes frequency response, not directivity.
The cure for this room curve is a better loudspeaker, one with better directional consistency. Many loudspeakers suffer from combinations of both problems - faulty axial frequency response and inconsistent directivity - which makes life complicated.
Once a loudspeaker is in a room, there are no measurements that will enable us to separate - with high measurement resolution and accuracy - the direct, early-reflected and late-reflected sounds. Without detailed information on the loudspeaker, equalization within the room is a game of chance. However, if one has sufficiently detailed information on a loudspeaker, it may be possible to predict what may happen in a room. If a loudspeaker is properly designed, and strong early reflections are not spectrally corrupted, equalization might not be necessary above certain frequencies. This is a worthy objective.