Research at Meyer Sound provides information on how loudspeaker arrays interact and also serves as a "proof of performance" for our MAPP acoustic prediction program modeling software.
Early in 2000, we began to test arrays of Meyer Sound loudspeakers to investigate the potential for changes in frequency response caused by the cavities that are created when adjacent loudspeaker cabinets are splayed to achieve a desired sound coverage pattern. A loudspeaker system should provide very precise coverage to the seating areas alone, without directing sound onto the walls and ceiling. Systems designed and installed in this manner can yield very high measured (and perceived) clarity and intelligibility. In some cases, there will even be a reduction in the number of loudspeakers required.
Fig. 1. Three MSL-4's arrayed for optimum coverage. Note splay between loudspeaker cabinets.
In constructing such arrays, cavities are created between adjacent loudspeaker enclosures. Intuitively, it would seem that at lower frequencies there could be some cavity resonance, diffraction - or perhaps both. Should such behavior exist, the next question would be: How can this be treated and minimized?
A second and equally important question relates to the development of our MAPP software. MAPP employs data from high-resolution measurements of Meyer Sound loudspeakers to predict both angular coverage and frequency response of any arbitrary array configuration. The amount of computing power needed for such modeling is rather high. We have determined that low-resolution measurement masks important behavioral details that will affect how the loudspeaker system performs in the real world. Therefore, all of our sound reinforcement loudspeakers are measured at high resolution (1° increments in both axes, and 1/24th octave), and include phase as well as amplitude.
When measuring and modeling loudspeakers, it is necessary to do so in the far field of their coverage, where the shape of the loudspeaker enclosure is not a factor. Typically, this is at a distance that is at least two times the longest enclosure dimension. Verification may be obtained by measuring the inverse square law loss: the far field is where there is 6 dB of loss with doubling of measurement distance. Within the near field, there will be less than 6 dB of loss with a doubling of the measurement distance.
So, the question that we have been faced with is this: if we base everything on far field models are we missing any data components having to do with the shape of the loudspeaker enclosure that will adversely affect our computer predictions? The loudspeaker array measurements described below would serve the additional purpose of verifying the accuracy of MAPP modeling of such arrays.
We began by measuring the smallest of our loudspeakers that typically are used in arrays, the UPA and CQ series, in our anechoic test chamber. We first looked at single reference devices, then at two devices (both tight-packed and splayed) and finally at three devices, also tight-packed and splayed. The UPA's showed no appreciable measured differences at any of the splayed angles. The CQ's displayed a small amount of reduced level (less than 2 dB) in the 250-400 Hz range when they were splayed with approximately 8" between cabinet faces, and there was a slight improvement when panels were inserted into these openings.
Fig. 2. Rendering of the Meyer Sound anechoic chamber and loudspeaker positioner
Having found very little effect of splay angle on small loudspeaker array responses, we proceeded to measure three-box arrays of larger Meyer Sound loudspeakers outdoors on our parking lot and with a ground-plane measurement microphone orientation to reduce boundary reflections.
Fig. 3. Three MTS-4A's measured outdoors with ground-plane microphone
In measurements of arrays constructed with the Meyer Sound MSL-4, a horn-loaded system designed specifically for arraying, there was no appreciable change in frequency response as splays were formed between the three loudspeakers, regardless of the degree of spread. Taken together with the UPA and CQ measurements, this would suggest that cavity resonance plays little or no part in the response of splayed loudspeakers in arrays. Yet the idea has gained increasing currency in the audio community. Could it be that all those who advocate it are simply wrong?
In an attempt to discover whether some other mechanism might be responsible for response anomalies in splayed arrays, we measured three Meyer Sound MTS-4A's. (The MTS-4A is not designed for horizontal arraying.) Here, we saw a significant reduction in level from 200 Hz to 700 Hz when the units were splayed at 13" between the cabinet faces (see figure 4). There was a very slight change at 200 Hz when panels were inserted into these openings (see figure 5). Clearly, if the system in question is not well suited for arraying, splaying adjacent cabinets can cause response problems. But the data show that cavity resonance is not the primary cause.
Fig. 4. Three MTS-4A's tight-packed versus splayed with 13" openings at front faces
Fig. 5. Three MTS-4A's with open splays compared with panels inserted in openings
Once we had made these measurements, we modeled in MAPP the same loudspeakers under the same conditions. This modeling was conducted using data from a single speaker measured in the anechoic chamber at 4 meters. MAPP predictions displayed very similar changes in amplitude over the same frequency range as we had measured (see figure 6).
Subsequent experiments in MAPP revealed that by adding a small delay of 0.6 ms to the center MTS-4A, the loss in level between 250 and 700Hz can be eliminated (see figure 7). This result suggests that the primary cause of the phenomena observed with the MTS-4A (and perhaps with other arrayed systems) is the arrival time difference between the two outer units and the center unit (see figure 8).
Fig. 7. MAPP prediction of three MTS-4A loudspeakers splayed with 13" openings at front and with 0.6 mS delay on the center device
Fig. 8. Arrival time differences between center and outer units in splayed array
This research has shown that the cavities between the cabinets in an array have very little effect on system response. Indeed, when systems are properly designed for arraying, splaying them does not change the overall system response in any appreciable way. Even though some in the professional audio community insist that the fronts of arrayed cabinets must touch one another, this is not the reason for poor system response. Rather, improper coverage overlap among cabinets in an array can result in phase cancellations due to arrival time differences.
There are many ways to array loudspeakers, and these must be carefully analyzed to determine whether there are any appreciable consequences. We are confident that array conditions may be predicted with MAPP software and, therefore, may be discovered and accommodated during the design process.
We will continue to explore any means that we can to provide the system designer, installer and operator with more relevant information about how their loudspeaker systems behave, and practical techniques for realizing the best possible sound system performance.