Sound System Design Essentials

Know the methods used to collect data to understand what you can and cannot predict about sound system performance.

 

Most audio pros know that loudspeaker data can be used to predict important aspects of a sound system's performance at the drawing board stage of a project. But it's also possible to extend the concept to array design.

The array prediction process is embraced by some, shunned by others, and amazingly accurate when done within the limitations of data and prediction methods.

We need arrays because many venues cannot be covered adequately with a single loudspeaker. Yet implementing multiple loudspeakers can introduce a lot of problems. Ask any RF engineer what happens when you radiate radio waves into a space from two different antennas operating on the same frequency. His likely answer: “Drop outs. Don't do it.”

The same principles apply to loudspeaker arrays. When multiple antennas are used, either they are located in very close proximity to create a desired radiation pattern (such as the log periodic dipole often used with wireless mics) or they are spread out with minimum overlap to cover a large area (such as cell phone towers).
Loudspeaker arrays are like the first example. This seems easy until you consider that the ideal loudspeaker array must have the same radiation pattern for over 8 octaves, not at a single frequency. This means that the appropriate spacing for one octave band will be incorrect for another, forcing the array designer to vary loudspeaker placements and orientations as a function of frequency.

The tools for successful array design include array modeling software, accurate loudspeaker data, an understanding of the underlying physics, and an open mind. The last item is important, because array design is not intuitive. In fact, this is one place where following intuition alone can lead to disasters. The biggest mistake made in array design is assuming that the sound goes where the loudspeaker is pointed, and that all one must do to form an array is to cluster some loudspeakers and point them at different seats.

In this column in the May issue of Pro AV, I showed the coordinate system for how loudspeaker data is gathered and explained the far-field dependence of both the data and the predictions. Each loudspeaker is treated like a “point source with directivity” in order to cover modeling and SPL calculations. The larger the loudspeaker, the more difficult it is to get into the far-field for accurate measurements.

Arrays become especially problematic, because an array is really just a big loudspeaker made from smaller ones. In fact, arrays can be so large and the far-field so remote that it is not practical or even possible to measure full spherical data to describe their performance. The next best thing is to predict the performance using array modeling software.

Points to PonderThere are several key factors in modeling arrays, and failure to consider them will result in inaccurate predictions.

First, each array element must be measured properly as an individual device, giving us our “point source with directivity.”

Second, the array elements must be positioned accurately relative to each other in the array. Modeling programs let each point source be placed at a unique XYZ coordinate. In the upper octave bands, even fractions of an inch can affect the predictions and performance. This hypersensitivity to positional accuracy affects the performance of the modeled array as well as that of the physical array. Your array will perform no better than the tolerances of the loudspeaker rigging and its aiming features. For these reasons, the highest octave band that can be predicted with reasonable accuracy is 8kHz.

Last, array predictions must include relative timing information between array elements. This is because the radiation pattern of the array is determined by the relative phase relationships between the individual elements, just like an antenna. When several “point sources” are placed in three-dimensional space, the result is a non-point source with an entirely different radiation pattern that is both frequency and distance-dependent.

Array modeling software can be used to predict the shape of the new pattern as a function of frequency, as well as account for the distance-dependence of the response. The numerous variables and their interdependence make array design impossible without the right software tools.

All arrays will exhibit lobing because of phase interaction between the elements. The perfect array that behaves like a single loudspeaker simply doesn't exist. But some array designs are vastly superior to others and can produce adequately smooth coverage to large audience areas.

Pick Your TypeIn most line arrays, the loudspeakers are relatively small and designed to be tightly packed and aimed in the same direction, allowing their interaction to produce the desired coverage pattern. Line arrays have constraints on minimum and maximum length that let them do their thing. Many line arrays have succumbed to value engineering that resulted in an array that was too short for its intended application.

With the point array, the idea is just the opposite: the loudspeakers are physically large, and the resulting aggressive pattern control is used to minimize interaction between adjacent devices, even when spaced very close together. Each array element is aimed at a different seating area, and overlap is minimized.

Exploded arrays break up the audience into zones and attempt to place a loudspeaker on each one. They are often positioned along an arc with several box dimensions between each element. Unlike line arrays and point arrays, exploded arrays are less dependent on the characteristics of the individual array elements to work. This makes them an attractive approach when conventional point-source loudspeakers must be used to cover a large room.

So, step one is to determine which array type you need. It is also possible to use a hybrid approach, such as using the line array for low frequencies and the point array for high frequencies.

Low-frequency array design is, perhaps, the trickiest, because at low frequency, most of the array elements are quite non-directional. This means that the array's radiation pattern is not determined by where you point the boxes. Instead, it is determined by the complex phase interactions of the individual elements, and it easily can be the opposite of what is expected and desired.

This presents a very fundamental problem for many full-range, trapezoidal loudspeakers when used in overhead arrays. If they are positioned for proper HF coverage, for instance, side by side, they are in the wrong orientation for proper low-frequency coverage. The “bass alley” heard in many concert systems is an example of what can result from this interaction, as does the low-frequency build-up often experienced on stage from an overhead array.

The ramifications of a bad array design are significant. They include uneven coverage, poor gain before feedback, reduced speech intelligibility, improper imaging and seat-dependent frequency response — in short, most of the problems that people complain about concerning sound systems.

Bad arrays are often overequalized in an attempt to correct them, which can make things worse. Some hope that the answer to these common system problems is more signal processing, but in most cases, DSP is just a Band-Aid thrown on a bigger, underlying problem.

The loudspeaker system is a filter through which all sound heard by the audience must pass. The array design may prove to be the most significant single factor in determining the quality of the sound experienced in the space, other than the room's acoustics. Designing good arrays is a monumental engineering task that requires powerful software, good data, and a good understanding of electro-acoustics.