So you need a loudspeaker. You intend to sit down and compare the specifications to determine which one is better. Depending on whom you ask, “better” can mean: louder, flatter response, higher power, better coverage, goes lower and/or higher in frequency, heavier, bigger, more drivers, or heavier-duty enclosure. Specifications usually cover all of these attributes. The question is: are these numbers useful? Are they even correct?
When the lights go down and the sound comes up, all that matters is how a loudspeaker sounds. How can you tell how a loudspeaker will sound from its specifications? The short answer is: you can't. The long answer is: you cannot. The simple reason is there are no standards for measuring the performance of complete loudspeaker systems or for creating specifications from those measurements. This means that loudspeaker manufacturers have been largely on their own in determining how to measure performance, to analyze the measured data, and to present the results. Because of this, one must be extremely wary when making comparisons between specifications.
What can specifications tell you? Very little that can be considered definitive. The best use of loudspeaker specifications is to classify various products as being somewhat similar in power handling, directionality, and output level, as well as for driver complement and size. Their worst use is to base buying decisions on differences in the numbers that only appear to be insignificant or that are simply “much better” when compared to similar products.
In spite of many scientific looking, often esoteric specifications, there are only three specifications that directly concern the sonic performance of a loudspeaker: frequency response, maximum output, and coverage. For the purposes of our argument, we will concentrate on these three most fundamental, widely observed, and, at the same time, most problematic specifications.
Frequency response is specified two ways. One way is a frequency response graph. This (or a table of values) is the only way that meets the scientific definition of “frequency response”: the variation in output with frequency in response to a constant input.
Many frequency response graphs are “smoothed.” Smoothing is a way to average the levels at different frequencies. The graph at the right shows two frequency response curves for the same loudspeaker. One is the data as measured and one is a “smoothed” version of the same data. For the latter curve, the levels were averaged by octaves. You can see that smoothing obliterates certain details, making the response curve appear much better. You can usually spot smoothing because the response curve will invariably appear, well, smooth. Graph smoothing does not make the loudspeaker sound better.
Below are four numerical frequency response specifications.
|A||Freq. Response (-3 dB) |
Freq. Range (-10 dB)
|62Hz - 15kHz |
50Hz - 16kHz
|B||Freq. Response (-3 dB) |
Freq. Range (-10 dB)
|60 Hz - 18 kHz |
37 Hz - 20 kHz
|C||Frequency Response||40 Hz to 18 kHz|
|D||Frequency Response (1 W @ 1m)||+/-3 dB 80 Hz to 18 kHz, -10 dB 60 Hz|
These specifications are not really frequency responses. Rather, they list the lowest and highest frequencies between which the loudspeaker is designed to operate. The dB numbers indicate how much the output might vary (e.g. +/- 3dB) within that range or the frequencies at the ends of that range where the sound level drops off by a certain amount (usually -3 dB or -10 dB). Numerical frequency responses might seem good to use for comparisons. They are not. The way in which the frequency response is measured and/or stated can be grossly misleading at face value.
Loudspeakers B and C would appear to have the best low frequency responses. However, the fine print reveals that, unlike the others, loudspeaker B was against a reflective surface such as a wall or the floor for the measurement. This will increase the low frequency output by up to 6 dB. If the low frequency response for B were measured like the others, it would likely be more similar to loudspeaker D. For loudspeaker C, are those the frequencies where the sound drops of by some amount and, if so, by how much? Because it is not stated, you should not compare it to the other frequency responses.
So how is one to know just what the frequency response is? You do if it is graphed and you know the measurement conditions. Otherwise, until there is a recognized and accepted standard for measuring and specifying frequency response, making comparisons between different products can lead to erroneous conclusions.
Maximum output specifications are normally calculated from two other specifications: sensitivity and power handling. Sensitivity states what the output level (SPL) is, usually with a 1-watt input. However, the measurement distance and how the data is post processed can significantly affect the results. Such variations include using the peak SPL in the frequency response, using the SPL measured at a particular frequency (like 1 kHz), or taking some sort of average of the SPL over some frequency range. These choices leave ample room to come up with different results.
Power handling states how much power the loudspeaker can handle before it is damaged. This is normally determined using a test signal and turning it up until the loudspeaker is physically or electrically damaged. However, the frequency range, peak to average ratio, duration of the test, and how the power limit is defined can be varied. These choices leave ample room to come up with different results.
Further complicating the above is the fact that sensitivity and power handling mean almost nothing by themselves. To see why, examine this chart of example specifications:
|Loudspeaker||Sensitivity||Power Handling||Calculated Maximum Output|
|A||95 dB||200 watts||118|
|B||98 dB||100 watts||118|
|C||108 dB*||10 watts||118|
|D||89 dB||800 watts||118|
*This is typically a specification for a high frequency, horn-loaded driver.
In spite of quite different sensitivity and power handling specifications, all have the same maximum output. Thus, do not be fooled big wattage numbers or high sensitivity numbers. It is the combination of the two that determines maximum output.
The most important point is that maximum output specifications do not tell you the quality of sound at that level. The truth is that, if operated at their specified maximum outputs, most loudspeakers would sound bad. The actual maximum usable output will invariably be less. In real use, this depends on a combination of program material, signal processing (such as equalization), available amplifier power, and operator competence.
Coverage defines the directions where the sound is projected from a loudspeaker. Most professional speakers are designed to project sound more in certain directions than in others. This reason is to help control the sound from bouncing off room surfaces where there is no audience. Such reflections can severely impair intelligibility, especially with difficult acoustics. Directional specifications have various names, all intended to mean the same thing. These include: beamwidth (the only correct term), dispersion, directivity, directionality, and coverage.
Here are some typical directional specifications:
|A||Horizontal Coverage |
|B||Horz. Coverage Angle (-6 dB) |
Vert. Coverage Angle (-6 dB)
|95° averaged 500 Hz to 16 kHz |
70° averaged 500 Hz to 16 kHz
|C||Dispersion||40° H by 90° V|
|D||Nominal Coverage Angle/-6 dB points (degrees)||Horizontal 90 Vertical 45|
The angles specified are centered on a line perpendicular to the center of the loudspeaker's face. Understand that the sound is not contained within those angles. In fact, most loudspeakers project sound all around themselves. These specifications really tell you that the best quality and loudest portion of the sound is directed primarily within these angles. Outside those angles, the sound is likely to be too low in level or the frequency response too poor to be useful.
In spite of the appearance of the above specifications, the directionality of most loudspeakers varies with frequency. The physics of sound make it much easier to control the direction of high frequency sounds than low frequency sounds. This means that the lower the frequency the wider the loudspeaker's coverage angle will usually be. Most loudspeakers become omni-directional at lower frequencies.
Over what frequency range is a directional specification valid? It can be stated, as it is for loudspeaker B, or better, it can be graphed as horizontal and vertical “beamwidth” curves or as polar plots. For many full-range loudspeakers, horizontal directionality is only maintained down to the 500 Hz to 1500 Hz range. Vertical directionality is usually only maintained down to the 1000 Hz to 2500 Hz range. However, these ranges do include the 2 kHz to 5 kHz range critical to speech articulation and intelligibility. Directional control at lower frequencies is typically achieved using multiple low frequency drivers, making a large loudspeaker with a midrange and, sometimes, a bass horn, or by using arrays of loudspeakers.
Directionality is a function of what you need for your application. Basically, you draw lines from the loudspeaker location to the extreme left, right, up, and down edges of the audience. The angles between the lines are the approximate horizontal and vertical coverage angles you need. However, you will rarely find an exact match to those angles in a product. It is usually necessary to pick something close or use more than one loudspeaker to provide wider coverage angles.
Because there are no standards for measuring and specifying complete loudspeaker systems for professional use, measurement conditions, data post-processing, and analysis can and do vary among manufacturers. This means that comparing specifications from different manufacturers can lead to inaccurate and often erroneous conclusions. At best, specifications should be used to make rough comparisons.
Here are ten sobering realities acknowledged (at least privately) among audio professionals about professional grade loudspeakers:
Most will sound good to an audience, if operated properly.
Loudspeakers always sound better playing certain things. What these are varies with the loudspeaker.
When making comparisons, the results usually change with the type of program being reproduced.
The sound operator has far more to do with sound quality that the loudspeaker designer.
You should expect to see similar specifications if the drivers, horns, enclosures, etc. are similar.
No one has designed anything close to a perfect sounding loudspeaker.
No one has invented anything close to an objective test of sound quality.
Differences of up to 3 dB or 4 dB in specifications are rarely very significant.
Large differences in frequency, such 15 kHz and 18 kHz, may actually span only a few musical notes.
The only way to determine sound quality is to listen.
And finally, here are four rules you must follow to make valid listening comparisons between any two loudspeakers:
A loudspeaker's apparent sound quality will vary with the type of program material. Listen to a wide variety.
Listen at soft and very loud levels. A loudspeaker should sound good at both levels.
Both must be very close to the same listening level. Otherwise, the louder one will always sound better.
Walk around to determine the coverage. When you start to notice an obvious loss in the higher frequencies, you have reached the edge of useful coverage.
Chuck McGregor is the technical services manager for Eastern Acoustic Works. He can be reached at Chuck.McGregor@eaw.com.