18.2 MEASURING THE ESSENTIAL PROPERTIES OF LOUDSPEAKERS
Frequency response is the single most important aspect of the performance of any audio device. If it is wrong, nothing else matters. That is a statement without proof at this point in the book, but that will come.
It is interesting to consider that for as long as anyone in audio can remember, all electronic devices had basically flat frequency responses. No manufacturer of an amplifying device, a storage device, or a music or film distribution medium would even momentarily consider a frequency response specification that was far from what could be drawn with a ruler from some very low frequency to some very high frequency.
Yet, when we come to loudspeakers, it is as though we threw away the rule book and suddenly tolerances of ±3 dB or more are considered acceptable. The measurements in Figures 17.2 and 17.3 show a few loudspeakers from the 1960s.
Some of these needed all of that tolerance, and more, to embrace even the on-axis digressions from flat. Yet, two of them, over substantial portions of the frequency range, behaved quite well. It could be done. Still, bad habits are hard to shake off, and the industry is still burdened with that embarrassingly inadequate descriptor for the most important specification.
20 Hz to 20 kHz ±3 dB is meaningless without seeing the curve that it describes. It could be a horizontal straight line that simply falls off sharply at the upper and lower frequency limits (perfection), or it could be a line that undulates randomly between +3 dB and - 3 dB, a 6 dB range (absolute rubbish).
But even worse than the uselessness of that description of frequency response is the fact that it is often assumed that a single curve is sufficient to describe the performance of a device that radiates sound in three dimensions - in all directions - into a room. When a manufacturer shows a specification for a loudspeaker frequency response in numerical form only, and the tolerance is more than about ±0.5 dB, ignore it. If a curve is shown, but there is only one, it might be correct, but by itself, it is not enough data.