Over the years many have debated the relative merits of “looking at screen traces” on an analyzer versus using human hearing to determine how a loudspeaker, or an entire system, actually performs.
While both practices are of course valid, it’s extremely difficult for the human ear to detect, characterize, and correct small deviations in frequency and phase response, and even harder to characterize driver distortion in a meaningful manner.
Likewise, the ear is often stymied when attempting to typify transient response across the spectrum. This is particularly true in the lower frequencies where slow cone acceleration and deceleration can produce a bigger, warmer sound – dependant of course on the program material, the room acoustics, and the listener’s perception.
By listening, we all can tell if a system sounds ragged and distorted; we know when a system is bass heavy or too trebly; we usually know when a note or two stands out unnaturally. But unless we play long passages of music in many different key signatures, it’s all too easy to miss the problem of uneven frequency response, especially if it’s confined to a narrow bandwidth.
And unless we’re willing to subject our ears to varying levels of distortion at high sound pressure levels (bad idea!), it’s nearly impossible to characterize the distortion products of a given driver, or loudspeaker system, in relation to other drivers and other loudspeaker systems.
Finally, if we haven’t trained our ears to a super-human level, we’re going to have an extremely hard time identifying gradual shifts in phase response, let alone what to do about them. This doesn’t mean that we aren’t able to hear the negative effects of distortion, frequency, and phase irregularity; what we’re saying is that hearing the problem is not the same as quantifying the problem and then developing a solution to correct it.
While it’s never a good idea to ignore what you hear in favor of what you see on an analyzer display, the information that an audio analyzer can provide will help enormously whether your task is to optimize a sound system that’s already in place, EQ a simple pair of loudspeakers on sticks, or design a superlative loudspeaker system from scratch.
And as I’ve pointed out in the first two articles in this series (here and here), an analyzer with narrow-band resolution is vastly more useful than a simple octave or third-octave portable device.
Let’s look at acquiring a distortion measurement as one important example of what an analyzer can do for you beyond just displaying a frequency response curve. Some analyzers offer distortion measurements as a built-in function, notably certain models manufactured by Audio Precision (Figure 1).
Figure 1: An ultra-advanced AP 2700 Series Audio Analyzer. (Computer not included.)
Distortion, in its many forms, is virtually always a function of signal level, but not always in a linear fashion. A linear device, such as a good quality pre-amplifier or power amplifier, will exhibit harmonic distortion products that remain at the same ratio to the test-tone fundamental signal – until the device becomes non-linear due to excessive level, at which time the harmonic distortion increases at a non-proportionate rate (Figures 2 and 3, next page).
The distortion products of a well-designed loudspeaker may remain linear at a wide range of power levels, but it’s a safe bet that as maximum power handling is approached, most loudspeakers will exhibit non-linear, even runaway distortion response.
I’ve investigated low- and high-frequency drivers for many years; in some cases, as the driver comes close to burnout, the second and third harmonic distortion products actually exceed the level of the fundamental frequency that’s exciting the driver for the test. This equates to more than 100 percent distortion, counter intuitive as that may be.
Incidentally, when running such tests, it’s strongly advised to wear serious hearing protection, and even then, the sonic signature is anything but pretty.