Research Work
Several breakthrough research works conducted over the past 20 or so years analyzed the nature and identified causes of nonlinearities in loudspeakers based on new approaches, techniques, and tools. We make reference to several major research works here, but there is a number of others that point to similar results.
In articles that Dr. Wolfgang Klippel wrote for the Audio Engineering Society (AE5) [1], [2] and many other related articles and presentations, he specifically points out that nonlinearities associated with voice coil inductance and modulation of flux B in magnetic gap are predominant distortion mechanisms in speakers.
Those distortion components manifest themselves in much wider frequency ranges than nonlinearities related to displacement dependent nonlinearity of suspension stiffness Kms(x} and motor force B|(x), which mostly dominate below and slightly above fundamental driver resonance
Fs. Nonlinearitv of such transducer parameters as Le(i) inductance vs. current, Le{x) inductance vs. displacement, and BL(i) motor force, play major roles in generating distortions above driver resonance and at all signal levels even at the lowest ones. For many dome tweeters, however, the Kms(x) nonlinearity is inherent, even at low levels since they have a single-roll suspension that is always
asymmetric and its nonlinearity is more prevalent since it is not compensated by spider dominant compliance (usually more linear) at small displacements.
Alex Voishvillo—in his extensive range of papers [3],[4} and other presentations on nonlinearity in speakers, their audibility, and the latest methods of measurements and assessment - clearly shows distortion that manifests itself at low and moderate signal levels is easily noticeable and the
most objectionable. Similar conclusions are reported by Earl Geddes and Lidia Lee [5].
This has to do with the fact that music, unlike sine wave test tones, has signal amplitude probability distribution with peaks occurring rather infrequently, while lower-level signals are present in program material most of the time. In other words, the lower the level of the signal, the higher
the frequency of its occurrence. That is why distortion components that we observe only at higher levels are very often benign (up to a point), since they are masked by higher-level signals that generate those distortions in the first place. Those are typically generated by suspension
stiffness Kms(x} limiting {some dome tweeters have Kms(x) nonlinearity at low level as well) and the voice coil getting out of the magnetic gap, generating B|{x) nonlinearity.
On the other hand, distortions related to Le(i}, Le(x), Bl(i), and cone/dome mechanical break—up are generated at all signal levels and in a very wide frequency band. Such distortions, without the presence of strong masking signals, often extend beyond the masking band (difference, parametric, higher-order nonlinearities, and many intermodulation components). That’s when they become
very audible. The key here is that the masking effect depends on the level of the masking signal. The lower the signal, the weaker the masking effect and the more acute our ability to perceive distortions.
Our ear perceives low-level signals with much greater acuteness and resolution. This is a very strong survival mechanism that humans developed over a very long time. Besides, low-level signals and associated distortions are very easy to discern since there is no masking effect present.
This explains why having low distortion at moderate and low levels is crucial for sound quality.
For years, audio experts and enthusiasts paid attention to low level signal resolution and signal detail retrieval. In fact, this quality is one of the main criteria for high-end systems’ differentiation from budget and mass products. From the early days of digital audio, disappointing sound quality of
CDs was associated with quantization error and low-level digital noise. We were promised the perfect sound forever and there were claims that distortions would be vanishingly
small to non-existent, and yet, many early CDs sounded uninvolving, to say the least. Later, dither was introduced to reduce low-level noise and distortion. Even earlier, before digital, audio professionals noticed that transistor amplifiers, claiming an extremely low percentage of zero- crossing distortion, sounded very unpleasant.
On paper, they may have had rather low total harmonic distortion (THD} levels, but since those distortions were generated at low level input signals, they were very noticeable. No wonder, that many high-end amplifier manufacturers religiously choose Class-A topology even though it is very inefficient and limited in power. A Class-A amplifier has an extremely linear performance at low level
input signals and this is what really matters for sound quality.