Distortion

Distortion technically refers to any kind of change to a waveform, but we typically are concerned with specifically nonlinear distortion. Nonlinear distortion is usually analyzed as two main types: harmonic distortion and intermodulation distortion (IMD).

For an input signal with a single fundamental frequency f, harmonic distortion would appear as additional energy at frequencies 2f, 3f, 4f, etc. Typically, these distortion products exist above the fundamental and do not extend below. Intermodulation distortion arises from the interaction of multiple frequencies in the input signal, and often creates distortion products both above and below the original frequencies. If the input has two frequencies, IMD appears at the sum of the two frequencies and the difference between the two frequencies as well. These new components often cause their own harmonic distortions and further intermodulate, cascading across the frequency range. IMD is measured and expressed as the percentage of distortion energy at all of these new frequency components relative to the energy at the input frequencies.

Harmonic distortion measurements are typically expressed as total harmonic distortion(THD), the percentage of distortion energy at the harmonics relative to the energy at the input frequency. This is a useful metric for evaluating systems that should be transparent, as lower THD+N means better clarity and accuracy. Hidden within the THD figures, however, are differences in how audible distortion can be. Harmonics at even multiples of the fundamental (even-order harmonic distortion) are exact octaves up from the fundamental and so are easily masked by the fundamental. On the other hand, odd-order harmonics are not as easily masked and tend to be much more audible.

Harmonic distortion is also differently audible for different frequencies. In the highest frequencies, above ~10kHz, the distortion products are so high that they are not audible to the human ear. This is not to say it doesn’t matter though, as the high frequency harmonic distortion energy can induce intermodulation distortion. Musically, different kinds of distortion in different frequency ranges are one of the most distinctive signifiers of how we describe a piece of equipment as “sounding.”  Vintage transformer coupled preamps for example tend to exhibit high degrees of third order distortion, but predominantly at low frequencies. Single Ended Triode and other class A tube designs, produce predominantly second order distortion. Broadly speaking, an aggressive, fuzz pedal style distortion is likely dominated by odd order, where as a great deal of what gets described as warmth is usually even order.

Distortion arises from nonlinearities in the signal path, which for a modern playback system includes conversion, amplification, and finally the speaker drivers themselves. In converters, nonlinear distortion typically comes from clipping and aliasing. Clipping is when the signal being converted has levels that exceed the limits of what the digital hardware can handle. This is determined by the bit depth of the converter; with more bits there is more headroom and less distortion from clipping. The earliest converters used 16-bits (8-bits if you go far enough back, but those were not really usable for much), while modern converters are typically 24-bits at a minimum and more commonly 32-bits. This all but eliminates unintentional clipping distortion. Aliasing occurs when the converter is asked to reproduce frequencies beyond its capability. This is governed by the sample rate, how many samples per second are used to represent the waveform. For a sample rate fs the maximum frequency that can be accurately captured is fs/2, also called the Nyquist frequency. Converters have to include what is called an anti-aliasing filter to remove energy above the Nyquist frequency to prevent this distortion. In the earliest days of digital audio, the standard sample rate was 44.1kHz, meaning frequencies above 22kHz would cause significant aliasing distortion. Modern high quality converters operate at much higher sample rates, dramatically reducing the possibility of aliasing distortion. Our speakers feature AKM VelvetSound converters with 32 bits of precision and 192kHz sample rate for inaudibly low THD and IMD.

Entire books have been written about distortion in amplifiers and ways to reduce it. For a deeper explanation of amplifiers, see [LINK to amplifier topologies paper]. Our speakers are powered by Hypex Ncore (Link: https://www.hypex.nl/p/technology/ncore/) amplifiers, state of the art Class D amplifiers that achieve some of the lowest distortion measurements available in commercial amplifiers.

The largest contributor to distortion in a speaker system is the transducers. Transducers have to physically move air to turn the electrical energy into acoustic energy. There are several approaches to achieve this and we will focus on moving-coil transducers, as these are the type used in all of our speakers. To be brief, moving coil transducers have two main parts: 

• a motor structure composed of a permanent magnet that remains stationary and an electromagnetic coil, also called a voice coil, in front of the permanent magnet that is allowed to move only front to back
• a diaphragm attached to the voice coil that moves air as the coil moves

Transducers produce distorted output when they are not moving pistonically, i.e. precisely front-to-back. Magnet assemblies that have uneven fields can force the voice coil to move outside of the pistonic axis, for example twisting or rocking side to side, which then introduces distortion into the acoustic output of the driver. Larger transducers often have a spring, called a “spider”, attached to the coil to help it move back into position. Uneven spring force in the spider can also introduce or exacerbate these problems, as can misshapen voice coils. Furthermore, the coil and magnet have a limited range where they are interacting linearly. If the coil moves out of the field of the magnet, their interaction breaks down and the acoustic output becomes heavily distorted. Our speakers feature transducers designed and built by SEAS (Link: http://www.seas.no ), capable of linear operation across a comfortable range of frequencies and volumes.

The diaphragm contributes distortion if the surface begins to “break up.” Break-up is when the diaphragm is beginning to resonate with the frequencies it is trying to reproduce, creating what are called “drum modes.”

When a diaphragm is in break-up, parts of the surface may be moving pistonically but much of it is moving erratically as well, distorting the acoustic output. The geometry of the diaphragm affects the distortion behavior, for example the smaller the diaphragm the higher the frequency it can reach before break-up. This is one of the reasons tweeters tend to be smaller than midranges and woofers. Different materials will also affect the distortion introduced by the diaphragm. Without going into too much detail, more rigid materials allow a diaphragm to reach higher frequencies without breaking up and highly damped materials reduce the amount of distortion caused when break-up does eventually occur. Our transducers feature advanced materials that are both rigid and well damped and, when combined with the motor structures and optimized diaphragm geometries by SEAS, achieve vanishingly low distortion levels.