To get the most enjoyment from our music collections, we need to get the least distortion from our audio systems. Distortion is not just an important subject for audio, it is THE subject. It is often not well understood, too limited in scope, and rarely covers all the various forms of distortion effecting sound reproduction. In some cases, it's conveniently ignored. After searching the internet for a comprehensive list of the various forms of audio distortion, I came up with next to nothing. Very surprising for a topic that is fodder for endless argument and disagreement among audiophiles. How can we begin to get the max from our music without a solid understanding of the obstacles standing in the way of achieving realistic sound?
Since I never found a single source outlining the various forms of audio distortion, this page is aimed at filling the need with a concise summary. It doesn't go into great detail. For depth, see the links to references below. Some of the links contain technical jargon and mathematical equations. Not to worry, the bulk of it is straight forward. You can easily pull out the essential information necessary for understanding the concepts without fully grasping the math.
It is very important to keep in mind that all distortion is not equal. Different forms have varying audibility levels and therefore some get in the way of the music much more than others. This is a complex subject in which too little knowledge, gaps of knowledge, or placing too much emphasis on a single part taken out of context can cause wild misunderstanding and misleading interpretation. With this in mind, let's start with a basic definition.
Distortion is any change in the content of an electrical signal or the shape of a sound wave during its transmission. It could also be described as information either lost or added relative to the original signal. Audio distortion can be placed in two broad categories.
- Type 1 — linear, altered amplitude content.
- Type 2 — nonlinear, added frequency content.
Type 1 is also known as frequency response, which involves both amplitude (changes in the volume) and phase errors. These distortions can be corrected with signal processing.
Type 2, in its most common form, is harmonic distortion (added frequencies). Type 2 can not be corrected after the fact. Each of these types can be subdivided into more specific forms.
Linear Distortion (Type 1)
When all audible frequencies are transfered at the same level, a system is called linear. If specific frequencies are output at different levels, the system is distorting. With a little experience this can be detected by ear if it exceeds +/- 2 dB, and sometimes less depending on the bandwidth of the deviation. Linear distortion is more easily perceived as the deviation spreads across a wider band of frequencies. The narrower the bandwidth and amplitude range of linear distortion the better. Over the entire audible spectrum less than +/- 1.5 dB is desirable. More than that is like watching TV with dark and light spots in the picture or an off-color cast.
Bandwidth Distortion (Type 1)
Bandwidth distortion is another aspect of amplitude distortion. It's about how much of the audible spectrum the system is able to produce. Although a reasonably flat response may be produced in the middle of the spectrum, if the high or low end falls off by more than 3 dB, those frequencies are no longer in balance and therefore not part of the useful bandwidth. A truncated bandwidth is as distorted as a lumpy one. Published specs often show bandwidth without specifying the tolerance. Consider those specs meaningless. Some specifications differentiate between the bandwidth, total frequency range of the system; and the frequency response, that portion of the bandwidth defined by the -3 dB roll off points of the low and high end frequencies. Specs often fail to mention that linear deviations between those points sometimes exceed +/- 3 dB.
Possibly important are the frequencies we can't hear. On the low end is the infrasonic. This is felt as vibration which adds to the total experience of music. At the other end of the spectrum is the ultrasonic. Many instruments produce harmonics beyond hearing range, some large amounts. One study has suggested that there may be some perception of ultrasonic frequencies, however, the importance of reproducing them has not been well established. On top of this, reproducing the infra and ultrasonic extremes is an extreme challenge. Most recordings don't have the content anyway. How far are we willing to go to create a virtual reality? Perhaps it's best to set the audible spectrum as the practical limit.
Harmonic Distortion (Type 2)
Harmonic distortion (HD) comes in two forms, symmetrical (odd order) and asymmetrical (even order). Harmonic distortion is the addition of related frequencies, e.g., whole number multiples of the fundamental. Given a fundamental tone of 250 Hz (approximately middle C on the piano), the 2nd through 5th harmonics would be 500, 750, 1000, 1250 Hz. Lower order harmonics, the 2nd, 3rd & 4th, are less audible as distortion because they are more consonant, closer in pitch and more easily masked by the fundamental. Low order harmonics may be tolerated at surprisingly high levels, 3-5%. (Tolerable, but not inaudible. 3rd is audible at 1%.) Higher order harmonics become audible at lower levels. Very high order harmonics, 9th and higher, are audible at very low levels, tenths of a percent.
Intermodulation Distortion (Type 2)
Intermodulation (IM) is the interaction of two or more frequencies. This interaction creates new frequencies that are the sum and difference of the reference tones, e.g., 100 &1000 Hz, may produce sidebands of 1100 and 900. Intermodulation adds non related, highly dissonant frequencies that are far more objectionable than harmonic distortion, and audible at levels lower than any other nonlinear distortion. Music is made up of many simultaneous frequencies making IM distortion a major concern. It is consistently agreed that IM is more detrimental than other forms of nonlinear distortion.
Dynamic Distortion (Type 1)
There is more to amplitude distortion than absolute linearity. There is the amplitude from the softest to the loudest levels, the dynamic range. These contrasts are an important part of music. Strong spikes, usually from percussion instruments, give live music impact. Soft passages make the loud more impressive. Unfortunately, many recordings have had the highs and lows of the dynamics compressed into a narrower range and/or the transients chopped off by limiters (sometimes it's done more subtly by gain riding). Some pop recordings have been highly compressed. There's nothing we can do about it, except try to avoid compressed recordings (good luck). At home, we need to concern ourselves with the playback system's contribution. Thermal compression by the drivers or inadequate power or power handling will steal the dynamic life out of music. Without the full dynamic contrasts and unclipped transient spikes the music will have lost the quality of live sound.
Temporal Distortion (Type 1 & 2)
Temporal distortion comes in numerous forms. Two related forms are phase angle shift and group delay. Group delay is time lag that varies by frequency. Group delay is not the same as phase distortion, but it results in phase shift. Phase distortion may or may not have a time delay, yet in the end the effects are similar. Group delay and phase distortion can arise from mechanical or electrical sources. Everything, without exception, starting with the recording process, produces some time/phase angle distortion. Fortunately, both are inaudible with music programming even in relatively high amounts. Over 1440°(more than four full cycles) of phase shift has been determined inaudible, and this is more than double the amount any typical audio system causes. This bears repeating : phase angle distortion is inaudible. See the links on phase angle distortion in the references below.
Another type of temporal distortion is resonance. This one is critical. Let's call it time dilation because it stretches the signal out for a longer duration than in the recording, in other words, lengthening the decay, or ringing. Stored energy that is not dissipated as heat is reradiated as unwanted sound. Drivers with poor damping and cabinets that don't completely absorb or block the internal vibrations of the enclosure radiate delayed acoustic energy into the room. This is a major contributor to the perception of, “It doesn't sound live; it sounds like speakers.” It is measured by the cumulative spectral decay, CSD.
Note on CSD : Measurements show strong correlations between amplitude, CSD, and harmonic distortion. Higher amplitude response at specific frequencies will usually presage higher resonance/longer CSD at those same frequencies, which in turn will presage higher levels of harmonic distortion. The reverse is also likely. If you look at a distortion chart that has a peak in a certain range, it will probably indicate a peak in frequency response and lengthened CSD in the same range. At very low frequencies the correlations partially break down. They continue to show a strong parallel between resonance and harmonic distortion, but not amplitude. The low end amplitude may fall off rapidly while the HD and CSD increase rapidly.
Group delay, phase shift, and resonance distortions also come into play when sound waves are altered or redirected after the transducer, under the control of the electrical signal, has produced them. Resonant cavities and various means of controlling dispersion result in alteration of the waveform. It's best to let the transducer disperse the sound without interference.
Analog magnetic tape and LPs are subject to their own set of temporal distortions, speed, wow, and flutter, not applicable to digital. When the playback speed varies from the recording speed it causes a change of pitch and tempo. Wow is a slowly wavering deviation from the proper speed, and flutter is a fast wavering. Wow & flutter are usually well controlled and mostly inaudible. Turntables need to be periodically checked to assure they are turning at exactly 33.33 RPM.
Noise Distortion (Type 2)
Noise is any addition of non specific, unrelated, and often broad spectrum frequencies, e.g., electrical noise, tape hiss, vinyl surface noise. Sometimes it is frequency specific in the form of 60 or 50 cycle hum and its harmonics. Good electronics add minuscule, yet slightly audible amounts. Grounding problems can add large amounts. Digital forms of noise include aliasing, jitter, and quantization noise. Digital forms are well understood and normally kept to extremely low levels, typically below the noise floor of the electronics.
Acoustic Distortion (Type 1 & 2)
Room acoustics as distortion? Certainly, it can be regarded as added new and amplitude modulated frequency content not in the original recording. It could almost be compared to noise. An anechoic chamber is the only way to eliminate this distortion, but most claim that listening in an anechoic chamber or very heavily damped room is not good. (Headphone/earphone listening is another option for eliminating AD, yet not without its own set of issues.) It is best to have rooms that are neither too lively, nor too dead. But no matter how the acoustics of the room are behaving, the reverberant field passively created by the room won't sound natural unless all of the acoustic energy radiated into the room by the speakers is balanced. It's advisable for the refections to be moderately absorbed and/or diffused and also slightly delayed before they reach the listener. Delay gives the ear/brain time to distinguish the direct sound from the reflected sound. It may also be important to keep the direct to reflected ratio at 1:1 or greater, however, this is difficult to do in practice.
(II) UPDATE 21-01-10 : It appears there are dissenting opinions about anechoic listening. The logic is solid. It makes me reconsider the importance of the room and that a more damped, absorbent room may be better. But the more one understands room acoustics, the more one realizes the near impossibility of doing it effectively.
Deliberate Distortion (Type S)
Distortion is not always avoided. Sometimes it's added deliberately by musicians to alter the sound of their instruments, or by engineers to modify their recordings. Numerous kinds of analog and digital signal processors are used to create distortion that musicians willing exploit for novel, interesting effects. On the production side of music, this is good. On the other hand, for the reproduction of music, distortion is the bane of high fidelity. Every step of the recording and playback process adds its share. In the end, the sound reaching our ears has a substantial accumulation of all types of distortion. Kept at a minimum, we can achieve some relatively realistic results. Yet many audiophiles aren't satisfied with minimal distortion on the reproduction side. Some have a strong preference for adding small amounts of just the right forms. A little extra low order harmonic distortion produces a pleasant fullness and depth that has an easy-on-the-ears quality. The attractiveness of this effect clearly explains the renewed interest in vinyl and vacuum tubes. Both add moderate amounts of the right stuff of distortion. (Plus some resonance and IM along with low order harmonics.) Not too little, not too much, just enough to add some warmth and smooth out the hard edges inherent in the sound of live acoustic instruments and human voices. The stark reality of neutral, clean, analytical, precise digital sound is not for them. I see the controversy between analog and digital similar to the bickering, back in the 1980s, over the merits of Ektachrome and Fujichrome (color slide film). Most photographers preferred Fuji for its wonderfully rich, saturated colors—its colors were hyper-real. Purists stuck with Kodak Ektachrome for its accuracy, color fidelity and realism. Ultimately, Fuji won and Kodak introduced new high saturation films to compete. Digital is for purists—analog for idealists.
Mental Distortion (Type F)
This is a very curious type of distortion. It's audible only when the music is off—exactly when we're least likely to be listening for it. It's an all too common form of distortion perpetuated by deceitful companies that feed on credulity and naïveté. They purport to make amazing products which are based on obscure, unprovable science, gross omissions of relevant facts, and their own twisted theories that are barely plausible to anyone who has graduated high school, and make absolutely no sense to anyone with a degree in a hard science. Their white papers and websites are so outrageously full of bewildering gibberish it's hard not to snicker, giggle, chuckle and finally conclude that they've got to be parodies. Yeah, they are funny, but there is a not so funny flip side. These scam operations are laughing right back in your face, harder and louder. How do you know when you've run across one of these cheats? It's easy. If it gets you asking yourself, "Is this for real?" You know. That twinge of doubt in your gut is telling you. (The goofy names of these companies and their products are another clue.) Their rambling sales pitches are all about hooking your curiosity, raising your hopes, using promises as bait, and baffling you with mental distortion just before reeling in your money.
Stay focussed and you'll soon learn to easily recognize and avoid this hideous form of distortion.
The more I learn, the more I realize the importance of supporting the content of this site. Links are added frequently. More can be found on other pages.
Links to the references for this page :
Start with a good Audio Glossary.
A link to information about Testing Procedures and Specifications.
Covers HD and negative feedback in amplifiers : Distortion & Feedback
Brüel & Kjær : Audio Distortion Measurements
Amplifier distortions : Valves vs Transistors & Harmonic & Intermodulation Distortion.& Output Impedence
A four part series on the properties of hearing and its correlation with distortion : Human Hearing - Part 1 and Distortion Audibility - Part 3
More on thresholds of Distortion Audibility
In addition : Phase angle distortion audibility
Waveforms & Harmonics : All About Circuits, Spectrum Analysis
Two links about ultrasonics : Ultrasonic Harmonic Content of Musical Instruments.& Promastering.com/pages/techtalk
An electrical engineer on LP/CD distortion : . . . orders of magnitude lower
More on LPs and related links : The Skinny on Vinyl
Some like it distorted : Analogue Warmth
A decade old article from an industry insider : High Density Audio Formats
Bits about jitter & aliasing : Oversampling vs Upsampling : Differences Explained
Often ignored : Audio Compression : The Silent Distortion.
A short video on Audio Compression.
Link to a short article about Dispersion & Directivity.
Rarely do I find a forum thread useful or informative, yet the following has an interesting discussion about room acoustics : Diffusion vs Absorption
For more about rooms see : Omnidirectional Loudspeakers
And in case you're wondering : What's with the Tube Hype?
Is the world really Analog or Digital?