Audio Distortion: What is It and Why Should You Care?
Image: Copyright Charles Rodrigues estateReaders of any number of home theater and audio review magazines have surely seen reference to the term distortion, for example: “This component has lower distortion than that component.” The topic has been a major focus of audio performance measurement and, unfortunately, has become the basis for a lot of desktop drag racing competitions. That is, my component has 10-fold less distortion than your component, as yours is .001% THD and mine is .0001% THD. Or, my speakers are flat to 15 Hz, yours are only flat to 20 Hz. My sympathies to anyone that’s engaged in this kind of unhelpful desktop drag racing. Many of you may be wondering: What’s THD? What’s distortion? What is linear vs non-linear distortion? And does any of it matter? Here’s a hint: it’s a numbers game that’s generally meaningless to anything other than a measurement device.
Let me start by making this article very short for most of you with one simple statement: Today’s audio and theater systems have distortion levels that are so low, they’re a complete non-issue. Believe it or not, that statement is generally true as far as distortion is concerned (the exception being linear distortion in the low frequencies). You may now disregard the rest of this article, take heart in the good sound of your system, and sit back and enjoy the music. For the rest of you, let’s dig a bit deeper into distortion and explain why my statement is true. This article is intended as a high-level overview of distortion described in lay terms. Many definitions were changed to be more easily understood by the average enthusiast, not by scientists and engineers.
When we talk about distortion in sound reproduction, we typically break it down into two buckets: Linear and Non-linear. But, before investigating each of these terms, let’s define distortion, which is any alteration of sound from its intended presentation. Distortion is a change, and we must agree upfront that any change is undesirable, as this was not what the sound engineer, artist, or producer intended for us to hear.
Linear distortion is an easier concept than non-linear distortion, and the one that gets the most attention. Linear distortion is simply an alteration to the amplitude of the frequency or phase of the signal (it doesn’t add any new frequencies). Imagine we are playing the telephone game. I whisper a message to you and you repeat it, but you do so with a different pitch (in a deeper voice). The message didn’t change. Nothing new was added. It was just stated in a lower voice. This, at its heart, is linear distortion. And when we talk about distortion, we typically are not talking about linear distortion despite it being one of the most egregious and audible distortions in our system. As you can probably guess, speakers and our room are the biggest contributors to this. Our amplifiers, preamplifiers, receivers, CD players, DACs, computer sound cards, and phones all generally have a ruler flat frequency response with only small variations starting above and below what we can hear; speakers and rooms contribute the most to linear distortion in our sound systems. A future article on room acoustics will help explain the impact of a room. For now, we will treat speaker and room as one. We will also ignore phase issues, as this is a more complex topic to display and explain, but is built on the same principles frequency variation (phase distortions are also far less audible).
Below is a picture of the frequency response of my left speaker and subwoofers. Anything that isn’t completely flat is linear distortion or a change from the original input signal.
This reflects all the linear distortion that my room and speakers have imposed on the signal, and you can see that it is quite high. Most of these variations from the flat line are significant enough to be very audible to even the casual listener. Yet this is actually a very good response for a speaker in most rooms; we often see much worse.
Here we see an example from the recent AV NIRVANA speaker evaluation event. We see the effects of severe linear distortion caused by the room on the tower speakers under test. Note that this distortion has little to do with the speaker itself and instead reflects what the room is doing (Note that these speakers were not used with subwoofers which is the reason for the less smooth bass response).
Now let’s look at this linear distortion differently. The problem with the above graphs is that while measured distortion is quite high (and certainly audible), visible variations overstate audibility by quite a bit. Our ears are very good at filling in the dips you see in that response and we don’t tend to hear them. Our brain can turn that rocky mountainous response into a smooth highway of blissful sound. At high frequencies, our ears don’t even hear most of the distortion we see in a measurement because our ears are also quite good at filtering out reflections. As such, its illustrative to look at a response that focuses on higher frequencies only, but with smoothing. This lets us better see what our ears hear at those frequencies. The next image compares EQ correction to no EQ correction from 500hz to 20khz, EQ is a method to reduce linear distortion.
As you can see, both look good. As I mentioned, the response was quite good to begin with. The Blue line is the equalized response, and we can see that it’s a bit flatter overall, with less big dips. The Blue line has lower linear distortion. The worst linear distortion in a room is typically at low frequencies, and the biggest cause is the room. Because this topic would take us into room acoustics, I’ll save further explanation of low frequency linear distortion for a future room acoustics article.
Non-linear distortion is a minor problem in the vast majority of scenarios; linear distortion tends to dominate most systems. That isn’t to say that most systems don’t have inaudible non-linear distortion, but most modern audio systems have such low levels of non-linear distortion that it’s no longer a major concern (you simply can’t hear its effects). Non-linear distortion is the addition of new frequency content that was not present in the source. Non-linear distortion is also the kind of distortion we think of typically when we hear the term distortion.
Harmonic distortion is the existence of new frequencies in sound that do not exist in the source. Harmonics refers to the harmonic multiple of the fundamental frequency (fundamental is the sound of the original source frequency before any new frequencies were added). Harmonics are like the symptoms of an underlying condition. This condition is a non-linearity in the reproduction system (i.e, source, amplifier, or speakers). This is a very complicated concept to illustrate when considering musical content, so let’s focus on pure tones. If you had a pure tone of 1 kHz, a 2nd harmonic would be 2 kHz. A 3rd harmonic would be 3 kHz. These harmonics are sometimes called overtones. In musical instruments, these overtones make an instrument sound unique, but in audio systems overtones are typically rather undesirable changes in sound. Imagine listening to a trumpet and then playing it through a fuzz pedal. The trumpet will sound different (perhaps interesting), but it won’t sound like a trumpet anymore. A stereo or home theater system, however, isn’t intended to alter sound. We want it to reproduce content as the artist intended it to be heard. As I said at the start, no distortion is good, it is just that some is less bad than others.
Let’s look at distortion measurement captured using Room EQ Wizard (REW) and my home theater system, taken in room at the primary listening position. The distortion is that of the entire system, including its interaction with the room.
REW displays any distortion that lies below the noise floor as light grey, because the room is the limiting factor. In this measurement, the cursor lies at 30.5 dB’s and the fundamental is 78 dB, this reflects 0.4% distortion. Each colored line within the graphic is a harmonic, shown because THD stands for total harmonic distortion (an equally weighted composition of the harmonics). However, the audibility of harmonic distortion increases with higher order distortion. That means as the harmonic level increases, we become more sensitive and can hear it at lower volume levels.
Ok, I just tossed several numbers around and you’re probably asking: what’s the bottom line? Is 0.4% THD audible? Unfortunately, the answer is rather complicated. Total Harmonic Distortion is not a measure of the audibility of distortion. In other words, there’s no definitive level of THD that is or isn’t audible until we get to fairly high levels. This concept is thrown around a lot, but unfortunately THD is very poorly associated with distortion audibility. We don’t hear all distortions the same (nor is harmonic distortion the only kind of non-linear distortion of concern). Low levels of higher order distortions (e.g., 5th order or 6th order) are very audible. In fact, very high levels of low order distortion (e.g. 2nd order) may be audible, but we might find it pleasant, and it will likely be masked. If the presence of high distortion (such as a quick dynamic) is very brief, our ears might not hear it. If it is persistent for a long time, we are more likely to notice it. I warned you, it’s complicated! Unlike linear distortion, which we can improve with EQ or room correction, non-linear distortion is far more complex in origin and so harder to reduce.
Masking refers to the notion that a musical signal with complex distortion, the fundamental will make some of that distortion unnoticeable during normal listening. Masking is caused because the fundamental is louder than and near the harmonic frequency, and so it takes precedence in our hearing. We hear it, but we don’t really hear its harmonics. Auditory masking is a hugely complex topic itself deserving of its own article. Here’s a link to an article on Mpeg-1 compression that discusses the role of auditory masking. Audio compression systems remove important information from musical signals in a manner that we don’t notice readily as a degradation by taking advantage of some of our sound sensitivity issues, such as auditory masking. Auditory Masking
Other Non-Linear Distortions
There are quite a few different kinds of non-linear distortions. This topic is too complex for this article nor does REW measure these distortions. It is very likely that forms of non-linear distortion have a negative audible effect on our sound system, but unfortunately the research into this topic is too sparse to give simple explanations. The effect of these distortions on sound is not well understood and most of the better research contradicts current standards. Here is what we can say: these distortions are not a significant cause of our perceptions of a system’s sound quality. It just isn’t something to lose sleep over.
Causes Of Non-Linear Distortion
Technically speaking, any portion of the signal chain can cause non-linear distortion. However, in a modern audio system the primary cause of significant audible non-linear distortion is either amplifier clipping or the speaker itself. Harmonic distortion increases with level, and as a system’s volume begins to exceed the range at which it can perform as intended, distortion rises until eventually it rises drastically. With amplifiers, this is the point at which the amplifier is exceeding its maximum output. With a speaker, the causes are complex, but typically marks when a speaker exceeds its mechanical limits. For most systems, our goal should be to create a linear range which exceeds the maximum volume at which we ever intend to listen. There is no metric for this, nor is there a simple way to say if a system will exceed your dynamic needs. There are a lot of causes of non-linear distortion that I’ll address in a future article.
Concluding Thoughts: What’s Next?
So, we’ve discussed the basics of distortion. Now what?
It’s important to understand the concept of distortion, both linear and non-linear, because it helps to guide room design, the value of EQ and room correction, and what parameters matter with certain components upstream from the speakers. Over the last 30 years, advancements in electronic and transducer design have allowed components (even inexpensive ones) to have very excellent measured performance. In addition, modern speaker drivers, amplifiers, preamplifiers, and DACs have negligible levels of non-linear distortion. In fact, non-linear distortion levels have become so low that other issues matter far more. Rather than chasing that last few dBs less distortion in a component, we really need to pay a lot more attention to a room’s characteristics and speakers, treating them together as a system. A speaker that measures anechoically flat (i.e., flat response when there are no reflections) will measure anything but flat in an actual room. Our main concern should always be to address the biggest detriments to good sound first. By far the biggest factor effecting the sound quality of any system is the smoothness of frequency response, or its linear distortion. Future articles will address how to measure a room, assess a room and speakers as a system, understand what measurement data means, and how to use that data to make informed decisions on improving the sound of a system. As you’ll see, gathering data and assessing it is easy, and room improvements are often simple, affordable, and highly effective.
- James Larson’s Excellent paper on the audibility of low frequency distortion
- Audioholics article on the audibility of distortion
- Earl Geddes paper on Linear Distortion
- Earl Geddes on an alternative Non-Linear Distortion Metric 1
- Earl Geddes on an alternative Non-Linear Distortion Metric 2
- Klippel’s paper on speaker distortion
- Klippel's paper on causes of speaker distortion