[SoundStage!]Max dB with Doug Blackburn
Back Issue Article
December 1998

Seeing Through The Fog

Most audiophiles would like to believe that everything they read that purports to describe important design elements of audio components really are important. Most of us would also like to believe that the reasons for the superior performance of these design elements are the reasons given by the manufacturer (or reviewer).

Here at SoundStage! we are doing our best not to make up our own explanations for technical aspects of product designs. In our reviews you will find that we clearly will say that design/performance claims made for components are clearly those of the manufacturer and not of our reviewers’ invention. We generally won’t comment on manufacturer claims in reviews unless there is something clearly amiss or noteworthy. However, here in the SoundStage! monthly columns we can further explore manufacturer claims from a general perspective and discuss how they might be important to the sound of a product.

This month we’ll look at some of the over-used, under-used and misunderstood performance parameters in audio components. The goal here is to help audiophiles, new and old, get a better grip on what’s important and what’s not. I’m also going to try to stamp out incorrect use of the terms phase and polarity. Wish me luck!

Distortion -- meaning and perspective

Most discussion about distortion focuses on a single type of distortion, harmonic distortion (THD for total harmonic distortion). Harmonic distortion is simply the amount of additional harmonic content a component adds to a musical signal. One percent total harmonic distortion at xx watts means when the component is reproducing a pure tone, say 400Hz, it will reproduce 99% 400Hz and 1% byproducts. Those byproducts are harmonic distortion -- in this example the byproducts will be other frequencies, mostly 800Hz with a smaller amount of 1200Hz, even less 1600Hz, even less 2000Hz, etc. When harmonic-distortion claims are made for amplifiers, the manufacturer may select test conditions that vary from the industry-standard test conditions. When a manufacturer is trying to hide poor harmonic-distortion performance, he will typically drive only one channel of the amplifier, giving the power supply almost completely to the one driven channel (assuming this is a stereo amp) and the wattage the test is run at may be on the low side.

Single-ended triode amps are often subject to very gentle total-harmonic-distortion testing because they are usually low-powered amplifiers with the poorest total-harmonic-distortion performance of any common audio amplifier being made as new products in the late 1990s. Test conditions for THD measurements can be made in such a way as to camouflage the fact that a low-powered amplifier may be delivering 10% THD even when operating at 10% to 50% below full rated output. This means that the signal coming out of the amplifier is 90% original and 10% byproducts, which do not exist live and which were not captured in the original recording.

If you find yourself liking the sound of a single-ended tube amplifier, that’s OK. You just need to be aware of what is going on. The sound you like isn’t more accurate or less distorted than the audio signals generated by other components. So the single-ended-amplifier listening experience should never be characterized as some radically more accurate presentation -- ever. It can and should be appreciated for what it is, a modification/presentation of the original recording that is not reproducible by other types of amplifiers. Characterizing single-ended amps as having high purity of tonal quality is a mistaken attribution on the part of someone (fan, owner, manufacturer or reviewer) who does not understand the signal that goes into the amplifier and how it differs from the signal coming out of the amplifier. The exiting signal has a bunch of extra stuff traveling along with the music signal, stuff that was not captured on the original recording. Again, this is OK as long as everybody understands it and accepts the reality of what is happening with one of these amplifiers.

THD specs that are made with both channels driven at an output level that is some significant fraction (1/4 or higher) of full rated power are the most meaningful and useful measurements. Furthermore, letting the amp get off easy by reproducing only a single frequency will favor the amp to a significant degree when specs are established. The best testing method would subject the amplifier to a wide-band audio signal that represents the complex content of music. Electronic components and loudspeakers often perform very differently when tested with a complex music-like signal versus a single-frequency sine wave from a signal generator. Imagine a car magazine doing acceleration and top-speed tests downhill with a 60mph tail wind and you get some idea of how THD measurements can be fudged to make a component look better than it really is.

Time domain distortion

Unfortunately, because of emphasis by manufacturers and the audio press on total harmonic distortion and almost complete lack of consideration of other types of distortion, audiophiles are apt to forget about other common forms of distortion. Frequency response is usually stated for loudspeakers and for some audio components, but it does not relate directly to distortion, or does it? Frequency response can be reduced to amplitude and time components. Amplitude distortion is covered under the general total harmonic distortion heading. But the time element of frequency response is also important. Phase shifts and time delays as the audio signal passes though electronic components and loudspeakers affect what you hear in your room, although in ways that are not as obvious as THD. Time distortions are more likely to show up more in your emotional response to what you are listening to than as some identifiable tonal difference. Audiophiles often overlook this important performance parameter in their headlong rush to achieve maximum detail and transparency.

What phase and phase shift (or phase error) really are

You are in the 100-yard dash with seven other runners and all of you run with identical times. You are waiting for the starting pistol to sound, but the other runners have already left the blocks, having already heard their starting pistol. Two strides later you hear your starting pistol and take off. The other runners all finish two strides ahead of you. You were just phase shifted. When your system is reproducing music, if every tone, every detail, started at exactly the correct time, exactly as it was captured on the original recordings, there would be no phase shift in the system. However, if some details (or frequencies) are shifted in time, their phase is distorted in comparison to the original recording.

There is some discussion in the audio community about how audible phase shifts really are. Since these phase shifts are frequency dependent in electronic components and loudspeakers, they are, at least, consistent. By this I mean that when phase error exists in a component, that component treats every specific frequency the same way (i.e. 1,000Hz is always shifted the same amount, 200Hz is always shifted the same amount, etc.). If phase were shifted dynamically, it would be very audible. If you are a fan of rock music, you have no doubt heard the space-y effects that dynamic phase shifting makes -- they are very audible and very wrong-sounding, although they can be quite entertaining in some of recordings. Producers use dynamic phase shifting to create odd spatial effects, and it works exceptionally well.

Not all that many electronic components have their phase performance described. Loudspeakers are the primary source of phase distortion in audio systems. Crossover designs that employ second-, third-, fourth- or higher-order slopes can’t help but introduce phase distortion. These phase distortions are static in that they do not change while we listen, so they are not as obvious, again, as tonality or transparency or detail. Because of that, many audiophiles overlook phase shifts in their systems. Remember the 100-yard dash? An audio system with phase shifts has some sounds exiting the loudspeaker at various time delays compared to when the sound should be leaving the loudspeaker. The audible results of these phase shifts are also going to be discovered in the emotional/artistic side of the brain -- the opposite side from the one many audiophiles use when they listen, the analytical/logical side. Because of this, the effects of phase shifts in audio systems are often missed until later in the ownership cycle when it gets increasingly difficult to enjoy the system emotionally while finding it exceptional-sounding from an analytical point of view.

There is a body of research and a bunch of loudspeaker manufacturers who claim that time delays introduced in loudspeaker crossovers (1ms up to several milliseconds) are below the threshold of audibility. And surely they are below the threshold of audibility in the analytical side of the brain. However, I’m not so sure about that other side of the brain that we forget about while we are selecting speakers and getting them placed in the right locations. Six months later when we’re done with all the analytical listening, we just can’t forget the speakers and enjoy them. Even when reviewing loudspeakers, how many reviewers live with the loudspeakers long enough (nine months or more) to get over the analytical phase so they can explore the performance of the loudspeaker in the other side of the brain? Not many. This is one reason you should be skeptical of reviews to a certain extent. Note here that I am directing your skepticism to the reviews, not the reviewers. I think most reviewers are accurately reporting their experience with whatever product they are reviewing. But reviewers tend to be perpetually locked in analytical mode because they always have one or two or more things in the system that are under review. This means there is little opportunity to kick back with a given product to see how emotionally it grabs you when you are not in analytical mode.

Audio amplifiers can have large phase-shift problems too. Tube amplifiers, as a group, are worse in respect to phase performance than solid-state amps in general; however, many designers of solid-state amps are either oblivious to designing for minimum phase shift or are convinced it doesn’t matter, so they don’t bother to fix or prevent phase shifts in their designs. The designers who aren’t bothering with minimizing phase shift aren’t mentioning phase performance in the specifications either. Amplifiers that do specify phase performance and that claim less than 10 degrees of phase shift across the 20Hz-to-20kHz audio spectrum are going to deliver you a system with lower levels of total distortion compared to poorer performing amplifiers -- all else being equal of course.

Square-wave response is the easiest way to get an idea about how a component performs as far as phase error goes. Proper square-wave performance with near-zero phase shift means that the plateaus (tops) of the square waves are flat. The more tilted the tops are (angled down to the right), the more severe is the product’s deviation from minimum phase-error performance. However, any given square wave only documents the component’s general phase-error performance at the single frequency represented by that specific square wave. You’d need to see square-wave response of many frequencies in order to form an accurate picture of the component’s phase-error performance. If you do see square-wave performance for any given component, you’ll be lucky to find waveforms for more than two frequencies. But even two frequencies would be instructional, especially if one was, for example, 1000Hz or higher and the other was 100Hz or so. 10,000Hz is sometimes used as the higher-frequency square wave.

How bad can phase performance in an amplifier be? First you need to understand phase. Each cycle (a full positive/negative cycle) contains 360 degrees. A 180-degree phase shift is a 1/2 cycle phase shift. A 60-degree phase shift is 1/6 of a cycle phase shift. Cycle is a variable thing in audio though. At low frequencies, a single cycle reproduced in air is quite a long distance -- 100Hz (Hz is an abbreviation for Hertz which means "cycles per second" a measurement of frequency), about 11 feet (depends on atmospheric pressure, here I used 1,100 feet/second as the speed of sound). This means a 60-degree phase shift at 100Hz represents 1/6 of 11 feet, about 1.8 feet. This means in your 100-yard dash, the 100Hz "runner" would get off 1.8 feet after everybody else, putting him slightly out of sync with them for the entire race.

You may see phase-shift specifications that are very good at low frequencies, but which deteriorate dramatically at high frequencies. What does this mean in terms of what is happening in your system? Well, at low frequencies where wavelengths are long, the phase-induced delays will be minimal if the low-frequency phase error is close to zero. But at high frequencies you might see as much as 720 degrees of phase shift, two full cycles. If we consider a 720-degree phase shift at 11,000Hz, the distance involved is the distance of two cycles. At 11,000Hz, a cycle is just 1.2 inches long. So two cycles would be 2.4 inches. In this case, there would only be a 2.4-inch delay at 11,000 cycles shifted 720 degrees compared to the 1.8-foot delay at 100Hz shifted only 60 degrees. What this means in your system is that a component that introduces a 720-degree phase shift at 11,000Hz is equivalent to repositioning the tweeter in your loudspeaker so that it is 2.4 inches farther from the listener. Many people spend months moving speakers around over distances that are 2 inches or smaller, and they hear differences. Whether the distance change is physical or electrical in nature, the end result is ultimately audible.

The preceding paragraph is intentionally simplified a bit for two reasons: (1) to make the point clearly and (2) to avoid another four pages of explanation of loudspeaker crossovers and compensation that can be applied to account for driver position. In the real world, people are not screaming bloody murder about how terrible the majority of loudspeakers and electronic components that introduce significant phase errors sound. There may be many reasons for this. Manufacturers would prefer that you believe it is because phase/time performance below a certain threshold is inconsequential and inaudible. I’m not necessarily positive that it is audible, but there is a sense that something is responsible for people being unhappy with equipment -- that "I just can’t get into it like I could with my old ________." I can’t prove the importance of or necessity for very low phase/time-error performance. However, let’s assume that the noble goal (my opinion of the noble goal anyway) of this high-end avocation is putting together a system that does as little damage as possible to the recorded signal and that provides you with a direct emotional connection to the music. All other things being equal, it seems to me that selecting the component with better time/phase performance only moves you closer to the elusive noble goal.

The problem with music is that an electrical music signal is rarely symmetrical for any length of time. The positive and negative cycles change radically in their content. This makes pinpointing any particular frequency impossible without a spectrum analyzer. This also means that the musical signal will be fairly significantly altered when some portions of it are shifted, for example, 40 degrees while other portions (determined by frequency, remember) are shifted by different amounts. These alterations will change the audio signal from the ideal reproduction of what was originally recorded. Whether you detect the changes or not depends on what you are listening for and your listening mode (analytical or emotional).

System electrical polarity

As long as we’re talking about phase, let’s introduce electrical polarity too. Phase and electrical polarity are incorrectly used interchangeably by many less technical people who write about high-end audio. They aren’t the same thing. It’s time to get electrical polarity and phase correct.

First, consider your camera, which operates on AA batteries. You will note that on the cover or somewhere nearby there will be + and - designations to show you the right way to install the batteries. The + and - represent the electrical polarity of the batteries. The batteries are either installed correctly, or they aren’t. Electrical polarity has only two states. For this discussion, I’ll call the two states correct and reversed.

If you are thinking ahead, you’ll have already realized that phase is a continuously shifting time-based thing. Phase is not an electrical thing -- at least not for the purposes of this discussion. There is electrical phase, but it is an AC-voltage issue that has meaning in that environment. For purposes of this discussion we are going to forget AC electrical phase and talk only about polarity of the audio signal and phase shift of the audio signal.

If you are reproducing a continuous sine wave that is not varying with time, you could shift the phase 180 degrees and achieve the same thing as reversing its electrical polarity. However, music is not a continuous sine wave. It varies asymmetrically with time. Shift a music signal 180 degrees and you fundamentally change when sounds occur and whether the first sound to arrive at the listener is a compression of air or a rarefaction of air (this is how sound travels though air). Reverse the electrical polarity of the audio signal and you change how the sound is propagated into your room. On a drum whack, speaker cones or panels should move forward first, compressing the air, then backwards. Reverse electrical polarity and the speaker cone or panel moves inward (away from the listener) first, causing an initial rarefaction of air. Shift the phase of the drum whack and you move when it happens in time without affecting the polarity. A phase-shifted drum whack will always be positive-going first. The only way to make the drum whack backwards is to reverse the electrical polarity.

Setting up your system so that it has the correct electrical polarity all the way though is important and something worth keeping in mind as you change components. Preamplifiers, especially tube preamplifiers, will often reverse electrical polarity. To make sure your system has correct polarity so that the speaker cones or panels move outward on pressure waves when the preamplifier reverses electrical polarity you will usually reverse the speaker leads when you connect them to the loudspeakers. Two polarity reversals in the system makes the electrical polarity correct for the system (so would four, six or eight polarity reversals).

Going back to the 100-yard dash metaphor, electrical polarity is analogous to which foot your runner starts on, the left or right. Phase shift is analogous to delaying the start time for your runner compared to the other runners. You can see that delaying your runner 1/2 second does not affect which foot he starts with. Polarity and phase are two separate concepts when taken on the context of an audio signal.

Power response (full load versus part load)

Earlier I touched on testing equipment at a significant fraction of its full rated power and using a wide-band music-like signal rather than a single-frequency sine wave. There are sound reasons the more difficult test tells you more about the performance of the component. Power supplies are modulated by the signal the component is reproducing. If a single frequency is used in the test, the power supply could be optimized to minimize the effects of modulation at that frequency. The power supply could perform much worse when providing power to a typical music signal.

As much as designers would wish that single-frequency testing tells the full story, it just isn’t so. Music is a very complex signal with little repeat energy within the timeframe that is critical in power supplies in audio components. What we humans detect as repeating elements in music (the beat for example) are quite spaced out in time compared to what is happening in the power supply within the timeframe critical for good audio performance. If a strong dynamic peak sucks a lot of energy from the supply before the line voltage goes high enough to allow the supply to recharge again, the supply can wimp out on another event immediately following the dynamic peak. You hear this clearly when listening to music, but you’ll miss the problem completely when analyzing the performance of the amplifier using simple constant-amplitude sine waves. This, in fact, is one of the factors that cause listening tests to produce different results than bench testing might lead you to expect. It’s also one of the factors that can make printed specifications completely incapable of predicting the subjective sonic performance of audio components.

Wrapping up

Lower TOTAL distortion is preferable to higher TOTAL distortion in an audio system. OK, call this a preference of mine if you must, but the recording process doesn’t remove anything that needs to be added back in by the audio system, and even if it did, there’d be no way for the audio system to know the right amounts to add back in for each different recording. Distortion means something has altered the input signal; to my way of thinking, this is not a good thing.

THD is only one kind of distortion. There are other types that are not often mentioned on spec sheets and that many audiophiles have great difficulty observing and understanding.

Specs are inadequate predictors of TOTAL distortion present in audio components.

Phase is a time-related concept, continuously variable (as applied to electrical audio signals).

Polarity is right or wrong, positive or negative, correct or reversed.

You won’t be able to analyze systems and components for time- or phase-related errors the same way you analyze tonality, transparency, detail or other audiophile performance parameters. You will need to get out of analytical mode and into a strictly emotional mode to do that.

Manufacturers often quote specs that minimize obvious problems with certain components. It’s not that they are necessarily misstating the facts, but they can optimize test conditions to minimize poor performance in the numbers that are generated by the tests.

Audio systems cannot reproduce live performances! The first obstacle is the recording. Recordings themselves are not accurate records of the live event. Therefore, the best we can ask of our systems is to reproduce recordings with as little error (i.e. distortion) as possible. To ask the system to reproduce a live performance means the system will distort the input audio signal. Because every recording would have to be distorted in different ways to sound live, the audio system would have to have continuously variable and self-correcting distortion levels. This is impossible as of 1998, so again, the best bet sure seems to be a system with the lowest levels possible of every type of distortion.

Some people prefer the sound of audio components with high levels of distortion. This is OK as long as everybody is on the same page of the same book and understands what these high-distortion components are doing to the recording. There should be no public deception or self deception about what these high-distortion components are doing to the audio signal. If you, the owner/user, understand this and still prefer that kind of sound -- great, that’s why the whole pantheon of audio components exists. All audiophiles should be able to find something that fits their personal ideals of what they want to hear in their systems.

Rather than pursuing the unachievable goal of selecting an audio system that reproduces a live musical event in your listening room, it is more realistic and less frustrating to assemble a system that reproduces the recordings with the least amount of editorial comment (distortion). We can do only as good as the recordings.

...Doug Blackburn


[SoundStage!]All Contents
Copyright 1998 SoundStage!
All Rights Reserved