The "Inherent Linearity" of Class A Amplifiers?

I'm what they used to call an audiophile. What with folks listening to music on cell phones, low bit-rate MP3's and the like, it's kind of out of fashion these days. That is, unless you're into music production. In that case, you can buy all manner of interesting goodies, including microphone preamps that will set you back a kilobuck and loudspeaker cables that could pay a semester's tuition. Unfortunately, just as it was true for hi-fi enthusiasts 30 years ago, there's a lot of misinformation floating around in the semi-pro or "prosumer" music field regarding audio circuitry.

I was reading the January issue of EQ Magazine the other day and came across an article entitled "A Touch of Class" by Vincent Miraglia and Rich Tozzoli (full disclosure: many years ago I wrote several articles for EQ's executive editor, Craig Anderton, when he was piloting Electronic Musician Magazine, although I never wrote one regarding the topic following). The article is an explanation of different types of amplifier classes (most notably class A and class B) and how this affects sound quality. I think it's great that the authors are trying to impart some technical knowledge to an audience who generally do not have engineering or technical backgrounds, thus enabling them to get more out of their equipment or make more intelligent purchases. Unfortunately, the authors are guilty of spreading misinformation. Consider this outtake:

What makes a Class A amp great, you may ask?
The sound quality is pure
The design is inherently linear, because the output varies in direct proportion to the input
It introduces no crossover or zero-crossing distortion

Are you getting the idea that class A amps are the cat's posterior? The authors go on to say that the downside includes inefficiency, heat, and relatively low output power. Class B amplifiers, while they are more efficient and offer higher output power, suffer from the afore-mentioned crossover distortion.

So, what's my problem? Well, there's absolutely nothing about class A topology that implies inherent linearity or "pure" sound quality. Class A operation simply means that the amplifying device is active for 360 degrees of the output waveform, whereas class B means that an amplifying device is active for only 180 degrees (and thus, a need for two devices to fill out the entire waveform cycle). Miraglia and Tozzoli state as much, and even go so far as to say that many different amplifying may be used, such as tubes, JFETs and MOSFETs, but don't seem to grasp the concept that these devices are themselves non-linear. (Oddly, they do not mention BJTs, the very common bipolar junction transistor, although they do mention ICs which I find very odd as ICs are themselves normally made of BJTs and FETs.)

Basically, a class A amplifier idles its output devices at one-half of maximum current. As the input voltage swings back and forth, it causes the output current to increase and decrease around this idle point. Precisely how this happens is dictated by the voltage-current characteristic of the device, technically known as its transconductance curve. If the curve is a straight line, then by definition, we have a linear input-output characteristic and supposedly, "pure" sound quality. Note that linearity is a function of the amplifying device, not of the circuit topology being class A. Unfortunately, no amplifying devices in use are perfectly linear. For example, bipolar transistors have a logarithmic characteristic while FETs have a parabolic (square law) characteristic. Obviously, if the voltage-current characteristic is non-linear, then the output signal will not follow the input perfectly, thus indicating distortion (it should also be noted that the output will vary in direct proportion to the input, that is, a larger input will produce a larger output, but it will not do so perfectly, thus their second statement is a non-starter).

OK, so you're thinking "How bad can this distortion be? Is it all just academic and class B's are worse anyway?" Nope. A class A amplifier can produce prodigious distortion. In fact, one of the things I do in my linear electronics lab is to show my students how to take a class A amplifier with heavy distortion and, through minor circuit modifications, reduce the distortion by an order of magnitude. The authors are absolutely correct when they say that class A amps don't suffer from crossover distortion while class B amps do, but this doesn't necessarily mean that class B amps have higher overall distortion inherently. (Crossover distortion, which can be mitigated through proper design, is the result of the two halves of the waveform being "stitched together" by two amplifying devices. Basically, one takes the positive polarity and the other handles the negative. They both idle at zero, thus promoting high efficiency. When conduction is handed off from one device to another, there will be a small glitch, and that's crossover distortion.)

Remember, class B amps use the same sort of devices as do class A, and they'll suffer from the same non-linearities, but there's an interesting twist. As a class A amplifier swings above and below its idle point, the curvature of the transconductance characteristic tends to elongate one half of the waveform while squishing the other half. In contrast, Class B's tend to elongate both halves. This means that practical class A amplifiers produce waveforms without halfwave symmetry while the class B's maintain halfwave symmetry. If a waveform does not have halfwave symmetry, then the distortion products must include even as well as odd harmonics (with symmetry, there will only be odd harmonic distortion).

Ultimately, it all comes down to the details of the individual designs. It is entirely possible for a class A amplifier to produce far more distortion than a class B amplifier. And to keep matters more interesting, if an amplifier is being used by a guitarist to produce a certain sound quality, rather than as part of a "sonically transparent" mixdown system, distortion might be a desired commodity.