There are a few additional thing to mention. Class A does not always imply single ended operation, you can have a push-pull Class A amplifier, and the Yamaha almost certainly is. There are popular single ended Class A designs. Nelson Pass has made a career out of them, and they are very popular amongst those than can afford them.
I wrote this bit 18 years ago as a posting on rec.audio.high-end. I have cleaned it up a bit and fixed the worst bits.
What are the amplifier classes and what do they mean for my sound?
Class A, B AB C(?) new classA, super Class A, opticalA, digitalA etc.
To understand the design options it is worthwhile to return to absolute basics. Sound is a series of compressions and rarefactions of air. To reproduce these speakers are designed to move both forward and backwards about some mean point producing acoustic waves. Consequently the designer of an amplifier must build a device that supplies current to the speaker to drive it in both directions. In general the output of a modern amplifier looks like this.
-------------------------------------------------- + Volts
|
|
/ \
A ----------- | | Output device a
\ /
|
|
|
|
|
+------+ speaker +--------- 0 Volts
|
|
|
|
|
/ \
B ---------- | | Output device b
\ /
|
|
----------------------------------------------------- - Volts
The devices “a” and “b” can be turned on by applying drive signal to the corresponding points “A” and “B”. This signal is forward current for bipolar transistors or a voltage for MOSFETs (and also tubes (otherwise known as valves), but tubes need other changes to the circuit and also need an output transformer in most cases which, I haven’t shown. But the principles are the same.)
As one can see applying drive to A results in current flowing from the positive supply to through the speaker. Consequently the speaker will move one direction. Conversely applying drive to B will result in the speaker moving in the other direction.
It turns out to be possible to connect both A and B together and apply drive, positive movements of the drive result in the delivery of positive current flow to the speaker, negative drive the converse. We now have what appears (at least at first sight) to be a perfectly satisfactory current driver for our speaker.
However life is no so rosey. We are assuming that the transfer of drive to forward current through the output device is linear. For the most part this is a reasonable assumption with the majority of devices used, but only in the main operating area of the devices. At the point of turn on, the relationship from drive to output is far from
linear. This results in quite noticeable distortion, especially at low volume levels. This goes by the name of crossover distortion. It is a sad fact that the majority of an amplifier’s life is spent at quite low signal levels even with quite loud music. On a 100w amplifier a signal 20db below the absolute maximum sound level deliverable only equals 1 Watt. Hence the crossover distortion dominates the sound.
A worse complication is that most devices won’t actually start to turn on until they receive a minimum level of drive. For transistors this is about 0.6 volts. This means that in the simple circuit above you actually have a dead zone in the middle of the audio waveform which results in quite appalling distortion. An easy fix for this dead band is to add circuitry that hold the points A and B apart by the voltage needed to push the output devices to the point they are just at turn on. In this case twice 0.6 volts. This is typically done by placing a permanent drive “bias” between the points A and B. The input signal is applied to the bias network which now rises and falls as a unit with the drive signal. The magic number of 0.6 volts is a single semiconductor junction forward drop, and you can use another semiconductor with a similar forward drop to act as the device that provids the needed voltage offset. Rather usefully, a pair of simple diodes (one each side) can do this. This is class B.
However the output devices still run though a non-linear operating area and some distortion remains. The problem clearly is in allowing the output devices to run into these non-linear transfer regions. The solution is found in increasing the bias to the inputs, turning the devices so far on that neither device is ever driven into it’s non-linear region. So a permanent drive taking them to half way on. One can see that turning each device on will result in a lot of current running straight through both devices, and the output at the speaker terminal staying at mid way between them (0 volts.)
Applied signal therefore results in one output device turning on harder whilst the other turns off by a similar amount. The unfortunate consequence is that a large amount of power is being lost pushing current through the output devices even when they are delivering no power to the speaker. However we have eliminated crossover distortion and the amplifier sounds quite a lot better. There are schools of thought that suggest that other attributes of bipolar transistors are improved when a healthy current is run through them, and this is could be another reason for improved sound.
Designs that bias the output so that neither device ever reaches it’s non-linear regions are termed class A.
Typically however, a compromise is used. Designers elect to apply some bias to the output devices so that at small signals the outputs never reach their non-linear regions. Once the signal becomes large enough that they do reach the non-linear part the distortion will be dominated by the wanted signal and will not be nearly as noticeable. In general negative feedback is quite able to control the remaining distortion to acceptable levels. This is class AB.
In truth this is a very good compromise. At small signal levels the ampifier operates as in class A mode anyway and high power peaks are delivered in class B where the distortion is much less noticable relative to the high power level. This is a very successful design idea and accounts for the vast majority of pre-digital (Class-D and similar) amplifiers.
An important fact to notice is that an amplifier can only remain operating in Class A into a given minimum impeadance. Once the load drops below this impeadance either the power supply gives up and cannot keep up with the current demand and the amplifier clips producing huge amounts of distortion, or more commonly, one of the output devices is driven into its non-linear operating region and then turns off. Effectively the amplifier is really only class AB, albeit with a very large operating bias. This is part of the reason why there are such large differences between class A amplifiers. Many only really deliver class A into 8 ohms, and will revert to class B when driven at full power into lower impeadances. Others that seem to offer the same basic power levels are an order of magnitude bigger and costlier but are capable of delivering power into much lower impeadances whilst still maintaining class A operation. They run considerably hotter as a consequence too. To truely call an amplifier class A it must run in class A mode at full power into the minimum impeadance it will operate with. Otherwise it is really class AB.
On some amplifer design forums you can provoke an argument that can rage without end about the exact semantics about the difference between class A and AB. There are some precise definitions, and modern designs that don’t fit the basic ideas behind those definitions, and it all gets rather tiresome.
Some designers give the owner the option of selecting the bias level. The Yamaha “Auto Class A” does nothing more than provide a switch that selects between two bias levels. The higher bias level resuls in a hotter running amplifier and improved performance. It is a misnomer to claim that the amplifier automaticly switches between class A and Class AB. The amplifier is Class AB all along, running as class A (neither output device moving into it’s non-linear region) up to some small power level and then as the output rises above this level, one output device runs into it’s non-linear region and then turns off altogether, and the output stage operates in class B for as long as the signal is above that level.
However the story is not over. We noted that in order to eliminate cross-over distortion it is enough to never let either device be driven into its non-linear transfer region. Why not dynamicly vary the bias so that it is low when low signals are needed and higher when large signals are wanted? Carefully done it should be possible to eliminate cross-over distortion without the huge penalties in power dissipation and large power supplies. Thus is born New Class A, Super Class A, and a whole lot of other almost-but-not-quite-really class A designs. It turns out to be quite a challenge to design a circuit to dynamicly vary the bias whilst still applying signal drive to the bias circuit. One innovation is to use an optical isolator to solve some of the problems, and so was born Optical Class A.
Finally, Pure Class A is nothing more than the original Class A trying to reclaim it’s birthright back from all the pretenders to the throne.