Interesting Article in PS Audio website.
What's the difference between Class A and AB?
Most people understand that Class A runs hotter, generally sounds better, and is typically more expensive.
The term class A refers to biasing of the output stage (in a transistor power amp). Bias is an engineering term that actually means what it implies (i.e. Leaning in one direction). By that, we refer to the amount of current flowing through the output stage.
Class A is when both devices conduct at all times while Class B is when only one device is on at any instant. These are both limiting cases and there is thus a setting between these two limits where, for small signal levels, both devices conduct but for larger levels, only one conducts. This is termed Class AB.
Class AB is not a true class, but is a very common term applied to the biasing levels used in most audio amplifiers. The output stage is biased to carry a quiescent current significantly less than half the maximum output current, (as needed for full Class A) but sufficient to keep both devices running in Class A for small output signals. However as the output signal increases, the amplifier becomes Class B with one device cutting off on each half cycle. This biasing scheme effectively moves the linearity curves toward one another resulting in a transfer curve that is more linear as it passes through the origin. This scheme approaches the efficiency of class B yet offers class A distortion levels (nearly) especially for small output levels where class B suffers most.
It is particularly significant that the distortion is reduced for small signal levels, as it is at low levels that the human ear is most sensitive to distortion.
With the exception of Nelson Pass's (Pass Labs) amplifier that essentially has one 'sex' (polarity) of output device handling the entire signal, all transistor amplifiers have (in essence) two output devices. One output device handles the positive (plus) half of the signal and the other handles the negative half of the signal. If you can envision such an arrangement, a possible problem might come to mind ? the transition area. Like a relay runner handing off the baton to the next runner, the first half of the two-device output stage 'hands over' the signal to the second half of the output pair ;this transition is the problem. As the signal goes from the positive device, to the negative device there is a moment when neither device is handling the signal, creating a 'gap' in the music.
Class A biasing in its truest form means that there is a lot of current running through these two devices all the time. When you run current through the devices, it creates heat, and the more you run through the stage the greater the amount of heat. A true Class A amplifier runs the same amount of current through its output stage as it is expected to deliver to the load (the speaker). So, for instance, a 100 watt (rated) class A amplifier draws 200 watts per channel at all times (even when there is not a signal). When the amplifier is asked to deliver 100 watts of power to the loudspeaker, 100 watts (or half of the current) goes into the load (speaker) and the other half continues to go through the output stage.
One interesting fact is that under full power delivery conditions (when the amp is putting its full 100 watts into the speaker), the Class A power amp actually runs cooler. This is because only half of the power is being converted into heat in the amplifier, while the other half is busy driving the loudspeaker.