http://www.hardwareanalysis.com/content ... amplifier/
By: Sander Sassen
Despite the fact that class-A operation is rather inefficient, generating lots of heat and requiring a large heatsink there are some compelling arguments in its favor. One of these is the absence of cross-over distortion, making it a good starting point for designing an amplifier with nearly immeasurable distortion levels. The DIY amplifier presented here offers just that, and can be scaled up to 100-watts in 8-ohm, which should be enough for the majority of applications. So without further ado, lets take a closer look at the design of this amplifier and what it is capable of.
Generally speaking there are two methods for setting the output stage bias current of a transistor amplifier to guarantee sufficiently low distortion levels. One is "optimal class B", a setting where the transconductance of the output devices at idle is the reciprocal of the degeneration resistors. At this bias setting, open loop output impedance and gain are maximally constant and distortion is the minimized. Amplifiers that use this approach are commonly referred to as "class-AB" amplifiers. The other means of setting the bias current is class-A. Here the output stage transistors are biased quite heavily, hence the bias current is substantially larger than with the optimized class-B setting. With this setting both output transistors conduct unless the output current exceeds the bias current. An output stage can be said to be operating in class A until that point. When an amplifier is said to "be" class A, this merely means that under normal operating conditions it is not supposed to go into class B.
For this design a class-A bias setting was preferred, for a number of reasons. First and foremost because there is no simple way to guarantee that an optimized class-B will also maintain its optimum setting. Especially a quick modulation of the die temperature of the output transistors caused by a changing dynamic load will cause for the setting to shift. The effects of this shift will not immediately show up in harmonic distortion measurements, but will however clearly be visible in intermodulation distortion measurements. Secondly, although the distortion of an optimized class-B can be kept low, it will consist mainly of higher harmonics, whereas class-A has mainly 2nd or 3rd order harmonics. Finally there is the issue of PCB routing and layout. Since the power supply currents are not distorted, magnetic coupling between the rails and the input stage is less likely to elicit distortion, making it easier for the novice builder to design a circuit board.