Musical Fidelity AMS100 power amplifier
I was talking last winter to Musical Fidelity's Antony Michaelson, who had been enthusing about his forthcoming stereo amplifier, the AMS100. It would be physically enormousalmost a yard deepand commensurately heavy at 220 lbs. Despite its bulk, its maximum rated output would be just 100Wpc into 8 ohms. It would also be expensive, at $19,999. And to cock a snoot at environmentalists and their concerns, the AMS100's output stage would be biased into class-A up to its rated 8 ohm power, meaning that, even when not playing music, it will draw around 10 amps from a typical US wall supply of 120V. This also means that it will run very hot, making the amplifier impracticable for summer use in homes without central air-conditioning. Like mine.
Antony dismissed my concerns with a snort. "The AMS100 realizes the full true potential of class-A sound. The AMS100 has been designed for an elite band of audiophile purists who want the ultimate class-A amplifier ever made. We don't expect to sell many. Those lucky few who get their hands on one will be fulfilled."
For a while, I became one of those lucky few.
Class A class-A
The very first amplifier I ever reviewed, for the August 1983 issue of the British magazine Hi-Fi News & Record Review, was a class-A design: the Krell KSA-50. Small-signal amplification circuits are almost universally run in class-A, in which the transistors conduct current continuously. But this is impracticable for an output stage because of the sheer amount of bias current required, which must be half the peak current at maximum power. The KSA-50's output transistors, for example, continuously passed 1.8 amps at idle, which allowed a maximum power of just 50W into 8 ohms before the upper and lower output transistors turned off at, respectively, the signal's maximum negative and positive voltage excursions in each cycle of the signal.
When the transistors are alternately turned on and off in every cycle, this is called class-B operation; if there were no standing bias current, there would be severe distortion every time one transistor turned off and the other turned on. This is called crossover distortion, and results from the fact that a finite gate or base voltage (the "cut-in" voltage) must be present before the transistor will begin to conduct current. Crossover distortion consists of high-order harmonics and is very audible. It also differs from nearly every natural phenomenon in that, instead of being monotonic ie, proportional to levelcrossover distortion is independent of level. As the signal decreases, the distortion increases as a percentage of that signal. Designers cope with this by arranging for just enough bias current to be present at all times to keep the transistor operating in the linear portion of its transfer function, which is called class-A/B operation. This significantly reduces crossover distortion, but the circuit still depends on loop negative feedback to eliminate it altogether.
There are advantages to class-A output-stage operation. As both transistors are conducting current throughout the entire signal cycle, there's no crossover distortion, which means less need for negative feedback. Both the current gain and the cut-in voltage of a transistor are dependent, in a nonlinear manner, on the transistor's junction temperature; if that fluctuates, then the current amplification of the transistor will be modulated by the change in temperature. With class-A operation, the transistors are handling the same average current at all power levels. Those transistors are therefore in thermal equilibrium and are not being operated anywhere near the cut-in voltage. With class-A operation, the power supply is under constant stress, whether or not the signal is present. As long as the maximum signal-voltage swing remains below the troughs of the rectifier ripple, the power supply is effectively regulated.
By contrast, with class-B operation, the demand on the power supply is signal-related. If the power supply is regulated, or at least of a low enough impedance across the audioband to minimize any such effects, then there should be no problems. But if, as would appear to be the case, the power supply is the first area of an amplifier to be compromised during the design phase, in the need to keep costs downwhy go to the expense of a transformer, capacitors, etc., capable of giving the current required at maximum signal voltage, if that current will only rarely be required?then signal modulation of such factors as power-supply impedance may well occur. With a class-A designwhere, as Gramophone magazine's Geoffrey Horn once put it, "the output [devices] dissipate more watts when silence reigns than when the entire London Symphony Orchestra lets fly with all they have"if the power supply is compromised, it just can't cope: hum and noise join in not only with the LSO, but also during silences. Class-A operation thus mandates well-sorted power-supply design and implementation.
But, as I mentioned earlier, there is a price to pay for all this good stuff. Because a class-A circuit conducts all the time, even when there is no signal, it is very inefficient, and the wasted power is dissipated as heat. This requires a power supply larger than would be dictated by the demands of the signal alone, and demands an effective method of dissipating the heat. The theoretical maximum efficiency of a class-A stage handling signal is 50%. A class-B circuit, however, wastes no power when there is no signal, and can reach a maximum efficiency of 78.5% when handling a signal, meaning that such amplifiers can be lighter (and cheaper).
The continuous power requirement of class-A operation places severe demands on the specification of the output transistors. The transistors of a class-A amplifier capable of 10W output must dissipate 20W (assuming the theoretical 50% efficiency). If two transistors are used, each therefore must be capable of continuously dissipating 10W. If the same two transistors are used as a class-B push-pull pair to give the same 10W maximum output, the maximum power dissipation in the transistors occurs at about one-third full power; in this case, around 4W. For the class-B amplifier, each transistor need dissipate only 2W and can be less highly specified. Alternativelyand more realisticallyfor the same investment in transistors, a class-B amplifier can be designed to be capable of some five times the output power of an equivalent class-A design.
It's not surprising, therefore, that the output stages of 99.99% of commercially available power amplifiers are run in class-A/B.