What is Class A?

Amplifier classes of operation have been with us from the beginning of "Amplifier Time." The first amplifier engineers of the 1920s developed the terms Class A, B, and C to denote the operating bias of the output tubes in power amplifiers. Low level (preamplifier) stages are always Class A as their output power is nil, and a single-ended voltage amplifier is most easily made in Class A. Our modern preamplifiers are Class A, as are the drivers in most power amplifiers. It is only in the output stage where we find the "Class" distinction, and what a distinction it has become. Everyone wants a Class A amplifier. It has become a seal of approval, and the audio industry is obliging consumers. The term is emblazoned upon advertising, literature, and even on the front panel of some units. What follows explains Class A technically, and how to easily spot amplifiers that fraudulently claim to be something they are not.

First, let us define the classes of operation as the authors intended. I will use vacuum tube terms, although this is applicable to transistors as well. Class A is a single-ended or push-pull amplifier where the grid is negatively biased so that it is about half-way between 0 volts and the more negative value, which will cause the tube to "cut off" the flow of plate (output) current. Thus, the tube is about halfway between full-on and full-off and can equally swing around this center point with minimal distortion. The ratio between grid swing and plate swing is not linear, so there is some distortion.

The main advantage of Class A is in the large overlap in conduction of the push-pull pair of tubes. The upward non-linearity of one tube is somewhat balanced by the downward non-linearity of the other. To work well, the tubes must be matched for bias and transconductance. Unfortunately, the non-linearities do not cancel completely and there is some distortion. The appeal of Class A is that it is the lowest distortion of all the classes, though only by a factor of 2-to-4 as a relative increase over Class AB1, and at great expense of heat dissipation and tube life.

Class B is defined by biasing the tubes at a cutoff so there is no overlap in the push-pull pair. This results in less linearity and some crossover distortion as the pair goes through zero. This was the largest fault of early transistor amplifiers until designers realized that transistors needed positive bias to be somewhat cut-on. Stabilizing this small, critical voltage was a problem for a great number of years, evidence that a new technology takes time to mature. This partial cut-on bias and its resulting idling current takes us into Class AB1, that ill-defined continuum between Class A and Class B. However, before expounding on the AB1 issue, let us take a brief look at Class AB2 and Class C.

Class AB2 is biased just as Class AB1, but at peak signal the grids are driven above 0 volts (positive). Tubes intended for this class of service will conduct more current, thus giving more output power. Tube types 6L6 and 807 are intended for Class AB2 service and achieve more than twice their AB1 power (47watts vs. 18 watts) with identical plate and screen voltages. Thus, Class AB2 amplifiers have low distortion and high efficiency, yielding the most watts per pair of output tubes. For example, the efficiency of the 6L6 is 65% for Class AB2 and only 36% for Class AB1. In addition, the dissipation in Class AB2 is 12.5 watts vs. 16 watts for Class AB1, so it runs cooler while putting out more power. Tubes optimized for Class AB1, such as EL-34, KT-88, and 6550, generally achieve around 50% efficiency in circuits with 450 volts B+ and proper loading. Every circuit is different, but this is a good rule of thumb. Unfortunately, Class AB2 amplifiers require positive grid drive, which must supply power to the output stage. This usually is achieved by a driver transformer (a second transformer smaller than the output transformer) and is generally undesirable for audio applications. These amplifiers were popular pre-WWII and were often found in radio transmitter modulators and Wurlitzer jukeboxes.

Lastly, the amplifier class of operation known as Class C is where the tubes are biased well into cutoff. Due to this the distortion is enormous. As such, it is unused in high fidelity audio. The main application of Class C is the final amplifier in a radio transmitter, where the tube supplies pulse to the resonant tank circuit, which then oscillates at the broadcast frequency.

Having defined Class A, we can now explore a proper design of that output stage. The toughest aspect of the design is satisfying the high dissipation required by Class A bias. The RCA data gives a 6L6GC design where the idle dissipation is 36 watts, and the output power is 17.5 watts with 42% efficiency, which is about what we would expect. What is interesting to note is the idling dissipation is twice the maximum undistorted power output. In measurements of a typical ultra-linear output stage like the Class AB1 Music Reference RM-9 at 100 watts per channel, the following was measured:

1. At an idle dissipation of 55 watts (the recommended RM-9 setting which achieves 10,000 hour tube life), the amplifier output stage was 47% efficient at full output.

2. Maximum Class A power, as observed by noting where cathode current in the tube going toward cutoff approached 0, was around 12 watts.

3. At a bias setting of 100 watts, efficiency dropped to 46% and Class A power rose to 50 watts. However, dissipation of 100 watts is 25 watts per tube and would shorten tube life to 1000-2000 hours.

4. The highest level of idle tested was 160 watts, at which point the EL-34's were straining under the heat, producing only 42% efficiency and 90 watts which was full output as the power supply voltage had dropped due to the high current draw.

All that was gained by this in the RM-9 was to reduce the 1KHz, 90 watt distortion from .27% at 55 watts idle to .14% at 100 watts idle, to .09% at 160 watts idle. Keep in mind that the RM-9 was designed to sound good and have low distortion with 55 watt bias (Class AB1), whereas other amplifiers may need higher bias to sound good.

Out of the above study we can make a generalization from the last two data points that Class A dissipation at idle is about twice the maximum Class A power. Data point #1 is not considered as the point of Class A departure, as it is too far below the maximum output power. A supporting argument is the one for constant current draw by the output stage defining Class A. Using this and the typical efficiency of 50% for the output stage, the idle power must be 200 watts, so that at full power 100 watts can be delivered to the speaker load (Pout-Pin x Efficiency). This argument is so simple and easy to remember, yet so unknown. You can have lots of fun at audio designer/enthusiast cocktail parties with this one. You may be surprised at how little the issue of dissipation has been considered. The grid cutoff argument is the original definition of the classes but is more complex.

Now we can get to the bottom line of the question of who is and who is not really Class A. At my 10,000 hour rating of 12.5 watts per EL-34, 20 watts per Sylvania 6550, and 15 watts for other 6550s, we can easily calculate the number needed for a 100 watt amplifier. The 200 watt idling dissipation requires 16 EL-34s, or 10 Sylvania 6550s, or 14 other 6550s. I know of one British Class A amplifier using 12 EL-34 output tubes that does achieve 80 watts with 220 watts idle dissipation, very Class A. Unfortunately, the tubes idle at 18.4 watts each. Their published maximum rating is 12 watts, and my long-life rating is 8 watts. I could almost see the gettering disappear as it played. I was not surprised when they did not last very long and caused severe damage to the amp when they failed. It would not be wise to base your purchasing decisions on the unfounded claim of Class A, but surprisingly many people do. I hope that this look into the real origins of class distinction in audio has shed some light for you.