Jim McShane wrote:
Also, it is not unusual for a push-pull amp like the Mk III to have one tube not functioning at all yet sound normal. This is because at lower outputs the single tube will often have the entire signal waveform applied to it and will drive the speakers with the full waveform - only when the tube is driven hard(er) can you tell something is wrong. It's a long story as to why, but it has to do with how a "push-pull" output stage and the phase inverter work.
Depends on the class of the push-pull output stage.
Class B: each half of the push-pull circuit is designed to handle half the output waveform. There is (virtually) no sharing between the halves of the signal, which is split in the driver stage and "reassembled" into a whole for listening; in a transformer-output amp, this happens in the output transformer; in an output transformerless amp, the two halves are connected together at the output to the loudspeaker. Most transistor output stages are virtually Class B; there is a TINY bit of sharing only to extremely low power levels, because if the "hand-off" between the two halves is discontinuous, there is a particularly annoying form of distorion that occurs, and it's mostly audible at low volume where you actually listen most of the time and the sound isn't loud enough to mask it. This distortion was why many transistor amplifiers sounded bad.
Class A: each half of the push-pull circuit handles the entire output waveform. All the signal is amplified by both halves of the output stage. The advantage is that there is no chance of any discontinuity at "hand-off" of the signal between the halves of the circuit, because there is no hand-off. Class A amplifiers are inefficient, however. They consume a lot of electrical current for their power output, wasting a lot of that power, which is dissipated as heat.
Class AB: As the name implies, this is partway between Class A and Class B, and it's the mode in which Dynaco tube amplifiers operate. Each half of the push-pull circuit handles somewhat more than half the output waveform but less than the complete waveform. There is a significant overlap between them, where both are amplifying the signal. This minimizes the audible effect when the signal is "handed-off" because when it does occur, it is happening at higher power than in a Class B amplifier, and the relatively small amount of discontinuity can be masked by the louder signal. A Class AB amplifier uses less current and dissipates less heat than Class A, but more current and dissipates more heat than a Class B. The point at which the two halves of the circuit stop sharing the signal is varied by the bias adjustment; the higher the bias current, the more "sharing" there is of the output waveform, but the more current is used and the more heat is dissipated (more sharing: the Class AB amplifier goes farther toward Class A).
There are subclasses: Class AB1, AB2, etc., that have to do with whether the output tube grid draws power from the preceding driver stage.
Most hi-fi power amplifiers are Class AB, a compromise between low distortion and efficiency/heat dissipation (most preamps are Class A). A single EL34 (which would be in Class A by definition; no hand-off of the output signal) at its maximum rated dissipation is capable of about 10 watts output at 5% distortion. A push-pull pair of EL34s in Class A, about 20 watts. A pair of EL34s in Class B can produce 100 watts at 5% distortion but probably would sound bad. The compromise Class AB is good for 35 to 60 watts at 5% distortion, depending on how the output stage is biased.
5% distortion sounds high, and for hi-fi specification purposes it is. But you may be surprised as to how little effect it really has on the sound, since it's only at maximum output and is mostly low harmonics (the distortion created by improper hand-off in a Class B amplifier is high-harmonics...very annoying). It is also an open-loop distortion rating; in a real life amplifier low-harmonic distortion is reduced substantially by negative feedback...which is another topic.