Signals always involve at least two paths to complete a circuit. Send some electrons down the center lead of a guitar cable, for example, and you get an equivalent number back on the shield. Because the shaft of the plug is connected to ground at the amplifier jack, we tend to think of its voltage as always zero. The voltage at the tip of the plug, on the other hand, varies with the signal. This conceptual understanding works fine in most amplifier work.
When the voltage at the grid of a 12AX7 triode rises relative to ground we can say with equal authority that ground drops in voltage relative to the grid. The fact that one statement seems scientifically reasonable and the other a sign of imminent psychosis is because of the substantial physical differences between the two reference points. One is just a tiny pin at the base of the tube socket. The other is a massive steel chassis connected by the power cable to a kilometer of big green wires in the wall. The building's multi-ampere distribution system connects the power cable to an electrical service panel bonded to a big copper rod that is thrust deep into the earth. So in effect our planet earth connects your chassis to my test bench, a Vox AC30 warming up in Tokyo, and thousands of other working amps.
The send and return path for the signal, in this case the tube pin and the chassis, are physically and electrically different. The impedances driving them are different. The circuitry connected to them is unsymmetrical. We call this type of signal transmission "unbalanced."
A simple triode preamp is unbalanced. So is a single-ended power amp. When the circuit is symmetrical, like for a pair of power pentodes in push-pull, and when both phases are driven by equivalent impedances, then the circuit is said to be "balanced." The power tube grids vary with respect to ground in an equal and opposite way. The signal is represented by a difference in voltage between identically configured, mirror-image circuits.
Balance has its advantages. For example, AC ripple in the power supply tends to change the voltage of both signal phases in the same direction. Both phases either increase or decrease together, and since the voltage difference between the phases remains the same, the effects cancel. For guitar amps there are important harmonic implications. Single-ended amplifiers remain unbalanced through the entire signal path from the guitar to the output transformer. In each stage the amplification factor varies as the signal swings positive and negative, producing second harmonic distortion.
Traditional push-pull amplifiers remain unbalanced through the preamp stages but shift into balanced-mode at the power stage, where equal distortion occurs for positive and negative swings. This tends to cancel even harmonics and create odd harmonics. Balance produces substantially different tonal characteristics.
In a push-pull amp, the transition from an unbalanced to a balanced signal path is facilitated by the phase inverter. It takes an unbalanced signal referenced to ground and creates two balanced signal paths of opposite phase. By moving the phase inverter all the way forward to the guitar input jack, we transform the unbalanced guitar signal into a balanced amplifier signal and create an opportunity to preserve balance through the remainder of the signal chain. Let's see how this would work in a low-power design.
For my low-power example circuit the EL84 power amp screens are operating at lower voltages than typically seen in a push-pull guitar amp and the plate and screen dissipations are well below their limits. I'm using an 8k plate-to-plate output transformer like a traditional Class AB amp, even though this design is biased to operate in pure Class A mode. In a high-power, high-fidelity world that would call for reducing the impedance to 4k.
My intended DC operating point is shown on the transfer characteristic curves by the red dot:
The difference between cutoff and saturation is about 12 volts and the DC operating point is halfway between (for Class A operation), so to drive the power amp to full power we need an input signal amplitude of 6 volts. Total preamp amplification for a 20-millivolt guitar signal is 6/0.02 = 300. This requires more than one stage, so I use a long-tailed-pair phase inverter followed by a differential amplifier.
The phase inverter is just a differential amplifier with one of its phases connected to ground and an added tail to keep the sum of the plate currents nearly constant.
The input signal drives only one tube. The other tube's signal input is shorted to ground by a coupling capacitor. So the inverter gets only about half the voltage gain of a differential amplifier that is driven by both phases of a balanced signal. Typical gain for a 12AX7 preamp (unbalanced) is around 60. For a 12AX7 differential amplifier: 60. For a 12AX7 long-tailed-pair phase inverter: 30.
For this design two 12AX7 stages creates too much gain, so I use a 12AY7 (unbalanced gain about 12) for the phase inverter and a 12AX7 for the differential amplifier, creating a total gain of (6)(60) = 360. This puts the input sensitivity at a respectable 17 millivolts. The resulting gain is going to be a bit less anyway, because of the lower-than-usual plate supply voltage.
Like in the Bassman 5F6-A's first preamp, the differential amplifier's triodes share a common cathode resistor. A key difference is that there is no need for a cathode bypass capacitor - the DC grid bias remains constant because the two triodes operate on opposite phases of the same signal. Because the cathode resistor passes the DC plate current of two tubes, its value is half of what is used in a single-triode preamp.
My intended plate supply voltage is 125 volts, which allows for a sufficient drop from the power stage to allow for easy ripple filtering and decoupling. It should be noted, however, that these concerns are not as significant as they are in a traditional, unbalanced amp. Here is the load line and DC operating point for the 12AX7 differential amplifier when the cathode potentiometer is set to minimum:
With the control set to minimum the preamp runs very clean. Crank the guitar's volume control to maximum, hit a big power chord, and only the power amp breaks into overdrive. We get a clean preamp that is well capable of driving the power amp to full power and beyond.
Increasing the cathode potentiometer setting pushes the 12AX7 closer to cutoff, reduces its headroom, and increases its contribution to distortion. With just a slight turn of the knob the 12AX7 breaks into overdrive at the same volume level as the power amp. At higher settings the potentiometer acts like a master volume control, enabling the 12AX7 to be overdriven while the power stage runs clean. With an input sensitivity of less than 20 millivolts, push-pull breakup, either in the differential amplifier or the power stage, is easily achieved. The cathode potentiometer determines the volume level at which this occurs.
Since the long-tailed pair is operating on the low-amplitude signals from the input jack, it doesn't need a lot of headroom, so we can afford to put a lot of voltage across the tail for improved balance. For this design I'm aiming at 75 volts across the tail and only 50 volts across the triodes.
Here is my intended load line and DC operating point:
Each triode has an idle plate current of 0.4 milliamps, so 0.8 milliamps flows through the tail resistor. For a voltage drop of 75 volts the resistor value is 93k. I use 100k.
Here is the schematic of the complete amp for my bench tests:
The voltages shown are for my particular tubes.