In fact, the drive voltages are balanced and, ultimately, identical in all cases. It is just that in the last circuit, one tube's cathode is tied to ground, while the other tube's cathode floats. But in all circuits, each grid sees the same voltage swing relative to its cathode as its pair. If we think primarily in terms of transconductance, rather than amplification factor, this makes perfect sense. To perform cleanly, each tube in a push-pull amplifier must work as hard as the other. By varying the cathode-to-voltage, the current flow through the tube is varied and, ultimately, so too the current through the speaker is varied. Therefore, each tube in a push-pull amplifier must see the same grid-to-cathode voltage swing.

Circular/Bridge/Balanced output stage allows flexibility in ground referencing, both the input driver stage are sensitive to where ground falls, as the shield of the input jack in a stereo amplifier must attach to ground.
   What is needed is a way to sample the output voltage at each cathode, and then to use these voltages to allow automatic adjustments of the voltage swings into each grid for an equal grid-to-cathode voltage per tube.
  Transformers are the conceptually clean way to go. As the windings are only AC coupled, primary and secondary can be DC voltage referenced to hundreds of volts of potential difference (the breakdown voltage of the wire insulation used in the making of the transformer is in the thousands of volts). By attaching one set of leads of the secondaries to the cathodes, the other leads will float with the cathodes and deliver an equal grid-to-cathode voltage swing.
   To work, the primary does not need to see a push-pull signal as it provides its own phase splitting (just reverse the leads on one of the secondary windings). In fact, the primary could load just a single tube input stage, if the transformer's step up ratio were high enough.     
   Notice that in the schematic below, the placement of the ground connection is very flexible. If we were willing to risk a safety hazard, the connection could even be made to one of the power supplies' B+ terminals or left floating with no connection other than the stray capacitive coupling between transformer windings. (In other words, do not do this, unless you know what you

  Attaching two scope probes to points within the amplifier will display the signals amplitudes shown at the left. (The voltage swing across the speaker is slight because the triodes exhibit a much weaker transconductance compared to solid-state devices.)

  Switching to the Circular/Bridge/Balanced amplifier configuration does not alter the need to have each tube in the amplifier see the same cathode-to-voltage swing. If the ground is referenced to the voltage midpoint between the output cathodes, then, obviously, each grid will see an equal voltage swing relative to ground. If the ground is referenced to one of the output cathodes, then each grid will see a greatly different voltage swing relative to ground, but not in relative to its own cathode.

Circular/Bridge/Balanced amplifier with ground referenced to the midpoint between cathodes

are doing.)
    The problem, of course, with this very rosy picture is that good transformers are hard to find. Furthermore, the higher the turns ratio, the worse is the transformer's high end response. (Besides, it would be odd to build an OTL that needed an internal signal transformer to work.)
   Fortunately, it is easy to mimic the transformer's functionality with circuitry. What is required is some means to ensure equal grid-to-cathode voltages. A Split Load phase splitter or a Differential amplifier can be worked into a  Circular/Bridge/Balanced amplifier easily

   We have seen that the output stage is indifferent to where we decide to place the ground connection, as long as the grid-to-cathode voltages are equal. The conclusion:  we may place ground where we please.

Driver Circuits
   Connecting the front end of the amplifier to the output stage can be tricky when the output stage uses floating power supplies. Tricky, but not impossible. While the

NEXT >