Ultra-linear operation requires something like 33% of the plate's movement to be fed back into grid 2. This can be accomplished by the same means as was used in the ultra-linear line stage amplifier. The top tube, on the other hand, works with a fixed plate voltage and a cathode the swings up and down with the output of the amplifier. How is ultra-linear mode to be applied to the top tube? Before tackling the schematic, a review of how ultra-linear mode works is needed. Ultra-linear mode uses a portion of the plate-to-cathode voltage movement to define an input signal for grid 2 that is in anti-phase to the signal being applied to grid 1.
 Totem-Poled, ultra-linear, OTL amplifier output stage
 Because the ultra-linear signal is phase inverted to the input signal, the gain is reduced along with the distortion and the output impedance. In other words, it is a form of negative feedback. If the ultra-linear signal were in phase with the input signal, the gain would increase but the distortion and output impedance would increase as well; in fact, the amplifier could easily become an oscillator.    Both grid 1 and 2 are referenced to the cathode: present either with a greater positive voltage and the conduction from cathode to plate increases. So what is needed is an ultra-linear signal for the top tube that will work in anti-phase to grid 1 input signal. The key here is to reference all voltage relationships to the cathode. If a two resistor voltage divider bridges output to ground and its division point is set 66% relative to the output of the amplifier, a +10 volt swing at the output would equal +6.6 volts at the voltage divider's output. Since a +10 volt swing at the output means that the top tube's cathode has also swung +10 volts, the +6.6 volt increase in the grid 2 voltage is actually equal to -3.3 volts relative to the cathode.
 Supplying the actual voltage may help make this clearer. In the absence of a drive signal the cathode is at 0 volts; the plate, +100 volts; grid 1, -15 volts; and grid 2, +100 volts. Thus, the cathode-to-plate voltage equals +100 volts; the cathode-to-grid 1 voltage, -15 volts; and the cathode-to-grid 2 voltage, +100 volts. In the presence of a +14 volt pulse at grid 1, the resulting voltage relationships would then be that the cathode is at 10 volts; the plate, +100 volts; grid 1, -1 volts; and grid 2, +106.6 volts. Now, the cathode-to-plate voltage equals +90 volts; the cathode-to-grid 1 voltage, -1 volts; and the cathode-to-grid 2 voltage, +96.6 volts. The voltage division ratio of 66% effectively becomes 33% relative to the cathode.    A string of three equally valued resistors gives us all the sampling of the output we need to make a totem-poled, ultra-linear, OTL amplifier. Two Cathode Followers and a few capacitors and resistors complete the output stage. Each Cathode Follower must be biased so that its cathode voltage is roughly equal to plate voltage of the output tube it will service. Thus, the Cathode Follower's grids need coupling capacitors and bias resistors.
 pg. 5