On the other hand, the cathode side connection can eliminate the need or any extra components, as the output is usually at ground level, which allows for DC coupling. Which circuit to use depends largely on which power supply noise canceling trick is used.
    As both topologies stand in isolation, the easiest path the power supply noise takes to the output is from the limitations of either the power supply shunting capacitor used in the plate side connection variation or the limitations of the voltage regulator or signal coupling capacitor used in the cathode side variation. Should any of these fail to isolate the power supply noise from the output stage's inputs, noise will leak into the output signal.
    But neither topology is used isolation: thus the input stage often provides a connection to the power supply noise. Of course, we can strive to eliminate the power supply noise from polluting the first stage's output by using large chokes and filter capacitors or by even using regulation.     Or we can use the output stage's common mode rejection ratio to our advantage.
    (The following technique works best with a Class-A operation mode for the output stage, but it is still useful with Class-AB output stages.) Remember that push-pull output stages need to see a pair of balanced drive signals, i.e. out-of-phase signals. When the push-pull output stage is presented with in-phase signals, the output stage should ideally provide zero amplification. Thus our goal is to ensure that whatever amount of power supply noise that leaks through must be presented to each output device equally in both  phase and amplitude. 
    The input circuit below realizes an equal power supply noise distribution by halving the power supply noise at the first triode's plate, as one half noise subtracts from one noise to yield one half noise at both outputs. This trick requires that Rk equal (Ra-rp) / (mu + 1), as this is the only ratio that halves the power supply noise. Here is a case where the circuit is device specific, as only a triode will work in the first tube's position, as only the triode has a low rp.

    If a pentode or FET is used as the input tube, then a 50% voltage divider is needed. Two equal valued resistors will work. The trick will be to find the value for these resistors that is not so low as to excessively load the first stage and not so high as to limit the high frequency response of the signal leaving the split-load phase splitter. Understand that this voltage divider will also divide the audio signal from the first tube. Still eliminating noise often comes at a price.     
     (A key point here is that both of these front-end circuits rely on the first stage seeing the same amount of noise as does the split-load phase splitter. Paradoxically, this requirement might require the removal of some preexisting power supply filtration circuitry in order to realize a quieter amplifier.)
     This noise rejection optimized front-end has a wider scope than just hybrid and OTL amplifiers; it should be used whenever a grounded cathode amplifier cascades into a split-load phase splitter; such as is found in the first half of the Williamson amplifier and even the Dynaco ST-70 and Mark-3.

< PREVIOUS

NEXT >