The way out of this problem is to imagine what would happen if the bottom triode's cathode did not see the negative power supply noise. Then the noise would halve. And if this cathode saw the positive rail noise, then the noise would greatly reduce, as both grid and cathode would move in unison. But how do we couple the noise from the positive rail to the bottom triode's cathode? Easily enough actually, we connect the bypass capacitor to the cathode and then the to the positive rail. Since the positive rail's impedance should equal that of the negative, the bottom triode will be equally bypassed. But as the grid and cathode now see almost the same noise signal from the power supply, the noise will not be greatly amplified.
    To help us see what is going on here, imagine the White Cathode Follower stage working into a dead short to ground. If both power supply rails remain noise free, then the ground will neither source nor sink any current into the output of the White Cathode Follower. But if the positive rail bounces up 1 volt, the net current increase from the top triode will equal 1 / (Ra + rp) and this current will flow from ground into the top triode's cathode. On the other hand, the bottom triode now sees less voltage across its cathode-to-plate, as the positive bounce is relayed by the bypass capacitor. Thus this triode draws less not more current. In greater detail, if the positive rail bounces up 1 volt, the net current decrease from the bottom triode will equal 1 / rp and this change in current flow will have to be made up by the flow from ground into the top triode's cathode. Although this scenario is much better than the one where the bottom triode's cathode is bypassed to the negative rail, it would be better if we could find a way to make both triodes increase and decrease in current phase so that the output would be at the null point.
     When the increase in the bottom triode's current equals 1 / (Ra + rp), we will have achieved noise nulling at the output. But how? Originally, the problem was that the bottom

Subject: Article Request
      Many thanks for the great magazine! I would like to make the following request for an article or design idea:

Noise nulling techniques for a (bipolar) White CF.

     I am currently designing a headphone amp using this simple circuit. (The diagram is attached. It is influenced by your Oct-99 and Apr-00 articles plus the articles on bipolar PSUs). However, as far as I can see, with no feedback applied, the noise at the output would be almost twice the noise on either rail! Is there a cure for this, or is heavy filtering/regulation the only option?
Regards,

Morgan

     First of all I think your circuit shows much promise. It is well thought out and efficiently laid out. I particularly like the voltage doubler scheme for deriving the input stage's higher power supply voltage. And using the EL86 is a great choice, as this muscular little tube can put out a lot of current with very little cathode-to-plate voltage across it. You are right, however, about the PSRR not being optimal as the output stage is presently arranged.
     Like an accordion being pulled apart, as the positive rail pulses positively, the negative rail pulses negatively. The plate resistor barely attenuates the positive rail noise before it is relayed to the bottom triode's grid, while its cathode is being pulled in anti-phase by the negative rail noise. Thus we have twice the noise of a White Cathode Follower in a monopole power supply. (Actually a monopole power supply would hold twice the noise of any one rail of the bipolar power supply; therefore, the noise would be effectively the same.).