From the plates of the 12AX7 the balanced signal travels through high voltage coupling capacitors down to the 6SN7, whose plates are at ground potential. A common cathode resistor is used both to bias the 6SN7 to 5 mA per triode and to help balance the output signal. The plate loads for the 6SN7 are made up of two 30k resistors in series.
   The power supply provides positive and negative 300 volts. It is built around a 480 volt center-tapped power transformer with a 6.3 VAC heater winding and a 5 VAC tube rectifier winding. The last two windings are voltage doubled and regulated to supply the 12AX7 and 6SN7 separately, as the cathodes are 285 volts away from each other. Solid-state diodes are used in a full-wave bridge center-tapped circuit. High wattage dropping resistors are used to bring the raw 330 volts down to 300 volts and to reduce the power supply noise.

Improving the Push-Pull
  Add some high voltage regulators was my first  thought, a thought shared by a friend of mine who was making a copy of my amplifier. He wanted to start a regulator race to see who could design and build the lowest Zo regulator. My competitiveness was tempered by growing distrust of complex circuits and the understanding that he would surely win. Not because he had a better design, but because he could source any part on earth. He was like Milo in Catch 22, he could convince any manufacturer to send him for free their top secret new IC, MOSFET, or voltage reference.
  So, the question was how could the amplifier  forgo the regulation and reduce the noise at the output, but not necessarily at the power supply rails. In spite of using high value electrolytic capacitors in the resistive pi filter, a couple of volts of noise rode on the power supply voltages. I found that the power supply noise on each rail was out of phase with the other. There was my answer: make sure the output of the amplifier was at the null point of the noise.

   Remember, with power amplifiers always design backwards. If the right valued, unbypassed cathode resistor were used, then triode's effective rp would equal the plate resistor's value and the anti-phase noise would cancel at the midpoint. The effective rp equals:
   rp' = rp + (mu + 1 )Rk.
Solving for Rk,
   Rk = (rp' - rp) / (mu + 1).
In this case, the effective rp for each triode must equal 60k, which is the plate resistor value. Thus:
   Rk = (60000 - 11000) / (21 + 1).
   Rk = 2200,
which when put in parallel equals 1100 ohms. Easy enough, except that this scenario assumes that the output tube's grids are, in AC terms, tied to the negative rail along with the cathode resistor. But the connection is made to the plates of the 12AX7. The work around is to force the 12AX7's plate to match the noise content of the negative power supply rail. 
    My first thought was to decouple the B+ connection for the 12AX7 and bridge it to the negative rail via a capacitor, a capacitor that could withstand at least 700 volts. Even if this capacitor could be found, the trick would not work. The 12AX7's grid are effectively grounded, which means its cathode is mimicking the absence of signal on ground. The other end of this cathode resistor is bouncing up and down with the noise present there. This variation of voltage across this resistor will cause a varying current to flow through the resistor and the 12AX7 and its plate load resistors. Since this current variation is in phase with the negative power supply noise, this noise developed at the 12AX7 plates will exceed that of the negative rail, as the in phase noise will add to the preexisting negative rail noise from the decoupling capacitor. Connecting the capacitor to ground instead, somewhat paradoxically, reduces  positive power supply noise by the ratio of the cathode resistor over that of parallel plate resistors.

< PREVIOUS

pg. 6

NEXT >

www.tubecad.com   Copyright © 1999 GlassWare   All Rights Reserved