A third advantage lies in the safety feature of the output tube being projected from excessive current draw at turn-on. Pulling the MOSFET positive turns off the output tube. So when the input tube is still cold and not conducting, the MOSFET will be at some high power supply voltage, which will descend as the input tube warms up. This arrangement even helps when the 6DJ8 tube is yanked from its socket (or just wiggled) while the amplifier is in use, as any increase in cathode voltage only diminishes the conduction of the output tube.

   The following circuit looks good on paper, but seldom works well in practice. The aim was to avoid coupling capacitors and simplify biasing arrangements. The first stage consists of a Long Tail phase splitter which then cascades into a Differential amplifier.
  The problem with this circuit lies in the assumption that the two triodes that make up the first stage are perfectly matched and will continue to remain matched over time. Only a 2% discrepancy in idle current will latch up the second stage, as one tube will see 2 more volts on its grid than the other tube.

  A fourth advantage to this circuit is that an auto-bias trick, much like the ones used for the Loften-White amplifier mod outlined in last month's issue, can easily be added to this amplifier. Connecting the MOSFET's drain to the 6DJ8's cathode is all that is required. 

  If the output tube draws too much current, the 6DJ8's cathode voltage will increase, which will decrease its conduction, which in turn will push its plate voltage higher, thus, pushing the output tube's cathode higher and decreasing its conduction. Conversely, if the output tube draws too little current, the 6DJ8's cathode voltage will decrease, which will increase its conduction, which in turn will pull its plate voltage lower, thus, pulling the output tube's cathode lower. We have a simple DC servo loop that can also be used as an AC feedback loop, if the bypass capacitor is removed.
   This hybrid topology is a sizeable improvement over the generic Loften-White circuit which presently is all the rage, as it is intrinsically a much safer way to DC couple the output stage.

   The solution to this problem is to add a DC servo loop to the first stage so that the voltage at each plate matches the other. This will require adding a coupling capacitor to the input, but at least it is only one capacitor and this capacitor is outside the feedback loop, which makes for better stability and preserves amplifier power, as subsonic signals will not be amplified. Additionally, two small valued resistors must be added, but as these resistors are small in value they do not compromise the gain too greatly and only serve to make the triodes more linear.

< PREVIOUS

pg. 8

NEXT >