Another solution (shown right) is to modify the servo by adding a zener diode from its output to the tube's cathode and referencing the servo's input away from ground to the same input that the triode sees. So arranged, the servo offers greater protection, as even if the tube is pulled from its socket, the servo's output has a DC path to output MOSFETs' gates. This is a new servo design (as far as I know, this circuit is original and it could easily have been patented, instead it is another free idea to be used freely in an increasingly un-free intellectual and commercial environment). This circuit works on the assumption that the triode's grid never becomes more positive than the cathode under normal use. Consequently, the diode and the zener only conduct if something goes wrong. In other words, even if the triode is pulled from its socket, the servo will find a DC current path through the diodes to the output MOSFETs' gates or even if the tube becomes gassy and begins to conduct excessively, the Op-Amp can find a DC current path to the MOSFETs.
     The remaining troubling issue is then the time constant set by the servo's capacitor and input resistor, which may be too slow to allow the servo to work at preventing damage to the loudspeaker. But just how often do tubes fail or pulled from their sockets or jiggled? Effectively, something like this happens each time the buffer is turned on, as the tube has a warm-up time before it can conduct, which will force the buffer's output negatively until the servo catches up. In other words, we should not set the transition frequency to anything below 1 Hz, as doing otherwise could damage the speakers.
     A further potential problem is the added voltage swing requirement made on the Op-Amp. The Op-Amp's output must equal the buffer's input voltage swing. High-voltage Op-Amps are one possible solution. (Another solution might be to use a floating power supply for the Op-Amp. This floating power supply could then be referenced [that is its ground] to the buffer's output, which would carry the Op-Amp with the output swing.)

     Additionally, replacing the tube's cathode resistor with a tube-based active resistor helps overcome the warm up problem. Given that both triodes come from the same single envelope, the odds are good that both triodes will heat up equally. Even the act of pulling the tube from its socket will be less likely to damage the loudspeaker, as the input to the output MOSFETs will float near ground potential until the servo regains control.
     One interesting aside is that the bottom triode's cathode resistor should not be bypassed, as doing otherwise will allow power supply noise to make its way to the output. The idea here is that if the impedance that the top triode presents to the positive rail voltage equals that which the bottom triode presents to the negative rail noise, then the two will cancel out. Not bypassing the cathode resistor would greatly magnify the bottom triode's effective impedance, unbalancing the noise cancellation at the midpoint. The disadvantage to using the extra triode is the need for a high-voltage negative power supply rail and the possible exceeding of the tube's heater-to-cathode voltage ratings.

< PREVIOUS

NEXT >