I know that a few readers are laughing to themselves: "A 12AX7 and 12AU7, what a fool. Who would actually use those lame tubes? " Well, the secret is that it is the topology, not the tube types that make a new circuit. 300Bs or KT88s or 845s or 6C33s can be used in this topology.

The message that came across, however, was that while different tubes could be used, the same plate and cathode resistor values would stay fixed. No, simply no. No such thing was implied (well, at least not intentionally by me). Different tubes, different resistor values. In fact, the same tubes with different B+ voltage, different resistor values.

For example, look at the data table above in the 6H30Pi section and notice how none of the values for rp and mu matches the one listed in the tube manual for these triodes. How is that possible? A triode's rp, mu, and gm are not chiseled in titanium plates sealed in a vacuum; instead they are a fluid blend from one extreme to another extreme. In other words, it makes little sense to specify a triodes rp and gm without also specifying the cathode-to-plate voltage and cathode current at which these characteristics were obtained.

Getting the rp and transconductance right is critical to obtaining the best performance from the Janus shunt regulator. For example, the rightmost triode voltage divides the power supply noise in the fashion as a two-resistor voltage divider would, except that the triode’s rp takes the place of a bottom resistor.

In the equivalent schematic above, we see the power supply noise reduced to a fourth of its B+ value; thus, the series resistor’s value must equal four times the inverse of the leftmost triode’s transconductance; in this example, 0.002A/V, as 2k * [100k/(100k + 300k)] = 500 ohms, the inverse of which equals 0.002. Had the triode's gm been 10mA/V, then a 400 ohm series resistor would be needed, as  1/0.01 = 100, which against (100k + 300k)/100 equals 400 ohms.

By the way, one problem the above circuit would face in reality is the pass tube’s high-input capacitance, due to Miller-effect enhancement. The 300k voltage divider resistor in parallel with it 100k brother equals an output impedance of 75k, which working into the magnified grdi-to-plate capacitance of the pass triode can too greatly limit the shunt regulator's high-frequency bandwidth. A simple workaround would be to bypass the voltage divider resistors with small capacitors, as shown below.

These added capacitors reduce the voltage divider output impedance at high frequencies. (Note how the larger-valued capacitor goes on the bottom, not the top, as is the case with the resistors.)

Pentode Driver Tube in the Janus Regulator
An interesting variation on the Janus shunt regulator would be to use a pentode instead of the a triode as the driver tube (the rightmost tube). Pentodes offer some interesting benefits, for example, high gain (much higher than the comparable triode could summon).  This is due to the fact that the pentode’s transconductance, unlike the triode’s, is not shunted away by the plate resistor, as the screen shields the grid from the plate’s movements. Second, the pentode’s grid number 2 can see a voltage higher than the plate. For example, in an ultra-linear power amplifier, because of the secondary’s DCR, the plate sees a lower voltage than grid number 2 does. This small oddity can be exploited by using the pass tube’s cathode voltage to drive the pentode’s grid number 2, thereby adding a DC feedback loop, which will help keep the DC operating points in line, as the tubes age or are replaced.

Now that you have had a minute to digest this circuit, let’s move on to the important missing details. At startup, when the cathodes are cold, but the B+ voltage is fully realized, do we really want the pass tube’s grid to be at +400V, while its cathode is at 0V? Adding the solid-state diode takes care of that problem, as it only conducts when the grid is more positive than the cathode, which it never is once the tubes are hot and conducting. Next, the all-important grid-stopper resistors are in place to prevent wild oscillations at 40MHz. Finally, high-quality bypass capacitors make up for clunky electrolytic capacitor’s failings.

Wait minute, where are the tube types and part values? I think I’ll leave them out this time (think of working the values out on your own as homework). One place that I would start looking for pentodes is in triode-pentode tubes, such as the 6AN8, 6BM8, 6AU8A, 6AW8… All of the tubes so far mentioned are 9-pin types, as octal triode-pentode tubes are rare; and the few that do exist have connections on the top of the glass envelope (eight connections are not enough, once the heater’s two are subtracted). Or, if discrete triodes and pentodes are going to be used, I would look into 6AU6, 6BY7, and the 6SJ7GT for the pentode. The key pentode feature is sharp-cutoff and it would not hurt to have an audio-use history. The triodes can start at the 6C4 and 6J5, move up through the 12B4 and 2A3, and end at the 300B and 845 (of course, triode-connected pentodes, such as the EL34 and EL84 can—and perhaps should—be used).

One last thought: as I have said in the past, when you see a pentode, think cascode instead.

Cathode resistors Rk1 and Rk3 voltage divide the pass triode’s cathode voltage, making the DC connection to the cathode voltage possible. (In the cascode circuit, the top triode's grid should never be at a voltage more positive than than its plate.)

Both the cascode and pentode variation share a common attribute: the pentode and cascode offer an extremely poor PSRR, which means that almost all the power supply noise will be relayed to the pass tube, which in turn means that the series resistor’s value will be much closer to just the inverse of the pass tube’s transconductance; in other words, smaller in value, which may mean too much output voltage, depending on the current drawn by the load. Another bad result stemming from a smaller-valued series resistor is that too little voltage gain may be be developed by the pass triode. On the other hand, the driver gain will be so much higher that the lower-valued series resistor may not hinder the shunt regulator’s overall performance with signal-induced perturbations on the output voltage. As you can deduce, it quickly gets complicated, but then that's what makes for interesting homework.

Next Time
I know that I promised phono stages, but it will have to wait until next time, as I have a lot to write about.

//JRB