Lately I’ve been looking into drive topologies for Single Ended Triode (SET) power stages. The major benefit of power triodes is of course how great they sound. The major down side, from the perspective of the amplifier designer, is the fact that their power sensitivity is very low. What this means is that it requires a large voltage swing to drive these tubes. Some examples from my recent designs: a 6AS7 SET design that requires ±75v, a 6336 SET that requires ±81v, and a triode strapped 807 “SET” that requires ±72v. All from a line level inputs of between 1v and 2v peak.

I like to design my amps with enough sensitivity so that I can drive them to full power with an input voltage of not more than 1v RMS or 1.41v peak. This means, for those amps mentioned above, my drivers need voltage gains of between about ≈51v/v and ≈58v/v. And there are some power triodes which require even more drive to get to full power. I also think that it’s rather absurd to ruin the sound of a nice SET power stage by driving it with a pentode, SRPP, or actively loaded stage. This means that passively loaded triode drivers are the rule for these fine sounding amps.

So there are a couple of driver configurations to investigate. The first is a single triode with a resistive load. As a rule of thumb, it is reasonable to expect to get an amplification factor of about 2/3 of the tube’s amplification factor, µ, from this configuration. Applying this rule means that we need a tube with an amplification factor of between 77 and 87 to get the required gain. There are not a lot of triodes with this level of gain. There is the 6SL7 with a µ of about 70 and the 12AX7 with a µ of around 100. Unless the drive level is increased this leaves the 12AX7 as the only single triode driver for these designs.

The 12AX7 clearly has the gain required to provide the needed amplification. There are some compromises to be made however. Most large power triodes have a limit on the maximum grid resistance, usually in the neighborhood of 500kΩ, but sometimes much lower. The driver load is usually limited to about 1/3 to 1/4 of this value. Any larger and the driver distortion can get excessive. So assuming a maximum load on the 12AX7 of 200kΩ and setting B+ to about 250v the driver will provide both the voltage swing, ±75v, and the gain, 67v/v, required. The sacrifices made in getting the gain and voltage swing out of the 12AX7 are twofold. The first is that its input capacitance is going to be very high. In this case about 115pf. The second is that the design is forcing the 12AX7 to swing almost to cutoff thus increasing total distortion. Neither of these are desirable.

The distortion under this configuration is high, coloring the sound of the output tube. In addition, the frequency response of the system is limited by the driver tube input capacitance and the output impedance of the signal source. In this case, this value of input capacitance limits the input load to 69.2kΩ for a 20kHz upper frequency -3dB point. If you want to use a larger grid resistor, drive the amp with a higher impedance source, or use long drive cables (with their own parasitic capacitance), the upper end frequency response will suffer. There must be a better solution to this design problem.

The answer is to use multiple triode stages to get the required gain. Instead of requiring one very high gain triode and pushing it to the edges of it’s performance envelope, multiple stages allows the selection of triodes from a much larger set of medium amplification factor tubes and provides many more design options. Additionally, by taking this approach, the final result will have better high frequency response, better distortion overall, and it will preserve the distinctive sound of the SET topology.

Here is what Herbert Reich has this to say in “Theory and Applications of Electron Tubes” on the subject of increasing amplification of audio amplifiers:

Increasing the amplification by increasing the resistance r₁ or r₂ results in a lowering of the high-frequency limit of uniform amplification. Furthermore, in a multistage amplifier, increase of the amplification of each stage affects the high-frequency response of the preceding stage adversely because of increase of effective input capacitance. For these reasons it is usually better to increase the amplification by increasing the number of stages rather than by raising the amplification of each stage.

The reason that the input capacitance increases is because it is proportional to the gain of the stage. If the gain can be kept smaller by using multiple stages, then the input capacitance of each stage will be lower and the high end will not rolloff so early.

I have two examples of drivers of this type from recent work. The first is a cascaded 12AU7 driver. This design not only provides over 76v/v of mid band gain, but has an input capacitance of only19.6pf, and a high frequency -3dB point greater than 60kHz (-0.36dB at 20kHz). The second is a cascaded 6SN7 driver with unbypassed cathodes. This design provides over 58v/v of midband gain, has an input capacitance of 37pf, and a high frequency -3dB point of almost 30kHz (-1.42dB at 20kHz). Both of these drivers are far better design solutions than what could be provided by any single triode.

Personally I think the trade off of adding a stage to get far better performance is well worth it. Often times people will assert that the fewer stages in an amplifier the better it will sound. However, when asking those stages to do extraordinary feats, sometimes it’s better to have them share the load rather that have a single triode go it alone.

Thoughts?