The EF86 is a small pentode commonly used in preamp stages. It is designed for voltage amplification, not power production, and thus has a relatively low plate dissipation limit. It can therefore be driven to maximum power at far lower volume levels. I've used it in practice amp designs, and it creates phenomenal tones in overdrive at television-set volume levels. Maximum power output is quite suitable for a hotel room or an urban apartment setting. This tube is a strange choice for a power amp, I will admit, but for a small practice amp it represents an irresistible alternative to an under-driven EL84 or 6V6 design. In the paragraphs that follow I'll crunch some numbers for a specific design and point out some of the amp's unique quirks along the way.
I like to start a design with all the optional parts, forcing me to make a conscious decision to eliminate the ones that I don't want:
Cathode bias makes sense here, because the plate and screen supplies are low-voltage and the DC bias is very small.
These are sharp-cutoff pentodes meant for voltage amplification, but they are being forced to perform power-amp duties. Like most preamp tubes they are high-voltage, low-current devices that match well to a high-impedance load. Here are the characteristic curves for the EF86-equivalent EF806S at a screen voltage of 140 volts.
I've drawn a load line that shows a voltage swing of 135 volts and a current swing of 6.6 milliamps. According to Ohm's Law this represents an impedance of 20k. For Class B operation the plate-to-plate primary impedance is quadruple this value: 80k. For Class A we need 40k. For Class AB perhaps something in between. Center-tapped transformers with impedances anywhere in this vicinity, however, are extremely expensive.
The highest impedance push-pull output transformer available at a reasonable price is 10k plate-to-plate. This is not enough to bring the load line over to the knee of the zero-grid-voltage plate characteristic curve, so the power amp won't operate at its full potential. For a very low-power design, however, this is not a particularly troubling problem.
Screen dissipation is certainly a consideration. After all, EF86 pentodes don't normally operate at their theoretical power limits. When the plate swings to a low voltage it doesn't attract as many electrons, thus encouraging more of them to be absorbed by the high-voltage screen. With a relatively low transformer primary impedance we get high plate current swings and low plate voltage swings. The sub-optimal output transformer impedance thus reduces average screen current and screen dissipation. This represents a silver lining to an otherwise troublesome circumstance.
For what it's worth, I used a Hammond 1609A for my bench tests. It has a 10k plate-to-plate impedance and is rated for 10 watts, which is far more than what a couple of EF86 pentodes can create.
We need relatively low plate and screen voltages and the overall filament demand is low, so only a small power transformer is called for. I use a Hammond 369AX rated 125-0-125 (250 volts RMS, center tapped). This drives an RC filter between the EF86 plates (VPP) and screens (VSS) as shown here:
This is nothing unusual except for the lower voltages that are produced. (They are still quite lethal, however, so I always include a bleed resistor to drain the capacitors when the amp is turned off, thereby adding at least a slight margin of safety to this high-voltage electronic death machine! At 1M the RC time constant is 22 seconds, so if the circuit works properly the power supply reaches safe voltage levels about two minutes after the amp is unplugged, although I wouldn't bet my life on a 1-cent, non-MILSPEC part.)
Theoretically a 125-volt RMS transformer produces a peak voltage of 177 volts. The transformer source resistance at these voltages is typically fairly low, because there are fewer windings in the secondary compared to higher-voltage supplies. The current load of the power tubes is also very low, so the transformer should be able to achieve 177 volts (125 times the square root of 2) without much sag. If we drop the screen voltage to 140 volts then we have a healthy drop across the first ripple filter resistor.
For this particular design the DC plate voltage is 177 volts. (Measured voltage was much higher.) A 10k plate-to-plate primary impedance corresponds to a primary impedance per tube of 2.5k. This means that a plate current swing from zero to 8 milliamps creates a plate voltage swing from 177 to 157 volts, so here is the rather steep load line:
The less-than-optimal primary impedance produces more current swing and less voltage swing. The tubes go into saturation at about 160 volts, 7.6 milliamps, representing a swing from DC of 17 volts, 7.6 milliamps. For a square wave signal the power output is thus (17)(7.6mA) = 129 milliwatts. For a sinewave we divide by the square root of 2 and get 91 milliwatts RMS. This is plenty of volume for a low-power practice amp, and yet not enough to bother the neighbors when fully cranked.
I set the supply voltage for the preamps at 125 volts. This creates another healthy DC drop across the second power supply resistor, making it easy to RC filter the AC ripple with modest-size capacitors.
The preamp and phase splitter need to supply only a few volts of swing to overdrive the power-amp, so 125 volts is plenty to create a conventional design.
Based on bench tests, which produced some healthy odd harmonics and a nice transition to overdrive, I biased the amp a little warmer than I would for big power tubes, placing the DC operating point a bit closer to Class A: -2.5 volts at the grids. There is a second reason to warmly bias this amp: reproducibility. Tubes vary considerably, even from the same vendor. In particular...
You want what?! A matched pair of preamp tubes? Not a very common request for tube vendors. I have four EF806S pentodes and the variance between them is quite significant, more than I usually get from conventional power tubes. A mismatch causes a generous helping of second-harmonic distortion, which is not necessarily bad, but a bit surprising if you're not expecting it. With the tubes in my arsenal some permutations sound much better than others.
Here is the final design: