by Fred Nachbaur, Dogstar Music ©2002

Part 1:
Part 2:
How It Works
Part 3:
Part 4:
Part 5:
Parts List
Part 6:
Sound Sample


I'll assume that you are familiar with the original SET "MiniBlok" in the discussion that follows. As in the original single-ended design, the push-pull MiniBlok II uses the relatively inexpensive 13EM7 "dissimilar dual triode" (or its variants the 10EM7 or 13EM7/15EA7). Using the same tube with essentially the same operating point makes for a more equitable comparison between the single-ended and push-pull topologies.

However, it should be fairly obvious that two of these are required for the push-pull version. As before, one of the small triode sections is used as our preamplifier. The component values are, in fact, exactly the same. The two power triode sections form our push-pull output stage, again with essentially the same operating point. That leaves one small voltage amplifier triode, which will form our phase-inverter for the "paraphase splitter".

The full schematic diagram is shown below, and a circuit explanation follows:

MiniBlok II schematic


As before, there are essentially three separate power supplies required: The first ("A" supply) is to heat the filament of the tubes, the second provides the high voltage for the plate circuits (the "B" supply), and the third provides a portion of the negative grid bias for the power amplifier ("C" supply).

Incoming power from the 120 VAC power line is switched by on-off switch S1, and thence applied to step-down transformer T1. Unlike the replacement-grade transformer used in MiniBlok, however, MiniBlok II uses a 24VAC, 40VA "Class 2" transformer. This kind of transformer is commonly used to provide low-voltage AC in house wiring, and is supplied with a large mounting nut suitable for installing in standard circuit boxes. While these are more expensive than the usual run of experimenter transformers, they are of higher quality and therefore capable of supplying more power for a given physical size. Although a 24V, 40VA unit is not much larger than the 12.6 volt at 2 ampere unit used for T1 in MiniBlok, it is capable of nearly twice the output power.

As before, the output from T1 supplies the power for the tubes' heaters. The two tubes' heaters are connected in series, resulting in a bit over 12 volts for each filament. (Though this is about a volt lower than the tubes' ratings, it works just fine.) Again, we don't bother converting to DC, in this application an AC heater is just fine because 1) we're not dealing with extremely high gain, and 2) the 13EM7's indirectly-heated cathode is naturally highly immune to AC hum.

At the same time, this 24 volt AC supply is connected to the network consisting of diode D4 and capacitors C5. This is a simple half-wave rectifier that provides about 35 volts DC at the junction of D4 and C5. This could have been used as our C supply (grid bias) for the output stages, but we're going to go an extra step to help insure that our two output tubes are balanced. Part of our grid bias is derived from cathode resistors R15 and R16. Assuming that the tubes will draw about 40 mA in operation, this means that the voltage across R15 and R16 will be (330 * 0.035)V = 13.2 volts. This means that we only need about 15 volts of additional grid bias to set the tubes near our desired operating point. The network consisting of resistor R14 and zener diode D3 provides this grid bias voltage.

During operation, mismatches between the two output tube sections will be significantly reduced by this combination biasing method. The reason is that if one of the tubes should draw more current, it will cause a higher voltage drop across the cathode resistor; this in turn increases the grid-to-cathode voltage, tending to buck the increase in current. In the prototype, there was a 50% difference in plate current between the two tubes I had on hand, using fixed-bias only; this dropped to a mere 10% difference using the joint cathode-grid biasing scheme. At the same time, it only "wastes" half of the supply voltage drop that a straight cathode biasing scheme would require.

An optional pilot lamp network consisting of LED (light-emitting diode) D5, resistor R5, and diode D6 can be also connected to the 13 volt line. Resistor R5 limits the maximum current to a safe level for the LED, and D6 prevents excessive reverse voltage during negative half-cycles.

Finally, the secondary of T1 is also connected to the low-voltage winding of an identical transformer T2 (24 volts at 40VA, or approximately 1.7A). T2 steps the 24 volts back up to approximately 115 volts AC. From there, a voltage doubler generates the high-voltage DC required by the plate circuits, exactly as in the single-ended version. However, because of the higher efficiency of the Class 2 transformers, the final B+ voltage will be significantly higher, approaching 240 volts in practise. Even with the loss of B+ across the cathode resistors, the total plate-to-cathode voltage will be a bit higher than it is in the SET version.

For convenience in comparsion, the size of the tank capacitors is the same as the SET version. It should be noted that you could get away with quite a bit less, given the push-pull circuit's greater inherent supply rejection. You could use 220 uF capacitors for C1 and C2, and as little as 47 uF for C3, without significantly degrading the amplifier's hum rejection.

An additional stage of filtration, consisting of R17 and C4, is provided for the preamplifier and phase inverter stages. This is necessary because of a shortcoming in the paraphase splitter scheme; residual hum at the preamp output is essentially amplified by the phase splitter. While it is theoretically possible to work up a paraphase in which the hum components cancel, the additional trouble far exceeds the single added resistor and capacitor required to kill the hum before the fact. (That being said, there probably was a time when electrolytic capacitors were quite expensive, and such an "after the fact" hum reduction method would have been practical.)

Again, another optional pilot lamp consisting of resistor R7 and LED D7 senses the voltage drop across R1. Once the filament reaches operating temperature, the plate circuits begin to draw current and the lamp starts to glow.


The preamplifier stage is identical to the SET version, except that the 47k plate load resistor is split into two sections (36k + 11k). This forms our voltage divider (attenuator) for AC signal to the paraphase inverter stage. As before, the output of the preamplifier is coupled to one side of the push-pull pair, via capacitor C7.

MiniBlok II schematic


Now we need a signal of opposite phase, but approximately the same amplitude (level) for the other member of the push-pull pair. The signal-triode section of V2 accomplishes this, along with its associated resistors R9 - R12 and capacitor C9. Let's step through the circuit to see how this works.

The mu (µ) or "Amplification Factor" of the *EM7 signal triode is approximately 64. This can be thought of as the maximum theoretical voltage gain, with an extremely high value of plate load resistance. In practical circuits, the actual voltage gain will be approximately the µ times the ratio of plate load resistance (Ra) to total circuit resistance (Plate Resistance Rp + Ra). Since Rp is about 40k ohms, and Ra is the parallel combination of R12 and R13 (43k) it follows that our stage will have an open-loop gain of about 64 * (43k/(40k+43k) = 33. In other words, a 1 volt change in grid-to-cathode voltage will cause a 33 volt change in plate voltage.

However, note the large unbypassed cathode resistor consisting of R10 and R11, for a total of about 8.32k ohms. This is essentially in series with the tube and the plate load resistor, such that a change in 33 volts on the plate will cause a change of 33 * (8.32/43) = 6.4 volts on the cathode. Add the one volt grid-to-cathode voltage we started with, and we see that it takes a 7.4 volt change on the grid to effect a 33 volt change on the plate. In other words, our stage gain is about (33/7.4) ~ 4.5.

So that means that the input has to be divided down by a factor of 4.5 in order to end up with an equal-but-opposite signal on the grid of the right-side power amplifier tube. This can be accomplished with a divider either on the plate load of the preamp, or on the grid resistor of the left-side power amp. I chose the former; because of the lower values, grid resistor R9 has much less effect on the calculation and can be ignored. Keeping the total plate load the same (47k) and choosing the closest 5%-tolerance values, we come up with 36k and 11k, for a division ratio of about 4.3 -- within 5% of our predicted stage gain.

Note that the large unbypassed cathode resistor represents a negative feedback element. This has two major effects:
  1. stabilizing the stage against variations in gain, and
  2. greatly reducing the stage's added distortion components (by about 17 dB).
The first point can be demonstrated by seeing what happens if the µ of the tube decreases; for instance, if it should drop to 50 due to aging (a drop of 22%), the stage gain will only decrease to 4.3 (less than 5%). What's more, this will actually get it closer to our attenuator ratio, which is why I chose the standard values on the low side of the projected ratio.

It could be argued that if some negative feedback does some good, won't more do more good? The answer is "yes." In fact, if the plate and cathode resistors are made equal, the stage gain is reduced to slightly less than unity, and we wouldn't even need an input attenuator. This would, in fact, be essentially the popular "cathodyne" phase splitter, where the output signals are taken from the plate and cathode of such a symmetrical, unity gain amplifier stage. However, the difficulty with the cathodyne in this application is that we would need a total peak-to-peak output swing approaching 140 volts, in order to accomodate the low-mu output triodes' drive requirements. A look at the required load line on the plate characteristics curves shows that, at a supply voltage of only 220 volts, this output drive would just barely be possible; however, it would 1) push the tube right to the edges, resulting in poor linearity, and 2) leave us absolutely no headroom. That's why, rather than complicate the power supply by adding a tripler or something to that effect, I chose the paraphase with moderate cathode feedback.

However, I did make one minor concession to the headroom issue. Since we already have a -15 volt supply for the PA grid bias, we can use this for the paraphase stage cathode return. This essentially offsets the B+ headroom loss imposed by the cathode resistor, and ends up placing the paraphase stage at exactly the same operating point as the preamp. No additional parts are required, and it doesn't affect performance in other ways, so why not?


After all that, the power amplifier stage is quite simple.

MiniBlok II schematic

As mentioned earlier, the grid bias for the output tube sections is derived from a combination of fixed grid bias, and cathode bias, combining some attributes of both. It has improved balance and stability over the straight fixed-bias approach, with less headroom reduction (B+ loss) than a straight cathode-biased stage. Capacitors C11 and C12 bypass the cathode resistors for signal, affording maximum possible gain and keeping the amplifier a "no NFB" design.

As before, blocking capacitor C7 (also called a "coupling capacitor") allows the preamplified AC signal to pass from the plate of the preamp to the grid of the power amp, while blocking the DC difference between these two points (over +100 volts on the plate of the preamp, and -15 volts on the grid of the power amp).

Note that the coupling capacitors into and out of the paraphase circuit are of higher value than the "in-phase" capacitor C7. This helps to insure that the paraphase will approximately track in the bass region also.

Finally, the plate current of the power amp is allowed to flow through the primary winding of output transformer T3. However, the transformer is center-tapped, with the ends going to the output plates, and the center going to B+. Since the DC currents in the two sections are equal-but-opposite, the net magnetization of the core is approximately zero. We can therefore use the more efficient interleaved core (no air gap) transformer design.

The secondary of the output transformer is connected to the speaker -- and our amplifier is complete! As shown, the amplifier's output resistance (after the transformer) is on the order of about 6 ohms. It will therefore work fine with either 8-ohm or 4-ohm speaker loads. Distortion will be marginally less when using 8-ohm speaker loads, however.

Back Home Next