some translating machines do better ...
quote (from dutch site):
Now the circuit in more detail. To make it myself easy I will start at the output and from there work myself back to the input. The desired power is about 25 watts. This is not all that easy to wring from a Single Ended design. That is why I selected a balanced circuit, in English it's called ' push-pull '. The automatic is negatively here obtained by 'black workers' (aka 'sand' = silicon or transistors) which also improves the symmetry of the circuit because small differences in the tubes are compensated. The two KT88's are connected as ' long tailed pair ' for us peasants known as difference amplifier. So we get a nice symmetrical circuit with good linearity and thanks to the fine tape wound core / split tape core / toroid (?) Schnittbandkern-) transformer a more than adequate bandwidth. The power tubes have to be controlled in anti phase. There are many good phase inverters, but many have the property that the outputs do not have in all respects equal characteristics. This is often considered unimportant, but a real symmetry strikes me as its advantages. So here, too, I opted for a ' long tailed pair ' configuration. In the conventional buildup with a cathode resistor voltage symmetry is only obtained by different values for the anode resistors.
The total current through both tubes is not constant because some of the input voltage on the common cathode resistance falls and this flow so more or less varies depending on the input voltage. Actually not such a perfect situation. The solution is in the constant current source. An infinitely high value for the common cathode resistance is theoretically a good solution however, with some practical disadvantages. A different solution is a current source in place of the cathode resistor. So either a pentode or again some 'sand '. The option with a pentode has the disadvantage that here quite some voltage must fall to work properly and then you run quickly short of supply voltage. With sand this is not the case with the added advantage that a much higher internal resistance is feasible and thus the ideal better approached. Symmetry is for each other.
Only the low output impedance is not yet a fact, that still needs to be worked on. The simplest one is of course to take on anode resistors with low resistance value. Result is that the gain therefore is low and we need to substantially control the phase to provide enough output voltage from KT88's to get to the full. It seems to me that this is not a good plan. We need to solve this otherwise. The first to come to my mind was to build up as mu-stage. A beautiful plan ..... Unfortunately I had to use an extra tube. Another 4 triodes or pentodes. The top tube in a mu-stage is simply used as a current source for the lower tube and cathode follower as seen from the exit of the mu-stage. A triode as a current source is not really a big party so that's not going to be it.
A pentode does all much better, especially as the transconductance increases. But yeah double pentodes are not up for grabs and four additional tube sockets do not fit on the chassis. We need a better solution. Take a few steps back and let the mind run its course. What we need is a current source with a high Ri for the best linearity of the circuit. It should also present a low output resistance. And should have a minimum of tube sockets. A tube without socket with huge gain of a few amperes per volt smells like black magic. And that smell is very real in a power MOSFET. And a minimum of tube sockets, too. The challenge taken and so tried. I don't want to exaggerate but I feel it works perfect, high gain, good linearity and low output impedance.
And all without sockets. In the end that is what went into the amplifier. Then we now have the main features of the amplifier. It consists of a phase shifter and a push-pull output stage. A not mentioned advantage is that you can make the best of course in this configuration of the amplifier from a symmetrical source input. Simply remove the ' ground ' in the triode stage and use it as a second entrance to the balanced pair. I myself do that not because I have no source with balanced output, but for the enthusiast, there is the possibility.
unquote (no guarantee...)