KT-88 Push Pull with CFB

[Kashmire]

I'm warming up my soldering iron (again). My parts bin has a pair of big Plitron Push-Pull output transformers with "specialist" cathode feedback windings. So behold, my next project.

This one opens up for critique. The output stage is flexible. The attached schematic shows output stage with cathode feedback (CFB), along with unusual screen grid connections. If you don't like this, pretend UL, Pentode, or Triode connections.

More interestingly, the preamp (12AX7) and driver (12BH7) receive power from the KT88 plates ("exolinear"). My EL84 amplifier uses the same configuration, and it works well. The actual amplifier splits between B+ and exolinear through a resistor divider, so the amount of exolinear feedback is adjustable.

Next is the AC bootstrapped driver stage. The 1uF capacitors AC couple the output of the 12BH7 driver back to the plate of the 12AX7. This cancels the voltage swing across one of the plate resistors, making that resistor carry a constant current. This dramatically reduces the 12AX7's current swing, a major source of nonlinearity of the driver. The 1M resistor creates a "virtual AC ground". For aggressive shunt action, the 1M resistor can be decreased down to 10k.

The parts values are dependent on configuration. As configured, the 12BH7's 23k cathode resistor creates a voltage swing that must be mirrored on the 12AX7's 75k plate resistor through the 1uF AC coupling. The attached schematic isn't perfect, but close. It reduces the 12AX7's current swing by about 90%.

The 12AX7's 600 Ohm shared cathode resistor forces about 1.75mA (a bit high?) so 100k of plate is needed at 350V to achieve 150V output bias point. This DC-couples to the 12BH7 grid for 8mA (a bit low?) with 23k cathode resistance. The total load is coupled to the high-voltage side through 10k of resistance, dropping about 100V.

The 12BH7 cathode drives the KT-88 grid with plenty of dynamic headroom. In this design, the KT88 runs out of voltage before the driver stage clips. This is partly due to the high impedance (4k) of the output transformer. I would have done better with a lower impedance OPT with this design.

Comments and critique welcome!

[DF96]

I'm not sure what is going on here. The "exolinear" feedback will be largely undone by the bootstrapping, although some feedback will still get through via the 12BH7. The output screens appear to be cross-connected, which means positive feedback.

[Frank Berry]

Cross-connecting the screens will possibly destroy the output tubes.
In this circuit, when audio brings the plate voltage down, the screen voltage goes up.
The screens may draw too much current.

[HollowState]

Originally Posted by Frank Berry

Cross-connecting the screens will possibly destroy the output tubes.

Not according to John Broskie. Scroll down about halfway to the Dynaco ST70 and start reading.
5687 Aikido Kits and More Circlotron Power Amplifiers

[pieter t]

I would say a bad implementation of a McIntosh 275.
My advice is to skip the UL connections and go the Mc route when the output transformer allows (you would need the CFB winding separate, not as a "tap"); you can get quite some power this way.
However the transformer will not be wound as a Mc one, so it will at its best approximate the unity coupling technique which works so nicely in class B.
When not possible, still skip UL and make a nice class A amp.

[Kashmire]

Seems like I made a mistake leading with this output stage schematic. The output stage was too distracting.

If you don't like this, pretend UL, Pentode, or Triode connections.

The inverted UL tap topology is specifically for this Plitron CFB transformer, designed by Menno Vanderveen (schematic attached). The CFB is dominant and adds lots of negative feedback and significantly lowers output impedance, mostly because the cathode drive is low-impedance. The inverted UL offers a tad of positive feedback without significantly adding output impedance. So I recoup some gain without sacrificing a lot of damping factor. Plus, it seems to bring the amplifier back to life.

Originally Posted by Frank Berry
Cross-connecting the screens will possibly destroy the output tubes.

Not if the transformer ratios are correct. This trick won't work without proper cathode feedback windings. The Plitron transformer was designed for this.

Now, to more important matters.

The "exolinear" feedback will be largely undone by the bootstrapping

Correct. I wanted the bootstrapping to linearlize the 12AX7 (by stabilizing the quiescent current), but I also wanted a bit of multistage feedback that included the transformer's primary. The exolinear provides local feedback to the 12BH7 stage, and a little bit to the 12AX7.

The ratio of resistors control the bootstrapping vs. feedback ratio. The feedback can also be completely eliminated by connecting to the B+ instead of the KT88 plates.

[DF96]

The 12BH7 will be a fairly poor CF in this circuit, as the feedback to its anode will raise its cathode output impedance. My hunch is that this is essentially an FX box, masquerading as an over-complex power amplifier.

[HollowState]

Originally Posted by DF96

The 12BH7 will be a fairly poor CF in this circuit, as the feedback to its anode will raise its cathode output impedance.

I don't think there is any feedback. The anode is at signal ground because the B+ feed point is right there. Unless the power source is decoupled, any signal will be swamped out by the supply's output capacitor. (Assuming it has one)

Edit: Actually, the drawing is incomplete since the B+ feed is only to the upper tubes.

[DF96]

There is feedback, as the 10k comes from the output anode. The 12BH7 anode has a voltmeter attached, not a power source. The diagram is not incomplete as everything gets its supply from the 450V attached to the OPT centre tap.

[HollowState]

OK, I see it now. I'm too old school because we don't make voltage notations like that. Looked like a source.