Paraphase Inverter Revisited

[K Teacher]

INTRODUCTION
I've been looking, for some time now, for a 6V6 push-pull amp, driven by a paraphase inverter. This configuration was very common on many amp designs used from the mid-40s to mid-50s, before they were replaced by other circuit types. However, surviving amps of this vintage, that are in reasonable shape, have an extremely expensive price. Also looked for a 1969-1971 Traynor YBA-2, 2-knob, 6V6 power output tubes to no avail.

For these reasons, the idea to build an amp clone, becomes appealing. I looked on the web for circuit schematics and technical information on these types of amps. However, technical info and design procedures on this type of phase inverter is scarce.
Additionally, looking at several commercial schematics of yesteryears, it is noteworthy that for similar (or the same) type tubes, the phase inverter circuit components values, vary widely.

That called my attention, because the key factor for the paraphase design is the gain of the inverter tube circuit, which depends mainly on tube type and plate resistance value. I wonder, if these differences were on purpose, to generate higher second harmonic content, or if it was just bad design practices (like copying resistor values from other schematics with different tube types).

BASIC PARAPHASE
The basic paraphase inverter circuit is shown below:




HOW IT WORKS
The basic paraphase inverter working principle is simple. The driving signal is sent to one of the output tubes (OUT 1). This same signal is also sent to an additional signal amplifier tube, which by its nature, inverts that signal, and then sends it to the other output tube.

Now, considering that an additional gain stage was introduced on the second output tube path, it is necessary to have a voltage divider (R15, R14) in front of this additional amplifier stage, in order for both circuit outputs to have the same amplitude. In other words, if for example, this extra tube provides a gain of let's say 48, the voltage divider needs to provide an attenuation of 1/48, for the circuit to give the same (or almost the same) output voltage for both output tubes.

DESIGN PROCEDURS
As previously indicated, the most important step is to determine the inverter tube voltage gain, in order to calculate its voltage divider to get a net gain is 1.
Depicted below is the circuit's DC & AC load lines and bias point for the basic paraphase circuit shown previously.

It is outside of this description's scope, to get in the details of how this is done, but there are few articles on the web explaining it step by step.



Once the stage gain is determined (in this specific case 48 approx.), then R_div. (R14 on our circuit), is equals to Rg_nxt (R17/ (Vg – 1)), or approximately 4,700 ohms.

SOME OBSERVATIONS

• The parahase circuit, considering all else the same, has the highest gain and voltage swing among of the popular tube inverters, which makes it ideally suited to drive big output tubes, like the 6L6. Just for refence, the split-load inverter, delivers half of the max. possible voltage swing and the LTPI, depending on the tail resistor, produces about 30% less output (B+ voltage reduction due to the tail resistor) and half of the voltage gain for the tube type utilized (shared input voltage signal).
• The second harmonic distortion for the OUT 2 of the paraphase IT'S NOT THE SUM of V2b and V2a distortion values. Why? Looking at the “Plate Average Characteristic Curves” (shown above), it's evident that the plate voltage swing is slightly less when the control grid voltage goes more negative (from the bias point), than when it varies to a more positive value of the same magnitude, resulting in an output signal with the positive semi-cycle compressed to a small degree. It can be mathematically demonstrated that such signal can be decomposed into its fundamental frequency and even harmonics.
  
  However, when this signal is fed to the inverting triode (V2a), due to the voltage inversion, the same happens to the other semi-cycle and this effect somewhat cancels the second harmonic distortion introduce by the first tube (or in this example, the other 12AX7 half).
• The inverter shared cathode resistor (R10) is "perceived" by each half of the 12AX7 as double of its physical value; that is why it is noted on the graph as 2x 1,000 ohms. It is important that this resistor stays unbypassed. If all is balanced, the cathode current increase on one side will by the same amount that it decreases on the other, producing a net current variation of ZERO and consequently no voltage variation on that resistor, so a bypass capacitor will have no effect. Nonetheless, any unbalance will produce a local NFB which will somewhat counterbalance the initial difference. This effect will be negated if a bypass capacitor is in place.
• This is regarding the circuit's LOW frequency response. On tube audio, with few exceptions, it's common practice to have R-C coupled amplifier stages. Every R-C network introduces some signal phase shift, like the tremolo phase shift oscillator circuit. The lower the frequency, the larger the phase shift on the signal path.
  
  Because of this, and the additional gain stage on one of the output tube's path (and consequently an additional R-C coupler), at lower frequencies, that circuit leg will have a larger phase shift than the other side. As result, when both halves of the signal are put together by the push-pull output transformer, they won't be perfectly aligned, causing a distortion like the cross-over distortion.
  
  This phenomenon limits the lower frequency amp's response, which is really a problem for Hi-Fi amps, because they need to go as low as 20Hz vs. 82Hz for guitar amps. This issue can be lessened by simply making R17 (on the above circuit) larger (e.g., Valco-Supro S6420 Thunderbolt), or C8 larger (e.g., Magnavox 9301/ 9302 chassis – the larger the capacitance, smaller its reactance). Either way, the phase shift for that leg will be reduced.
  
  Additionally, to help this issue, some designs have a small value resistor between pin-3 of V2b and the shared cathode bias resistor node, so it's possible to introduce NFB at the pin-3 (e.g., Ampeg M-12).
• The following resource was very helpful to improve my understanding of this subject:
  https://dalmura.com.au/static/Phase%...aft%201948.pdf

So, I am posting these notes here, for those who want to learn more about the paraphase inverter circuit.
To the experts, please feel free to correct or add any detail that I might have missed.
That's all I have for now... Cheers.

[Helmholtz]

This is a good roundup of the essential paraphase properties.
Yes, gain is about twice that of a LTPI using the same tube.
Directly comparing to the gain of a cathodyne PI (a little below 1) wouldn't be fair as this only needs one triode and the other one is typically used as a gain stage before the cathodyne, resulting in roughly the same total gain as with a paraphase.

Still the paraphase provides the largest output voltage swing of all with the same B+.
Question is if that's needed, as too much voltage swing at the power tube grids can result in ugly blocking distortion.

Other variants of the paraphase are descibed here:
https://www.valvewizard.co.uk/paraphase.html

[mozz]

From back in the days of 6SC7, 6j6, with joined cathodes.

[Helmholtz]

From back in the days of 6SC7, 6j6, with joined cathodes.

Now, that won't allow to feed global NFB to one of the cathodes only.

[K Teacher]

Helmholtz, you are correct… the paraphase may be too much output for the 6V6 or 6BQ5/ EL84 output pairs. A cathodyne PI will be just right (e.g., Tweed Deluxe 5E3, Peavey Classic 20 or Park 20W).

Note that the Peavey Classic 20 and the Valco-Supro S6420 Thunderbolt, both have output tubes grid stoppers of 47K… I think this is to take care of the possible blocking distortion…