Disbalancing paraphase inverter?

[Tobyk]

Did anyone try disbalancing a paraphase phase inverter?
And if so, how did it turn out in terms of overall tone, distortion and harmonics?

[R.G.]

It's a well explored topic - at least in the 1950s. I was just reading a book titled "High Fidelity Circuit Design" by Norman Crowhurst and George Cooper from 1956. It had a whole chapter on phase inverters, which covered as bad things (for their purposes) what happens when the PI is unbalanced. Unbalancing the signals to the output tubes can be done by tinkering the resistors is done in many amps using long tailed pair PIs. Fender uses a 100K and an 82K plate resistor pair in their PI to make up for the difference in gain in the two halves of the PI. This is also a good place to experiment if you're so inclined.

The short answer is that with the PI unbalanced, one of the output tubes runs at a higher or lower effective gain compared to the other. This introduces a variant of half wave rectification, where one half of the AC cycle(s) shrinks, even down to zero. There are effect pedals that do exactly this, with a variable amount of asymmetry. There was at least one amp that would deliberately turn off the signal to one output tube entirely to fake a single ended amp.

Feedback from the output of the power amp works to reduce this asymmetry to the extent that it can. Crowhurst and Cooper went through this a bit.

Best suggestion? Go try it. From your earlier questions, you know my take on the words "tone", "distortion", and "harmonics". You have to define what you mean these words to mean to have the answers make sense. "Tone", to an engineer, means frequency response, perhaps with phase response added. "Tone" to a guitarist may mean "sounds like lasers and flying saucers" or anything else. "Distortion" is something you measure with scopes and meters. "Harmonics" is something you measure with a spectrum analyzer. You will spend a lot more time scouring the internet for answers that make sense to you for how it sounds (and get less meaningful answers!) than if you spent a Saturday afternoon tinkering the PI resistors to see how it sounds to you.

As to how it sounds to me - to my ears it adds a slight "sweetening" in low doses, going to a more and more prominent "octave" effect and more prominent distortion as you turn up the imbalance. Your ears may like different things.

[roberto]

It's a well explored topic - at least in the 1950s. I was just reading a book titled "High Fidelity Circuit Design" by Norman Crowhurst and George Cooper from 1956.

I was reading it too few weeks ago.
I think that studying and experimenting Hi-Fi helps in viewing new path in instruments amplification, and studying and experimenting with guitar amps helps to find nice overdriving Hi-Fi amps (instead of clean-clean-clean-wall of distortion).

An easier option to evaluate the effect of an unbalance of the PI on a LTP, is to disconnect the nfb and add a 500 Ohm trimmer between the cathodes of the PI with the wiper to the LT.

[wpaulvogel]

Slowly removing NFB from a LTP is not going to show you anything about the imbalance of the phase inverter outputs unless it’s already fairly imbalanced to begin with and the feedback is helping to correct the balance. It WILL remove the NFB and show you more gain. Replacing one plate resistor with a pot configured for variable resistance and removal of NFB will allow you to experiment with imbalance of the LTP.
In the paraphase inverter you can make the voltage divider for the second triode adjustable to create imbalance.

[Tobyk]

It's a well explored topic - at least in the 1950s. I was just reading a book titled "High Fidelity Circuit Design" by Norman Crowhurst and George Cooper from 1956.

I was reading it too few weeks ago.
I think that studying and experimenting Hi-Fi helps in viewing new path in instruments amplification, and studying and experimenting with guitar amps helps to find nice overdriving Hi-Fi amps (instead of clean-clean-clean-wall of distortion).

An easier option to evaluate the effect of an unbalance of the PI on a LTP, is to disconnect the nfb and add a 500 Ohm trimmer between the cathodes of the PI with the wiper to the LT.

Replacing the voltage divider on a paraphase with a pot would be easy enough, I believe. Seems noone here tried it though. Which is why I need to do it – I think I saw a thread on 18watts.com where Merlin stated disbalanced paraphases creates more 2H than LTP or Cathodyne.

[roberto]

Hi wpaulvogel,

yes, absolutely, you need to unbalance the PI to unbalance it. It's self-evident.
R.G. was suggesting to do it from the top of the LTP, I suggested from the bottom, but the concept it's the same.

IMHO Paraphase should always have a trimmer to balance/unbalance it.

[roberto]

Replacing the voltage divider on a paraphase with a pot would be easy enough, I believe. Seems noone here tried it though. Which is why I need to do it – I think I saw a thread on 18watts.com where Merlin stated disbalanced paraphases creates more 2H than LTP or Cathodyne.

I tried, but I don't like paraphase. That's why I talked about the LTP that I like more and I have more experience with.
Every asymmetrical waveform has even harmonics in it. It's not just 2nd, but 4th, 6th, etc...
And it is not the paraphase that has more than others, every PI can be unbalanced how much as you want, but as we were saying, you will hear more if you disconnect the nfb, otherwise it will try to fix the unbalance and will add more upper harmonics due to the injection of a strong 2nd harmonic correction on the original sound.

[Tobyk]

Replacing the voltage divider on a paraphase with a pot would be easy enough, I believe. Seems noone here tried it though. Which is why I need to do it – I think I saw a thread on 18watts.com where Merlin stated disbalanced paraphases creates more 2H than LTP or Cathodyne.

I tried, but I don't like paraphase. That's why I talked about the LTP that I like more and I have more experience with.
Every asymmetrical waveform has even harmonics in it. It's not just 2nd, but 4th, 6th, etc...
And it is not the paraphase that has more than others, every PI can be unbalanced how much as you want, but as we were saying, you will hear more if you disconnect the nfb, otherwise it will try to fix the unbalance and will add more upper harmonics due to the injection of a strong 2nd harmonic correction on the original sound.

Yeah, well LTP is less interesting to me since it’s everywhere. I’d like a more old school lo-fi sound, which maybe the paraphase is better at. Why do you dislike them? So you’re saying all PI’s will render the same effect when unbalanced by the same amount?
I wasn’t thinking of using NF at all, by the way.

[roberto]

I absolutely agree that the best solution is to experiment with different circuits and then see what we prefer.
Yes, you can unbalance every PI you want in different ways. The hard task would be to keep it always symmetrical.

What I like to use is a simple CCS as tail of the LTP, because you can DC couple the previous stage to the PI input and you increase the dynamic of the stage.
On top of that, you can select the current you want flowing through the triodes, and you can use the same anode resistor value for both anodes as the tail is generally a very high value (if you use a transistor with a high Hfe).
In Hi-Fi I cascode two transistors for the tail, and I balance them through a trimmer between cathodes. Not something needed in a guitar amp.

It has been done so much work on amps, that it is just a matter of finding all possible variation, or invent new ones.

[R.G.]

Just clarifying a couple of items.
- The only reason I mentioned negative feedback is that negative feedback hides imperfections - like any introduced by unbalancing the PI. Large amounts of negative feedback are very effective at hiding imperfections and nonlinearity. This is the reason that hifi amps are squeaky clean right up until they distort massively. The feedback hides any funny stuff right up until the devices just can't provide the currents and voltages needed for "clean" any more. So if you want to see what imbalances do to sound, you need to remove the NFB so it doesn't hide what you're testing.
- There are always imbalances. As Roberto mentions, the LTP without a constant current source cathode drive is inherently imbalanced. This is the reason for the unbalanced plate resistors; they offset the built-in imbalances. Beyond that, the output tubes are potentially imbalanced not only for DC bias point (which we correct for with individual bias adjustments) but for AC gain per tube, which varies slightly per tube; and for tolerance imbalance in the resistors in the PI (plate resistors in LTP, plate vs cathode resistors in paraphase); and for imbalances in the two halves of the OT primary, even for resistor tolerance in grid leak resistors. These are generally small, but they're always there, and they will mess with any conclusion you draw about amps in general.
- I haven't read any of Merlin's books, but the comment about distortion sounds right-ish. Comments about second harmonic, third harmonic, etc strictly only apply to sine wave testing. If you take a sine wave and alter the gain on one side, the resulting waveform will have some degree of second harmonic distortion. If you run a sine wave through a specially-prepared distorter which has only signal-squared distortion, you get pure 2x frequency as a distortion product. This was the basis of my MOS Doubler effect circuit, as MOSFETs happen to have a range over which they introduce only square-law distortion. Non-square-law distortion, as in unbalancing a PI, you get 2H, 4H, 6H, 8H, etc.
This is only true for pure sine waves. If you have a signal that has, say frequency F1 and frequency F2, you get some amount of 2xF1, 2xF2, which are the second harmonics you are notionally trying to get, but also F1+F2, F1-F2, 2xF1+/-F2, F1+/-2xF2, etc. as smaller intermodulation products. There is no pure second harmonic generation possible with analog electronics of a music signal.

Sorry for wandering on. As I mentioned, there's a lot of pre-existing work on the effects of distortion and unbalanced P-P.

[roberto]

Going on with methods to unbalance other PIs, with concertina you can simply set the working point of the previous DC coupled stage on the warmer or colder side, and change the behaviour of the concertina.
You can also set a specific working point and then dc couple the concertina with a split load on the anode of the previous stage (stages are still dc-coupled, but at a different voltage).

[roberto]

If you want to see a new phase splitter (I've never seen any like that before), take a look at the schematic below.
It's an Hi Fi PP amp I'm developing (born as SE where I can get 26 Wrms at 5% THD with one single KT88) fully DC coupled.

You can see the cascoded CCS as the tail, the balancing trimmet on cathodes and the two triodes of the PI.
They work at a fixed voltage (kept fix by the transistors on their anodes), so the voltage on their grid changes the current that flows through them.
They basically work on a vertical loadline.

The difference between the fixed amount of current that flows through the resistor at its anode and the current that flows through the triode, will flow through the 33k resistors, and being V= I x R, you will get a signal.
More than that, you'll get a signal that can be higher than mu. It is an evolution (or adaptation, as you prefer) of Rod Coleman's shunt cascode. The good thing is that with the output tubes' configuration I can directly DC couple the PI to the output tubes (it is required a positive bias voltage, not a negative one), and swing all the needed 230Vpp to reach 105 Wrms at 5% THD.

In this case you can unbalance the PI through the trimmer on the cathodes of the PI or using a different value for one of the 33k resistors.

Roberto Hi-Fi 12AT7 6550.jpg

[Tobyk]

If you take a sine wave and alter the gain on one side, the resulting waveform will have some degree of second harmonic distortion. If you run a sine wave through a specially-prepared distorter which has only signal-squared distortion, you get pure 2x frequency as a distortion product. This was the basis of my MOS Doubler effect circuit, as MOSFETs happen to have a range over which they introduce only square-law distortion. Non-square-law distortion, as in unbalancing a PI, you get 2H, 4H, 6H, 8H, etc.
This is only true for pure sine waves. If you have a signal that has, say frequency F1 and frequency F2, you get some amount of 2xF1, 2xF2, which are the second harmonics you are notionally trying to get, but also F1+F2, F1-F2, 2xF1+/-F2, F1+/-2xF2, etc. as smaller intermodulation products. There is no pure second harmonic generation possible with analog electronics of a music signal.

And by ’pure 2x frequency’ you mean ’pure one octave above’ added? I’ve actually heard something at least close to a pure octave above, from the grid leak biased mic channel on a 50’s GA-20 Gibson. Do you have any idea what could cause that in an old analog amp like that?

[Tobyk]

Going on with methods to unbalance other PIs, with concertina you can simply set the working point of the previous DC coupled stage on the warmer or colder side, and change the behaviour of the concertina.
You can also set a specific working point and then dc couple the concertina with a split load on the anode of the previous stage (stages are still dc-coupled, but at a different voltage).

I think on a cathodyne, altering the PI cathode tail resistor would be the easiest way to imbalance it, no?
Btw, you didn’t answer why you don’t like paraphase inverters
Congrats on your invention, most impressive!

[roberto]

Btw, you didn’t answer why you don’t like paraphase inverters

I can be biased by the fact that most overdriven sounds come from LTP based amps and the way it behaves, but I prefer the way the LTP sounds.