Floating paraphase stage

[Alan0354]

The 5E9-A schematic shows a poorly designed PI for all the reasons previously stated. I'm surprised ole Leo didn't know better at the time.

Remember he designed in the 50s. We have the 20/20 hindsight with the extra 60 years of knowledge base. Like people said, you have to judge by historical perspective. Like I have a lot of respect towards RDH4, even though some does look a little funny. BUT those are the fore front where without them, we won't have nothing to improve on.

Like Sir Isaac Newton. Most of his theory has been overthrown already. But that does not mitigate his greatness and deserves all the respect.

In fact, of the books I read on tubes, I like RDH4 the best so far. It has a lot of information without the editorial. It gives the important characteristics so I can think and digest. I don't want the coddling the newer books try to do, that potentially can be less than right!!

[Dave H]

What do you think of this PI Alan with its unity gain tube "opamp"


When I posted the above in the original thread. It was only supposed to be a small diversion to show Alan how the deficiencies of the Paraphase PI had been addressed by the Floating Paraphase. I never expected this behemoth to develop!

I’m not quite sure what we are supposed to be discussing in this new thread so I’ll describe how I think the LTPPI and the two circuits I posted work (or don’t work) and leave it at that for now.

A positive going signal applied to the left grid of the LTPPI increases the current through the tube causing the plate voltage to fall and because the cathodes are not grounded the cathode voltage rises by cathode follower action. The right grid is grounded so the rising cathode voltage reduces the current through the right triode and its plate voltage rises. The cathode of the right triode is its input.

How is the Floating Paraphase different? Superficially it looks like a duck (LTPPI) because of the common 470R cathode resistor but does it have the other two attributes of a duck? Again a positive going signal applied to the left grid of the FPPI increases the current through the tube causing the plate voltage to fall but this time the right triode is configured as a unity gain inverting opamp (as I called it) so the current and voltage of the right triode exactly mirror those of the left. The left triode current has increased by x (say) but the right triode current has reduced by x meaning there is no net change in current through the common cathode resistor therefore no change in voltage across it. It’s effectively a virtual ground. You could nail it to ground by putting a BFC (capacitor) across it and it would make no difference to the operation of the circuit which is what I was trying to show by separating the cathodes and adding bypass caps on the second circuit I posted. I’ve simulated the two circuits and you can’t get a Rizla between them, for all intents and purposes they are the same. The FPPI may look like a duck but it doesn’t walk like a duck or talk like a duck. It is not a differential amplifier like the LTPPI. It’s still a bird (PI) alright but it’s a different species from the LTP.

What about converting the FP into a differential amplifier by removing the “unnecessary” components and grounding the right grid as Alan recommends? What species do we have now? It’s a duck alright but it’s the runt. It has a vestigial tail and a crippled right wing. It can flap its left wing up and down 50V but only manage 12V with its right (just as nickb predicted). As a phase inverter I’d say it’s a dead duck.

[jazbo8]

What looks like a duck but doesn't quack like one Well it's the weekend so we can use some diversion... if you remove all the "un-necssary" parts and ground V2's grid, then you have the classic Mullard LTP, well as least very close to it. One thing worth noting is the high output capability of the FP, when you need the grunt, it is still a good alternative. If concern about the output balance remains, it can be addressed with build-out resistor to make the output impedance of the two halves identical, and/or use a cascode per half at the cost of even more parts!

[Alan0354]

What do you think of this PI Alan with its unity gain tube "opamp"


When I posted the above in the original thread. It was only supposed to be a small diversion to show Alan how the deficiencies of the Paraphase PI had been addressed by the Floating Paraphase. I never expected this behemoth to develop!

I’m not quite sure what we are supposed to be discussing in this new thread so I’ll describe how I think the LTPPI and the two circuits I posted work (or don’t work) and leave it at that for now.

A positive going signal applied to the left grid of the LTPPI increases the current through the tube causing the plate voltage to fall and because the cathodes are not grounded the cathode voltage rises by cathode follower action. The right grid is grounded so the rising cathode voltage reduces the current through the right triode and its plate voltage rises. The cathode of the right triode is its input.

How is the Floating Paraphase different? Superficially it looks like a duck (LTPPI) because of the common 470R cathode resistor but does it have the other two attributes of a duck? Again a positive going signal applied to the left grid of the FPPI increases the current through the tube causing the plate voltage to fall but this time the right triode is configured as a unity gain inverting opamp (as I called it) so the current and voltage of the right triode exactly mirror those of the left. The left triode current has increased by x (say) but the right triode current has reduced by x meaning there is no net change in current through the common cathode resistor therefore no change in voltage across it. It’s effectively a virtual ground. You could nail it to ground by putting a BFC (capacitor) across it and it would make no difference to the operation of the circuit which is what I was trying to show by separating the cathodes and adding bypass caps on the second circuit I posted. I’ve simulated the two circuits and you can’t get a Rizla between them, for all intents and purposes they are the same. The FPPI may look like a duck but it doesn’t walk like a duck or talk like a duck. It is not a differential amplifier like the LTPPI. It’s still a bird (PI) alright but it’s a different species from the LTP.

What about converting the FP into a differential amplifier by removing the “unnecessary” components and grounding the right grid as Alan recommends? What species do we have now? It’s a duck alright but it’s the runt. It has a vestigial tail and a crippled right wing. It can flap its left wing up and down 50V but only manage 12V with its right (just as nickb predicted). As a phase inverter I’d say it’s a dead duck.

That's where you miss it. Using your theory, tell me what mechanism that cause the right triode to follow with gain of -1? What is the input of the second triode?


The input of the right triode is the cathode of the right triode!!! In order to have a gain of -1 where the plate of the second triode swing negative for x amount. You need to swing the cathode of the right triode positive by x amount!!! Remember it is suppose to be a unity gain inverting amp......a voltage amp?

It is absolutely not equal to the circuit having a bypass cap on the two cathodes. That's where the coupling of the two triodes if you don't have the cap.


Anyway, the question pretty much being resolved. As Jazbo8 already confirmed the original FP patented by RCA has the two triodes completely separated. This is confirmed by the working product of Fender Tremolux, that there is a big cap bypassing the two triodes.

At this point, I believe the diagram of Merlin is a typo. This happens very often as the publisher make mistake in reproducing the material in the book all the time, It is to the point I buy 4 books on every subject I study. When things don't make sense, I go to another book and verify.

[nickb]

Been busy all day so rather late to the party...

<snip>
My main gripe about paraphase is the way they get the opposite phase is using an inverted stage of unity gain to generate the 180 deg off phase signal. Problem is the extra stage has propagation delay in form of limited frequency response. I tested out a typical common cathode stage with 100K plate resistor. The -3dB corner frequency is about 20KHz. At that frequency, the phase shift is 45 deg.
<snip>

My view is have to look at the whole thing from a historical perspective.

Now I'm not certain of the actual timeline but it seems reasonable to me that the first attempt to get a push pull drive (ignoring the cathodyne) was the paraphase where the input to the second stage is attenuated by it's gain to give a net unity gain. This, as you point out, does suffer from the HF rolloff.

I imagine that the next step to solve this was the 'unity' gain buffer where local feedback was applied to the second stage to give unity' gain and have the benefit of moving that troublesome pole way out and also stabilizing the gain. Soon, Mr Clever Engineer realized that since the currents in the first and second stages were about equal but complementary, a single common cathode resistor could be used saving a resistor and two electrolytics, yet still get the same performance. Just for fun, I simulated the two cases and the gain difference is only 0.05dB. So the floating paraphase was born and he earned a place in history, a pat on the back from his boss for the cost saving and a debating topic for evermore for the rest of us for his efforts.

I do wonder if he was looking as the FP design one day and the idea for the true differential PI occurred to him or not, but it seems possible.

I don't worry too much about the deficiencies that these designs have. It is important to know about them but at the same time relish the subtle differences in sound that they make. I think of particular note is the higher second harmonic distortion.

[Dave H]

That's where you miss it. Using your theory, tell me what mechanism that cause the right triode to follow with gain of -1? What is the input of the second triode?


The input of the right triode is the cathode of the right triode!!! In order to have a gain of -1 where the plate of the second triode swing negative for x amount. You need to swing the cathode of the right triode positive by x amount!!! Remember it is suppose to be a unity gain inverting amp......a voltage amp?

It is absolutely not equal to the circuit having a bypass cap on the two cathodes. That's where the coupling of the two triodes if you don't have the cap.


Anyway, the question pretty much being resolved. As Jazbo8 already confirmed the original FP patented by RCA has the two triodes completely separated. This is confirmed by the working product of Fender Tremolux, that there is a big cap bypassing the two triodes.

At this point, I believe the diagram of Merlin is a typo. This happens very often as the publisher make mistake in reproducing the material in the book all the time, It is to the point I buy 4 books on every subject I study. When things don't make sense, I go to another book and verify.

I tried really hard to get it all down unambiguously but I’ve failed again. You’ve misread it. I separated it into paragraphs, one paragraph for each circuit.

The paragraph for the LTPPI ends with the statement “The cathode of the right triode is its input.” That only applies to the LTPPI where the two triodes have similar gain (25?)

In the next paragraph it’s the right triode of the FPPI which has a gain of -1 and its input is not the cathode but the grid. The cathode doesn’t swing anywhere. The input is to the grid through a 1M from the plate of the first triode. There is a second 1M feedback resistor from plate to grid of the right triode making it a unity gain inverting opamp. I tried to make it look more like an opamp on the schematic of the second circuit.

The question is not resolved and Merlin‘s circuit Works fine. It is not a typo. He is too conscientious for that. He is holding back publication of his power supply book because he is not satisfied with its accuracy.

Please read my previous post again with the above in mind and hopefully it will make more sense.

[nickb]

Alan0354 wrote: "This is a differential pair, I don't think you can isolate the two triode."

In hindsight, perhaps the above sentence sums up why the discussion went off in the direction that it did... the circuit shown on Merlin's page is a modified form of the RCA's floating paraphase circuit (I mistakenly said Mullard in my earlier post). The RCA FP did not have a shared cathode resistor, so it could not have been construed as a differential amplifier stage at a quick glance. In any case, given the relative small resistance used in Merlin's FP example, it would not make a good differential amplifier as nickb pointed out on this post.

The first mention of the "self-balancing phase-inverter", aka the "floating paraphase" that I can find is in RDH4's footnote on phase-inverters and it dates back to 1938! If anyone has a copy of the original RCA application note, please share it!

For detailed descriptions on how the circuit works, we can refer to two patent filings by RCA (way after the application note came out);
I think this is the actual patent: RCA patent 1
One kinda related: RCA patent 2

Here is an article from 1940 showing the RCA circuit:
[ATTACH=CONFIG]27889[/ATTACH]

A geek with nothing better to do...
Jaz

Nice research Jazbo8. +1

[Alan0354]

I tried really hard to get it all down unambiguously but I’ve failed again. You’ve misread it. I separated it into paragraphs, one paragraph for each circuit.

The paragraph for the LTPPI ends with the statement “The cathode of the right triode is its input.” That only applies to the LTPPI where the two triodes have similar gain (25?)

In the next paragraph it’s the right triode of the FPPI which has a gain of -1 and its input is not the cathode but the grid. The cathode doesn’t swing anywhere. The input is to the grid through a 1M from the plate of the first triode. There is a second 1M feedback resistor from plate to grid of the right triode making it a unity gain inverting opamp. I tried to make it look more like an opamp on the schematic of the second circuit.

The question is not resolved and Merlin‘s circuit Works fine. It is not a typo. He is too conscientious for that. He is holding back publication of his power supply book because he is not satisfied with its accuracy.

Please read my previous post again with the above in mind and hopefully it will make more sense.

It is not true, the moment you apply a +ve signal to the grid to the first tube, the cathode of the first tube immediately follow before the plate goes down, then pulling the 1M down. You have to follow the physical order!!! The cathode of the first tube will pull the cathode of the second tube up first.

Then the plate current of the first tube increase as the current of the second tube decrease at the same time. The junction between the two 1M resistor does not even move as the voltage cancelled each other. That is the virtual ground you mentioned in your very first post before. That you can put a bypass cap on the junction of the two 1M and the circuit works the same.

They absolutely interact, you have to follow the event in order!!!! I tested these common cathode stages, the -3dB is only about 20KHz, there is a definite delay from grid to plate due to the plate capacitance and other factor. The cathode on the other hand, follow the grid much faster as the cathode follower has much higher frequency response!!! You have to look at this in the picture. The second tube is acting like a common grid stage and response faster as it does not have the Miller effect as it is being driven from the cathode.

At best, you can say they interact with each other as I said in the very beginning. It works because both ultimately are doing the same thing.

This is not the same as FP from RCA as Jazbo8 posted and Fender Tremolux.

[Alan0354]

Nice research Jazbo8. +1

Definitely big improvement from the original paraphase.

[Dave H]

It is not true, the moment you apply a +ve signal to the grid to the first tube, the cathode of the first tube immediately follow before the plate goes down, then pulling the 1M down. You have to follow the physical order!!! The cathode of the first tube will pull the cathode of the second tube up first.

LTSpice says the cathodes go nowhere. I’ve put spice plots of Merlin’s circuit, Merlin’s circuit with cathode bypass cap and my split cathode circuit below. The green flat line just above zero is the cathode voltage. It doesn’t move in all three cases. You have to zoom in to the mV level to see any movement. All thee circuits produce the same output. They were smart back in the day. It must be as nickb said. They noticed that they could save three components by using an unbiased cathode resistor so they did.

Then the plate current of the first tube increase as the current of the second tube decrease at the same time. The junction between the two 1M resistor does not even move as the voltage cancelled each other. That is the virtual ground you mentioned in your very first post before. That you can put a bypass cap on the junction of the two 1M and the circuit works the same.

I don’t think you can put a cap from the virtual earth to ground. It’s not a very good virtual earth because the “opamp” only has an open loop gain of 50 so if it has a signal of 50V at its output there will be a 1V signal at its input (the virtual ground). If you short it to ground with a big C there will be no input and the plate voltage will flat line. Even a real opamp must have a signal between its inputs equal to its output signal divided by the open loop gain.

[Alan0354]

Well, there is common sense in circuit design, just ask yourself which path is faster.

1) a cathode follower (V1) driving a common grid stage (V2).

2) a common cathode (V1) stage through the plate that hamper by the plate capacitance, then driving through 1M voltage divider. The junction is hampered by the Miller capacitance of the grid of V2. Then from the grid of V2 to the plate.

I tested these common cathode stage, the -3dB frequency is about 20KHz, that is phase lag 45 deg at 20KHz. This is 1/f=50uS delay, phase shift can be look at as time delay!!! There is a big delay from grid to plate!!!! You care to conclude what path arrive first?

this is common sense, in digital design, we look in terms of propagation delay. #1 runs at RF speed, the other run at audio speed!!! I don't need simulation program to tell me. In digital design, this is how people look at the delay of different paths. Last I check, this is every bit analog electronics applies to tubes.


If you run the simulation with a few hundred Hz sine wave, you might not see it as it is slow. Try put a square wave with fast rising edge, if you don't see a glitch at the cathodes in the simulation, something is WRONG with the program. This will show the propagation delay between the two path.

I am not saying your stage does not work for guitar amp, I am saying this is not what you think how it works. Don't just read from Merlin, think about it, look at the two path and judge it yourself.

[Alan0354]

I am not trying to be difficult, This circuit will work on guitar amp, it just does not make sense. This is theory section and this is my thread. I think it is legitimate to ask this question and challenge the establish theory that does not make sense. This is what theory is about. Not just try and see!!!

[printer2]

It is not like we need response out much past 5 kHz, who cares what happens at 20 kHz? You can put a guitar signal though a Mcintosh, Audio Research, Quad and it will sound pretty boring.

[Alan0354]

It is not like we need response out much past 5 kHz, who cares what happens at 20 kHz? You can put a guitar signal though a Mcintosh, Audio Research, Quad and it will sound pretty boring.

Right. Absolutely.

BUT you also have to consider the theory behind when you use a circuit also. It just happen that the opamp is doing the exact job of the differential amp and you don't see a difference. But if you try an asymmetrical plate resistor, the problem is going to be exaggerated. You do try to do it right by theory too.

This is what theory is all about.

[Alan0354]

I forgot, remember when you clip the circuit in the high gain channel, the waveform into the PI is square wave. Square wave has 3rd harmonics and 5th harmonics. Say you input a 2KHz from guitar ( low enough?). When clipped, you generate the 3rd harmonic = 3X2KHz=6kHz, 5th harmonic = 5X2KHz=10kHz. Now that is not low audio frequency. That you definitely have to look at phase shift delay through the PI stage.

Could this be the reason people say the sound is the same with clean, the difference is when clipped? True Differential pair is symmetrical, it does not affect by this. Both the Paraphase and FP have extra delay on the second signal. When you increase the freqency due to clipping, the sound changes as the two signals from the PI is no longer opposite phase.

Sue, you don't hear those high frequency through the speaker, but it does have effect on the intermodulation inside the power amp. That's when the color, the "breath" of the sound comes in!!!!