Floating paraphase stage

[nickb]

Lot's to get through I'm afraid. I've added plots and schematics to try to illustrate.

Let's start with a simple comparison of the Floating Paraphase with common cathode resistor (FP) and Floating Paraphase without common cathode resistor (P) Bode Plot. I left off phase to make it clearer. I also set the source impedance to zero to simplify things.

The thing of note here is that there is only a 0.2dB difference between the P and the FP. This is contrary to the suggestion that the FP behaves differently from the P due to the 'differential amplifier' effect. I'll come back to this. Also note that there is a 1dB difference in gain between the A and B (i.e inverting and non inverting) sides.



Next, we can fix the gain imbalance:

I changed the value of the input resistor of the second stage to correct for gain imbalance by changing it from 1Meg to 886K. Something to try on the next one you come across but consider the effect of other component tolerances as they may be 5% or even 10% resistors which will unbalance the whole thing pretty darn quickly. Sorry about the blurry screen capture. I was trying to save myself some time and also I wanted to combine the schematics and plots into a single image.



It's time to widen the X axis frequency out to 1Mhz:

Now you can see the extra HF roll off on the 'B' output that has been mentioned. It's 45Khz so you'll be starting to get see about 5 degrees of phase shift at 4.5Khz. Not a lot, but a good case has been made that this might color the sound. You will get, I think, increasing even harmonic distortion with frequency although I'm having a hard time to see how to approach this bit analytically.



Not to worry, we can fix that HF roll off:

I just fixed the P side so you can see the difference. A teeny 2.7pF cap fixed it. Now that's kinda scary. Before you rush off and modify your amps think about stray capacitance. Just running the grid wire next the the plate wire could easily give that much coupling. Layout could be significant here.




Lastly, why in blazes do the FP and P versions behave almost identically? To be sure the differential effect is real, even with a small 470 ohm resistor. But why doesn't it have any effect in the FP case? The answer lies in the magic feedback network around the second stage. As has been pointed out, the second stage is bit like a common grid stage with a low input impedance to the cathode. The feedback network lowers the input impedance so the first stage (with a much higher output impedance) now has very little effect on the second stage. The plot shows that input impedance is only about 5 ohms.





I don't know what value there might be in experimenting the gain and frequency balancing but to be sure, there's some flexibility there. Of course you can do the same kind of thing to a fully differential PI too...

[Alan0354]

Wow, it's too much to read after a few drinks and dinner. I'll look at it later or tomorrow.

[Alan0354]

I looked at the first 3 cases, I don't know what to comment or supposed to comment!!! This is another variation of the circuit again. I don't know whether I want to get into it again!!! Others are welcome to join in and take over. I don't mind being hijacked!!!!

[jazbo8]

nickb's simulations showed how the imbalance and HF concerns could be addressed, but the simulations assumed perfectly matched tubes with nearly ideal characteristics, so in real amps, the performance is always worse depending on the tubes, but even if you hand-pick matched tubes, the balance will still drift with tube aging; that was the main reason RCA came up with the floating paraphase, i.e., the circuit is self-balancing.

[Alan0354]

nickb's simulations showed how the imbalance and HF concerns could be addressed, but the simulations assumed perfectly matched tubes with nearly ideal characteristics, so in real amps, the performance is always worse depending on the tubes, but even if you hand-pick matched tubes, the balance will still drift with tube aging; that was the main reason RCA came up with the floating paraphase, i.e., the circuit is self-balancing.

I think I just about having enough of the paraphase and FP already. I think I make my analysis very clear and I stand by my conclusion. I really have no interest of using it as I explained many times. I am happy with the differential PI and it's time for me to move on.

Of cause, if someone tell me in definitive that it sounds better than the differential pair, then I can be interested in it.

[jazbo8]

I don't think anyone would claim the FP sounds better, after all it is used in so few amps, and it really comes down to personal preference anyway. For the theory part, we only looked at small signal analysis, and have not even touched on at the large-signal performance which is what truly differenciate the various PIs, but large-signal analysis are much harder to do. Save it for a rainy date perhaps...

[soundmasterg]

I think I just about having enough of the paraphase and FP already. I think I make my analysis very clear and I stand by my conclusion. I really have no interest of using it as I explained many times. I am happy with the differential PI and it's time for me to move on.

Of cause, if someone tell me in definitive that it sounds better than the differential pair, then I can be interested in it.

It sounds different...better or not is in the ear of the beholder. It is nice to get experience with it and other circuits so you know for a fact what they sound like rather than coming to some conclusion based totally on the theory. The LTP has a distinctive sound to it that benefits an amp that is overdriven a lot. The FP and cathodyne also have distinctive sounds, though the cathodyne has issues when it distorts that Merlin addressed in his books and his site. The FP actually has the capability of more drive to the power tubes compared to the other two, and depending on your application this might be just the thing to use. Say you have an output stage using KT88's or some other larger tube like an 813 that needs a lot of drive. The FP might be a better choice here than the LTP or cathodyne. Anyway, don't just assume that the theory is all you need to understand a circuit and how it will perform. When we are talking about audio, it can work fine on paper or in sims but the end result is how does it sound, and for that you need to build it and test it.

Greg

[ReadyTeddy]

The LTP can easily be configured to provide more drive. The key is to use a solid-state current source connected between ground and the common cathode connection. I've seen the LM334 used with a 12AX7 for this purpose. The LM334 has a burden voltage down to 1V, so you have virtually the whole B+ voltage available for the plates. It would be interesting to see how this would sound in a guitar amp--the example I saw was used in a hifi amp.

[Alan0354]

The LTP can easily be configured to provide more drive. The key is to use a solid-state current source connected between ground and the common cathode connection. I've seen the LM334 used with a 12AX7 for this purpose. The LM334 has a burden voltage down to 1V, so you have virtually the whole B+ voltage available for the plates. It would be interesting to see how this would sound in a guitar amp--the example I saw was used in a hifi amp.

I think Leo Fender attempted to make the long tail constant current by feeding back from 820 ohm to the bottom of the long tail and use the 100 ohm to develop the voltage so the voltage across the current setting resistor (22K+470) is kept approximately constant to get constant current.

[nickb]

nickb's simulations showed how the imbalance and HF concerns could be addressed, but the simulations assumed perfectly matched tubes with nearly ideal characteristics, so in real amps, the performance is always worse depending on the tubes, but even if you hand-pick matched tubes, the balance will still drift with tube aging; that was the main reason RCA came up with the floating paraphase, i.e., the circuit is self-balancing.

That's rather missing the point. The point was to demonstrate that both the FP (i.e. withe shared cathode resistor) and the P (no shared resistor) are the same to all intents and purposes. True, the simulation uses ideal components but it makes no difference - both are 'self balancing' and are very tolerant of mismatched tubes or ageing. OTOH, both are equally intolerant of an error in the two feedback resistors that set the gain of the second stage.

I do seem to have misappropriated the "FP" and "P" designations hope that hasn't caused too much confusion. I should have called them "Shared resistor" and "Non shared resistor", I suppose.

[jazbo8]

Nick,

The paraphase circuit you used for the simulation is not its "standard" form, there is an extra 1M resistor from the plate to the grid on U4. The standard paraphase (see below) does not have any feedback mechnism, thus can not be self-balancing. You actually improved it

[nickb]

Nick,

The paraphase circuit you used for the simulation is not its "standard" form, there is an extra 1M resistor from the plate to the grid on U4. The standard paraphase (see below) does not have any feedback mechnism, thus can not be self-balancing. You actually improved it

Quite so. I see my misuse of the terms has indeed caused confusion. Sorry about that.

Forget where I use "paraphase" or "P" and substitute "non shared cathode resistor floating paraphase" and change "floating paraphase or "FP" to "shared cathode resistor floating paraphase". So what I'm comparing is these two topologies. It was argued at one point that these have quite different behavior yet in fact they are very similar.

[printer2]

I think I just about having enough of the paraphase and FP already. I think I make my analysis very clear and I stand by my conclusion. I really have no interest of using it as I explained many times. I am happy with the differential PI and it's time for me to move on.

Of cause, if someone tell me in definitive that it sounds better than the differential pair, then I can be interested in it.

Better? More like just another flavor. Me, I want both.

[jazbo8]

Got it, it was a long and winding thread, I lost track many times too...

[jpfamps]

We use the paraphase PI in most of our amps, and it works very well (in my opinion).

The issue I had, that confused my for some time, was unpleasant distortion on the decay of the signal when the amp was overdriven.

After some investigation, this was due to the inverted signal drive line going into grid conduction before the non-inverted (or should that be re-inverted). Once the phase inverter was rebalanced to avoid this the problem disappeared.

I think the paraphase (and indeed floating paraphase) splitter went out of favour due to hi-fi driving the state-of-the-art and the paraphase inverters lost out to the concertina (in terms of balance), and the LTP (in terms of distortion cancellation).

My guess is that up to clipping in the PI and/or grid conduction by the power valves there will be little audible difference between the various PIs, however their behaviour under clipping or at grid conduction will vary.

Concerns about audible effects of phase shifts at HF in the PI in a valve guitar amp are in my view pointless.