Getting a better understanding of push pull, amplifier class, and tube output stages

[mikepukmel]

Hi All,
Reading up
http://www.valvewizard.co.uk/pp.html and a few others, came up with a few questions about push pull output stages. First, various articles say about the same thing, (Hope Merlin doesn't mind the quote from his content).

"An output transformer designed for push-pull operation has a centre tap. The HT is applied to the centre tap and either end of the primary winding is connected to the anode of a power valve. In this way, current flows in opposite directions through the output transformer."

I get the idea that the center tap is used to feed HV DC to the plates of the output tubes. And, since its DC, it doesn't pass through the transformer, only the AC signals.

What Im not quite sure about is whether the center tap is actually required for a push pull amp. E.g. suppose there was no center tap, but the HV DC was fed through, say plate resistors back to the power supply. The DC voltage would travel to the plate, across the OT primary winding to the other tube plate. Wouldn't this effectively be the same or similar circuit?

The AC signals having been mirror imaged in the phase inverter, the output tube plates would still have signals 180 out of phase from each other, so when the signal is causing charge to flow out of one plate, it is flowing in to the other plate. Or something like that.

Is the center tap needed for the CLASS A/CLASS AB/ CLASS B stuff. Still not quite there on all that, but I understand in a descriptive way.

Thanks.
MP

[Helmholtz]

What Im not quite sure about is whether the center tap is actually required for a push pull amp. E.g. suppose there was no center tap, but the HV DC was fed through, say plate resistors back to the power supply. The DC voltage would travel to the plate, across the OT primary winding to the other tube plate. Wouldn't this effectively be the same or similar circuit?

While this might work in principle (think a LTP with an OT having the floating primary connected between the plates), the plate resistors would eat up most of the availabe output power.

[mikepukmel]

Thanks Helmholtz. A related question: So, the plates are held up to some HV DC potential through the primary OT windings. in class AB, with no input signal, current is not changing in the OT primary so no audio output. If a pulse, suppose very simple one half a sine wave cycle. The tubes are biased so that one tube amplifies the signal. Suppose voltage increases on the plate of one tube, so that charge flows into this tubes side of the OT. If the other tube is "off" so that it is not amplifying, where does the charge go? Since the other tube is off, the charge can't go back through that tube to ground. Or can it? Does 'off' just mean that it is not amplifying its input? If the other tube, the one that is amplifying, pushes charge, then does that charge go through the 'off' tube to ground? Trying to get a sense of how the halves of the primary windings work. I think these amps always use the entire primary winding (no matter what the class of the amplifier), is that correct?

[Dave H]

If a pulse, suppose very simple one half a sine wave cycle. The tubes are biased so that one tube amplifies the signal. Suppose voltage increases on the plate of one tube, so that charge flows into this tubes side of the OT. If the other tube is "off" so that it is not amplifying, where does the charge go? Since the other tube is off, the charge can't go back through that tube to ground.

Remember all current comes from and returns to the power supply. Current leaving the +ve terminal of the reservoir capacitor must return to its -ve terminal. Current flows from the cap +ve through half the OT primary then through the 'on' tube to ground which is cap -ve.

[Helmholtz]

If the other tube, the one that is amplifying, pushes charge, then does that charge go through the 'off' tube to ground?

No, it can't. You are right, current (moving charge) always flows in loops, meaning that there always must be a return path, as charge can't get lost. In class B operation primary current can only flow through one tube and primary half, alternatingly. So the active tube works like a single ended amplifier with transformer load (but zero idle current), see: http://www.valvewizard.co.uk/se.html

The fact that there is only current in one primary half doesn't prevent the AC voltage to get mirrored in the other half ( by means of coupling), resulting in doubled voltage across total primary.

[Old Tele man]

There's Class-A push-pull, Class-AB push-pull, and Class-B push-pull...all push-pull output circuits require center-tapped OT.

Pardon the 'barn yard' humor, but push-pull operation is like 'two-handed' cow milking: as one hand pulls down (conducts), the other rests (in cutoff), and vice-versa, but there's only one cow source (B+ into OT centertap).

[Helmholtz]

I may help better understanding the PP stage by drawing it as 2 singled ended stages (each loaded with a primary half) side-by-side (or just rotate the typical PP drawing by 90° ccw.)

[soundmasterg]

I may help better understanding the PP stage by drawing it as 2 singled ended stages (each loaded with a primary half) side-by-side (or just rotate the typical PP drawing by 90° ccw.)

With Old Tele Man talking about cows, when I first looked at your statement I saw "or just rotate the typical PP drawing by 90° cow" and I was thinking about what you meant by a 90 degrees cow...haha.

Greg

[The Dude]

Would that be considered cow tipping?

[Old Tele man]

Today, it's more likely to be a bovine "ME TOO" movement infraction (wink,wink).

[R.G.]

"An output transformer designed for push-pull operation has a centre tap. The HT is applied to the centre tap and either end of the primary winding is connected to the anode of a power valve. In this way, current flows in opposite directions through the output transformer."

I get the idea that the center tap is used to feed HV DC to the plates of the output tubes. And, since its DC, it doesn't pass through the transformer, only the AC signals.

What Im not quite sure about is whether the center tap is actually required for a push pull amp. E.g. suppose there was no center tap, but the HV DC was fed through, say plate resistors back to the power supply. The DC voltage would travel to the plate, across the OT primary winding to the other tube plate. Wouldn't this effectively be the same or similar circuit?

The quote you mention is actually a pretty boiled down result. There's a lot hiding under the rock.
First: Our tubes only use electrons, not positrons. That is, tubes only work with negative charge carriers, and a positive plate. There are no "PNP" tubes. That means that tubes can only control current in one direction, through the plate to the cathode.

If we want push-pull, there are are only two ways: somehow floating an "active load" tube above a grounded-cathode tube and concocting some drive scheme to make the upper tube pull "up" (i.e, more positive) while the cathode-to-ground tube pulls down. There have been amps constructed this way, but they have some ... er, difficulties. The other way is the way we think is "normal" - two cathode-to-ground tubes whose plates only pull "down" toward ground, and an OT that makes this be a see-saw effect into a secondary.

Given that see-saw setup, there needs to be a CT to supply HV to the two so each can pull "down" from it. This setup lets you make each half of the OT be magnetically opposite to the other, so DC through BOTH of them effectively cancels in the core's magnetic field, and that leaves the core's whole flux capability usable for positive and negative swings, unlike SE, which can only use 1/4 of the core's flux swing.

Push pull says nothing about bias or class of operation. Bias can be class A, Class AB, Class B, or Class C, although there are implications about each of these in designing the OT.

[R.G.]

OK, I've recovered from that last spate of typing. More about OT CT and bias class.

The OT CT lets both tubes only pull "down" from a high voltage middle voltage. The OT combines the two pulls down into one AC signal and feeds this out the secondary. Class of bias is all about whether both tubes ever pull down at the same time, and by how much they overlap in conduction.

The phase inverter is there to feed equal-and-opposite signals to the output tubes in support of this push-pull see-saw operation. Each output tube gets fed a full signal, both positive and negative-going parts. The output tube's grid is biased somewhere between cutoff (where the bias turns it completely off for no-signal conditions) and Vgk=0, where the tube is conducting as completely as it can.

Imagine that the phase inverter makes a signal that has a peak to peak voltage equal to the cutoff voltage of the output tubes, and that the DC bias on both tubes is about half of the cutoff voltage. If that is true, each output tube is fed a signal that just barely runs it from cutoff to full-on (Vgk=0), so neither tube ever goes into cutoff. The OT combines these equal-and-opposite signals into a single signal in the secondary. This is Class A bias. Neither tube ever turns completely off.

The screaming advantage to push-pull Class A compared to single ended Class A is that you get about four times the audio output power from the same DC power supply, and that the OT can be about a fourth as much mass as for SE of the same output power. That is, it's still Class A, but costs less and weighs less for the same output power.

From there, you can hare off down the path of more audio power out for the same amount of DC power supply. By increasing the bias voltage to turn both power tubes progressively more "off" in no-signal cases, you can increase the amount of "pull" down on each tube at the expense of an increasing "off" period where only one tube is active but the other is in cutoff. Again, the signal level from the PI is adjusted so the "on"-est tube just hits Vgk=0 on peaks. So by increasing the negative bias on the output tubes, the no-signal power wasted in the output tubes is reduced, the amount of audio power out for a given DC power supply size goes up, and the OT gets smaller and the audio power you get out of the same tubes increases. The price you pay for this Class AB biasing is trickier biasing, and the introduction of crossover distortion because now there are periods of every full-size signal swing where only one output tube is on.

If you turn the bias even more, you can reach a point where both output tubes are just fractionally "on", nearly turned off, only infinitesimally on, you have reached Class B. In Class B, both tubes are sitting there, off, or nearly so, and with the slightest signal in the on direction, they pull on the OT. Each tube is on for notionally only one polarity of signal, and the OT combines these alternating half-cycles into a glitch-free composite in the secondary. Ideally at least. Crossover distortion gets a bit worse the closer to true Class B you set the amp, so the name of the game is to bias the two output tubes just a bit into Class AB. How big "just a bit" might be is the subject of never-ending debate.

If you go further, biasing both output tubes significantly off for no signal, then each tube is only turned on in blips. Obviously, turning on signal only on peaks sounds terrible. It's not meant to be listened to. Class C is largely reserved for RF applications where L-C filtering removes the ugly distortion products.

The progression from Class A to AB to B is a continuum of using the DC power supply ever more efficiently for conversion to audio power, wasting less power in the output tubes just sitting there with no signal, getting more audio power out of the same output tubes. The details of the OT change at each notional bias point because you need less wire and iron to deal with the overlapping conduction and current as you progress from A to AB to B, and you need tighter coupling between the part-primaries and secondaries to minimize the effects of crossover.

[bea]

The progression from Class A to AB to B is a continuum of using the DC power supply ever more efficiently for conversion to audio power, wasting less power in the output tubes just sitting there with no signal, getting more audio power out of the same output tubes. The details of the OT change at each notional bias point because you need less wire and iron to deal with the overlapping conduction and current as you progress from A to AB to B, and you need tighter coupling between the part-primaries and secondaries to minimize the effects of crossover.

On the price of more and more THD.

[R.G.]

To some degree, yes. Looked at another way, not so much. And not so much THD as crossover. The cost in distortion is not linear, and not always clearly stated.

To be sure, Class A is what you want for lower intrinsic distortion. You pay a price for that, especially in tube amps. It's vastly inefficient of both heat and the active devices' ability to make audio power.

Audio enthusiasts - and not just distortion-happy guitaristas - volted with their pocketbooks long ago for the higher power ability of Class AB.

It turns out that the continuum of higher power and higher distortion is not a linear relationship. You get a lot higher power out of the same tubes for very modest increases in distortion up to some critical point where crossover distortion starts increasing rapidly. Or, put another way, the ugly crossover distortion present in Class B decreases very rapidly with modest biasing into Class AB until it's largely unobjectionable.

That's the whole point of rebiasing Class AB amps - you get to much lower crossover distortion long before you start cooking everything with Class A excess heat.

It is a tradeoff - more power for more distortion; but it's not a linear tradeoff, or the most popular amps would all be Class A, not Class AB.

[SoulFetish]

Should we clarify that the need for a CT in a push pull output stage applies to the specific example of an output Transformer coupled output stage (which Mike was asking about). But a “push pull” output stage can be achieved a number of ways and doesn’t require an output Transformer or even Tubes for that matter. But however it is designed, it pretty much operates in some fashion based on the principles RG touched on.