Thanks Wa2ise

This is the concept I unsuccessfully proposed in my earlier post. I see the advantage of this circuit as providing as close to identical amplitudes independent of component and valve characteristics. My "unity gain inverter" reference was a flashback to using op-amps to do just that.
My grey cells aren't what they used to be!

Harold

I can appreciate seeing V4A as a unity gain inverting op-amp: the voodoo, for me, is that if the output drives are equal and opposite, the summation of them in the symmetrical T would be zero signal - or worse: what if the resistors' tolerances are such that 4A's output signal dominates back at its grid, trumping the 'in-phase' signal from V5?

Second question: What is the advantage of this topology versus the more common version (Magnavox etc.) where R34b is simply grounded?
Is it that this circuit is 'self balancing' and/or lower distortion?

As a layman, here is another "problem" I see with the circuit: If drive to V4B is 'x' amplitude, then signal at grid of V4A would be 0.5x due to divider R34a/R34c. Then since V4A is constrained to unity gain as a hardwired op-amp, its drive to V5b will also be only 0.5x (an undesirable outcome.)

I didn't make a good job explaining my thoughts in my earlier post, so I would like to add some more thoughts to Wa2ise's post and hope this helps. I also feel I shouldn't have referred to V4A as "unity gain", this was me seeing it as similar to an operational amplifier. Sorry.

The condition of equal amplitude at both anodes cancelling out to zero at the junction of R34A and R34B cannot exist, as no signal at V4A grid means no signal at its anode. What is happening is that a very small signal is present at V4A grid, the "error" signal Wa2ise referred to and I think the level of this is determined by the gain of V4A amplifying it to nearly the same level as at V5As anode when the circuit is in balance.

The different cathode resistors in both halves is probably significant, however I cannot offer a precise reason for this.

Harold

Like Wa2ise said, it's probably best to think of it as an op-amp. The error of my thinking in post 18 is not considering the high open-loop-gain of V4A available for it to match (and invert) input from V5A. The high cathode resistor of V4A might be to promote stability/degeneration as they say using local feedback within a global feedback arrangement can lead to problems. I note on voltage table that both cathodes end up at 1v, so would have the same biasing for linearity.

Following this circuit's explanation in Wireless World, it states: http://www.keith-snook.info/wireless-world-magazine/...phase splitter.pdf

At the centre, where the grid of
the valve is connected, we have
the fulcrum of the see-saw, at
which the up-and-down movement
is nil. In other words,
the valve is delivering its output
without any grid input, which of
course is absurd.
But there is no need to be too
downcast about this fiasco...

Called a "see-saw" because of the envelope shape of dynamic voltages across the top of the 'T'.
I have to admit, I cannot fully follow his explanation!

To perhaps explain the operation of the phase splitter discussed here, this is an extract from G A Briggs's (Wharfedale Wireless Works) book, "Amplifiers".
He describes this type of splitter as the "best all-round type" with a very large degree of self balancing due to the negative feedback employed. It also provides phase unbalance that can be held to 1%.
Harold

Photo uploaded to Post 22. I apologise for the delay.

‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
A valve a day keeps the transistor away...

The whole subject of phase splitters in valve amps is a fraught one. The Anode Follower has issues when you start putting respectable amounts of feedback around the amp. I am currently modelling PSs in LTSpice looking for the best solution for a high-end valve amp design I'm working on - yes, very retro for me! So far, the best all-round performer is the classic long-tail pair, using a large cathode resistor and a 240v negative rail. But my heaters are running on clean DC so I don't have the heater- cathode problem.

I definitely will try the Kriesler circuit tomorrow and tell you what I find. Would have done so today but VR was down and I didn't have the info!

"I definitely will try the Kriesler circuit.."

I want to build & test the Kriesler phase splitter circuit also.

QUOTE: V4A tube's grid is similar to the "virtual ground" on an op-amp

As a self taught layman, I struggle with some engineer-level concepts like 'virtual ground', also I would view V4A's grid as a high impedance point (not very ground/power-rail like.)

But I do now see a way to resolve Kieth Snook's dilemma:
("the valve is delivering its output without any grid input, which of course is absurd..") - quoted in post #21

This epiphany was triggered by looking at Brigg's drawing (uploaded since I last posted.) In this drawing he has the inverter placed downstream - sequential - so due to the transition times through V4's amplification then feedback, a signal appears on V4's output before being "cancelled" by nulling feedforward/feedback symmetrical network, solving Snook's "Chicken-Egg dichotomy".

So, all's fine with the world. But what about in a perfect world where the speed of conduction/amplifiers was instantaneous: there would be instant cancellation - no signal at input, back to the dilemma.

I was a bit concerned by transit times through the stage.

I did get around to simulating the Kriesler circuit. Not impressed, I couldn't get more than 6 or 7 watts RMS out of it at the onset of clipping, which was asymmetrical.

Since then I tried a few more ideas, because I wanted better than 15 watts from a pair of 6BQ5s.

Simulations showed the big problems happened as soon as the 6BQ5s started to draw grid current on peaks, i.e. as soon as I headed into Class B territory. The bias would be pushed back and crossover distortion appeared.

So eventually after trying a lot of things, I came up with the circuit in the PDF I'll send to Brad. I'll send the LTSpice file too - maybe someone can pick me up on a mistake. Tthe simulation performance is impressive, maybe a bit too impressive, but I am allowing for realistic transformer losses.

It delivers an easy 20 watts RMS to the load at the onset of clipping. But better still, its clipping behaviour is benign. The way it handles gross overdrive, almost AGC-like, I reckon it would make a great little guitar amp.

Cathode Follower (PDF)
Cathode Follower (ASC)

QUOTE: Since then I tried a few more ideas, because I wanted better than 15 watts from a pair of 6BQ5s.

This is very intriguing; looking forward to seeing these experimental designs!

I like the cathode follower buffers to prevent loading effects on long-tail-pair splitter, which itself has no delay for inverted signal!

Yes that arrangement simulated better than anything else I tried. It was particularly stable with a large amount of negative feedback and square wave input.
I was particularly interested in what happened as the amp approached clipping. The cathode followers avoid the crossover distortion that happens when the outputs are driven into grid current. They make a huge difference to the available output power.