Ian Iveson wrote:

"John Byrns" wrote

As near as I can see the Quad II uses an ordinary paraphase phase
inverter, not a "floating paraphase" phase inverter? I also can't
see any
large use of positive feedback in the Quad IIcircuit. I see a
very small
incidental amount of positive feedback that appears to be a side
effect of
the way the loop feedback is connected.


Hmm, maybe, just hovering perhaps. Certainly not "ordinary", because
the two valves are coupled in a host of different ways. Just been
looking it up. I was wrong about the "local". Here's the circuit in
case anyone doesn't know it:

http://www.triodeel.com/quad2.gif

I forget which end of the floating paraphase is supposed to float. I
think it's something of a misnomer anyway.

The circuit shows unbypassed 680 + 100 ohms common to the coupled
cathodes. Feedback from the secondary winding is applied to the
junction of those two resistors, so it appears as -ve to valve 1 and
+ve to valve 2.

Looking at the input of the power stage, the grids are connected via
680k to either side of a single 2k8. There is no central ground but
rather this 2k8 floats, with one end connected to valve 2 grid, and
the other to same cathode junction as the feedback. OK, so that's
not floating much, because there is a low resistance path to ground
via the feedback winding in parallel with the 100ohms, but the
floating paraphase is never really floating is it?

When the splitter is balanced, the signal either side of that 2k8
resistor must be equal and opposite. It may seem that the +ve fb to
V2 cathode is matched by the -ve fb to its grid.

But how do you explain how come 2k8 is such a small proportion of
680k? What is the gain of V2, and how does it compare to V1?
Remember that the grid signal for V2 is developed across just half
of that 2k8, since ac ground is half-way along it.

Looking for some words from my paltry library, I find JL Hood
"...gain of V2...rather more than twice than that which would have
been expected". Expected, that is, by someone not seeing the +ve fb,
he says. Irritatingly, he fails to mention the link between the
feedback and that 2k8 resistor. It requires a fair amount of maths
to work out the whole explanation, but the gain of V2 is
indisputably much higher than it could be without +ve fb...that fact
doesn't take too much calculation.

Oh...and the screens are floating with no reference at all to
ground. Shouldn't I get half a point for that?

cheers, Ian

The Quad has a paraphase PI with each EF86 with a gain of around 200.

The instataneous AV at various points around the circuit are as
follows:-

V3 anode, +160v, grid -40 = anode signal of V1,
V4 anode, -160v, grid +40 = anode signal of V2,
Feedback signal at top of R10 = +1.2v
Signal at top of 2.7 k = +1.04 approx.
Vg-k both V1&2 = approx 0.2v,
so input voltage g-g = 0.4 v,
so total input voltage = v at R10 + 0.4 = approx 1.6 v.

Only about 0.16v is across the 2.7k.

Global FB is applied to the second grid of the PI,
and fully to the bottom of the common Rk of R4, 680 ohms,
thus bootstrapping the input stage, or allowing
whati is the normal grounding point of the of whole of the input stage
including the
bias R for the output tubes to be in series with the AV input signal.
In effect, the FB is applied in common mode,
but the differential signal between input and fed back signal is all
that is amplified.
I think the cuicuit would work OK without the 2.7k, but
then you have a balanced short tail pair, which would be slightly
unbalanced, if driven more at one side than the other, so
to compensate, there is that 2.7k.

Once there, the 2.7 k helps the selfbalancing of these sorts of stages.

The pentode inputs are in class A, and the 2H screen currents cancel,
and so no need to reference the screens to the common cathode, although
this could be done.

A more modern "pure" way of doing this circuit is to have a CCS
tail for the EF86, and take the FB
straight to the second input grid of V2, and have the bias R for the
output tubes taken to ground.

I think the modern way would be better, and easy to implement,
with a single fet CCS, but in 1953, there were no fets.
And when there were fets, people were throwing out tubes.

Patrick Turner.