On May 31, 1:06*pm, John Byrns wrote:
In article ,
*Patrick Turner wrote:
On May 29, 3:47*am, John Byrns wrote:
In article
,
*Patrick Turner wrote:
Well, I can't explain it any better than I have. If circuit gain is
increased by feeding back a fraction of an output signal of one stage
to an earlier stage and the gain is increased, THD increased,
bandwidth reduced, then I'd say PFB exists. We see feathers, a ducks
bill, wings and webbed feet, a ducks tail that wags and we say there's
a duck.
Unfortunately this is a completely incorrect description of how a generic
paraphase inverter operates, so the duck analogy doesn't apply.
Oh well, I can agree to dissagree
The concertina "throws away" most of its gain so that +19Vin to a 6SN7
produces -17Va and +17Vk, with the +2Vgk needed to make the total
34Vak output, so open loop gain = 17, but closed loop gain is 34 / 19
= 1.79, and this has beautiful low THD because the 19dB current NFB
present.
So the concertina requires 1.12 Volts in for the specified 34V grid to grid
output drive vs. 1.0 Volts in for the paraphase and 2.0 Volts in for the
LTP. *
1.12 Volts is close enough to 1.0 Volts for Government work, at least it is
a
lot closer to 1.0 Volts than it is to 2.0 Volts.
The paraphase tends to add gain "artificially" for a given number of
tubes, moreso than merely cascading them.
"artificial" gain is the best kind, it can be made so pure that its sound is
simply sublime.
If only it was true. Paraphase and PFB usually have higher than
possible THD without the paraphase PFB. The sound of such stages
cannot be assumed to be sublime, and wherever higher THD is allowed to
breed unecessarily the sound drifts to mud.
And in the case of Quad-II, the KT66 grids are biased via 680k Rgs,
and one has a 2k7 at its bottom from which the signal to V2 pentode
grid is wrought.
Now that 2k7 is just what Walker chose, but it could be a range of
values; if larger, there's more PFB and more gain and weird things
happen when you increase that
2k7 too much because there is a definite limit to the amount of PFB
which can be applied without terrible instability.
I don't know if 2k7 is the correct value for this resistor, I assume that Quad
choose correctly here, there is only one correct value for this resistor where
the circuit will function properly, other values are not an option! *
Don't be so sure about this. Try building an input stage identical to
Quad-II in some amp you have laying around. You don't need pentodes; a
single 12AX7 will do, and where the Quad has 680k plus 2k7 try putting
a 5k pot to replace the 2k7, then add some NFB and make some
adjustments with the pot. After awhile the penny drops about the PFB,
too much is hell, just enough is all that's needed - about 6dB.
You'll soon find that probably optimal R value is where there is
minimum signal voltage across the 680 ohm common Rk to the two EF86,
or 1/2 12AX7, so the trick of the circuit is that the EF86 tend to
have a fake CCS connected to common cathodes. One may analyse the
effects of a distortion signal fed back along the signal path and what
happens to it in terms of gain. But the PFB increases the amount of
NFB applied beyond what it appears to be. There is more to Quad-II
input drive amp than 90% of people may realise.
Changing
the value of this resistor will change the amount of PFB as you say, thereby
changing the gain of V1. *The problem with changing the value of this resistor
away from the correct value, to some other value, is that in addition to
changing the amount of PFB, it also unbalances the phase inverter so that the
two output tubes don't receive the same amount of drive voltage.
Very few Quad-II amps have well balanced drive. But I recall the 2k7
was not critical for balance. I suggest you try examining a Quad-II
circuit.
What I can say is that a pair of EF86 set up as a TRUE LTP with
existing load values, but with large R value for common Rk taken to a
rail at -400Vdc, and NFB applied to V2 EF86 grid, then the drive amp
produces 1/2 the THD as the original Quad-II circuit. But sensitivity
halves and twice the grid 1 input signal is needed. Quad couldn't push
pentode gain higher so they needed the trick thay have used. Quad
didn't think of a way to bootstrap the RL of each pentode anode to
increase gain. That would have needed a twin triode for a pair of µ
followers in a balanced pair but the balance of such pairs is poor
because so little signal Ia exists and balance in LTPs depends of R
load equality.
The better way is to use an SET input triode followed by low µ twin
triode for LTP, 12AU7, 6CG7 is ideal. or have LTP input with 12AU7 and
CCS tail followed by balanced amp with long common R tail to -120Vdc;
methinks this is the king of input/driver stages for PP.
Next time you play around with a paraphase stage, try tinkering with R
diver values and drive from V1 to V2. Interesting stuff happens, with
not much of it of real value.
What exactly are "R diver values"? *There is only one correct value for the
"drive from V1 to V2", you may find other amounts of drive interesting, I'm not
sure why though?
Divider values. Nobody dives anywhere. 680k and 2k7 give Quad's wanted
signal for V2 grid, and you can't guess the voltages.
Patrick Turner.
--
Regards,
John Byrns
Surf my web pages at, *http://fmamradios.com/