I have just posted 3 schematics at abse showing various input and LTP
driver arrangements with comments.

Fig 1 is for a basic 6SN7 input tube with direct coupled LTP,
with a common Rk.

I have calculated the loads and working voltages as best i could,
but expect some variation in practice.
In actual use, this basic driver circuit must have R10 varied until
balance
of drive voltages is obtained.
Once set, it won't change much, and although there is imperfect
cancellation of 2H in such an LTP, it works and sounds ok.
Output voltage is restricted by the Ea for the LTP, and
Ea = 260v - 106v = 154v only, and a a large anode v swing is
impossible.


Fig 2 shows the similar input stage CR coupled to the LTP which has
got a CCS instead of a common Rk, and this allows automatic voltage
balance
since RLs can be exactly equal and Ea = 200v - 30v = 170v, which is
healthier
and allows a higher v swing and lower thd at the same low v swing as in
fig1.

Fig 2 can also be stretched to make it give much more, and R13,R14
can be increased, 550v used instead of 400v for the supply.
In one amp I have the supply for the LTP is at +600v, with paralleled
6SN7 for each half of the LTP, and max possible drive voltage is above
100vrms per anode
The same 300v can still be used for the V1 supply, by increasing R6.

The use of a paralleled 6SN7 for *each half* of the LTP is a good idea,
since it slightly increases v swing, bandwidth, and lowers thd.
I normally like to play with the value of R11, which sets the constant
cathode current,
and thus adjusts the value of Ea, and this is done with a big signal
present
and the Ik is adjusted for symetrical clipping ( with no output tubes
present ).
I like the LTP to be able to produce twice the required drive voltage
needed
to make the output stage clip.

Fig 3 with CT choke and R loading for the EL84 triodes of the LTP
has a response graph attatched showing
the reponse at each anode output...curve A.

Curve B shows the voltage at the winding ends of the choke, and this
shows
the variation of voltage across the choke compared to the anode output.

In this test I used the primary winding of horrible old OPT
suited for a pair of 6BM8, and which had a quite small amount of
primary inductance.

If the CT choke has 10 times the L shown in this case the curve for the
voltages across the choke
are much closer to the curve for the anodes, which would indicate
that the choke's impedance would be a lot higher down to a lower F.

Note the resonance dip and peak above 100 kHz, caused by
stray leakage and capacitances, and note how the 4.7k R
prevents the effects of such impedances from seriously affecting the
response
at the EL84 anodes.

Of course, all of what I have shown is a lot more complex in terms of
component count compared to what Lynn Olson has proposed
in his PP amps, which he prefers to SET by the way.

But I don't have to try so hard to get the bandwidth in my amps as I
would
have to with several cascaded transformer coupled stages.

The iron distortions which are not mentioned by Mr Olson are
not a problem in my designs.

Mr Olson and several others and including
Mr Sakuma of Japan with his restaurant full of tube amps, mainly SET
types,
all try to reduce thd as much as possible with wise loading of their
tubes.
Its not as if the measurements don't matter to these dudes, they surely
do,
and they all mention linearity and distortions often, as I do.
These other dudes my well wax lyrical about the sound, or complain about
the sound of a
stray cap here or there, and they surely hate capacitors.
I don't, and find capacitors are the least blameable reactive coupling
element,
since they are purer than almost any inductor, which users damned iron,
and which has odd resistance to windings and stray C and damned
resonances,
not to mention the appalling tendency to create crossover like harmonic
distortions.
But at least the iron has its better side, that of vastly increasing the
loads seen by the anode
circuits of voltage stages, and this reduces the triode thd,
and we only need to have Ra nice and low to naturally
reduce the iron thd to negligible levels.

I believe that if a triode amp meaasures
well, and has good spectral content, ie, freedom
from much high numbered odd harmonic products, it should sound OK,
and in fact its hard to make a bad sounding triode amp
if it has plenty of ceiling and measures OK.

All PP designs produce some 3H.

With the LTP, or in the case of most class A PP triode output stages
the 3H isn't just due to the 3H which is in the SE triode transfer curve

being uncancelled by the balancing in the LTP by the PP action.
The gm of the two halves of any LTP vary each side of the zero point.
As one tube turns on with more Ia, the gm tends to increase,
and the other triode which is turning off tends to have its gm reduce,
and
the change in gm isn't equal, so as the output voltage is increased the
net
differential voltage gain of the LTP ( or output stage ) tends to
decline a bit,
resulting in the 3H we see gradually increasing from low voltages,
and rapidly increasing near clipping.

But set up right, the LTP can be engineered to produce
astonishingly low thd figures at say 75 vrms with a couple of EL84 in
triode
with a choke as in Fig3.
Unlike an output stage which is forced to endure a large current swing
and voltage swing to make some power, the LTP should be engineered
to have to make as little power as possible, which means minimizing
current change in the triodes concerned, even with a high Vswing.
Thus the thd of the LTP can be kept to much less than the thd of the
output stage connected, assuming we have made the output stage
in triode, UL or CFB, and mainly class A, so its thd at clipping is only
1%, ( no global FB ).
Often it may be so easy to have 2% thd just because we have not
got a low thd LTP driver.

When I built my first 8585 with a Quad of 6550 to each channel
with 12.5%CFB windings on the OPT, I was able to get 0.7% thd at 40
watts class A without
any global NFB, and this was the straight measurement of the whole amp
without deducting the LTP and /or input stage thd.

A lot of the measurements quoted for triode or other PP output
stages may not be taking the thd of the LTP into account.

Since it is impossible to easily and naturally engineer any LTP
with a rise in 3H which is oppositely phased and of a similar rate of
increase
for most output RLs to the peak flattening 3H in nearly
all output stages, it is left to the designer to try to engineer the
most linear
LTP he can, since there isn't any natural other better way to minimise
thd.
Voltage distortion cancelation is impractical in PP amps, a somewhat
different
situation to SE amps where the much larger quantities of 2H
in drivers and outputs can actually cancel to good effect, providing
we don't deliberately set up a driver stage to produce lots of 2H
just to cancel the large 2H in the output, by means of adversly loading
the driver.
Its better to linearize the output stage to get its 2H minimized so that

the slight 2H produced in an SE driver will usefully cancel but
not produce many second order harmonic and imd products.

The simplest way, but by no means the cheapest way to linearize
amplifiers is to have a high power ceiling, so that a small amount
of class A is all we ever will ask of our amps, even if its a class AB
amp.


Patrick Turner.