[Ian Bell[_2_]]

I am thinking of paralleling two triodes in a CF to lower output
impedance and increase drive capability. Given that I uses a twin triode
tube (6SN7/6CG7 et al) how do I do this? I guess connecting the plates
together is OK but what about the grid and cathode. There are bound to
be some variations in the characteristics of the two triodes. Can I
safely ignore these or what do I do to mitigate them?

Cheers

Ian

[Ian Iveson]

Ian Bell:

I am thinking of paralleling two triodes in a CF to lower
output impedance and increase drive capability.

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Given that I uses a twin triode tube (6SN7/6CG7 et al) how
do I do this? I guess connecting the plates together is OK
but what about the grid and cathode. There are bound to be
some variations in the characteristics of the two triodes.
Can I safely ignore these or what do I do to mitigate
them?

Common cathode seems inevitable to me...they must both go
directly to the output, and that means they must be joined
directly together unless you use two coupling caps, which
could be the start of a nightmare.

For DC current balance, you could either adjust relative Va
with a bypassed resistor, or relative Vg with a potential
divider or whatever, depending on your biasing scheme. Can't
immediately see why DC balance is necessary in itself. If
there remains some AC imbalance, you might unbypass part of
the anode resistor, or attenuate the signal to one valve.

It seems likely to me that a bypassed anode resistor on the
keener of the two halves would achieve both AC and DC
balance. If you use the grid you will need to provide for
both signal and bias voltage attenuation.

Different heater currents would be and interesting
alternative maybe.

I don't think I've seen anyone bother with any of these
things.

Patrick's got lots of experience in connecting small triode
halves in parallel IIRC. Seems inelegant to me.

Ian

[Staring at old audio chassis as a hobby :)]

On Jun 1, 2:16�pm, Ian Bell wrote:
I am thinking of paralleling two triodes in a CF to lower output
impedance and increase drive capability. Given that I uses a twin triode
tube (6SN7/6CG7 et al) how do I do this? I guess connecting the plates
together is OK but what about the grid and cathode. There are bound to
be some variations in the characteristics of the two triodes. Can I
safely ignore these or what do I do to mitigate them?

Cheers

Ian

Hi RATs!

You could strap on a Blumlein Garter circuit

Happy Ears!

Al

[Patrick Turner]

Ian Bell wrote:

I am thinking of paralleling two triodes in a CF to lower output
impedance and increase drive capability. Given that I uses a twin triode
tube (6SN7/6CG7 et al) how do I do this? I guess connecting the plates
together is OK but what about the grid and cathode. There are bound to
be some variations in the characteristics of the two triodes. Can I
safely ignore these or what do I do to mitigate them?

Cheers

Ian

The anodes and cathodes can be connected together without worry. In some
CF or other parallel apps you can get spurious RF oscillations with high
gm triodes like 12AT7 or 6DJ8, so the use of 3k3 series R between the
source and each grid should be done to avoid the parasitic oscillations.

If the two halves vary a bit in characteristics it simply does not
matter. I've never found huge variations.

Patrick Turner.

[Ian Iveson]

Missed the word "safely" from OP. Yes it is safe to ignore
whatever the effects of imbalance might be. AFAIK Patrick's
amps are safe, ergo...

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Not as clear as I intended. I just mean that, if low
distortion, large amplitude, and low output impedance are
required, you need to be careful not to double-count (or
triple-count) the effect of feedback. For example, at the
edges of the bandwidth, distortion is relatively high, and
signal capacity relatively low, because the feedback is
fully occupied with bandwidth extension, er, if you see what
I mean.

A problem of feedback in general, perhaps sometimes
forgotten wrt the CF.

Ian

[Ian Bell[_2_]]

Ian Iveson wrote:
Ian Bell:

I am thinking of paralleling two triodes in a CF to lower
output impedance and increase drive capability.

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Given that I uses a twin triode tube (6SN7/6CG7 et al) how
do I do this? I guess connecting the plates together is OK
but what about the grid and cathode. There are bound to be
some variations in the characteristics of the two triodes.
Can I safely ignore these or what do I do to mitigate
them?

Common cathode seems inevitable to me...they must both go
directly to the output, and that means they must be joined
directly together unless you use two coupling caps, which
could be the start of a nightmare.

For DC current balance, you could either adjust relative Va
with a bypassed resistor, or relative Vg with a potential
divider or whatever, depending on your biasing scheme. Can't
immediately see why DC balance is necessary in itself. If
there remains some AC imbalance, you might unbypass part of
the anode resistor, or attenuate the signal to one valve.

It seems likely to me that a bypassed anode resistor on the
keener of the two halves would achieve both AC and DC
balance. If you use the grid you will need to provide for
both signal and bias voltage attenuation.

Different heater currents would be and interesting
alternative maybe.

I don't think I've seen anyone bother with any of these
things.

Patrick's got lots of experience in connecting small triode
halves in parallel IIRC. Seems inelegant to me.

Ian



I have seen several examples where 12AU7s for example are just wired
pint to pin but in another example, a Pultec mic pre that have a 470R in
one anode lead and a 1K between the two grids.

Cheers

Ian

[Ian Bell[_2_]]

Patrick Turner wrote:

Ian Bell wrote:
I am thinking of paralleling two triodes in a CF to lower output
impedance and increase drive capability. Given that I uses a twin triode
tube (6SN7/6CG7 et al) how do I do this? I guess connecting the plates
together is OK but what about the grid and cathode. There are bound to
be some variations in the characteristics of the two triodes. Can I
safely ignore these or what do I do to mitigate them?

Cheers

Ian

The anodes and cathodes can be connected together without worry. In some
CF or other parallel apps you can get spurious RF oscillations with high
gm triodes like 12AT7 or 6DJ8, so the use of 3k3 series R between the
source and each grid should be done to avoid the parasitic oscillations.

If the two halves vary a bit in characteristics it simply does not
matter. I've never found huge variations.

Patrick Turner.

Thanks Patrick, a good pragmatic answer.

Cheers

Ian

[Ian Bell[_2_]]

Ian Iveson wrote:
Missed the word "safely" from OP. Yes it is safe to ignore
whatever the effects of imbalance might be. AFAIK Patrick's
amps are safe, ergo...

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Not as clear as I intended. I just mean that, if low
distortion, large amplitude, and low output impedance are
required, you need to be careful not to double-count (or
triple-count) the effect of feedback. For example, at the
edges of the bandwidth, distortion is relatively high, and
signal capacity relatively low, because the feedback is
fully occupied with bandwidth extension, er, if you see what
I mean.


If you are talking about global feedback of which the CF is a part then
I agree. In this application there is no global NFB. The only NFB is
that within the CF itself.

Cheers

Ian

A problem of feedback in general, perhaps sometimes
forgotten wrt the CF.

Ian

[Patrick Turner]

Ian Iveson wrote:

Missed the word "safely" from OP. Yes it is safe to ignore
whatever the effects of imbalance might be. AFAIK Patrick's
amps are safe, ergo...

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Not as clear as I intended. I just mean that, if low
distortion, large amplitude, and low output impedance are
required, you need to be careful not to double-count (or
triple-count) the effect of feedback. For example, at the
edges of the bandwidth, distortion is relatively high, and
signal capacity relatively low, because the feedback is
fully occupied with bandwidth extension, er, if you see what
I mean.

A problem of feedback in general, perhaps sometimes
forgotten wrt the CF.

There are golden rules to follow when considering a cathode follower.

The main one is to always try to load the follower with a similar load
to that used in a normal common cathode gain stage, ie, Rl should be
above 4 Ra for best results.

This means the internal or open loop gain is high, and closed loop gain
reduction to just under 1.0 will result in the desired reduction of
THD/IMD, and transparency of sonics.

There will be asymetrical sharp clipping if a cap coupled load is too
low value, or less than the dc carrying RL.

Never expect a CF to work into a reactive L or C value which would
present a load to cause distortion.
CF don't mind working into L or C, but the undistorted voltage output is
reduced by the inability to maintain a flat voltage output and ever
increasing current in the L or C at extreme ends of BW.
eg, 1,000pF is a difficult load at 1Mhz, and only 159 ohms, so if the CF
Ia = 5mA, expect only 0.5Vrms of clean signal output.

But say a CF has Rout = 600 ohms. At low levels well away from serious
slew distortions below 1Mhz, the response would sag -3dB where ZC of
1,000 pF = 600 ohms, which is at 265kHz.

In fact most garden variety CF with Ia of 4mA, say halves of 6DJ8,
12AU7, 6SN7, 12AT7 etc, will give impressive audio BW at say 1Vrms
required to drive a power amp even if loaded with 1,000 pF. Tubes that
ain't so good with reactive loads are 12AX7, 6SL7, EF86, 12AY7, but they
are just fine when you have to work into a mainly R load with little
stray C or L.

As much as I have been attracted to a White CF, I have rarely ever used
one. I find it is simpler to parallel the same two triodes and use a CCS
from cathode to 0V or a negative supply, then cap couple the SE output
to the load.

In most instances, the results are fine.

Horses for courses.

Patrick Turner.




o

Ian

[Patrick Turner]

Ian Bell wrote:

Ian Iveson wrote:
Missed the word "safely" from OP. Yes it is safe to ignore
whatever the effects of imbalance might be. AFAIK Patrick's
amps are safe, ergo...

If these are two different objectives, watch out you don't
double-count the effectiveness of the feedback.

Not as clear as I intended. I just mean that, if low
distortion, large amplitude, and low output impedance are
required, you need to be careful not to double-count (or
triple-count) the effect of feedback. For example, at the
edges of the bandwidth, distortion is relatively high, and
signal capacity relatively low, because the feedback is
fully occupied with bandwidth extension, er, if you see what
I mean.


If you are talking about global feedback of which the CF is a part then
I agree. In this application there is no global NFB. The only NFB is
that within the CF itself.

Cheers

Ian

In a KORA DAC I had to fix recently, there was a 1/2 6922 common cathode
gain triode with a the other 1/2 used as a direct coupled CF.
Then they had shunt FB from CF to input to reduces the OLG to 3 from OLG
of about 28. THD/IMD and Rout is all quite blameless and extremely low.
The CF is included in the shunt FB loop. Now many ppl would say its
wrong to do this because the shunt FB is applied as global NFB around
two tube stages. And of course the R network is often 100k + 330k where
a gain of 2 to 3 is wanted, so Rin for the shunt FB network is low
compared to say a normal stage with 470k grid resistance.

And high resistance/impedance sources increase the FB loop because they
are in series with the FB network, so the answer is to use a CF input,
so that could make 3 triodes. Its over design unless you really want low
THD and with some gain.

Shunt NFB derived from a 2 R network is not used as much as it could be
and series voltage NFB is the one mostly used and because its effective
if the BW is wide as it raises input impedance and reduces input C. In a
Williamson, the CPI has a lot of series current NFB to give total gain
of nearly 2, ie, for 1V at g1, there is about 0.9V at anode and 0.9V at
cathode. So a tube with gain of 16 has had its gain reduced to 1.8,
which is 20dB of NFB.
Then the W had 20dB of GNFB because it could rely on the low phase shift
of its OPT and the tube line up. The GNFB aound a stage which includes
an internal stage with 20dB didn't stop the W from having wide BW and
good stability **if** you follow all the rules W laid down about OPT
design. The reputation of NFB suffered badly at the hands of the
fuctards and bean counters who laughed at Williamson and refused to
allow amplifiers with adequate and thus expensive OPTs to ever see the
light of day.

Many designs are overly complex because designers were chasing zeros.
ARC reckoned they could sell more preamps if THD was 0.001% instead of
only 0.015%, so they constructed nearly every gain stage with 3 triodes
and some mosfets/j-fets and had oddles of GNFB. The buyers listened in
vain their expensive purchases, silly really, when they could have tried
listening to the music.

Patrick Turner.





A problem of feedback in general, perhaps sometimes
forgotten wrt the CF.

Ian