[Patrick Turner]

The 6550 has to be one of the most popular tubes used for hi-fi and
guitar amps.

I am not so interested in using such a tube in guitar amps where low
fidelity, ie, high distortion levels are welcomed because it enhances
the musicians efforts. I am concerned about extracting hi-fi from
6550.

In hi-fi amps, McIntosh use the 6550 in nearly class B conditions
where the tube has low bias current and the loading for a tube in a PP
pair of tubes while in class B working is only 1k. However, McIntosh
have the OPT primary set up so that half the Va-k of each tube appears
at the anode and at the cathode, so there is heavy local NFB and ß =
0.5.

With the addition of a lot of global NFB as well as the local FB in
the OP stage, the appalling non linearity of a beam tetrode in class B
with a low load value can be "cured" and the measured outcome at all
levels right up to clipping are acceptable, and rather good compared
to many other standard designs using UL type config with GNFB only.

My preference is for large amounts of class A working of output tubes
so that with PP circuits I would have a pair of 6550 coupled to a
speaker load with a an OPT with an 8k a-a load where Ea was say +500V.

I also don't like having any OP tubes idling anywhere close to the
maximim Pda rating of the tube. So I will try to have Pda 28W for
6550 even though the rating is 42W. This means that where Pda was say
25W, and Ea was say 450V, then Ia will be just above 50mAdc.

The data says 6550 will have Ra = about 18k, Gm about 11mA/V, and thus
µ = 198.
but gm and Ra are very variable and related to Ia, so that the only
time gm will be 11mA/V and Ra less than 18k is when Ia 120mA, and
there will never be any amp you or I will build which will have such a
high idle Ia dc.

Typically, all of us will bias at about 50mA or just above for most
projects.

So typically we will find a 6550 tube biased at 55mA and Ea = +450V to
have control grid gm = 5.5 mA/V, and gm of screen to be about 0.83mA/V
where Eg2 = +270V, ane the Ra will be 32k, way above the data figures
say it will be.

For pure class A operation, the load for maximum pure class A PO is
approximately 0.9 x Ea / Ia dc.
So for 55mA and 450V, RL will be around 6k to 7k, and for a PP pair a
class A load would be 12k to 14k.

Efficiency is up to 45% in such conditions. Linearity is not brilliant
at near clipping, but its the first few watts everyone needs to
consider.

During the extremes of each wave cycle the gm and Ra vary
considerably. BUT, where the RL is fairly high, the Ia change is less
tha the Va change, so the gm and Ra change much less during a wave
cycle than were one to have Ea = 350V, and RL a lot lower, and Ia a
lot higher. This becomes especially true when you try to make an SE
amp with 6550.

So what is a simple equivalant model for a 6550?
I suggest you first have to set one up in a test circuit to confirm
your samples give some data you need to know to calculate the gains
and tube properties in a given circuit. The data you seek to
findoutabout can't be found in the data sheets or books.

This means you have a B+ applied from a low Z supply, and have a 100
ohm anode load, and a regulated Eg2.
Apply 1Vrms at 1kHz to the control grid, and measure the current in
the 100 ohms.
The gm is thus measured.
Then with the grid shunted to 0V, apply a 1Vrms signal voltage to the
screen from a low Z signal source, and measure the current in the 100
ohm anode load.

I used Ea = 420V ( Ek was 35V ) and Eg2 = 270V, and Ia = 55mA, and got
grid1 gm = 5.5mA/V, and screen gm = 0.833 mA/V
Ra was approximately 32,000 ohms.

The model for the tube is as follows :-

Ra is a 32k resistance between anode and cathode.
Grid 1 is a terminal controlling current between anode and cathode
which is a CURRENT SOURCE of 5.5mA/V. Also between anode and cathode
there is a second controlled current source controlled by the voltage
at screen grid2, and current produced a to k is 0.833mA/V.
The impedance looking into the anode sides of each current source is
an infinite amount.
So each current source can act to control current change at the anode
across loads or across the RA without interfering with the two gm
figures, and the anode current and Ra current is the sum of the two
current sources.
In pure beam tetrode operation, the voltage at the screen is kept
constant, so the screen exerts no change in Ia. So when you apply
1Vrms signal to grid1, you get a 5.5mA Ia current change applied to
Ra, and to the anode load which is in parallel with Ra. If there is no
RL, the 5.5 mA of Ia change can be said to operate within our
equivalent mathematical model, and you get a an anode signal = 5.5mA x
32k which is about 176, and 176 is the amplification factor of the
this tube in tetrode mode.
So the formula for all tube gain holds true, A = µ x RL / ( RL + Ra ).

But for all tubes, µ = gm x Ra, so we could replace µ in the last eqtn
with Gm x Ra.

How does one measure Ra? Well in the test set up above, we use a 30H
choke to supply dc to the 6550 under test, and at 1kHz the choke's
reactance is so high the reactance loading can be neglected. We can
then apply 1V to grid 1, and we will find that the anode signal will
be perhaps 176Vrms, ie, gain = µ = gm x Ra. But with 176 Vrms at the
anode the distortion will be high and it will affect our measurements
so we lower the grid input to leave about say 50Vrms max at the anode.
Then we connect a 7k load across the choke which loads the the anode.
You will observe the anode voltage falls a lot and you measure this
voltage and you calculate the current in the load.

To find out the Ra, Ra = ( Anode voltage with no load - anode voltage
with load ) / load current.

One can also use two different load values, and the Ra = change in
anode voltage / change in load current.

So what happens when triode connection is used? Well, you need to know
what the triode µ is.
So with g2 connected to the anode, and bias Ia adjusted for 55mA, you
measure the voltage gain without any load except the choke, and you
should find gain = µ = 6.7 or therabouts.
Ra with triode connection can also be measured, and will be found to
be about 1.2k

So imagine the model we have, and with triode connection, and with no
load.

We have +1V at the grid1, and there is -5.5mA produced by current
source1. So a -Va signal of -6.7V is produced because we have triode
connection. This whole -signal is applied to the screen controlled
current source, and there is a +current produced at the anode = +6.7 x
0.833 mA/V.
The two currents sum to produce the current across the Ra = 6.7V /
32k.
If there is a load also present, you can calculate the load voltage
and gain using the model.
if the tube operates in UL mode, or CFB mode, then the voltage applied
to the screen is a fraction of the anode voltage.

I leave you all to ponder this, and once understood, its somewhat easy
to draw a graph plotting Ra and µ for all values of UL tapping %.

Hardly anyone would bother to go to all this bother, they would just
hook things up and hope for the best with educated guesses if they
have a good gut feeling about the anode curves for any multi grid tube
which one wants to use.
The only trouble with the model is that the screen needs some input
current because of the screen current; its not like grid 1 where the
input Z is extremely high. So just to complicate the model which so
far was fairly simple in the world of electronic models, we need to
add a resistance between cathode and the screen grid control point of
its controlled current source.
This R could be found by measurement with a series R with the screen
while a small screen signal is applied. This R will form a loading on
the anode circuit if you have triode connection or UL partial triode
connection. But with triode connection, or UL connection the resulting
Ra which is far lower than for pure tetrode operation means that the
loading by screen drive input current is not going to greatly change
the calculated results and graphs we may draw about the tube's
behaviour.

Just using only triode connected 6550 or other multigrids does make
life very easy. Once we add in the effects of the screen upon Ia flow
in the case of UL then matters can rapidly become quite complex, so
any audio amateur can be forgiven for saying "Damn all models, let me
build, measure, and tweak it to perfection!"

One tip though is certain. For best hi-fi, the anode load value has
usually to be at least twice what many comercial makers use. The
cretins of commerce design for POWER, and don't care much about hi-fi.
So a typical OPT for a pair of 6550 or EL34 would be 10k:5ohms, not a
paltry 3.5ka-a seen i many designs. I happen to think the 3k8 a-a
loading in Quad-II amps is appalling.
Its forgivable though, and if one has 16 ohm speakers, then connect to
the 8 ohm setting, and the RLa-a then becomes 7k6, a far nicer loading
for a pair of KT66. There is less maximum power, but we shouldn't need
huge power...
Now when you try to use pure tetrode or pentode operation with such a
high RL, the linearity is poor at high levels and the Rout is
uselessly too high. If you use triode, the efficiency will be good for
triode where Ea is rather high above 400V and and Ia lowish at 50mA.
The reduction of gain from pentode to triode by means of triode
connection means you have a large application of NFB via the screen,
and linearity is vastly better. One will find that to extend the range
of anode signal without the limitations of grid current with triode,
one may use a very high ultralinear % tapping, and 66% is just fine,
so that instead of say 16W from a pair of EL34, you can have 25W, and
Ra-a is nearly as low as the triode case. So one just places the UL
tap just far away enough from the anode connection to allow a healthy
Va signal without grid current, and one can not adhere to the usual
mantra of always using the 43% tapping point for UL.
43% was convenient, and countless OPTs had 14 layers of primary wire
in their construction.
This meant that 3 layers out from the CT was where the UL taps went.
so the UL % came from the taps being at 3/7 along each half primary.
Leak always used 50%. This was also convenient, but taps at 66% can be
also considered especially where RL is high and we want class A and we
want the best hi-fi.
I built and sold an amp last Xmas where the OP tubs were EL34, and UL
taps were at 66%.
I got a very nice 25W max, but the guy had 16 ohm speakers of 96dB
sensitivity, and so the load was even higher than I made for the 25W,
so linearity was excellent. The guy said he struggles to hear the amp.
But he said he likes the the fact he cannot hear the amp, and can only
hear the music, and that its better than when piped through his sordid
state amp.


Patrick Turner.

[Ian Bell[_2_]]

Patrick Turner wrote:
The 6550 has to be one of the most popular tubes used for hi-fi and
guitar amps.

I am not so interested in using such a tube in guitar amps where low
fidelity, ie, high distortion levels are welcomed because it enhances
the musicians efforts. I am concerned about extracting hi-fi from
6550.

In hi-fi amps, McIntosh use the 6550 in nearly class B conditions
where the tube has low bias current and the loading for a tube in a PP
pair of tubes while in class B working is only 1k. However, McIntosh
have the OPT primary set up so that half the Va-k of each tube appears
at the anode and at the cathode, so there is heavy local NFB and ß =
0.5.

With the addition of a lot of global NFB as well as the local FB in
the OP stage, the appalling non linearity of a beam tetrode in class B
with a low load value can be "cured" and the measured outcome at all
levels right up to clipping are acceptable, and rather good compared
to many other standard designs using UL type config with GNFB only.

My preference is for large amounts of class A working of output tubes
so that with PP circuits I would have a pair of 6550 coupled to a
speaker load with a an OPT with an 8k a-a load where Ea was say +500V.

I also don't like having any OP tubes idling anywhere close to the
maximim Pda rating of the tube. So I will try to have Pda 28W for
6550 even though the rating is 42W. This means that where Pda was say
25W, and Ea was say 450V, then Ia will be just above 50mAdc.

The data says 6550 will have Ra = about 18k, Gm about 11mA/V, and thus
µ = 198.
but gm and Ra are very variable and related to Ia, so that the only
time gm will be 11mA/V and Ra less than 18k is when Ia 120mA, and
there will never be any amp you or I will build which will have such a
high idle Ia dc.

Typically, all of us will bias at about 50mA or just above for most
projects.

So typically we will find a 6550 tube biased at 55mA and Ea = +450V to
have control grid gm = 5.5 mA/V, and gm of screen to be about 0.83mA/V
where Eg2 = +270V, ane the Ra will be 32k, way above the data figures
say it will be.

For pure class A operation, the load for maximum pure class A PO is
approximately 0.9 x Ea / Ia dc.
So for 55mA and 450V, RL will be around 6k to 7k, and for a PP pair a
class A load would be 12k to 14k.

Efficiency is up to 45% in such conditions. Linearity is not brilliant
at near clipping, but its the first few watts everyone needs to
consider.

During the extremes of each wave cycle the gm and Ra vary
considerably. BUT, where the RL is fairly high, the Ia change is less
tha the Va change, so the gm and Ra change much less during a wave
cycle than were one to have Ea = 350V, and RL a lot lower, and Ia a
lot higher. This becomes especially true when you try to make an SE
amp with 6550.

So what is a simple equivalant model for a 6550?
I suggest you first have to set one up in a test circuit to confirm
your samples give some data you need to know to calculate the gains
and tube properties in a given circuit. The data you seek to
findoutabout can't be found in the data sheets or books.

This means you have a B+ applied from a low Z supply, and have a 100
ohm anode load, and a regulated Eg2.
Apply 1Vrms at 1kHz to the control grid, and measure the current in
the 100 ohms.
The gm is thus measured.
Then with the grid shunted to 0V, apply a 1Vrms signal voltage to the
screen from a low Z signal source, and measure the current in the 100
ohm anode load.

I used Ea = 420V ( Ek was 35V ) and Eg2 = 270V, and Ia = 55mA, and got
grid1 gm = 5.5mA/V, and screen gm = 0.833 mA/V
Ra was approximately 32,000 ohms.

The model for the tube is as follows :-

Ra is a 32k resistance between anode and cathode.
Grid 1 is a terminal controlling current between anode and cathode
which is a CURRENT SOURCE of 5.5mA/V. Also between anode and cathode
there is a second controlled current source controlled by the voltage
at screen grid2, and current produced a to k is 0.833mA/V.
The impedance looking into the anode sides of each current source is
an infinite amount.
So each current source can act to control current change at the anode
across loads or across the RA without interfering with the two gm
figures, and the anode current and Ra current is the sum of the two
current sources.
In pure beam tetrode operation, the voltage at the screen is kept
constant, so the screen exerts no change in Ia. So when you apply
1Vrms signal to grid1, you get a 5.5mA Ia current change applied to
Ra, and to the anode load which is in parallel with Ra. If there is no
RL, the 5.5 mA of Ia change can be said to operate within our
equivalent mathematical model, and you get a an anode signal = 5.5mA x
32k which is about 176, and 176 is the amplification factor of the
this tube in tetrode mode.
So the formula for all tube gain holds true, A = µ x RL / ( RL + Ra ).

But for all tubes, µ = gm x Ra, so we could replace µ in the last eqtn
with Gm x Ra.

How does one measure Ra? Well in the test set up above, we use a 30H
choke to supply dc to the 6550 under test, and at 1kHz the choke's
reactance is so high the reactance loading can be neglected. We can
then apply 1V to grid 1, and we will find that the anode signal will
be perhaps 176Vrms, ie, gain = µ = gm x Ra. But with 176 Vrms at the
anode the distortion will be high and it will affect our measurements
so we lower the grid input to leave about say 50Vrms max at the anode.
Then we connect a 7k load across the choke which loads the the anode.
You will observe the anode voltage falls a lot and you measure this
voltage and you calculate the current in the load.

To find out the Ra, Ra = ( Anode voltage with no load - anode voltage
with load ) / load current.

One can also use two different load values, and the Ra = change in
anode voltage / change in load current.

So what happens when triode connection is used? Well, you need to know
what the triode µ is.
So with g2 connected to the anode, and bias Ia adjusted for 55mA, you
measure the voltage gain without any load except the choke, and you
should find gain = µ = 6.7 or therabouts.
Ra with triode connection can also be measured, and will be found to
be about 1.2k

So imagine the model we have, and with triode connection, and with no
load.

We have +1V at the grid1, and there is -5.5mA produced by current
source1. So a -Va signal of -6.7V is produced because we have triode
connection. This whole -signal is applied to the screen controlled
current source, and there is a +current produced at the anode = +6.7 x
0.833 mA/V.
The two currents sum to produce the current across the Ra = 6.7V /
32k.
If there is a load also present, you can calculate the load voltage
and gain using the model.
if the tube operates in UL mode, or CFB mode, then the voltage applied
to the screen is a fraction of the anode voltage.

I leave you all to ponder this, and once understood, its somewhat easy
to draw a graph plotting Ra and µ for all values of UL tapping %.

Hardly anyone would bother to go to all this bother, they would just
hook things up and hope for the best with educated guesses if they
have a good gut feeling about the anode curves for any multi grid tube
which one wants to use.
The only trouble with the model is that the screen needs some input
current because of the screen current; its not like grid 1 where the
input Z is extremely high. So just to complicate the model which so
far was fairly simple in the world of electronic models, we need to
add a resistance between cathode and the screen grid control point of
its controlled current source.
This R could be found by measurement with a series R with the screen
while a small screen signal is applied. This R will form a loading on
the anode circuit if you have triode connection or UL partial triode
connection. But with triode connection, or UL connection the resulting
Ra which is far lower than for pure tetrode operation means that the
loading by screen drive input current is not going to greatly change
the calculated results and graphs we may draw about the tube's
behaviour.

Just using only triode connected 6550 or other multigrids does make
life very easy. Once we add in the effects of the screen upon Ia flow
in the case of UL then matters can rapidly become quite complex, so
any audio amateur can be forgiven for saying "Damn all models, let me
build, measure, and tweak it to perfection!"

One tip though is certain. For best hi-fi, the anode load value has
usually to be at least twice what many comercial makers use. The
cretins of commerce design for POWER, and don't care much about hi-fi.
So a typical OPT for a pair of 6550 or EL34 would be 10k:5ohms, not a
paltry 3.5ka-a seen i many designs. I happen to think the 3k8 a-a
loading in Quad-II amps is appalling.
Its forgivable though, and if one has 16 ohm speakers, then connect to
the 8 ohm setting, and the RLa-a then becomes 7k6, a far nicer loading
for a pair of KT66. There is less maximum power, but we shouldn't need
huge power...
Now when you try to use pure tetrode or pentode operation with such a
high RL, the linearity is poor at high levels and the Rout is
uselessly too high. If you use triode, the efficiency will be good for
triode where Ea is rather high above 400V and and Ia lowish at 50mA.
The reduction of gain from pentode to triode by means of triode
connection means you have a large application of NFB via the screen,
and linearity is vastly better. One will find that to extend the range
of anode signal without the limitations of grid current with triode,
one may use a very high ultralinear % tapping, and 66% is just fine,
so that instead of say 16W from a pair of EL34, you can have 25W, and
Ra-a is nearly as low as the triode case. So one just places the UL
tap just far away enough from the anode connection to allow a healthy
Va signal without grid current, and one can not adhere to the usual
mantra of always using the 43% tapping point for UL.
43% was convenient, and countless OPTs had 14 layers of primary wire
in their construction.
This meant that 3 layers out from the CT was where the UL taps went.
so the UL % came from the taps being at 3/7 along each half primary.
Leak always used 50%. This was also convenient, but taps at 66% can be
also considered especially where RL is high and we want class A and we
want the best hi-fi.
I built and sold an amp last Xmas where the OP tubs were EL34, and UL
taps were at 66%.
I got a very nice 25W max, but the guy had 16 ohm speakers of 96dB
sensitivity, and so the load was even higher than I made for the 25W,
so linearity was excellent. The guy said he struggles to hear the amp.
But he said he likes the the fact he cannot hear the amp, and can only
hear the music, and that its better than when piped through his sordid
state amp.


Patrick Turner.







First I am surprised at you Patrick, wanting a model for a tube. I
advise you go build and measure because any simulation, be it with a
computer or pencil and paper, is nothing like the real thing ;-)


Secondly, I notice from one data sheet that the UL distortion figures
are much higher then the class AB PP ones so it would seem the latter is
the better way to start. You should still get nearly 50W in AB which is
surely plenty for domestic use.

Cheers

Ian

[Bret L]

The best use for a 6550 in high fidelity service IMO is as a pure
pentode with a lower screen voltage, perhaps 300 volts, regulated or
just a low impedance supply, or in ultralinear mode at some ratio.
Rather than spend a lot of brainpower modelling just set the amp up
with differing load values and run a full performance eval at each.
You will quickly see which is "the best"..

No two 6550s from here on out are likely to be the same and those
changes will make a lot of difference.

McIntosh used the 6550 only in one unit, the MC60 and it is
inadvisable to run them in any other stock McIntosh amp. Because of
the operating points what applies to Mc circuits may not be
particularly applicable elsewhere. I thought we established you did
not favor the Mc circuit anyway, which is a defensible position to
take. The Mc's are not for every application or listener and do
present manufacturing issues, i.e. the transformers.

Were I building a commercial tube hi fi amplifier today I would
design them where possible to take any available eight pin audio power
type from the 6L6GC up. Provided you make maximum plate dissipation
and plate and screen voltage within limits for all types, provide a
wide range of bias voltage adjustability and plenty, but plenty of
heater power available that is no problem. As disgusting and
coprophagic as it is, the Toob Rollers are unstoppable and must sadly
be provided for.

For guitar I don't like 6550s at all. For bass, or amplifying
acoustic or steel guitar the output section is simply a re-form-
factored hi fi amp so they are fine for that. I actually prefer the
6V6 as a guitar tube,

[Fu Knee]

On Oct 23, 10:09�am, Patrick Turner wrote:
The 6550 has to be one of the most popular tubes used for hi-fi and
guitar amps.


Jeezus Fukk, Patrick,

Your clinging to every iota of pain you can extract from jamming an
incompetently debarked log of ill-selected truth up your own ass is
just freaking dead dull.

"Most popular tubes" may be your idea of critical insight, but, my
dear man, consider reality from a bit less preoccupied perspective:

Lucille de Ville may be the uglist bit of whoopee ever to darken the
doorway of the garden of Earthly delights, but, if she is the only one
present during a John's visit, she is perfection.

And knowing you have ten girls available who make her appear, quite
clearly, as a life long cure for lust is not a reflection on you.

It is the romance of retail consumption.

People buy ****. Each week, the sellers compare their net and declare
a winner. Of what? Responsible product management? Hardly. It is a
race for morons counting their take of the stupid flushings of
temporary wealth. On an arbitrary time frame.

Quit this beligerant posing as the victim of the Hell of hyphenated
wealth. That is a suckers game.

If you simply build whatever sounds the best, to you, you will still
be equally humiliated by the mindless statistics of counting of random
dollars, but, may, in good faith, put your head easily on the warm
pillow knowing you have made reality as best you honestly could, this
day.

______

Much of today's world is reeling from the effect of too many too
bright boys giving up on women because they could, statistically, be
more productive using their own, known hand.

For God's sake man, do not follow those imbeciles down into the sewer.

Get a grip.


Life ain't fair. That does not mean we cannot be.

It don't mean shee-it.

Happy Ears!

Al

[Patrick Turner]

On Oct 24, 9:01*am, Bret L wrote:
*The best use for a 6550 in high fidelity service IMO is as a pure
pentode with a lower screen voltage, perhaps 300 volts, regulated or
just a low impedance supply, or in ultralinear mode at some ratio.
Rather than spend a lot of brainpower modelling just set the amp up
with differing load values and run a full performance eval at each.
You will quickly see which is "the best"..

With pure tetrode opp the use of lower Eg2 below Ea is the best for hi-
fi because the load will always be higher than for maximum class AB
where load is lower, and you want to swing Ea low without the knee of
the Ra curve for Eg1 = 0V limiting the Ea swing.


*No two 6550s from here on out are likely to be the same and those
changes will make a lot of difference.

*McIntosh used the 6550 only in one unit, the MC60 and it is
inadvisable to run them in any other stock McIntosh amp. Because of
the operating points what applies to Mc circuits *may not be
particularly applicable elsewhere. I thought we established you did
not favor the Mc circuit anyway, which is a defensible position to
take. The Mc's are not for every application or listener and do
present manufacturing issues, *i.e. the transformers.


I'm not a fan of the McIntosh circuit because it flogs output tubes
hard with a low value RL to get the power, and to get away with the
high distortion local NFB is used in the output stage with ß = 0.5.
So the driver stage needs to make a max of about 150V at each grid,
and then you have the distortion of the driver stage to worry about.
So GNFB is used to clean up the whole mess and the final result isn't
any better than a regular UL amp with less total applied FB except
that the McI gets more power and a slightly better damping factor. So
to me the McI is 3 steps forward, and two backward. The EAR509 is
another high NFB amp with near class B working of the output tubes.
I had a couple here that I rewired after one had destroyed itself with
a fire inside which nearly burned a house down. When I got it working
just right it sounded worse than amps I make with less power max, less
FB but more class A.



*Were I building a commercial tube hi fi amplifier today I would
design them where possible to take any available eight pin audio power
type from the 6L6GC up. Provided you make maximum plate dissipation
and plate and screen voltage within limits for all types, provide a
wide range of bias voltage adjustability and plenty, but plenty of
heater power available that is no problem. As disgusting and
coprophagic as it is, the Toob Rollers are unstoppable and must sadly
be provided for.

I basically agree with all that.

*For guitar I don't like 6550s at all. For bass, or amplifying
acoustic or steel guitar the output section is simply a re-form-
factored hi fi amp so they are fine for that. I actually prefer the
6V6 as a guitar tube,

Just who likes what in guitar amps is never very clear.

But guitar amps all have very high distortion.

Once you move to hi-fi amps then distortion should be minimised IMHO,
and then class A becomes desirable and affordable because not a huge
amount of power is wanted. I find the 6550 for hi-fi sounds very well,
and my latest amps have 6 x 6550 in parallel for 60W pure class A,
single ended, with OPTs with 20% of local cathode FB, and with a fixed
screen voltage at about 300V, with Ek at 30V, Ea at 485V, and the
anode load for each is between 6k and 7k. The drive voltage max is 75V
from two paralleled EL84 in triode with CCS dc feed whichy allows the
grid bias R for the 6 OP tubes to be a bootstrapped 22k.

There is no real need for GNFB, but I have 10dB to reduce Rout to 0.29
ohms where the load match for maximum power is 5 ohms.

I find the OPT CFB winding with fixed Eg2 or "Acoustical Connection is
the best way to coax hi-fi from most beam tetrodes pentodes.

The model and analysis allows understanding if you look for it. The
model for any one tube is usually one of two types.
One is where you have a low output impedance voltage generator
producing µ x Eg1 of output, with a resistance of Ra in series. This
works fine where you have a tube working in pure beam, pentode or
triode.
The other model is where Ra is a shunt resistance and there is a high
output impedance voltage controlled current source shunting the Ra as
well as the load. The current into load and across Ra is generated as
Eg1 x gm. This second model allows one to insert the function of the
screen and its signal feed to a second shunt current source.

Some are bored stiff with analysis. I write for those who ask
questions and don't mind the analysis.

Patrick Turner.

[Patrick Turner]

On Oct 24, 7:59*pm, Fu Knee wrote:
On Oct 23, 10:09 am, Patrick Turner wrote:

The 6550 has to be one of the most popular tubes used for hi-fi and
guitar amps.

Jeezus Fukk, Patrick,

Your clinging to every iota of pain you can extract from jamming an
incompetently debarked log of ill-selected truth up your own ass is
just freaking dead dull.

"Most popular tubes" may be your idea of critical insight, but, my
dear man, consider reality from a bit less preoccupied perspective:

Lucille de Ville may be the uglist bit of whoopee ever to darken the
doorway of the garden of Earthly delights, but, if she is the only one
present during a John's visit, she is perfection.

And knowing you have ten girls available who make her appear, quite
clearly, as a life long cure for lust is not a reflection on you.

It is the romance of retail consumption.

People buy ****. Each week, the sellers compare their net and declare
a winner. Of what? Responsible product management? Hardly. It is a
race for morons counting their take of the stupid flushings of
temporary wealth. On an arbitrary time frame.

Quit this beligerant posing as the victim of the Hell of hyphenated
wealth. That is a suckers game.

If you simply build whatever sounds the best, to you, you will still
be equally humiliated by the mindless statistics of counting of random
dollars, but, may, in good faith, put your head easily on the warm
pillow knowing you have made reality as best you honestly could, this
day.

______

Much of today's world is reeling from the effect of too many too
bright boys giving up on women because they could, statistically, be
more productive using their own, known hand.

For God's sake man, do not follow those imbeciles down into the sewer.

Get a grip.

Life ain't fair. That does not mean we cannot be.

It don't mean shee-it.

Happy Ears!

Al

Al, stop worrying.

I build what sounds best **for other people**. What I think does not
count a great deal. I'm driven by demand from others, and not by how I
percieve the sound. To get what sounds best I ask many questions and
go for class A linearity.

Rather than waste time chasing monumentally useless women, or watching
vast amounts of crap on TV, or spending time bull****ting in pubs over
a beer, I do spend some small amount of time analysing the circuits I
build including the essential inner workings of the active devices.

Tomorrow I might ride 100km on my bicycle with friends, so I share my
existance with vigour.

Patrick Turner.

[Fu Knee]

On Oct 24, 7:58�am, Patrick Turner wrote:

Al, stop worrying.

I build what sounds best **for other people**. What I think does not
count a great deal. I'm driven by demand from others, and not by how I
percieve the sound. To get what sounds best I ask many questions and
go for class A linearity.


Hi RATs!

Oh, you have a real customer. Oops, sorry.

I forget other people still have lives.

All I have is some memories ... and this great hifi

Happy Ears!
Al

[Patrick Turner]

On Oct 25, 7:19*am, Fu Knee wrote:
On Oct 24, 7:58 am, Patrick Turner wrote:



Al, stop worrying.

I build what sounds best **for other people**. What I think does not
count a great deal. I'm driven by demand from others, and not by how I
percieve the sound. To get what sounds best I ask many questions and
go for class A linearity.

Hi RATs!

Oh, you have a real customer. Oops, sorry.

I forget other people still have lives.

All I have is some memories ... and this great hifi

Happy Ears!
Al

I stand upright and I am alive, so I am grateful for the Force Behind
Life who neither loves or hates or gives a tinker's damn about me.

I have a quite a few customers.

Not nearly as many as say Peter Walker had, but as you might think,
who cares about the numbers?

Meanwhile, I need to investigate screen grid drive for the 6550 ( or
any other audio heathen's favourite octal tube with a screen grid
within ).

While farnarkling around measuring the G2 Gm I found that you can
easily get a gain of about 5 into aa anode load of about 7k with a
6550 with Ea = 435V, Ia approx 55mA, and Eg2 at 270V.
The linearity in screen grid triode mode seems rather a lot better
than when you drive through the G1 control grid.

But the tube Ra is 30k, about the same as with pure beam tetrode
operation.

OK, so you can apply global NFB and you can easily get Ra effectively
down to about 900 ohms, and if the OPT Z ratio is 7k : 7 ohms, you get
Rout = 0.9 ohms at the sec, and a reasonable damping factor.

But with screen grid drive there will be a low distortion drive
voltage for the screens from a low output impedance source, say a
choke fed triode such as EL84, EL86, or EL34 etc.

A resistance divider can be made between the output tube anode and the
driver anode so that at the junction of the R divider the signal
voltage is say 0V where there is a 7k load. From this null point on
the divider you have a cap and biasing resistor to the control grid of
the output tube. This forms a local shunt FB network. The gain of the
output 6550 is about 32 with a 7k load when driven by G1 because it
acts as a pentode. ß of about 0.2 is possible, so say +Dn appearing at
the 6550 anode creates +0.2Dn at G1, and this gets amplified 32 times
to appear as -6.4Dn at the anode. This subtracts from what would be
+7.4Dn without the FB loop. Not a bad amount of distortion reduction.
And the Rout of 30k is reduced to less than 900 ohms, or about 0.9
ohms at the secondary.

I would hazard a guess that such a set up would be many times more
linear than a 300B, or normal triode connected 6550 which I have found
to be marginally more linear that the 300B.

So when some GNFB is added if one wants to, you'd only need 10dB to
reduce Rout to about 0.3 ohms, and reduce the already low THD to a
third.

The amplifier can still be a 3 tube affair where there are low µ input
and driver tubes driving what is a low µ triode output tube.

Ann alternative way of driving the screen is via a gain triode with
directly connected cathode follower to drive the screen and the FB
divider between CF output and output tube anode.
There would be other methods.

Load changes cause a change of gain in the screen grid driven 6550 so
much so that the gain varies the same way as with normal tetrode
operation. The changes of gain cause a small G1 drive voltages to
appear and the changes in gain are opposed by this feedback. The FB
delivery is via resistances outside the tube so the delivery network
is linear, and rather better than that of trying to put lots of local
NFB around a tetrode by connecting the screen to the anode.

If however you try to do things the other way and apply a normal G1
voltage and you have a divider from source voltage to anode voltage,
and you apply the error signal to the screen, then the amount of error
correction is miniscule because the screen has such low Gm.

There's food for thought to any thinkers out there.

Patrick Turner.

[Fu Knee]

On Oct 25, 3:51�am, Patrick Turner wrote:

Meanwhile, I need to investigate screen grid drive for the 6550 ( or
any other audio heathen's favourite octal tube with a screen grid
within ).

While farnarkling around measuring the G2 Gm I found that you can
easily get a gain of about 5 into aa anode load of about 7k with a
6550 with Ea = 435V, Ia approx 55mA, and Eg2 at 270V.
The linearity in screen grid triode mode seems rather a lot better
than when you drive through the G1 control grid.

But the tube Ra is 30k, about the same as with pure beam tetrode
operation.

OK, so you can apply global NFB and you can easily get Ra effectively
down to about 900 ohms, and if the OPT Z ratio is 7k : 7 ohms, you get
Rout = 0.9 ohms at the sec, and a reasonable damping factor.

But with screen grid drive there will be a low distortion drive
voltage for the screens from a low output impedance source, say a
choke fed triode such as EL84, EL86, or EL34 etc.

A resistance divider can be made between the output tube anode and the
driver anode so that at the junction of the R divider the signal
voltage is say 0V where there is a 7k load. From this null point on
the divider you have a cap and biasing resistor to the control grid of
the output tube. This forms a local shunt FB network. The gain of the
output 6550 is about 32 with a 7k load when driven by G1 because it
acts as a pentode. � of about 0.2 is possible, so say +Dn appearing at
the 6550 anode creates +0.2Dn at G1, and this gets amplified 32 times
to appear as -6.4Dn at the anode. This subtracts from what would be
+7.4Dn without the FB loop. Not a bad amount of distortion reduction.
And the Rout of 30k is reduced to less than 900 ohms, or about 0.9
ohms at the secondary.

I would hazard a guess that such a set up would be many times more
linear than a 300B, or normal triode connected 6550 which I have found
to be marginally more linear that the 300B.

So when some GNFB is added if one wants to, you'd only need 10dB to
reduce Rout to about 0.3 ohms, and reduce the already low THD to a
third.

The amplifier can still be a 3 tube affair where there are low � input
and driver tubes driving what is a low � triode output tube.

Ann alternative way of driving the screen is via a gain triode with
directly connected cathode follower to drive the screen and the FB
divider between CF output and output tube anode.
There would be other methods.

Load changes cause a change of gain in the screen grid driven 6550 so
much so that the gain varies the same way as with normal tetrode
operation. The changes of gain cause a small G1 drive voltages to
appear and the changes in gain are opposed by this feedback. The FB
delivery is via resistances outside the tube so the delivery network
is linear, and rather better than that of trying to put lots of local
NFB around a tetrode by connecting the screen to the anode.

If however you try to do things the other way and apply a normal G1
voltage and you have a divider from source voltage to anode voltage,
and you apply the error signal to the screen, then the amount of error
correction is miniscule because the screen has such low Gm.

There's food for thought to any thinkers out there.


Hi RATs!

I built a version of Tim de Paravicini's screen drive SV501 SE amp,
back in the salad days of this hobby.

The screen was driven by an SV83. It mave have been choke loaded.
"Retard Coils" were a personal favorite, early on.There was some tube
driving the SV83, but memory fails. I used to create engineering
notebooks on my computer with annotated schema and stuff, as it once
was my intention to join you angry mob of burning amp boutique
manufacturer-warriors. Things slip away. The PC's became quaint and
were filed in the storeroom. The new PCs did get useful or interesting
files from the old, but dwindling functionality of yrs trly never kept
anything alive for three generations of PC. My current Unit is a dual
core 64 bit with gigs of memory, I wanted to see Fractals generated on
the 1280x1024 screen, but the great, free software is lost to me. I
imagine it would be pretty cool... I know it was

To give some scale to this bout of Yuppie flu, my first upgrade while
horizontal - me, not the technology - was from a 386 to a 486.

My wife now has a quad core, It cost less than half what my first
8086, did. But then, that did have the full 64K of RAM 8*D

Our young son started his journey with a Vic 20, 5K of RAM! He is now
the senior VP of a financial sytems shop. Eyeblink after eyeblink, it
just keeps happening :')

Anyway, that screen grid drive amp was a great joy to listen thru I
think I just tied G1 to the K...

Do pursue the thought. It does not just measure well...

Happy Ears!
Al

[Fu Knee]

Errata, SV501 in prev post S/B EL509 (6KG6).

Al

[Patrick Turner]

On Oct 25, 9:05*pm, Fu Knee wrote:
On Oct 25, 3:51 am, Patrick Turner wrote:







Meanwhile, I need to investigate screen grid drive for the 6550 ( or
any other audio heathen's favourite octal tube with a screen grid
within ).

While farnarkling around measuring the G2 Gm I found that you can
easily get a gain of about 5 into aa anode load of about 7k with a
6550 with Ea = 435V, Ia approx 55mA, and Eg2 at 270V.
The linearity in screen grid triode mode seems rather a lot better
than when you drive through the G1 control grid.

But the tube Ra is 30k, about the same as with pure beam tetrode
operation.

OK, so you can apply global NFB and you can easily get Ra effectively
down to about 900 ohms, and if the OPT Z ratio is 7k : 7 ohms, you get
Rout = 0.9 ohms at the sec, and a reasonable damping factor.

But with screen grid drive there will be a low distortion drive
voltage for the screens from a low output impedance source, say a
choke fed triode such as EL84, EL86, or EL34 etc.

A resistance divider can be made between the output tube anode and the
driver anode so that at the junction of the R divider the signal
voltage is say 0V where there is a 7k load. From this null point on
the divider you have a cap and biasing resistor to the control grid of
the output tube. This forms a local shunt FB network. The gain of the
output 6550 is about 32 with a 7k load when driven by G1 because it
acts as a pentode. of about 0.2 is possible, so say +Dn appearing at
the 6550 anode creates +0.2Dn at G1, and this gets amplified 32 times
to appear as -6.4Dn at the anode. This subtracts from what would be
+7.4Dn without the FB loop. Not a bad amount of distortion reduction.
And the Rout of 30k is reduced to less than 900 ohms, or about 0.9
ohms at the secondary.

I would hazard a guess that such a set up would be many times more
linear than a 300B, or normal triode connected 6550 which I have found
to be marginally more linear that the 300B.

So when some GNFB is added if one wants to, you'd only need 10dB to
reduce Rout to about 0.3 ohms, and reduce the already low THD to a
third.

The amplifier can still be a 3 tube affair where there are low input
and driver tubes driving what is a low triode output tube.

Ann alternative way of driving the screen is via a gain triode with
directly connected cathode follower to drive the screen and the FB
divider between CF output and output tube anode.
There would be other methods.

Load changes cause a change of gain in the screen grid driven 6550 so
much so that the gain varies the same way as with normal tetrode
operation. The changes of gain cause a small G1 drive voltages to
appear and the changes in gain are opposed by this feedback. The FB
delivery is via resistances outside the tube so the delivery network
is linear, and rather better than that of trying to put lots of local
NFB around a tetrode by connecting the screen to the anode.

If however you try to do things the other way and apply a normal G1
voltage and you have a divider from source voltage to anode voltage,
and you apply the error signal to the screen, then the amount of error
correction is miniscule because the screen has such low Gm.

There's food for thought to any thinkers out there.

Hi RATs!

I built a version of Tim de Paravicini's screen drive SV501 SE amp,
back in the salad days of this hobby.

The screen was driven by an SV83. It mave have been choke loaded.
"Retard Coils" were a personal favorite, early on.There was some tube
driving the SV83, but memory fails. I used to create engineering
notebooks on my computer with annotated schema and stuff, as it once
was my intention to join you angry mob of burning amp boutique
manufacturer-warriors. Things slip away. The PC's became quaint and
were filed in the storeroom. The new PCs did get useful or interesting
files from the old, but dwindling functionality of yrs trly never kept
anything alive for three generations of PC. My current Unit is a dual
core 64 bit with gigs of memory, I wanted to see Fractals generated on
the 1280x1024 screen, but the great, free software is lost to me. I
imagine it would be pretty cool... I know it was

To give some scale to this bout of Yuppie flu, my first upgrade while
horizontal - me, not the technology - was from a 386 to a 486.

My wife now has a quad core, It cost less than half what my first
8086, did. But then, that did have the full 64K of RAM 8*D

Our young son started his journey with a Vic 20, 5K of RAM! He is now
the senior VP of a financial sytems shop. Eyeblink after eyeblink, it
just keeps happening :')

Anyway, that screen grid drive amp was a great joy to listen thru I
think I just tied G1 to the K...

Do pursue the thought. It does not just measure well...

Happy Ears!
Al- Hide quoted text -

- Show quoted text -

Indeed Parrot A'Chini did put out a screen drive design with high
GNFB. I have a circuit copy somewhere.I doubt he ever thought of using
the idea of using the output tube G1 as a live entry port for local
NFB,and I doubt he tried out the possibilities with enough more
commonly available tubes still being manufactured.

The screen drive requires a low driver output resistance which could
be a µ-follower. but then I'd need a 600V supply, so that the output
tube screen is at 300V and direct coupled to the follower.
I'm thinking of using screen drive in my next large customer project
using PP 13E1 tubes where the Eg2 is ideally quite low at 220V, and G2
Gm is quite high.

Patrick Turner.

[Bret L]

I'm not a fan of the McIntosh circuit because it flogs output tubes
hard with a low value RL to get the power, and to get away with the
high distortion local NFB is used in the output stage with ß = 0.5.
So the driver stage needs to make a max of about 150V at each grid,
and then you have the distortion of the driver stage to worry about.
So GNFB is used to clean up the whole mess and the final result isn't
any better than a regular UL amp with less total applied FB except
that the McI gets more power and a slightly better damping factor. So
to me the McI is 3 steps forward, and two backward.

The Unity Coupled output stage is simply a buffer. The work is in the
driver stage BUT said driver stage works under ideal conditions in
some ways,i.e., a constant load. Some refinement of the Mc circuit is
possible and Milojub Nestorovic is probably the guy who did the most.
I have seen schematics for Capt. Catchfire's stuff but never
Nestorovic's.

Separating the driver supply from the output stage supply is a big
win in this circuit as is a totally decoupled heater supply for the
whole works as the supply wavers with output level because they are on
the same core. More work remains to be done but IMO the potential of
this circuit is still great without the awkwardness of workarounds
like the Circlotron or various Norman Crowhurst designs. Norm was a
very excellent writer but his circuits were always not so practical.
His circuit, the Bereskin and the Circlotron were all workarounds IMO
of the Mc patent. Bereskin was under contract to Baldwin Piano who
initially tried to license McIntosh tech for their organs, as had
Allen Organ, and both were turned down. AO built later, what for hi fi
use is a superb amp with 6550s that ran them at 600 v B+ and 300 vdc
on the screens, Baldwin never did build a Bereskin design
commercially, but there is reputed to be a prototype bass amp built in
the days when they sold Burns guitars here that ran a pair of small
transmitting tubes and put out over 250 watts. That's notable because
the SVT (which is essentially a hi fi amp in its output section and
which you should study) "set whole new standards' with just under 300
watts from six 6146s or 6550s. That was two or three years later. Did
Baldwin take their Big Amp to NAMM and get gazumped?

[Patrick Turner]

On Oct 26, 3:43*pm, Bret L wrote:
I'm not a fan of the McIntosh circuit because it flogs output tubes
hard with a low value RL to get the power, and to get away with the
high distortion local NFB is used in the output stage with ß = 0.5.
So the driver stage needs to make a max of about 150V at each grid,
and then you have the distortion of the driver stage to worry about.
So GNFB is used to clean up the whole mess and the final result isn't
any better than a regular UL amp with less total applied FB except
that the McI gets more power and a slightly better damping factor. So
to me the McI is 3 steps forward, and two backward.

*The Unity Coupled output stage is simply a buffer.

Well, you could say that about any output stage. The output stage is
what puts amps of current behind a voltage signal.

I recall people marvelled how you could get 50W from apair of 6L6 from
an early McI.
And without a very high Ea voltage, which was the same as the Eg2
supply because screens were cross coupled to force the 6L6 to work as
pure beam tetrodes but with 50% of the signal fed back at the cathode.
So you could have very high open loop THD in the output stage which
was linearized with bucket fulls of local NFB so the output stage then
gave THD about equal to triodes without GNFB but you got 4 times the
triode power.
But where you run a pair of 6L6 in pure class A in beam tet mode, the
THD can be 2% at 25W, with very low THD at low levels even without any
FB at all.
So for 50W I would use a quad of 6L6. Since we want the amp for hi-fi,
the 50W would class AB at the lowest speaker Z envisaged, or at 3
ohms, so that at 8 ohms the PO would be 35W and definateIy all class
A.
If we add a CFB winding of say 12.5%, then its easy to get that 35W at
0.6%, and this is better than the McI, and we don't have to use such a
high drive voltage when the drive amp begins to make more THD/IMD than
the output stage.
Quad-II is an example, and its such a simple design compared to the
McI. Of course Peter Walker saw no reason to make Quad-II amps with 2
x KT88 for 40W, or 4 x KT88 for 80W. There was rarely any need for
such power in 1960.
And after 1960 most makers dived into Sordid State, boots and all.


The work is in the
driver stage BUT said driver stage works under ideal conditions in
some ways,i.e., a constant load.

Indeed it does. And with bootstrapped anode loads which give some
positive FB which luckily increases THD less than the reduction of THD
because the driver tubes see a higher anode load.

I'd prefer not to bootstrap and use a CT choke plus anode resistances
to get DC to the driver stage.
THD is a typical 0.3% at 75Vrms at each anode. And 75V is all I need
for an amp with 20% CFB in its output stage.
The extra benefits by going to 50% CFB don't exist IMHO.


Some refinement of the Mc circuit circuit is
possible and Milojub Nestorovic is probably the guy who did the most.
I have seen schematics for Capt. Catchfire's stuff but never
Nestorovic's.

*Separating the driver supply from the output stage supply is a big
win in this circuit as is a totally decoupled heater supply for the
whole works as the supply wavers with output level because they are on
the same core. *

I doubt having separate rail supplies makes any difference at all if
they are well filtered, ie, they use huge values of C compared to the
old PSU designs which were controlled by accountants.

Many old amps gave you the sound of an accountant removing the
quality.

More work remains to be done but IMO the potential of
this circuit is still great without the awkwardness of workarounds
like the Circlotron or various Norman Crowhurst designs. Norm was a
very excellent writer but his circuits were always not so practical.
His circuit, the Bereskin and the Circlotron were all workarounds IMO
of the Mc patent. Bereskin was under contract to Baldwin Piano who
initially tried to license McIntosh tech for their organs, as had
Allen Organ, and both were turned down. AO built later, what for hi fi
use is a superb amp with 6550s that ran them at 600 v B+ and 300 vdc
on the screens, Baldwin never did build a Bereskin design
commercially, but there is reputed to be a prototype bass amp built in
the days when they sold Burns guitars here that ran a pair of small
transmitting tubes and put out over 250 watts. That's notable because
the SVT (which is essentially a hi fi amp in its output section and
which you should study) "set whole new standards' with just under 300
watts from six 6146s or 6550s. That was two or three years later. Did
Baldwin take their Big Amp to NAMM and get gazumped?

I hear what you mean about the basic operation of McI and Circlotron
etc. All have 50% CFB in the OP stage and allow near class B op and
still get good measurements. Its all a result of high NFB application.

I had an Ampeg here for a re-design and and instaltion of an Oz made
PT with primary of 240V. The Ampeg surely had its faults. Indeed you
can get 300W from 6 x 6550, ie, at a rate of 100W per pair of 6550.
But it means its nearly all class B power with very low class A
content. So for 300W, I like to use a dozen 6550.

Although having Ea at 600V and Eg2 at 300V does allow a high PO with a
low load if you drive it class AB2, its still a bit nervous nelly time
for the tubes. I have never seen reason to have Ea higher than 500V
and this allows UL or triode operation where Ea = Eg2; the Pd of G2 is
kept within limits because of the Vac applied to G2.
If I wanted to use Ea = 600V, I'd always have the Acoustical
connection with CFB. The load vale for class AB1 is a bit high though
which means Vaa is high so the OPT needs many turns and / or a bigger
than normal core to avoid saturation at too high an F.

There is a point where the use of a high Ea means that Ia dc at idle
has to be kept real low, and the first watts are then created by tubes
which are in their cut off zone and you get early 3H production. With
Ea = 600V, and with idle Pda = 15W, Ia will be only 25mA, and its fine
in a guitar amp where THD/IMD is welcomed. The load line will rise
above the 42W Pda limit to get the high PO, and hence we see many
early tubes fail where high OP stage power way into clipping is
routinely used. For hi-fi I'd feel happier with 500V and Pda can be
25W and Ia can 50mA.
With a high RL you can get 45% efficiency in class A giving about 22W
of class A, about enough for anyone I know where they don't want bone
crushing levels, and they do want the very cleanest sound.
We mainly listen with the first 5 watts which covers all the signal
peaks in a hi-fi situtation.

But the high RL means the open loop tube gain is high, so when one
does use sensible amount of CFB, ie, 20%, this linearizes the tube
nicely, and better than having half the load, and pure tetrode with
CFB.
The fact you have a fixed Eg2 with 20% CFB makes the tube give much
better THD spectra than if you had the screens bypassed to the
cathode; I like the combined effects of CFB in the Acoustical
connection.

Where you have 20% CFB with a fixed Eg2, the stage works similarly in
terms of its open loop gain as youwould have in an OPT stage which has
20% UL taps. Just having 20% UL taps alone reduces Ra and the odd
order spectra very usefully. The 20% CFB combines screen FB and G1
feedback. You end up with an OP stage with closed loop gain of about 4
instead of a McI stage CLG of about 1.6, and needing a very high drive
voltage. ( In a McI, for 250V a-k at an OP tube you need about 150V
applied to its grid. )


Patrick Turner.