LOAD MATCHING 2. Last edited, 2011.
SINGLE ENDED OUTPUT BEAM TETRODES.

This page explains how to set up single ended output beam tetrodes in tube amplifiers
using the 6550 and KT88 working in pure class A beam tetrode mode or with 43% UL
screen taps, or with local cathode feedback windings which are a portion of the OPT
primary winding.

Operation of the beam tetrode with cut-away sketch by RCA.
Fig1. Graph of Ra curves for GE 6550A beam tetrode with screen supply = +350V.
Fig 2. Graph of Ra curves for GE 6550A beam tetrode with screen voltage = +350V
plus 3 load lines and calculated results for gain, power output and second harmonic
distortion.
Fig 3. Graph of Ra curves for GE6550A beam tetrode but with screen supply = +200V.
Fig 4. Graph of Ra curves for GE6550A with one load line for 3.5k and  calculated
gain, power and 2H.
Explanation of the effects of NFB application for 6550 beam tetrodes.
Formulas for NFB and output resistance.
Fig 5. Graph for power output for single 6550 beam tetrode vs anode load value.
Choosing the OPT ratio.
Fig 6. Graph of 6550 Ra curves for 6550 in UL mode with 43% screen taps.
Calculated gain, power output and 2H for 4 different load values.
Data for Ra, µ and gm for UL and comment on the effects of NFB
and distortion outcomes.
-------------------------------------------------------------------------------------------------

Let us consider the the 6550 which is a beam tetrode.
This tube has four electrodes which are cathode, k, control grid, g1,
screen grid, g2, Anode, a, and two beam forming plates near the anode,
with each connected to the cathode.

The sketch shows invisible electrons flowing from one side of cathode to anode,
but in fact electron flow occurs on both sides of the cathode.

The central cathode consists of a flat metal tube and inside this tube is a length
of folded filament wire coated with a special inert insulating metal oxide to
prevent current flow from heater to cathode. The filament has 3.5 ohms
resistance when hot with an applied 6.3Volts at 1.8Amps needed to raise
the temperature of the cathode to about 900C.

The small orange glow seen in tubes is the hot central cathode. All other parts are
designed to run at a much lower temperature without any glow of their own.
The cathode, k, is often nickel and coated with special metal oxides which promote
the emission of a cloud of electrons once the tube is warmed up. This emission is
only possible if there is a very pure vacuum inside the sealed glass tube structure.

The grid, g1, is made of special wire wound in a slightly flattened helix and wires
are spot welded to two support rods. All the grid support rods, anode and cathode are
supported in discs of mica sheeting which is inert and which rests against the glass to
locate all electrodes at their correct positions within small tolerances. The pitch and
size of the helical coil and and its distance from the cathode determine the tube
behaviour.
Grid 1 is usually set up with a bias dc voltage about -50Vdc below the cathode
potential. By varying the grid 1 voltage + and - with an ac signal, the flow of electrons
from cathode to anode is controlled fairly linearly. The grid 1 draws no current because
its negative charge repels negatively charged electrons.

There is a second grid of larger size, the screen grid, g2, placed outside the grid g1
and with its helical coil wires aligned to the g1 wires to allow about 90% of electrons
to flow past and be absorbed by the anode. The screen g2 usually has a potential
at +300Vdc above the cathode and it attracts 10% of anode current flow and the
heat generated is about 1/15 of the heat generated at the anode.

The TWO beam forming plates are connected internally and permanently to the
cathode. Their function is to repel electrons so that the electrons which are attracted
to the screen and anode are focused into two beams on each side of the cathode
The presence of the beams of concentrated electrons suppresses the effect of electrons
bouncing off the anode and returning to the next best positive thing, the screen, which
is called secondary emission. The beam forming plates were invented to counter the
need to pay royalties to earlier European inventors of the pentode which has a third
grid, g3, between screen and anode which functions similarly to supress secondary
emissions which otherwise spoil the function of pure tetrodes.

The plate, or anode as it should be called is a sheet metal box with open ends and
it absorbs about 90% of the electrons emitted from the cathode with the balance of
10% being absorbed by the screen. The anode has open ends to allow heat to radiate
from the cathode more easily up and downwards. The anode connects to the load
being driven by changes in Va and Ia. The anode is heated by the flow of electrons
and its heating is known as its plate dissipation, Pda, easily calculated as Ia x Ea.

The silver coating on the inside of the glass is called gettering, a very thin layer of
special metal which absorbs stray gas molecules to form compounds which are then
locked to the glass rather than re-cycling around the tube and destroying its operation.

The advantages of the beam tetrode and power pentode allow for about twice the
audio or RF output power to be generated when compared to triodes of the same
size and cost of production and same anode dissipation. The beam power tube or
power pentode can be used with its screen grid connected to the anode to make the
tube operate as a triode. In the case of the 6550, the maximum anode dissipation
rating is 42 watts, and the theoretical maximum amount of audio power from a single
tube in class A1 = 18 watts in beam power mode, but only about 10 watts in class A1
triode. In practice, nobody should ever have a single 6550 set up with Pda = 42 watts
at idle because the tube would run too hot and have a short life. Pda plus Pdg2 should
never exceed 35 Watts at idle, and in fact 28Watts is optimal for class A1, and perhaps
22W for class AB. With Pda = 25W, max power in class A SE beam tetrode = 11 watts,
and in triode about 7 watts. It is possible to extract more power in triode by operating
in class A2, and get perhaps 10 watts but I feel there are not many benefits because
of the higher THD and cost of extra circuit complexity.

Not many people will ever have access to a supply of NOS GE6550A or NOS KT88
which are very similar. Most old tubes made in the US before US factories or UK
factories closed have now been worn out. So tube supplies will be from Russia,
China or eastern Europe.

The New Sensor Corp sold me a large batch of EH 6550 tubes in Feb 2002.
My previous update of this website was in March 2006, and I have not had any
reports of any early failures tubes in that purchase. It is now April 2011, and I
still have no reports of early failures.

The working characteristics of these recently made Russian tubes appear to be
equal to NOS samples of the same tube type such as the renowned GE6550A,
and can be generally be used as replacements in all amps requiring 6550 or KT88.
I also recently measured Sovtek KT88, which gave identical test results to the
EH6550. Sovtek or EH 6550 or EH KT88  from Russia all appear to me at least
to have exactly the same internal physical structure and electronic beam tetrode
data parameters of Ra = 19,000 ohms, Gm = 10mA/V, and µ = 190 at Ea =
+400V, Ia = 90mA, and Eg2 = +350V.

But usually 6550 or KT88 are set up with Ea = 450V, Eg2 = +350V, Ia = 51mA,
Ig2 = 3mA. Then Ra = 32,000 ohms, Gm = 5.5mA/V, µ = 176. This is very
different to data quoted for the higher Ia condition which causes Ra to be lower
and Gm to be higher.

Anode curves for GE 6550A are similar to EH6550.....
Fig 1.

These Fig 1 curves are for GE6550A but typical for any 6550.

If we look along the Ia = 100mA line we see that for a change in grid voltage
of 1.5V there is an Ea change = 350V, so µ = 350 / 1.5 = 233.
If we look along the Eg = -17V line we see that for 350V of Ea change there is
an Ia change of 18mA, so Ra = 350 / 0.018 = 19,444 ohms.
Therefore gm = 12mA/V.

At Eg = -10V, the Ra = 10k and µ = 100 so gm = 10mA/V.

The anode curves are more crowded together vertically at the bottom of the
graph indicating a large change in Ra, µ and gm between Ia = 50mA and Ia
= 300 mA.

Here are some load lines and analysis for loads of 1.75k, 3.5k, 7k....
Fig 2.

Here we have the 6550 loaded with 1.75k, 3.5k and 7k. Some of you may
not know what the heck a load line is.

The load lines in the above Fig 2 are straight lines drawn across the data sheet
"anode characteristic curves" and such straight lines represent load ohm values.
For example, consider load line for RL = 1.75k. It passes through the voltage
axis at +512Vdc, and also through the Ia axis at 293mA. The "slope" of this line
represents a load resistance = Voltage / Current = 512V / 0.293A = 1,747 ohms,
or close enough to 1.75k.

The idle point of the 6550 is at Q where Vdc between anode and cathode
= +350Vdc. This a - k voltage is known as Ea. The idle Iadc for the 6550 is
at 90mA.
This means Pda in this example = Voltage x Current = 350 x 0.09 = 31.5 Watts.

The load line for a single ended tube MUST pass through the Q point. It will
intersect the curves for Ra at various values of negative grid voltage. You may
wonder how do such curves come to be, but its simple, really.

To determine 6550 beam tetrode anode curves, the anode is connected to a
low impedance signal voltage supply able to swing to perhaps + and - 600Vpeak.
Such a signal may come from a 50Hz 424Vrms transformer winding with one
and grounded and the other to the anode. The cathode is grounded, and grid has
a variable negative supply connected to give a range of between 0V and - 50Vdc.
There is a 50 ohm resistance connected between the 424Vrms winding and anode
to allow a second small transformer to transfer the Vac across the 50r to one trace
of an oscilloscope in X-Y mode so that the amount of current flow in the 50r anode
current sensing R is related to to the applied and varying anode voltage. A sample
of this voltage is applied second trace of the CRO. The curves may be seen as
those above for various values of grid bias voltage. Machines to automatically
draw such curves were developed many years ago for use in tube development
laboratories. So the curves are a way of depicting the dynamically changing anode
resistance, Ra, because the curves are simply describing a function of Ohm's Law,
where I = V / R.

For Q1 point, Pda = 31.5watts, one can calculate the 3 outcomes in Vg, Va, gain,
power and second harmonic distortion.

The grid -Vdc bias at Q1 = -18.5Vdc. If we look at the 1.75k load line, we can see
where it intersects the Ra slope at Ia = 270mA, and Ea = +40Vdc. The high slope
line close to Ia axis is the limit of linear anode voltage movement. Beam tetrodes
( and pentodes ) are quite queer because the Ra is initially very low and only 225
ohms for where Ea < +50Vdc, and for higher Ea voltages the Ra suddenly becomes
thousands of ohms. But the queerness is acceptable and it allows a good wide useful
Ea voltage swing.
At Ea = +40Vdc and Ia = 270mA, grid voltage at this point = -3Vdc. We have
changed grid voltage by +15.5Vdc. Load voltage has changed by -310Vpk. If we
change grid voltage -15.5Vdc to -33.5Vdc by applying a linear undistorted grid input
signal then we may read from the load line where the grid voltage = -33.5Vdc, and
it is where Ea = +495Vdc. Load voltage has changed by +145Vdc.

If the tube was perfect, and it surely is not, the changes in Ea and Ia would be equal
in both positive and negative directions. From the above we may calculate 2H distortion
as 2H %  = 100 x 0.5 x
difference of V swings / sum of V swings = 50 x 165 / 455 = 18%.
Output Load Vrms = Vpk-pk / 2.82 = 455 / 2.82 = 161Vrms and
PO = V squared / RL = 14.8Watts.
Class A efficiency % = 100 x PO / Pda = 100 x 14.8 / 31.5 = 47%.
This is a very poor THD outcome, but the 1.75k is a POOR LOAD MATCH, and
the lesson here is that we should try to get a better load match if we want better
sounding hi-fi!

So, with the Fig 2 Q point, we get......
1.75k :-  Vg = 11Vrms, Va = 156Vrms, gain = 14.2, power = 13.9 watts, 2H = 19%.
3.5k   :-  Vg = 6.7Vrms, Va = 189Vrms, gain = 28.2, power = 11.6 watts, 2H = 12%.
7.0k   :-  Vg =  4.7Vrms, Va = 216Vrms, gain = 45.9, power = 6.7 watts, 2H = 3.7%.

The "best" RL = 3.5k, and at 1 watt, at Va = 60Vrms, and THD is about 3%.

From these rather poor figures we can see that 2H is dreadful at low RL but reduces
as RL rises. In fact 2H will continue to reduce to 0% where there is a null in 2H
production and then the 2H will begin to increase as RL increases. But the relative
phase of 2H will be opposite to phase at low RL. The load lines will show that if
you care to interpret them correctly. The figures for 2H don't include most of the
other odd and even number harmonics which are significant and numerous. The
awful linearity of beam tetrodes and pentodes may be tamed with NFB, but first
we need to just consider loadlines.

Here are curves for GE6550 with screen voltage a lot lower = 200V :-
Fig 3.

Sometimes the beam or pentode tubes produce better curves so lower distortion
with screen voltage lower than anode voltage.
Here is a 3.5k load plotted on the above curves :-
Fig 4.

With a lower screen voltage = +200Vdc, The operating point can be Ea = 350V,
Ia = 90mA for Pda = 31.5 watts as in the case where Eg2 = 350V.  The load
is 3.5k.
The outcome at maximum power with no NFB :-

Vg =  6.7Vrms, Va = 185Vrms, gain = 27.6, power = 9.8watts, 2H = 8.3%.
At one watt, THD = 3%
If there was 20dB of global NFB we could expect THD at 9.8 watts to become
about 1% and 0.3% at 1 watt. Maximum power would be boosted to about 13
Watts because the anode V swing would be increased to about 600V pk to pk
from the 523V shown above. There is an increase in power output which is
proportional to the increase in V swing squared.

Use of a beam tetrode shows that the load that gives a healthy power output
also produces high THD even at low levels, because THD is approximately
proportional to output voltage. Where you have 9.8 Watts of output, Vo = 185Vrms,
and where you have 0.98 Watts Vo = 58Vrms. The change in THD will be from
8.3% to 2.62%.

I would never use any beam tetrode pentode in single ended mode unless I
could have ultralinear taps or cathode feedback windings on the output transformer,
or I just use plain triode connection.

Let us suppose the OPT ratio chosen was 5k : 6 ohms, or 833 : 1, so that if there
is a dip in speaker Z below the nominal 6 ohms to say 4.3 ohms, the load seen by
the anode = 3.5k as shown above. The high Ra of the tetrode = approximately 17k
ohms at the working point Q. The Ra of the tube measured at the secondary =
17,000 / 833 = 20.4 ohms = output resistance without NFB. The damping factor
= RL / Ra = 6 ohms / 20.4 ohms = 0.29, and atrocious, when what we want is
DF = 10 if possible.

Let us now explore the effect of 20 dB of global NFB on an SE tetrode amp.

Where 20dB of global NFB is used, the gain without NFB is reduced by a factor
= 0.1.

Consider the SE 6550 in tetrode in the above load line sample with 4.3 ohms
load on the OPT secondary. we have 6.7Vrms grid signal to make 9.8 watts into
4.3 ohms = 6.5vrms. Say have a 12AX7 used to drive the 6550 and it has a gain
of  say 67. Therefore 0.1Vrms is needed at the 12AX7 grid to produce 6.5 Vrms

Therefore gain without NFB, called Open Loop Gain, OLG, = 6.5 / 0.1 = 65.
Suppose we reduce the OLG by applying 0.9Vrms of NFB to the 12AX7 cathode
via a low resistance divider network from the OPT secondary speaker connection,
using a typical divider of R1 = 293 ohms plus R2 = 47ohms.

This will reduce the apparent over all gain with FB applied and is known as Closed
Loop Gain, CLG. CLG will become 6.5, 1/10 of the OLG of 65.

But let us use RDH4 terms, and OLG = gain A without NFB, and CLG = A' with
NFB applied.

ß, the fraction of the output fed back = R2 / ( R1 + R2 ) = 47 / 340 = 0.138.

The formula for working out A' where we know A and ß is :-

A'  = A / ( 1 + [ A x ß ] ).

Distortion with FB, Dn' = Dn /( 1 + [ A x ß ] ),
where Dn = distortion without FB.

A' here = 65 / ( 1 + [ 65 x 0.138 ] ) = 6.5.

The amount of NFB in dB due to FB is 20 x log (A / A') =
20 x log (65/6.5) = 20dB

We can say THD of 8.3% would reduce to 0.83%

But if there was no load on the tetrode output tube, open loop gain,
A = ( gain of 12AX7 x µ of 6550 ) / OPT turn ratio = 67 x 190 / 28.9 = 440.
If open loop gain was 440, and ß remained at 0.138, A', closed loop gain =
440 / ( 1 + [ 440 x 0.138 ] ) = 7.12. The amount of gain reduction is much
greater where A is much higher. The change in A' of between 6.5 and 7.12 is
not much, so tubes with varying gains will have much more even and constant gains
when NFB is used. So NFB with two channels of SE amps will keep the balance
constant and give us better stereo imaging.

With a load of 3.5k, the feedback resulting from ß = 0.138 gives 20dB of gain
reduction, but without any output load at all there is a gain reduction 440 to 7.12
= 61 times, = 36 dB of effectively applied NFB. This amount of NFB **will**
make the amp become unstable at both ends of the audio spectrum and is a reason
why steps must be taken with beam tetrodes and pentodes to use zobel networks to
reduce gain and phase shift in the open loop character below 20Hz and above 20kHz.

When 20dB of global NFB is applied with the 3.5k RL the open loop distortions
are also reduced by approximately 20dB. 2H, which is the main harmonic product
will be indeed reduced about 9 times to 0.92% at about 9.5 Watts and 0.3% at 1 Watt.
However, since the 2H is fed back there is a slight production of 3H by means of
intermodulation and there is greater spectral complexity of the harmonic content
after FB is applied then before, even though the level is much reduced.

The effect of intermodulation where only 10dB of NFB is applied around a tetrode
amp is all the more with an increase in 3H that could make the sound worse.
So when applying NFB, use enough where distortion is high to begin with as
is the case with a tetrode amplifier.

The calculation of output resistance is more difficult to calculate because we are
need to take into account the µ of the output tube.
To reference our calculations to the secondary connection,
µ of the output tube at the OPT secondary = µ / OPT turn ratio.
In this case it is 190 / 28.9 = 6.57.

The output resistance of a tube amp can also be calculated if you know
1, voltage gains of the input and driver stages,
2, µ of the output tubes,
3, turn ratio of primary to secondary which is the unloaded voltage ration between
P and S windings,
4, the anode resistance, Ra of one output tube at the Q point.
5, winding wire resistance of the primary and secondary windings of the OPT.

Rout of the SE amp with FB applied  =             Ra   +  Rw total
ZR x ( 1 + [ A" x {µ/TR} x ß ] )

Where Ra is for the one output tube,
Rw total is the sum of OPT primary winding wire resistance and ZR x secondary
winding wire resistance. Allow Rw total to be 10% of the rated primary load if the
Rw measured is unknown. TR is the turn ratio of the OPT, or unloaded P to S
signal voltage ratio at 1kHz, or square root of the exact known ZR. ZR is the
output transformer impedance ratio which is the turn ratio squared, A" is the
gain of the stages preceeding the output tube/s, ie, V at output tube grid / Vg-k
of the input tube, µ is the amplification factor of an output tube/s,
ß is the fraction of OPT secondary voltage fed back to be "in series"
with the input voltage to V1.

For example in this case we have :-
Ra-a = 17,000 ohms,
Rw = 10% of primary RL of 3.5k = 350 ohms,
TR =  28.86 : 1
ZR = 5,000 ohms : 6 ohms = 833 : 1,
A" = 67
µ = 190
ß = 50 / ( 50 + 600 ) = 0.138
Rout' = closed loop output resistance at the secondary output terminals

In this case Rout' =                          17,000 + 350                    =   0.34 ohms
833 x ( 1 + [ 67 x { 190 / 28.9 } x 0.138 ] )

This is a good result and could be compared to the result with the 6550
connected in triode where Ra = 900 ohms, µ = 7.33, and the OPT had the same ratio.
But for 20dB of applied NFB with 3.5k anode load, ß must be increased to 0.73
and with the same driver tube with its gain 67.....

Rout', triode, =
900 + 350                = 0.11 ohms, 1/4 of the tetrode case.

833 x  ( 1 + [ 67 x 7.33/28.9 x 0.73 ] )

We can conclude that about 17dB would be enough global NFB for the tetrode
amp to get Rout = about 0.5 ohms. The triode amp would need only 8 dB to
achieve the same low Rout.

Here is a typical power output vs load graph for 6550 in beam tetrode calculated
off the tube curves. When measured, the distortion isn't low unless a lot of NFB
is used, so the graph is a very rough guide to power output and valid only for the
Ea & Ia conditions.
Fig 5.

The power levels shown are without any NFB and at the anode with no OPT
losses, and could be approximately 20% more with about 20dB of NFB, to give
a net 10% increase allowing for an OPT with 10% winding losses. With NFB the
2H and other distortions are reduced to less than 2% and voltage swings on the

The OPT for the above Ea and Ia condition should have the lowest expected
speaker load matched to 3.5k, so if the speaker's nominal Z is 6 ohms, allow
for a dip in Z to say 4.3 ohms so the OPT impedance ration = 3.5k : 4.3 ohms
which is 5k to 6 ohms, ie, 833 : 1. The Rout before NFB is about 21 ohms and
with 20 dB of global NFB it will be approximately 0.33 ohms as calculated above.
A trioded 6550 will also give Rout = 2 ohms without any NFB but with 12 dB of
global NFB the Rout will be about 0.6 ohms.

Single Ended Ultralinear could be another option to use.
The data here is for a GE 6550A with 43% screen taps.
Fig 6.

RL for maximum power for SEUL or SE beam/pentodes = ( Ea / Ia ) - 280,
where 280 is a constant number of ohms which is based on the Ra limiting
line slope for Ra below the knee of the Ra curve for Eg1 = 0V.
The other traditional formula for SE pentodes or UL or CFB connected
tubes is RL = 0.9 x Ea / Ia.

One may wonder about the constant of 280 ohms. But look at the beam
tetrode curves above and the slope of the Ra lines to the far left side between 0V
and 50V have a 300mA current change, so the slope of the all the beam tetrode
lines below 50V = 50 / 0.3 = 166 ohms but I like to allow a little more, especially
for UL, because the UL curves are not so clear, hence the above formula and Ra
for beam tetrodes and pentode and for UL as seen above or for CFB need only
be 280ohms for OPT design. The actual E minimum swing in the above UL
graph is somewhat vaguely indicated between 0V and Ea = 75V.

If I were to start with an Ea = 400V and Ia = 80mA for Pda = 32 watts,
RL =  ( 400 / 0.08 ) - 300 = 5,000 - 300 = 4,700 ohms.
If the  operating  idle point is 350V x  90mA as in the above  graph, we
get RL = ( 350 / 0.09 ) - 300 = 3,588 ohms.

With 43% taps, the SE tetrode characteristics change to the following :-
Ra =  2.0k, µ = 15.4, gm =  7.7mA/V at Ea = 350V,  Ia = 90mA,
Pda = 31.5 watts.

The performance outcomes are :-

1.2k :-  Vg = 17Vrms, Va = 92Vrms, gain = 5.4, power = 7.1 watts, 2H = 15%.
2.0k :-  Vg = 17.7Vrms, Va = 131Vrms, gain = 7.4, power = 8.5 watts, 2H = 12%.
3.6k :-  Vg = 17.7vrms, Va = 170Vrms, gain = 9.6, power = 8 watts, 2H = 8.3%,
( 3.6k load line not shown )
4.7k  :-  Vg = 17.7Vrms, Va = 191Vrms, gain = 10.7, power = 7.8 watts, 2H = 5.5%.
11.7k :-  Vg  = 17.7Vrms, Va = 218Vrms, gain = 12.3, power = 4.0 watts, 2H = 1.7%
The "best" RL = 4.7k, and at 1 watt, at Va = 60Vrms, and THD is about 1.5%.
The use of global NFB increases the maximum power output up to 20% without
more severe THD.

The advantage with UL is that although the 2H remains fairly high, there is much
less other harmonics to worry about and the tube behaves much like triode but
without the limitation of the Eg = 0V limiting the Ea minimum as much.

Therefore an OPT would have a ratio of  say 5k : 6 ohms which has an impedance
ratio = 833 : 1. This allows for any dip in speaker load impedance to 4 ohms without
hugely increasing distortion. The UL Ra when measured at OPT secondary =
2,000 / 833 = 2.4 ohms. 12dB of NFB will reduce this further to about 0.8 ohm
and nearer to a triode with the same global NFB and OPT ratio.

If we plotted a load of 4,700 ohms through the Quiescent point Ea = 400V
and Ia = 80 mA, we would get a swing of about 0.9 x Ea = 360V peak giving 13.7
Watts maximum in theory. In practice we may get 13 watts when pushed into low
THD by  NFB application. If you plot low loads less than 2k the phase of the output
2H is like triode, and 2H diminishes to zero at some load above 2k then 2H increases
as load increases further but phase of 2H is opposite. The SEUL connected tube
has only slightly less 2H than triode or about 1% when power output is 1/10 the
maximum.

The use of distributed loading in the OPT with 20% of the primary turns in a cathode
feedback winding and with the screens taken to a fixed voltage of about +270V will
give much better linearity than triode or UL screen taps. Such an arrangement has
voltage NFB applied relationships between grid and cathode and between screen and
cathode. 20% CFB with fixed screen supply voltage Eg2, is the equivalent of having
20% UL taps but with about 10dB of series voltage NFB between cathode and grid
when a 4.7k load is used. Thus THD < 2% THD at 9watts. This oversimplifies what
is occurring in the tube, because there are two paths of NFB, one via the screen to
cathode interface and the other via the grid cathode interface. There is still some
odd numbered harmonic products in the THD but they are much reduced below the
levels occurring with pure beam tetrode. The results using EH6550 ( or EH6CA7 )
in my SE 35 Watt amps with CFB indicated that more could be achieved with CFB
than just UL screen taps and still have gain about equal to triodes. The SE 35 Watt
amps have remarkably low THD and low Rout and with only very minimal global
NFB. See my page on SE35 amps.

To Educational and DIY