More thoughts on 80Watt AB1 amps. 2014.

This page is about :-
Need for high quality driver stages for output stages using CFB.
Fig 1. Complete amp schematic 6CG7, 2xEL84, 4xEL34.
Fig 2. B+ Delay, Active Protection, Bias monitoring with LEDs.
Graph 1. Loadlines for one EL34 in PP amp, Ea = 400Vdc,Ia = 40mAdc.
Fig 3. OPT details for 75Watt OPT for 4k3 : 4r0 and 7r1.

Why 80W amps?
The 80Watts is a nominal maximum output power very easily and cheaply
achieved using a quad of octal tubes such as EL34, 6CA7, 6L6GC, KT66, 807, 5881.
One pair of EL34 can make 40Watts, so a quad can make 80Watts in class AB1
pentode mode with fixed screen Eg2, fixed grid bias EG1, Ea = +400V, idle Ia = 40mAdc,
idle Pda = 16Watts.

The power wanted for good domestic hi-fi will never be 80Watts and will usually
average only 1Watt or less from a single amp channel. Peaks in music might reach
30Watts and we want the first 15Watts to have very low THD and IMD so this power
should ideally be all class A1 power.

To obtain 80Watts in AB1 requires a low RLa-a loading better suited to PA amps or
guitar amps where distortion does not matter. Such loading yields very little initial
class A1 power and most power is class B1.

If we insist that all amp power must be only class A1 then amp efficiency cannot be
more than about 45%, which means that if we use 4 x EL34 with a total Pda at idle
of say 80Watts, then maximum Po is 36Watts.

And this Po can only occur for ONE load value with load ohms above or below
giving less power. My SE35 is a classic example of a pure single ended class A amp.

The same 4 tubes could be used to make a PP class A amp giving same 36Watts
maximum of class A at only ONE load value, but it could make more power when
load ohms are reduced but give less power when load ohms are increased.
The increased power with lower load will be class AB1.

The PP amp working in class AB will produce an initial amount of pure class A1
before tubes begin working in class AB. The power between end of class A
operation and clipping is class B power, only one of the pair of PP tubes produces
audio power.

As long as the initial amount of class A1 power is enough to cover most of what
everyone listens to when using an average of 1Watt, then 15Watts is usually than
enough class A and higher musical transients such as drumbeats can be well
reproduced by class AB action.

To get adequate class A1 power and good total AB1 power ceiling, the idle Ea can
be higher, Ia lower, Pda lower for the PP AB amp. In this case, total Pda at idle for
4 x EL34 = 64Watts. Maximum possible class A is about 30Watts but only if RLa-a
for the quad = 8k0. RLa-a may be reduced for class AB1 action to about 4k0, and
about 13Watts of pure class A is possible with about 62Watts total class AB1.

It is possible to get up to 120Watts in class AB1 with RLa-a below 1k4, but the THD
will be 6 times greater and tubes may easily overheat, and Eg2 should be +400V.

With 4k0 load for 4 x EL34, and with continuous sine wave at clipping the EL34 will not
overheat, and Eg2 may be +300V. Other tubes such as 6CA7, KT66, 6L6GC, 807, 5881,
may all be used with same idle conditions for EL34, although Eg1 may need slight
change to obtain the same Iadc.

6550, KT88, KT90, KT120 all can be used but Eg1 will need to be much more
negative than for EL34 for the same Iadc. These tubes have higher Pda
ratings and can idle with higher Ia of say 55mA to obtain more pure class A.

More class AB power is also available. See 8585 two channel integrated.

I believe the best output stage for a quad of PP EL34 has a fixed Eg2 and CFB
windings. It is known as the Acoustical connection, and was made famous by QUAD,
in the Quad-II monobloc amps. I have used it in my 8585 amp.

The CFB winding should be between 12.5% and 25% of total primary turns.
This always means the maximum drive voltage needed for the output tubes must
be between 50Vrms and 75Vrms, even higher than pure triode, but still low enough
to achieve very low THD if the driver stage uses medium µ triodes such as EL84
strapped in triode. The driver stage should not use pentodes in pentode mode.
Other suitables can be EL86 in triode or paralleled 6SN7, 6BL7, 6CG7, 12BH7
and ECC99. EL84 triodes give gain = 18, EL86 gives 11, and both have Ra < 2k5,
so Rout anode to anode of an LTP driver stage  is less than 5k0.

At another page about use of 4 x EL34 for Mr Zel, I show an input stage using 6DJ8
and a 6BL7 driver, both set up as a fully balanced differential voltage amp for following
EL34 output tubes with 25% CFB. The operation of 6BL7 was optimised with
“bootstrapping” the R supplying Idc between 6BL7 anodes and taps on OPT anode
windings.

However, the best optimisation for balanced amp or LTP amp for driver is done
using a CT choke and R between ends of choke and the two driver anodes.

Fig 1, 75Watt complete AB1 amp.
80W-PP-amp-6CG7-EL84-4xEL34-15August2014.gif
80W-PP-amp-6CG7-EL84-4xEL34-15August2014.gif

In Fig I have 4 x EL34 used with idle Ea = +400V, Ia = 40mAdc, Pda = 16Watts.
Fixed bias is about –29Vdc, Eg2 = fixed Vdc = +300Vdc.

Input V1a & V1b is 6CG7 set up as an LTP differential amp with commoned cathodes
connected to constant current sink using MJE340.

Input signal feeds V1a grid and GNFB feeds V1b grid. Differential gain about
16 and the stage has good common mode rejection, and much less THD than
any SET for V1 used as V1.

Driver tubes V2 and V3 are a pair of EL84 in triode and in a balanced amp.
C9&10 and R18&R19 give local dc current FB to best regulate the Ea of each EL84.
The cathode currents of both flow in “tail” R17 3k3 to –100Vdc rail. The “long” R17
value ensures good balance of the opposite phased Va from each anode.

I show the OPT with 12.5% CFB, and signal Vrms shown are at clipping with RLa-a
For quad of tubes = 4k3. Maximum drive Vac to each EL34 grid is about 50Vrms.

The gain of EL84 = 18 approx, so the Vg-g to EL84 must be 5.6Vrms, so V1 6CG7
needs Vg-g = 0.35Vrms, so Vin with 11dB GNFB is 1.47Vrms.

The amount of CFB can be easily increased to 25% by using OPT primary layers
slightly re-arranged. Then the Vg to each EL34 may become about 75Vrms, needing
Vg-g at EL84 = 8.4Vrms, and Vg-g at 6CG7 = 0.53Vrms, and if GNFB network R
values are left unchanged, Vin with GNFB = 1.58Vrms, so amount of GNFB = 9dB,
and "enough".

The L1 choke acts to increase the load bringing Idc to EL84 anodes so that the
anode load is dominated by 2 x 100k Rg of  EL34. So loading on each EL84 anode
is about 50k. 100k is a low value for Rg but this stops an excessive positive Vdc
across 100k when tubes age when a tiny input grid current begins to flow. There is
no bootstrapping of anode loads for EL84 with L1 so there is no mild application
of positive FB.

L1 choke with CT has no air gap and provides a high L reactance in the anode
dc feed circuits. It is in fact what is called “choke feed”, but is in balanced form in
this and other designs at this website. Typical properties of such a choke have
L = 820H at 50Hz, and µ = 5,000, and shunt C = 200pF. I show 3k3 in series with
each anode and ends of choke winding. Ra of EL84 is about 2k2, and 1/2 Rw = 250r.
Loading of each EL84 is mainly the cap coupled Bias R for EL34 = 50k.
But at very low F and high F the XL and XC decline towards 0.0 ohms, and lowest
load on EL84 is 3k3 + 250r = 3k5 approx and gain = 12.2. But this drop in gain is
only -3dB at below 1Hz and above 227kHz. The effects of phase shift caused by
XL and XC are avoided. At 1kHz, XL and XL are extremely high, and load is
dominated by 50k and gain = 19 approx. The loading of the EL84 by the choke
becomes quite negligible for the audio band. You can expect THD at 90Vrms to
be < 0.2%.

Choke details for L1 can vary considerably.
For my 300W amps I used GOSS E&I lams with T = 25mm, S = 32mm.
Lams were maximally interleaved with no air gap, iron µ max = 17,000.
Plastic bobbin has centre divider and 2,500t of 0.2mm Cu dia wire random wound
with slow traversing speed on each 1/2 bobbin, with CT brought out between the
two windings.

Max Idc allowed for each winding = 90mA. Low Rw is not required.

At low signal levels the iron µ = 5,000 assumed. Thus L at 50Hz = 820H, and
reactance XL = 256k. This rises at higher F to megohms. Shunt C or self
capacitance was about 200pF, and XC equals 100k at 8kHz. Because Ra =
2k2, phase shift is avoided. For my 8585 amp I used same 25T E&I lams but
only 10mm stack, and 0.125Cu dia wire which gave total turns = 8,000.
These chokes were then easily fitted under the chassis. Max Idc allowed each
winding = 37mAdc. With µ = 5,000, at 50Hz, L =  916H, ZL at 50Hz = 287k.
This choke with 8,000 turns can be carefully random wound with lathe turning
at 5 turns per second and wire fed on from one side of bobbin to other every
10 seconds, and while lingering at ditches in height to keep windings level
as bobbin is filled. A divided bobbin makes everything easiest. Just fill one
side with 4,000t and then fill the other. A turn counter is essential.

A series R at each end of winding MUST be used in case EL84 develop a short
between anode and 0V to limit high current in choke winding which would quickly
fuse the winding. The R value used must limit maximum Idc to less than Cu wire
rating based on 3A/sq.mm of wire section area.

Needless to say, this most wonderful sounding arrangement was never adopted
by any manufacturer because they always settle for a cheap solution.
Details of OPT are given below.

Fig 2. Delayed B+, active protection, bias condition indication.
schem-delay+protect+biasLED-4xEL34+cfb.GIF
Fig 2 has several splendid functions to ensure a tube amp will not cause expensive
repairs when something malfunctions or an owner connects a shorted speaker or
speaker cable.

Inrush current delay.
The inrush current at turn on is limited by R21 in mains neutral line to large PT.
This resistance is shunted by relay 1 after about 4 seconds, and when B+ has reached
about 2/3 its maximum level and input current has reduced. When relay 1 shunts 100r,
 the input peak current input increases slightly but is no more than limited initial peak
current at turn on. Without this delay circuit, a fuse of twice the current value is needed.
Q1&Q2 are a darlington pair of PN100 with high base input resistance.

R1 & C1 have a time constant of 7 seconds, and at about 4 seconds after turn on,
the Vdc across C2 has risen high enough to get Idc flow through 6V8 zener diode to
Q1 base, which turns on quickly enough to turn on relay 1 with a nice click heard after
turn on. R21 100r is in series with mains neutral line to the large PT primary. The mains
fuse really only protects against catastrophic faults in the amp PSU, such as electrolytic
caps or silicon diodes failing to become a short circuits, or some winding on PT shorting
to 0V or chassis, or there is a prolonged excessive rise of mains voltage due to fault in
the town supply.

If for any reason the amp is turned off, then back on again in less than 4 seconds, the
delay circuit will again work to limit inrush current thus avoiding excessive currents which
are all the greater when tubes are still hot from being on.

Bias setting without a voltmeter. There are 4 output tubes in each channel of this type
of amp. The best way to make any power amp with 4 x EL34 or other octal output tubes
is in monobloc form, complete with its own PSU and protection circuits. Monobloc use
keeps weight to less than about 20kg per channel.

Each channel has its grid voltages adjusted by 10k pots seen in Fig 1, VR1 to VR4.
These are 3W rated wire wound linear pots. The Fig 2 schematic is can be used to to
adjust the bias pots without a voltmeter, and to be guided by watching a pair of LEDs
for each EL34 while turning the pot shaft located near LEDs. After adjustment, the grid
bias voltage Vdc stays "fixed" until time comes for any re-adjustment.
All output tubes cannot ever be perfectly matched, and will never stay matched after use,
so individual bias adjustments are needed where there is no "auto-bias" which better
suits class A amps. The yellow and red LEDs should become equally bright when bias
is set correctly. This is easy to see because a small amount of turn of each pot will
make one of the LED glow brightly, and the other turn off.

Bias pots are best installed under the chassis facing upwards on sub-plate with 6.3mm
dia shaft protruding through chassis top by 3mm and with a sawn slot on shaft allowing
screwdriver adjustment, or use of kitchen knife by non technical minded owner.
The LEDs should be 5mm dia, facing upwards and nearly flush with chassis top, and
both pot shaft and 2 LEDs mounted close to each EL34. The arrangement means there
is no need to remove the amp from its equipment stand, no need to remove a cover
or use a voltmeter, and no danger of amp damage and no risk of electrocution while
probing around the circuitry which has many points at +400Vdc potential.
There is less likelihood of an owner becoming confused & electrocuted by what he
is doing, or being interrupted by phone call, when his 3yo son wanders over to the
amp.

The amp must be turned off before removing or replacing any tubes. Bias adjustments
must be done with amp turned on but without any signal present. Speakers need not
be connected if the amp is not likely to oscillate. When plugging in a new set of tubes,
all pots should be turned to mid position before turning amp on. If one tube is to be
replaced, turn off amp, wait 2 minutes, plug in new tube with bias set to mid position.
After turning on, the replacement tube bias is adjusted and the LED brightness
made about the same. The rest of the tubes should remain biased correctly if they
were before you replaced the tube. When biasing a new set of tubes, or one tube,
DO NOT allow yourself to be interrupted because some tubes may run too hot with
the incorrect bias setting, leading to possible damage to the tube. However, if the
active protection circuit is used this won't happen, but having the amp switch off
and needing re-setting during the biasing procedure is a PITA. The adjustments
need to be repeated as tubes warm up until all red & yellow LEDs have equal
brightness after amp has been on for 15 minutes.

Bias monitoring.
Just after turn on all red LED should remain unlit, with all yellow being bright.
This indicates tube current is low, because tubes have not warmed up. But as current
begins to flow after about 15 seconds, red and yellow LED will flicker and turn on/off
until tube current settles down and LED should all be lit with about equal brightness.
If a yellow LED is unlit, the red LED should glow brighter, and this means Ia is a little
too high, and if the red LED is unlit, the yellow LED should glow brighter. Bias pot
should be adjusted to equalize LED brightness. If the bias pot cannot be adjusted to
any position to give equal brightness for LED, and only red LED is lit, or only yellow
is lit, the nearby tube is faulty and MUST be replaced. But when many tubes fail
randomly and early in life, or after 5,000 hours, they often conduct enough excessive
Idc that the amp protection is triggered. Attempts to reset the amp by turning off, then
back on will fail to fix the problem. But after turning off, and allowing the amp to cool,
then turning back on, AND while watching which tube is first to have its red LED turn
on brightly, you should be able to work out which tube is faulty, without trial and error
by replacing each tube in turn with a new replacement.

If you don't like lots of LED, then the use of a voltmeter and test points is next best thing.
 you will need to mount accessible test points to measure the Ek of each EL34 at top
of C8,9,10,11 in Fig 2. The test points should 2mm chassis mount sockets with 4 red
for the four Ek Vdc, and one black socket connected to 0V rail. The meter probes
may be plugged in while probing each of the four Ek while turning the pot shaft to
get the correct Vdc. If you don't want LED to set bias and monitor Ek, then you will
not need the 4 differential amps to drive LED, but you will need the the protection
circuit because nothing will tell you that a tube is conducting too much current or not
enough.

About tube heat.
It is only natural that amp owners often ignore everything I am saying here, and never
watch the bias condition of their tubes via the LED, or measure any test point voltages.
All tube amps produce heat while doing nothing at idle. For each EL34, heater filament
power = 6.3V x 1.6A = 10 Watts. With each EL34 correctly biased, Ia = 40mAdc and
heat generated by anode = 0.04A x 400V = 16Watts.
The screen dissipates heat = 300V x 3mAdc = 0.9Watts. So Pda and Pdg2 heat
liberated from EL34 is 16.9Watts. With heaters, total heat = 26.9Watts. The filament
power is a fixed and unavoidable 10Watts. The combined anode and screen heat must
not exceed the EL34 ratings for maximum Pda + Pdg2 which is 28Watts. The heat is
called "anode and screen dissipation." Some EL34 made over last 20 years have
Pda+g2 ratings lower, and some higher. 6CA7 were made to be plug a plug in
replacements for EL34, and have same data as EL34, although most 6CA7 many
samples withstand higher Pda + Pg2 than original designs of EL34. With EL34
Pda+Pdg2 = 17Watts in the amp here, the EL34 are very happy and temperature is
tolerable for maybe 20 years. If an EL34 conducts say 75mAdc, its Pd+g2 = 33.75 Watts,
and the anode will begin to glow red and tube is too hot and and tube life is threatened,
and music turns to mud. We wish then that the amp automatically turn itself off.

LED Bias monitoring.
In Fig 1, each EL34 has a 22r between cathode and the 1/2 CFB winding which
probably has Rw < 4r0. The correct Ikdc for each EL34 is the sum of Ia and Ig2 =
43mAdc. The Vdc across each 1/2 CFB winding should be about 0.35Vdc.
There should be 0.95Vdc across each 22r in EL34 cathode circuit. The Ek - 0V
should be approximately 1.25Vdc. In Fig 2, consider K5 cathode point which is
at +1.25Vdc. This Vdc is applied to a low pass filter formed with R7 3k9 and C8
470uF. This filter removes the high signal Vac which occurs at the cathodes which
can reach 70Vrms at clipping. During normal operation we only want to monitor the
 Ikdc. Signal Vac below 1.0Hz will have very low amplitude because the amp
response LF pole may be at 7Hz and there is very little music signal at 1Hz.
The R7&C8 offers -20 db attenuation at 1 Hz, and -51dB at 30Hz, so the normal
signal operation does not affect operation of bias monitoring or protection circuits.

The Vdc at K5 causes some Idc flow in R7 3k9 and to the inputs to a diode and bjt
base input to Q11 buffer and to each base of each bjt differential amp. The current in
R7 3k9 is less than 0.05mA if the hfe of bjts is more than 100. Generic bjts like PN100
costing 10c each are OK to use. So the RC filter after K5 drives high input impedance
of following devices so it is easier to calculate and predict Vdc at all points in Fig 2
as I have indicated on the schematic.

Each bjt LTP amp has Vdc input from cathodes on right side bjt base. Each bjt base
on left side of LTP is kept at a reference voltage of +1.25Vdc. A small Vdc change
at any one or more cathodes is enough to change Idc flow in LEDs.

Automatic turn off with excessive Ikdc. Suppose K5 Vdc rises to +2.2Vdc because
Ia+g2 current in V5 ( Fig 1 ) has increased to 75mAdc. This rise of Vdc is
transferred to Q12 SCR gate to give +0.68Vdc which is enough to turn on SCR
 which turns the whole amp off. Before the amp is turned off, the rise in Ek also
causes a red LED to be turned fully on with 8mAdc and yellow is turned off.
So there is some indication something is wrong with tubes before automatic
turn off. The connection of a shorted speaker cable or speaker can cause rapid auto
switch off when the owner tries to turn up the volume with music. The turn off avoids
tube damage. After auto turn off, all yellow LED will remain alight. The blue “on”
LED turns off, and red “fault” LED turns on.

The differential amps in Fig 2 have their gain reduced with 22r emitter resistors.
These R should give less sudden changes of brightness of LED when setting
bias or when monitoring brightness after bias is set. Grid bias voltage should
not need to be adjusted more than once each 3 months. The correct range of
Idc is between say 38mA and 48mA. The LED will easily tell an owner if tube
bias is wrong by more than +/- 5%. Neither of these very slightly incorrect
conditions will lead to catastrophe, or poorer music quality.

The differential amps require +12Vdc and -12Vdc rails and I have shown these
produced by small 8VA mains transformer with 15Vac sec giving +/- 19Vdc
which are applied to 7812 and 7912 regulator chips which are easier to
arrange than having RC filtering +/- shunt regulation with 12Vdc zener diodes.
The -12Vdc rail is VERY important because it provides a near constant Idc to
emitters of bjts in diff amps. The emitter resistance of Q12 buffer supplies
1.86mA to Q11 emitter follower buffer. At turn on, the gate of SCR should
initially be negative, and slowly rise to -0.18Vdc depending on Ek of tubes.

The SCR gate has a LPF using R24 4k7 and C14 1uF. If the amp is turned off
automatically all power circuits are turned off except the 8VA aux trans and its PSU
and the devices it powers. The SCR gate voltage will subside to less than 0V when
 Ek reduce to 0V within seconds but SCR remains turned on and keeps Relay 2
open and main large PT turned off. This can be discomforting to an owner, but
may save him much expense later. The amp may be reset switching amp off then
back on after 3 seconds. When the amp is turned off at switch, power to auxiliary
PT is turned off. Therefore the +12Vdc rail falls rapidly to 0V because of current
through relay 2 and Q12 SCR quickly draining +12Vdc rail caps. When anode
of SCR is below +0.8Vdc, it then turns off. The time is usually 1second for this
to occur. So the amp can be turned back on and it will try to operate normally
unless the cause of excessive Ek of one or more EL34 remains.

Graph 1. Load line analysis for EL34.
EL34-6CA7-4load-Eg1=0V-Eg2=+250to+400V.GIF
Graph 1 shows basic load lines for one EL34. The Ra curves for EL34 do not have
multiple Ra curves for many values of Eg1 as shown on old data sheets. Such old
data sheets have so many lines they confuse many who use them. There is only
ONE Ra curve you need to know about and its for Eg1 = 0V and for a specific
 value of Eg2. I have FOUR Ra curves here, for 4 different Eg2 between +250Vdc
and +400Vdc.

I have drawn 6 loadlines for various class A and B RLa, all using idle condition
of Ea = +400Vdc, Ia = 40mAdc, and from there we may calculate expected
performance for 4 different Eg2 values, +250V to +400V.

The colored load lines represent loads :-
Dark blue line D-C = B RLa  = 666r. Dark blue line E-Q-D = A RLa = 1k33.
These 2 lines are used for RLa-a = 2k66. I show C on Eg2 = 350V curve
and theoretical max Po = 58Watts AB1. The theoretical Po assumes Ea and

Eg2 should theoretically remain well regulated but in practice both may drop 20%
with increasing Ia during AB operation, tube samples may have knee of curve
further to right, and winding losses in OPT may be 10% so at a secondary of a real
amp you might only see 42Watts. Use of UL or CFB can also slightly move Ra knee
right.

Brown line is for B RLa = 1k0 for RLa-a = 4k0,
Magenta line is for B RLa = 1k33, for RLa-a = 5k33,
Crimson line is for B RLa = 2k0, for RLa-a = 8k0.

Black straight line passing through Q is class A RLa = 9k3. This line shows the
class A load for maximum possible pure class A where max Ia change = +/- 40mApk.
For 2 x EL34, RLa-a = 18k6, and there is no AB operation, and Po = 14.9Watts, with
each EL34 producing 7.45Watts.

To allow maximum possible Ea swing with all loads the Eg2 would need to be +400V.
But analysis reveals using RLa-a = 2k66 gives high THD and tubes will overheat easily.
So let us never ever require the amp to give maximum possible AB1 Po with 2k66.

Consider Brown line for B RLa = 1k0 for RLa-a = 4k0. The ideal Eg2 would be +350V
because the Ra knee starts above Ia max at 340mApk. Class AB1 Po max = 57Watts,
class A Po = 3.2Watts for 2 EL34. Using lower Eg2, clipping would occur at a lower
Ea peak swing and lower Po. The Ra curve prevents voltages extending to the left
of the Ra line or aka the “diode line”. This is because g1 grid draws grid current and
Rin to g1 becomes less than say 2k0 and the driver amp output clips. EL34 and 6CA7
 do not like class AB2 operating conditions with grids being forced to go positive.
But 6L6GC do not mind, and with Eg2 = +300V, and Ea = +600V, some 80 Watts
is possible in AB2, but its unreliable and THD = 13%.

Consider Magenta line for B RLa = 1k33 for RLa-a = 5k33. Eg2 could be as low as
+300V. Class AB1 max = 45Watts, A1 = 4.2 Watts.

Consider Crimson line for B RLa = 2k0 for RLa-a = 8k0. Eg2 could be +250V but
better would be +300V, and class AB max = 33.2Watts. Class A = 6.4Watts.
The Crimson line represents the best RLa line for an EL34 for a class AB amp IMHO.
With a quad of EL34, RLa-a for the quad is 1/2 the RLa-a used for a pair of EL34,
and for the quad of tubes RLa-a could be 4k0. Maximum class AB will be 66Watts
and class A max = 12.8Watts. Because of low increases of Iadc and Ig2 between first
12Watts and 66Watts max, it is easier to rely on large B+ rail caps to keep rail Vdc
close to constant during musical transients.

There is another alternative configuration which some will say always sounds better.
The use of 12.5% CFB is retained, but instead of having a fixed Eg2 at say +350Vdc,
screens on each side of PP circuit are connected to UL taps on anode windings for
+/-64Vac. See connections 4 and 15 on Fig 1. Series 270r “screen stoppers” are
retained. The EL34 then work with the same gain as for plain UL with 37% screen
taps. The Vac applied to screens reduces all THD spectra to be similar to triode,
with most reduction of odd number H. The CFB still is very effective in reducing THD
and Rout, but overall the combined screen FB and CFB gives a better overall outcome
 with little change to Vg needed at EL34 grids. It is not universally true for other tubes
such as 6550 which I found gave less THD with fixed Eg2 = +330V and Ea = +480V.
 So merely having Eg2 lower than Ea does a lot to reduce THD. Feel free to make
your own experiments.

Fig 3. Output transformer details.
75W-PP-bobbin-CFB-4k3-4r-7r-15August2014.GIF
Here is a good recipe for a PP OPT meant for 4 x EL34 or other octal tubes.
The winding layers show the primary has 16 layers in 5 sections with 18 connections.
The numbers give the exact sequence of windings from one layer to next and arrows
show the direction of traverse width of winding wire across bobbin. Terminals all
should be turret type at least 10mm part and in rows along each side of bobbin
winding. Avoid having any one primary wire less than 5mm away from any other
wire to prevent arcing if these wires have high peak Vac or Vdc difference.
The primary has 12.5% Cathode Feedback winding with CT at 8 & 10 and cathodes
to 9 & 10. The anode windings have CT at 7 & 12, and anodes connect to 1 & 18.
Should 25% CFB be used, ALL the central primary section may be used.
Connections along anode windings may be used for Ultralinear screen taps.

The OPT has Np = 2,368t for RLa-a = 4k3, and sec = 4 parallel 72t windings for 4r0,
3 parallel 96t windings for 7r1. There could be 2 parallel 144t windings for 16r0,
which probably will never be needed. Fig 1 schematic shows two ways strap sec
windings to give 4k3 : 4r0 or 7r1. This load matching variability should suit 95% of
hi-fi listeners because 95% of all speakers made are between 3r0 and 10r0.
Those that are 16r0 are likely to be old and / or sensitive types which may be
connected to 7r1 and the 40Watts available will have lots of class A and sound
wonderful. Use of 8r0 speakers with 4r0 strapping will also give nearly 40Watts of
mainly pure class A. This OPT design is similar to others I have made for amps
producing 100Watts. Some previous designs have specified E&I wasteless
GOSS 44T x 62S x 66L x 22H. This is often hard to get, and this design above
uses the more common size of E&I with 50T x 50S x 75L x 25H.

The other possible E&I could be 38T x 100S x 57L x 19H. Afe = 3800, and Np may
be only 1,500t for same Fsat as 44T x 62S. You may find it a struggle to get all
the wanted turns in the small window, and keep winding loss % low.

Happy soldering,

Back to Power amp directory

Back to Index