Calculating and winding a Hi-FI push-

pull transformer for EL 84

From the book by Klaus-Peter

Hoffmann, "The HIFI tube amplifier",
published by Pflaum Verlag, Munich
1994, page 34ff
ISBN 3-7905-0691-5

A low-frequency amplifier can be built without great difficulty in such a way that it
transmits a large frequency range linearly, i.e. without phase rotation, with an
extremely small distortion factor. Bringing this frequency range via the output
transmitter is much more difficult.
A transformer that is supposed to transmit a frequency range from 20 Hz to 20
kHz in a straight line, for example, requires multiple nested windings and a
correspondingly large iron cross-section. If you want to keep the distortion level
small even at the lowest frequencies by means of negative feedback, or if you
want to make it practically inaudible, the transmitter frequency response must be
better than that of the amplifier connected upstream. This is the only way to
make the negative feedback independent of the frequency.
The industry usually avoided the high manufacturing costs of such a transformer
by using simpler transformers and compensating for their shortcomings with
complicated, frequency-dependent negative feedback networks. Their calculation
is not only very difficult, one also usually needs measuring devices that a private
person hardly has at their disposal. The reproduction of proven industrial circuits
is only successful if it is an extremely exact copy of the device in question.
It is therefore much safer to dimension the transformer in such a way that both
the required power and the desired frequency range can be transmitted with
certainty. This is easier to achieve with triodes in the output stage than with
pentodes, but the material and electricity costs are higher for the same output.
The transformer has the same size for triodes and pentodes with the same power,
only it requires less work for triodes.
In the following, a push-pull transformer for two pentodes EL 84 is calculated and
designed as an example. With a corresponding change in the number of turns and
the wire diameter, such a transformer can also be made for triodes.

The calculation

Roughly, one can calculate that with a lower limit frequency of 20 Hz an LF

transformer can still transmit about 1/20 of the power of a mains transformer of
the same size for 50 Hz without the distortion factor becoming inadmissibly high.
Based on this size, the required iron cross-section should first be determined (see
Telefunken laboratory book, Volume 1, Franzis-Verlag):

(N = power in W, fu = lower limit frequency)

The cross-section of 10 cm2 resulting from this calculation corresponds to the
closest standard core M102a made of Dyn. Sheet IV, if possible 0.2 ... 0.35 mm
thick, layered without an air gap.
The easiest way to calculate the number of turns required for the primary winding
is from the voltage that occurs at full power. The voltage results from the power
and the matching resistance as follows:

for two EL 84 tubes (prescribed matching resistance 8 kΩ)

and the number of turns required for this Npr=

For frequencies other than 20 Hz, the factor 115 must be changed accordingly,
inversely proportional to the frequency, i.e. at z. B. 40 Hz to half, at 10 Hz to
double.
The number of turns calculated in this way is absolutely sufficient to bring the
desired lower limit frequency over the transformer without phase rotation. The
upstream amplifier can then be designed for this lower limit frequency without
hesitation, provided that the voltage in the amplifier has dropped to a factor of
0.7 (compared to 1000 Hz) at the stated frequency if the factor 1 was used as the
basis for 1000 Hz.
If the transformer is manufactured according to these instructions, the upper limit
frequency is around 60 kHz. If the amplifier is designed so that its upper limit
frequency is 30 kHz (drop to 0.7), there will definitely be no phase shifts in the
negative feedback path.

Practical structure

In order to achieve this high upper limit frequency in the transformer, the primary
and secondary windings have to be interleaved several times; In AB mode, the
two primary winding halves also have to be nested below one another (B mode is
ruled out for hi-fi quality). In addition, the two primary winding halves must be
meticulously symmetrical, i.e. the same wire length, the same capacitance against
the secondary winding and also against the iron core. Therefore, the coil body is
to be divided into two exactly equal chambers with a central flange. If one of the
chambers is wound clockwise and the other in the opposite direction, the
requirements for the same wire length and the same capacity can be met. In
addition, if the primary and secondary partial windings are interleaved in such a
way that a secondary partial winding comes between each primary partial
winding for tube 1 and for tube 2, the requirement for interleaving the primary
windings is also met.
The primary winding is expediently divided into ten sections and the secondary
number of turns resulting from the transmission ratio into eight partial windings.
From the calculated number of turns, z. B. 5 ohm loudspeaker with 8 kOhm
matching resistor

Nsec = Npr: √Raa / Rload

The calculated number of turns cannot be precisely adhered to, but that is not a
cause for concern if only the ratio of the primary to the secondary number of
turns is retained. In the case of pentodes, a change in the transformation ratio
leads to an increased harmonic distortion; in the case of triodes, it only leads to a
change in power.
The total winding should take up a maximum of 80% (including paper interleaves)
of the available winding height; the wire diameter, which must correspond to the
currents occurring in each case, is a lower limit. A current density of 3 A per mm2
is permissible with the provided M102a iron core. Nevertheless, it is cheaper to
use thicker wire in order to keep the losses as small as possible. A wire diameter
of 0.25 mm for the primary and 0.65 mm for the secondary winding comes closest
to the demands made, each primary part winding then takes up just three full
layers, each secondary part winding two full layers.

After the coil body has been divided into two chambers of exactly the same width
by inserting a central flange, one chamber is first filled with two suitable wooden
blocks, which prevent the intermediate flange from slipping during winding. The
body is clamped on a mandrel and wrapped as follows:

First, three layers of 0.06 mm lacquer paper, feathered on both sides, are applied
to the body, then the first primary part winding follows, starting at the center
flange (see picture).
The beginning and end of the winding are covered with a suitable insulating tube
and led out through the holes in the outer flange. The beginnings and ends of
each winding are expediently numbered consecutively, i.e. 1 for the beginning, 2
for the end.
Between each layer of each winding there is once a feathered varnish paper 0.06
mm. The beginnings and ends of each winding are attached with adhesive film.
After the first winding of the primary section, three layers of lacquer paper must
be insulated, followed by the first winding of the secondary section. The
beginning is marked with 3, the end with 4. This is how the following windings are
identified consecutively.
Since the calculated 81 turns of 0.65 mm wire occupy just two layers, you now
start with the outer flange (don't forget the insulation between the layers!). The
secondary part winding is followed by three layers of lacquer paper and then the
second primary part winding.
In this way, a total of five primary and four secondary windings are continuously
applied; at the end, insulation is carried out twice with 0.1 mm insulating foil.
Now you pull the body from the winding mandrel, remove the sticks from the
unwound chamber and clamp everything again so that the empty chamber comes
to lie on the same side where the wound chamber was previously. This achieves
the necessary opposite winding direction.
The winding device is now rotated in the same direction as before, starting with
connection 19 for the beginning and ending with 20 for the end.
Once all partial windings have been made and the protective insulation has been
applied, the winding is removed from the mandrel and the primary winding is
wired according to the table.
Now a check follows:
After alternately tamping the coil with the core sheets, the ends 18 and 36 are
connected to the 220 V alternating current network via a mA meter. The current
consumption must be less than 10 mA, and the voltage between 1 and 18 must
have exactly the same value as that between 1 and 36.
If you now connect all the secondary beginnings (i.e. 3, 7, 11, 15) with each other
and then all the ends (4, 8, 12, 16), nothing should change in the current
consumption. If this is the case, the secondary numbers of turns do not match
and the error must be eliminated. Proceed in the same way with the secondary
part windings of the second chamber, at the end both secondary packages are
connected in parallel (15 with 34 and 16 with 33), even now the current
consumption must not change.

The finished transformer also meets extreme requirements:

If desired, connections 2 and 35 can be used for negative feedback from the
screen grid. The anodes are at points 18 and 36, and the anode voltage is at 1 and
19, respectively. Terminal 2 then belongs to the screen grid of the tube, the anode
of which is connected to 18.
For the final assembly neither iron screws nor iron brackets should be used, but
non-magnetic material, as magnetic material can reduce the quality of the
transformer at high frequencies. With this transformer, very high levels of
frequency-independent negative feedback can be achieved and, with negative
feedback over several stages, the distortion factor and internal resistance of the
amplifier including the transformer can be reduced to a fraction of a percent.
Now only the quality of the connected loudspeakers determines whether the
music played sounds natural or technically.
It turned out that the book author had taken over some errors from his template
book.
Kurt Schenk kindly provided the correct formulas and values:
In the first formula, the root that is shown here after the correction is missing.

√ ❑200 N
Fe(cm²) = fu

The mathematically correct result is obtained with the following equation:

U=√10Wx8kohm = 282.84V
In order to be able to calculate more easily, the "crooked" value of 282.84 V is
rounded down to 280 V.
The number of turns of the secondary winding is determined as follows:

Exemplu #1: 32W-25Hz-16cm² Raa:5000Ώ

U=390V

Primar: 2450spire/0.28mm CuEm

Secundar: 98spire/8ohm 69spire/4ohm

Exemplu #2 : 40W-32Hz-16cm² Raa:6000ohmi

U=490V

Primar: 2450spire/0.28mm CuEm

Secundar: 89spire /8ohm 77spire/6ohmi 63spire /4ohmi