n12) Understanding broad-band ferrite transformers used in solidstate Power Amplifiers
(by LA7MI Stein Torp - 1994)

The author is retired design engineer from A/S NERA BERGEN, a radio link manufacturer for over 50 years, and he has worked there for over 40 years. He has been a very important origin source for the ideas I have presented in several magazines over the last 35 years.

In general, the ferrite loss for small signal does not give the proper values for transmitters. So you may waste hours with math only to learn that the results were wrong, or perhaps you stick to the theory and insist that the math is right, but the material was wrong, and make a complaint to the manufacturer. But it is easier to try finding practical conclusions with experiments, then it is also easier to test any unknown material which may be offered at very low price, instead of being dependent on Amidon cores which in many cases are very overpriced. Philips wrote some document around 1970 showing curves which proved the problem that the small-signal theory is wrong for transmitters. The Philips document seems unknown to several designers we've asked, so it might be a thought to show it here later.

Have bought books by Helge Granberg, and also read the Motorola application notes, but I am surprised how much is described using obscure references to math which most readers may not understand, it is much easier than he describes, and I do not agree with many of the conclusions made in several articles. It is easier to find important parameters using some practical experiments, this opens for use of inexpensive sources of surplus materials, in particular the Philips core mentioned below has been available at no cost in larger quantities for many of my friends. I've also been to surplus rallies in Los Angeles and London and experienced that a lot of very useful cores have found little interest among the crowd. I am surprised that some readers have not access to curves which are shown in Philips data books as early as 1970, and have drawn the wrong conclusions with a lot of math. If you don't agree or like to exchange opinions I'll be pleased to receive a letter via post, address below.

Two types of transformers can be found in linear amplifiers:

The transmission line transformer, and the conventional transformer.

The transmission line transformer isu used for large bandwidth ratios (Fmax/Fmin). This transformer is difficult to make.
In modern RF-power amplifiers we find the conventional transformer type consisting of primary and secondary windings wound on a ferrite core

To build a transformer we must know a little about:
1) The turns-ratio to give the wanted impedance trasformation
2) loss in the ferrite core due to magneticfield (generates heat)
3) how many H for minimum operational frequency
4) coupling between primary and secondary to cover maximum operational frequency
5) effect of DC current in the windings
6) how to increase max. operational frequency with capacitive compensation

Ferrite losses
Let us start looking at an ordinary mains transformer.

The current in the primary sets up an alternating
magnetic field in the core. The loss is the core
depends on the strengths of the magnetic field.

When we connect a load to the secondary winding
the priary current will increase. Question: Will the
larger primary current set up a stronger magnetic
field in the core? <No>..

For all practical purposes the magnetic field is independent of the transferred power. This means we can investigate the loss in a ferrite core by winding a few turns, and connect the winding to a suitable signal source.

80m highlevel RF signal source

This 3.7MHz 0-1.5W RF output Colpitts oscillator using 6BQ5/EL84 tube is extremely useful for investigating loss in ferrite cores, and also as a higher level signal source for testing linear power amplifiers. (also see page m3).

See page-n15 for notes about DC saturation

How to test a ferrite core for power loss.

Increase the level and the temperature in the core will increase. Find the voltage for 20C temperature increase.
Example 8. This core can take 4 volts per turn.

Some actual ferrite cores for 2-30MHz PA's have permeability in the range of 100-1000

10.5 mm OD, 6.9 mm ID, 19.5 Long
Philips 3122 134 90783 Ferrite grade 4A4
permeability 500. 4v per turn at 80m band
gave 20C increase. 1 turn = 0.9H

Amidon FB 43-2401
Ferrite grade 43,
permeability 850.
1v/turn on 80m band
gave 20C increase.
1 turn = 0.35H

Formula to calculate the load impedance.

8W output and (Uce-Usat) = 10V gives R load = 6.25 ohm
The load resistance of the two transistors in push-pull is two times (not 4 times) the load of a single transistor. A push-pull amplifier with 16W RF from the transistors with 12V supply must see 12.5 ohm load between the collectors. The output transformer in the push-pull stage must transform 12.5 ohm to 50 ohm, the turns-ratio is 2.

16W RF Push-pull amplifier

The inductance in the secondary must be: 10H or more for frequencies down to 3.5MHz, 20H or more for frequencies down to 1.75MHz. (The reactance must be larger than 200 ohms at the lowest frequency of operation).

Lower frequency limit for a broadband transformer

Without any loss in the ferrite we could be satisfied with 10H inductance in the secondary (50 ohm load) for operation down to 3.5MHz. In the real transformer we have loss due to the magnetic field. As previously shown this limits the volts per turn. For the 16 watt output rating, we will have 28..3V across the secondary

Philips ferrite tube 3122 134 90783 (10.5 mm OD, 6.9 mm ID, 19.5 Long). Let us say max 3.5V per turn. We can use 1, 2, 4 etc number of tubes.
(We chosed this type only because it had been available here at very low cost in larger quantites, somewhere else another type is optimum).

Secondary winding (50 ohm load):

One tube, 8 turns, inductance 0.9 *(8) = 28.8H
Lowest useful frequency = 0.6MHz
Two tubes 4turns, inductance (0.9 +0.9) * (4) = 28.8H
Lowest useful frequency = 1.2MHz
Four tubes, 2 turns, inductance (4x 0.9)2 = 14.4H
Lowest useful frequency = 2.4MHz
8 tubes, 1 turn, inductance = 2x 0.9 = 7.2H
this combination gives lowest frequency = 4.8MHz

In the 16w push-pull amplifier we can use two or four tubes in the transformer. The winding wire must not give additional loss due to resistance and skin effect. With 16W RF output the current is 28.3/50 = 0.57A in the secondary winding.

What happens when we double the number of turns?

The transformer can now supply 56V RF.
Doubling the voltage gives four times more power (4x 16 = 64W). Since doubling the number of turns gives 4 times greater inductance, the low frequency limit is also divided by 4.

Power dissipation in a broad band transformer

A id4eal transformer has zero power dissipation. A real transformer has two types of dissipations:
1) Ferrite losses. Can be reduced by reducing the "volts per turn" rating
2) Ohmic losses in the windings. Can be reduced with heavier wire.

Upper frequency limit for a broadband transformer

In a conventional transformer (not a transmission line transformer) the upper frequency limit is dependent on the coupling between the primary and secondary winding.

How to test a transformer for leakage-inductance?

the primary
Measure the secondary inductance in "H"
= leakage inductance referred to the
secondary winding (50 ohm load)

limit without
limit with
capacitor across
the secondary (pF)
(50 ohm load)
0.25 H
0.5 H

For proper performance we must also put a capacitor across the primary winding. With a turns ration of 1:n this capacitor must be n larger than the secondary capacitor.
Note: We must subtract the transistor output capacitance.

Testing a broad-band transformer by VSWR measurements

By measuring the returnloss (or VSWR) we can optimize a broadband transformer. The primary must be loaded with resistors = load impedance per transistor in the push-pull stage.
Try to achieve better return-loss than 17dB or SWR better than 1.3:1.

DC current in a broadband transformer

Feed the supply via a 'centertapped balun'.
This method also improves the waveform at the transistor collectors

DC can reduce the inductance in the windings, it is best to avoid a DC magnetizing component

DC-saturation (how to measure DC saturation in a ferrite core).
For an Amidon FB43-4301 ferrite core the inductance drops by 50% for 1.2At. 3At for 75% reduction. One should keep the DC-current below the value of 0.5A to keep the reduction below 10%. In the example above this rule is used, but the best is to use push-pull arrangement and the DC magnitation is kept to a minimum level. The inductance reduction is stored and the effect is called remanence, cores used with direct current should have air-gap, and several such cores exists for low frequency power inverter applications.

As an illustration for measuring core saturation this picture could be used. A DC voltage is applied to one side, and an inductance meter to the opposite. Note the DC current voltage needed to reduce the inductance by 10%. Say the DC voltage is 1.25V, the current is 0.25A, and the ampere-turn = 2 * 0.25 = 0.5At (provided voltage drop over the winding is negligible).

Output transformer in the 16W amplifier

No DC magnetic field in the ferrite core

No DC magnetic fields in the ferrite cores

Un-edited circuit diagram (no available photo program)

The 16W power amplifier covering 3.5-10MHz
It is a plate to shortcircuit the two metal tubes at the left side of the driver transformer

Block diagram for the homebrew SSB transceiver in SMD technique
TX power 15W, RX NF 10dB

Document received from LA7MI in April 2004, and the note about saturation has been added in May,
the author has no email, but is reached by telephone nr
The postal address is: Stein Torp, Tollbodalmenning 34, N-5005 Bergen, Norway

Comparing data for Philips og Amidon ferrite core material:

Philips   Amidon  
Type Permeabilitet Type Permeabilitet



- LB8X

My comments about RF transformers where it is one turn consisting of a brass tube, is this really neccessary, could a coax braid be used just as well?
The ends could be solved, and you need no end plates (LA8AK)

Assessing unknown Ferrite Toroids (G3NYK)
See notes about bias regulators
and LA7MI 40W FET linear amplifier (n15)


Last update: 2005.01.08