Audio Transformers Chapter

Audio Transformersby

Bill Whitlock

Jensen Transformers, Inc.

9304 Deering Avenue

Chatsworth, CA 91311

This work first published by Focal Press in 2001 as

Chapter 11

Handbook for Sound Engineers, Third Edition

Glen Ballou, Editor

Copyright © 2001, 2006 Bill WhitlockAll rights reserved

Bill Whitlock Audio Transformers Page 1

Handbook for Sound Engineers, 3 Editionrd

1 Audio Transformer Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1 Basic Principles and Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.1 Magnetic Fields and Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.2 Windings and Turns Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.3 Excitation Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Realities of Practical Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2.1 Core Materials and Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2.2 Winding Resistances and Auto-Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.2.3 Leakage Inductance and Winding Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2.4 Winding Capacitances and Faraday Shields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.2.5 Magnetic Shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.3 General Application Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.3.1 Maximum Signal Level, Distortion, and Source Impedance . . . . . . . . . . . . . . . . . . . . . . 9

1.3.2 Frequency Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3.3 Insertion Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.3.4 Sources with Zero Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.3.5 Bi-Directional Reflection of Impedances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.3.6 Transformer Noise Figure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.3.7 Basic Classification by Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Audio Transformers for Specific Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.1 Equipment-Level Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.1.1 Microphone Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.1.2 Line Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.1.3 Moving Coil Phono Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.1.4 Line Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.1.5 Inter-Stage and Power Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.1.6 Microphone Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2 System-Level Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.1 Microphone Isolation or “Splitter” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.2 Microphone Impedance Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.3 Line to Microphone Input or “Direct Box” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.4 Line Isolation or “Hum Eliminators” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2.5 Speaker Distribution or “Constant Voltage” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.2.6 Telephone Isolation or “Repeat Coil” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.2.7 Telephone Directional Coupling or “Hybrid” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.2.8 Moving Coil Phono Step-Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Measurements and Data Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1 Testing and Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1.1 Transmission Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1.2 Balance Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.1.3 Resistances, Capacitances, and Other Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2 Data Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2.1 Data to Impress or to Inform? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2.2 Comprehensive Data Sheet Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4 Installation and M aintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.1 A Few Installation Tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.2 De-Magnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29



Figure 1 - Magnetic Field

Surrounding Conductor

Figure 4 - Inductive Coupling

Figure 3 - Coil

Concentrates Flux

Figure 2 - AC Magnetic Field

1 Audio Transformer Basics

Since the birth of audio electronics, the audio transformer has played an important role. When compared to modern miniaturized

electronics, a transformer seems large, heavy, and expensive but it continues to be the most effective solution in many audio

applications. The usefulness of a transformer lies in the fact that electrical energy can be transferred from one circuit to another

without direct connection, and in the process the energy can be readily changed from one voltage level to another. Although a

transformer is not a complex device, considerable explanation is required to properly understand how it operates. This chapter is

intended to help the audio system engineer properly select and apply transformers. In the interest of simplicity, only basic concepts

of their design and manufacture will be discussed.

1.1 Basic Principles and Terminology

1.1.1 Magnetic Fields and Induction

As shown in Figure 1, a magnetic field is created around any

conductor (wire) in which current flows. The strength of the

field is directly proportional to current. These invisible magnetic

lines of force, collectively called flux, are set up at right angles

to the wire and have a direction, or magnetic polarity, which

depends on the direction of current flow. Note that although the

flux around the upper and lower wires have different directions,

the lines inside the loop aid because they point in the same

direction. If an alternating current flows in the loop, the

instantaneous intensity and polarity of the flux will vary at the

same frequency and in direct proportion to the instantaneous

current. We can visualize this flux, represented by the concentric circles in Figure 2, as expanding,

contracting, and reversing in polarity with each cycle of the ac current. The law of induction states

that a voltage will be induced in a conductor exposed to changing flux

and that the induced voltage will be proportional to the rate of the flux

change. This voltage has an instantaneous polarity which opposes the

original current flow in the wire, creating an apparent resistance called inductive reactance. Inductive

L Lreactance is calculated according to the formula X = 2ðfL, where X is inductive reactance in ohms, f

is frequency in Hz, and L is inductance in Henries. An inductor generally consists of many turns or

loops of wire called a coil, as shown in Figure 3, which links and concentrates magnetic flux lines,

increasing the flux density. The inductance of any given coil is determined by factors such as the

number of turns, the physical dimensions and nature of the winding, and the properties of materials in

the path of the magnetic flux.

According to the law of induction, a voltage will be induced in any conductor (wire) that cuts

flux lines. Therefore, if we place two coils near each other as shown in Figure 4, an ac

current in one coil will induce an ac voltage in the second coil. This is the essential principle

of energy transfer in a transformer. Because they require a changing magnetic field to

operate, transformers will not work at dc. In an ideal transformer, the magnetic coupling

between the two coils is total and complete, i.e., all the flux lines from one cut across all the

turns of the other. The coupling coefficient is said to be unity or 1.00.

1.1.2 Windings and Turns Ratio

The coil or winding that is driven by an electrical source is called the primary and the other is called the secondary. The ratio of

the number of turns on the primary to the number of turns on the secondary is called the turns ratio. Since essentially the same

voltage is induced in each turn of each winding, the primary to secondary voltage ratio is the same as the turns ratio. For example,

with 100 turns on the primary and 50 turns on the secondary, the turns ratio is 2:1. Therefore, if 20 volts were applied to the

primary, 10 volts would appear at the secondary. Since it reduces voltage, this transformer would be called a step-down

transformer. Conversely, a transformer with a turns ratio of 1:2 would be called a step-up transformer since its secondary voltage

would be twice that of the primary. Since a transformer does not create power, the power output from the secondary of an ideal

transformer can only equal (and in a real transformer only be less than) the power input to the primary. Consider an ideal 1:2 step-

up transformer. When 10 volts is applied to its primary, 20 volts appears at its secondary. Since no current is drawn by the



Figure 5 - Excitation Current Figure 6 - Cancellation of Flux

Generated by Load Current

Figure 7 - Excitation Current and Flux

Vary Inversely with Frequency

primary, its impedance appears to be infinite or an open circuit. When a 20 Ù load is connected to the secondary, a current of 1

amp flows making output power equal 20 watts. At the same time, a current of 2 amps is drawn by the primary, making input

power equal 20 watts. Since the primary is now drawing 2 amps with 10 volts applied, its impedance appears to be 5 Ù . In other

words, the 20 Ù load impedance on the secondary has been reflected to the primary as 5 Ù . In this example, a transformer with a

1:2 turns ratio exhibited an impedance ratio of 1:4. Transformers always reflect impedances from one winding to another by the

square of the their turns ratio or, expressed as a formula: Zp/Zs = (Np/Ns) where Zp is primary impedance, Zs is secondary2

impedance, and Np/Ns is turns ratio (which is the same as the voltage ratio).

The direction in which coils are wound, i.e., clockwise or counter-clockwise, and/or the connections to the start or finish of each

winding determines the instantaneous polarity of the ac voltages. All windings which are wound in the same direction will have

the same polarity between start and finish ends. Therefore, relative to the primary, polarity can be inverted by either (1) winding

the primary and secondary in opposite directions, or (2) reversing the start and finish connections to either winding. In schematic

symbols for transformers, dots are sometimes used to indicate which ends of windings have the same polarity. Observing polarity

is essential when making series or parallel connections to transformers with multiple windings. Taps are connections made at any

intermediate point in a winding. If 50 turns are wound, an electrical connection brought out, and another 50 turns completes the

winding for example, the 100-turn winding is said to be center-tapped.

1.1.3 Excitation Current

As shown in Figure 5, when there is no load on the

secondary of a transformer and an ac voltage is

applied to the primary, an excitation current will flow

in the primary creating magnetic excitation flux

around the winding. In theory, the current is due only

to the inductive reactance of the primary winding. In

accordance with Ohm’s law and the formula for

E P P E Pinductive reactance, I = E ÷ 2ðfL where I is excitation current in amperes, E is

Pprimary voltage in volts, f is frequency in Hz, and L is primary inductance in Henries. In an ideal transformer, primary inductance

would be infinite, making excitation current zero. As shown in Figure 6,

when a load is connected, current will flow in the secondary winding.

Because secondary current flow is in the opposite direction, it creates

magnetic flux which opposes the excitation flux. This causes the

impedance of the primary winding to drop, resulting in additional current

being drawn from the driving source, which creates additional flux just

sufficient to completely cancel that created by the secondary. The result,

which may surprise some, is that flux density in a transformer is not

increased by load current. This also illustrates how load current on the

secondary is reflected to the primary.

Figure 7 illustrates the relationships between voltage, excitation current,

and flux in a transformer as frequency is changed. The horizontal scale is

time. The primary voltage Ep is held at a constant voltage as the frequency

is tripled and then tripled again. For example, the left waveform could

represent one cycle at 100 Hz, the middle 300 Hz, and the right 900 Hz.

Because of the primary inductance, excitation current Ip will decrease

linearly with frequency, i.e., halving for every doubling in frequency or

decreasing at 6 dB per octave. The magnitude of the magnetic flux will

likewise decrease exactly the same way. Note that the inductance causes a

90-degree phase lag as well. Since the slew rate of a constant amplitude

sine wave increases linearly with frequency, i.e., doubling for every

doubling in frequency or increasing at 6 dB per octave, the resultant flux

rate of change remains constant. Note that the slope of the Ip and flux

waveforms stays constant as frequency is changed. Since, according to the

law of induction, the voltage induced in the secondary is proportional to

this slope or rate of change, frequency response will be uniform or “flat.”



Figure 8 - Transformer Low-Frequency Parasitic Elements

1.2 Realities of Practical Transformers

Thus far, we have not considered the unavoidable parasitic elements which exist in any practical transformer. Even the design of a

relatively simple 60 Hz power transformer must take them into account. The design of an audio transformer operating over a 20

Hz to 20 kHz frequency range is much more difficult because these elements often interact in complex ways. For example,

materials and techniques which improve low-frequency performance are often detrimental to high-frequency performance and

vice-versa. Good transformer designs must consider both the surrounding electronic circuitry and the performance ramifications of

internal design tradeoffs.

A schematic representation of the major low-frequency

parasitic elements in a generalized transformer is shown

in Figure 8. The “IDEAL XFM R” represents a perfect

transformer having a turns ratio of 1:N and no parasitic

elements of any kind. The actual transformer is

connected at the “PRI” terminals to the driving voltage

source, through its source impedance RG, and at the

“SEC” terminals to the load RL.

One of the main goals in the design of any transformer is to reduce the excitation current in the primary winding to negligible

levels so as not to become a significant load on the driving source. At a given source voltage and frequency, primary excitation

current can be reduced only by increasing inductance LP. In the context of normal electronic circuit impedances, very large values

of inductance are required for satisfactory operation at the lowest audio frequencies. Of course, inductance can be raised by using

a very large number of coil turns but, for reasons discussed later, there are practical limits due to other considerations. Another

way to increase inductance by a factor of 10,000 or more is to wind the coil around certain highly magnetic materials.

1.2.1 Core Materials and Construction

Magnetic circuits are quite similar to electric circuits. As shown in Figure 11, magnetic flux always takes a closed path from one

magnetic pole to the other and, like an electric current, always favors the paths of highest conductivity or least resistance. The

equivalent of applied voltage in magnetic circuits is magnetizing force, symbolized H . It is directly proportional to “ampere-turns”

(coil current I times its number of turns N) and inversely proportional to the flux path length R in the magnetic circuit. The

equivalent of electric current flow is flux density, symbolized B . It is measured as the number of magnetic flux lines per square

unit of area. A graphic plot of the relationship between field intensity and flux density is shown in Figure 9 and is referred to a the

“B-H loop” or “hysteresis loop” for a given material. In the United States, the most commonly used units for magnetizing force

and flux density are the Oersted and Gauss, respectively, which are CGS (centimeter, gram, second) units. In Europe, the SI

(Système International) units amperes per meter and Tesla, respectively, are more common. The slope of the B-H loop indicates

how an incremental increase in applied magnetizing force changes the resulting flux density. This slope is effectively a measure of

conductivity in the magnetic circuit and is called permeability, symbolized ì. Any material inside a coil, which can also serve as a

form to support it, is called a core. By definition, the permeability of air is 1.00 and common “non-magnetic” materials such as

aluminum, brass, copper, paper, glass, and plastic also have a permeability of 1 for practical purposes. The permeability of some

common “ferro-magnetic” materials is about 300 for ordinary steel, about 5,000 for 4% silicon transformer steel, and up to about

100,000 for some nickel-iron-molybdenum alloys. Because such materials concentrate magnetic flux, they greatly increase the

inductance of a coil. Audio transformers must utilize both high-permeability cores and the largest practical number of coil turns to

create high primary inductance. Coil inductance increases as the square of the number of turns and in direct proportion to the

permeability of the core and can be approximated using the formula: L = 3.2 N ì A / 10 R where L = inductance in Henries, N =2 8

number of coil turns, ì = permeability of core, A = cross-section area of core in square inches, and R = mean flux path length in

inches.



Figure 9 - B-H Loop for Magnetic Core Material

The permeability of magnetic materials varies with flux

density. As shown in Figure 9, when magnetic field

intensity becomes high, the material can saturate,

essentially losing its ability to conduct any additional

flux. As a material saturates, its permeability decreases

until, at complete saturation, its permeability becomes

that of air or 1. In audio transformer applications,

magnetic saturation causes low-frequency harmonic

distortion to increase steadily for low-frequency signals

as they increase in level beyond a threshold. In general,

materials with a higher permeability tend to saturate at a

lower flux density. In general, permeability also varies

inversely with frequency.

Magnetic hysteresis can be thought of as a magnetic

memory effect. When a magnetizing force saturates

material that has high-hysteresis, it remains strongly

magnetized even after the force is removed. High-

hysteresis materials have wide or “square” B-H loops

and are used to make magnetic memory devices and

permanent magnets. However, if we magnetically

saturate zero-hysteresis material, it will have no residual

magnetism (flux density) when the magnetizing force is

removed. However, virtually all high-permeability core

materials have some hysteresis, retaining a small memory of their previous magnetic state. Hysteresis can be greatly reduced by

using certain metal alloys which have been annealed or heat-treated using special processes. In audio transformers, the non-

linearity due to magnetic hysteresis causes increased harmonic distortion for low-frequency signals at relatively low signal levels.

Resistor RC in Figure 8 is a non-linear resistance which represents the combined effects of magnetic saturation, magnetic

hysteresis, and eddy-current losses.

The magnetic operating point (or zero signal point) for most transformers is the center of the B-H loop shown in Figure 9, where

the net magnetizing force is zero. Small ac signals cause a small portion of the loop to be traversed in the direction of the arrows.

Large ac signals traverse portions farther from the operating point and may approach the saturation end points. For this normal

operating point at the center, signal distortions (discussed in detail later) caused by the curvature of the loop are symmetrical, i.e.,

they affect the positive excursion and negative excursion equally. Symmetrical distortions produce odd-order harmonics such as

third and fifth. If dc current flows in a winding, the operating point will shift to a point on the loop away from the center. This

causes the distortion of a superimposed ac signal to become non-symmetrical. Non-symmetrical distortions produce even-order

harmonics such as second and fourth. When a small dc current flows in a winding, under say 1% of the saturation value, the effect

is to add even-order harmonics to the normal odd-order content of the hysteresis distortion, which affects mostly low-level signals.

The same effects occur when the core becomes weakly magnetized, as could happen via the brief accidental application of dc to a

winding for example. However, the narrow B-H loop indicates that only a weak residual field would remain even if a magnetizing

force strong enough to saturate the core were applied and then removed.

When a larger dc current flows in a winding, the symmetry of saturation distortion is also affected in a similar way. For example,

enough dc current might flow in a winding to move the operating point to 50% of the core saturation value. Only half as much ac

signal could then be handled before the core would saturate and, when it did, it would occur only for one direction of the signal

swing. This would produce strong second-harmonic distortion. To avoid such saturation effects, air gaps are sometimes

intentionally built into the magnetic circuit. This can be done, for example, by placing a thin paper spacer between the center leg

of the E and I cores of Figure 10. The magnetic permeability of such a gap is so low — even though it may be only a few

thousandths of an inch — compared to the core material, that it effectively controls the flux density in the entire magnetic circuit.

Although it drastically reduces the inductance of the coil, gapping is done to prevent flux density from reaching levels which

would otherwise saturate the core, especially when substantial dc is present in a winding.



Figure 12 - Auto-Transformers

Employ a Buck/Boost Principle

Figure 11 - Magnetic

Circuits in Shell Core

Figure 10 - Core Laminations are Stacked and

Interleaved around Bobbin which Holds Windings

Because high-permeability materials are usually electrical conductors as

well, small voltages are also induced in the cross-section of the core

material itself giving rise to eddy currents. Eddy currents are greatly

reduced when the core consists of a “stack” of thin sheets called

laminations, as shown in Figure 10.

Because the laminations are effectively

insulated from each other, eddy currents

are generally insignificant. The E and I

shaped laminations shown form the widely

used “shell” or “double-window” core

construction. Its parallel magnetic paths

are illustrated in Figure 11. When cores

are made of laminations, care must be

taken that they are flat and straight to avoid

tiny air gaps between them which could

significantly reduce inductance.

A toroidal core is made by rolling a long thin strip of core material into a

coiled ring shape that looks something like a donut. It is insulated with a

conformal coating or tape and windings are wound around the core

through the center hole using special machines. With a toroidal core, there

are no unintended air gaps which can degrade magnetic properties. Audio

transformers don’t often use toroidal cores because, especially in high-

bandwidth designs where multiple sections or Faraday shields are

necessary, physical construction becomes very complex. Other core configurations include the ring core, sometimes called “semi-

toroidal.” It is similar to core of Figure 11 but without the center section and windings are placed on the sides. Sometimes a solid

(not laminations) metal version of a ring core is cut into two pieces having polished

mating faces. These two C-cores are then held together with clamps after the windings are

installed.

1.2.2 Winding Resistances and Auto-Transformers

If zero-resistance wire existed, some truly amazing transformers could be built. In a 60

Hz power transformer, for example, we could wind a primary with tiny wire on a tiny

core to create enough inductance to make excitation current reasonable. Then we could

wind a secondary with equally tiny wire. Because the wire has no resistance and the flux

density in the core doesn’t change with load current, this postage-stamp sized transformer

could handle unlimited kilo-watts of power — and it wouldn’t even get warm! But, at

least until practical superconducting wire is available, real wire has resistance. As

primary and secondary currents flow in the winding resistances, the resulting voltage

drops cause signal loss in audio transformers and significant heating in power

transformers. This resistance can be reduced by using larger (lower gauge) wire or fewer

turns, but the required number of turns and the tolerable power loss (or resulting heat) all

conspire to force transformers to become physically larger and heavier as their rated

power increases. Sometimes silver wire is suggested to replace copper, but since its

resistance is only about 6% less, its effect is minimal and certainly not cost-effective.

However, there is an alternative configuration of transformer windings, called an auto-

transformer, which can reduce the size and cost in certain applications. Because an auto-

transformer electrically connects primary and secondary windings, it can’t be used where

electrical isolation is required! In addition, the size and cost advantage is maximum when

the required turns ratio is very close to 1:1 and diminishes at higher ratios, becoming

minimal in practical designs at about 3:1 or 1:3.

For example, in a hypothetical transformer to convert 100 volts to 140 volts, the primary

could have 100 turns and the secondary 140 turns of wire. This transformer, with its 1:1.4

turns ratio, is represented in the upper diagram of Figure 12. If 1 amp of secondary

S(load) current I flows, transformer output power is 140 watts and 1.4 amp of primary



Figure 13 - Transformer High-Frequency Parasitic Elements

Figure 14 - Bi-Filar Windings Figure 15 - Layered Windings

Pcurrent I will flow since input and output power must be equal in the ideal case. In a practical transformer, the wire size for each

winding would be chosen to limit voltage losses and/or heating.

An auto-transformer essentially puts the windings in series so that the secondary voltage adds to (boosting) or subtracts from

(bucking) the primary input voltage. A step-up auto-transformer is shown in the middle diagram of Figure 12. Note that the dots

indicate ends of the windings with the same instantaneous polarity. A 40-volt secondary (the upper winding), series connected as

Sshown with the 100-volt primary, would result in an output of 140 volts. Now, if 1 amp of secondary (load) current I flows,

Ptransformer output power is only 40 watts and only 0.4 amp of primary current I will flow. Although the total power delivered to

the load is still 140 watts, 100 watts have come directly from the driving source and only 40 watts have been transformed and

added by the auto-transformer. In the auto-transformer, 100 turns of smaller wire can be used for the primary and only 40 turns of

heavier wire is needed for the secondary. Compare this to the total of 240 turns of heavier wire required in the transformer.

A step-down auto-transformer is shown in the bottom diagram of Figure 12. Operation is similar except that the secondary is

connected so that its instantaneous polarity subtracts from or bucks the input voltage. For example, we could step down US 120-

volt ac power to Japanese 100-volt ac power by configuring a 100-volt to 20-volt step-down transformer as an auto-transformer.

Thus, a 100-watt load can be driven using only a 20-watt rated transformer.

The windings of low-level audio transformers may consist of hundreds or even many thousands of turns of wire, sometimes as

small as #46 gauge, whose 0.0015 inch diameter is comparable to a human hair. As a result, each winding may have a dc

resistance as high as several thousand ohms. Transformer primary and secondary winding resistances are represented by RP and

RS, respectively, in Figure 8.

1.2.3 Leakage Inductance and Winding Techniques

In an ideal transformer, since all flux generated

by the primary is linked to the secondary, a

short-circuit on the secondary would be

reflected to the primary as a short circuit. In real

transformers, the unlinked flux causes a residual

or leakage inductance which can be measured

at either winding. Therefore, the secondary

would appear to have residual inductance if the

primary were shorted and vice-versa. The

leakage inductance is shown as LL in the model

of Figure 13. Note that leakage inductance is

reflected from one winding to another as the square of turns ratio, just as other impedances are.

The degree of flux coupling between primary and secondary windings depends on the physical spacing between them and how

they are placed with respect to each other. The lowest leakage inductance is achieved by winding the coils on a common axis and

as close as possible to each other. The ultimate form of this technique is called multi-filar winding where multiple wires are wound

simultaneously as if they were a single strand. For example, if two windings (say primary and secondary) are wound as one, the

transformer is said to be bi-filar wound. Note in the cross-section view of Figure 14 how the primary and secondary windings are

side-by-side throughout the entire winding. Another technique to reduce leakage inductance is to use layering, a technique in

which portions or sections of the primary and/or secondary are wound in sequence over each other to interleave them. For

example, Figure 15 shows the cross-section of a

3-layer transformer where half the primary is

wound, then the secondary, followed by the

other half of the primary. This results in

considerably less leakage inductance than just a

secondary over primary 2-layer design. Leakage

inductance decreases rapidly as the number of

layers is increased.



Figure 16 - High-Frequency Equivalent Circuit of Transformer

with Faraday Shield and Driven by a Balanced Source

1.2.4 Winding Capacitances and Faraday Shields

To allow the maximum number of turns in a given space, the insulation on the wire used to wind transformers is very thin. Called

“magnet wire,” it is most commonly insulated by a thin film of polyurethane enamel. A transformer winding is made, in general,

by spinning the bobbin shown in Figure 10 on a machine similar to a lathe and guiding the wire to form a layer one wire thick

across the length of the bobbin. The wire is guided to traverse back and forth across the bobbin to form a coil of many layers as

shown in Figure 15, where the bobbin cross-section is the solid line on three sides of the winding. This simple side-to-side, back-

and-forth winding results in considerable layer-to-layer capacitance within a winding or winding section. More complex

techniques such as “universal” winding are sometimes used to substantially reduce winding capacitances. These capacitances

within the windings are represented by CP and CS in the circuit model of Figure 13. Additional capacitances will exist between

the primary and secondary windings and are represented by capacitors CW in the model. Sometimes layers of insulating tape are

added to increase the spacing, therefore reducing capacitance, between primary and secondary windings. In the bi-filar windings

of Figure 14, since the wires of primary and secondary windings are side by side throughout, the inter-winding capacitances CW

can be quite high.

In some applications, inter-winding capacitances are very

undesirable. Their effects can be almost completely

eliminated by the use of a Faraday shield between the

windings. Sometimes called an electrostatic shield, it

generally takes the form of a thin sheet of copper foil

placed between the windings. Obviously, transformers

that utilize multiple layers to reduce leakage inductance

will require Faraday shields between all adjacent layers.

In Figure 15 the dark lines between the winding layers

are the Faraday shields. Normally, all the shields

surrounding a winding are tied together and treated as a

single electrical connection. When connected to circuit

ground, as shown in Figure 16, a Faraday shield

intercepts the capacitive current which would otherwise

flow between transformer windings.

Faraday shields are nearly always used in transformers designed to eliminate “ground noise.” In these applications, the transformer

is intended to respond only to the voltage difference or signal across its primary and have no response to the noise that exists

equally (or common-mode) at the terminals of its primary. A Faraday shield is used to prevent capacitive coupling (via CW in

Figure 13) of this noise to the secondary. For any winding connected to a balanced line, the matching of capacitances to ground is

critical to the rejection of common-mode noise or CMRR, as discussed in Chapter 37. In Figure 16, if the primary is driven by a

balanced line, C1 and C2 must be very accurately matched to achieve high CMRR. In most applications, such as microphone or

line input transformers, the secondary is operated unbalanced, i.e., one side is grounded. This relaxes the matching requirements

for capacitances C3 and C4. Although capacitances CC1 and CC2 are generally quite small (a few pF), they have the effect of

diminishing CMRR at high audio frequencies and limiting rejection of RF interference.

1.2.5 Magnetic Shielding

A magnetic shield has a completely different purpose. Devices such as power transformers, electric motors, and television or

computer monitor cathode-ray tubes generate powerful ac magnetic fields. If such a field takes a path through the core of an audio

transformer, it can induce an undesired voltage in its windings — most often heard as hum. If the offending source and the victim

transformer have fixed locations, orientation of one or both can sometimes nullify the pick-up. In Figure 11 note that an external

field which flows vertically through the core will cause a flux gradient across the length of the coil, inducing a voltage in it, but a

field which flows horizontally through the core will not. Such magnetic pick-up is usually worse in “input” transformers (discussed

later) because they generally have more turns. It should also be noted that higher permeability core materials are more immune to

external fields. Therefore, an unshielded “output” transformer with a high-nickel core will be more immune than one with a steel

core.

Another way to prevent such pick-up is to surround the core with a closed (no air gap) magnetic path. This magnetic shield most

often takes the form of a can or box with tight-fitting lid and is made of high-permeability material. While the permeability of

ordinary steel, such as that in electrical conduit, is only about 300, special-purpose nickel alloys can have permeability as high as

100,000. Commercial products include Mumetal®, Permalloy®, HyMu® and Co-Netic®.[1][2] Since the shield completely

surrounds the transformer, the offending external field will now flow through it instead of the transformer core. Generally



Figure 17 - Measured THD at 20 Hz and 40 Ù Source vs

Signal Level for Three Types of Core Material

speaking, care must be taken not to mechanically stress these metals because doing so will significantly decrease their

permeability. For this reason, most magnetic shield materials must be re-annealed after they are fabricated.

The effectiveness of magnetic shielding is generally rated in dB. The transformer is placed in an external magnetic field of known

strength, generally at 60 Hz. Its output without and with the shield is then compared. For example, a housing of 1/8" thick cast-

iron reduces pickup by about 12 dB and a Mumetal can by about 30 dB. Where low-level transformers operate near strong

magnetic fields, several progressively smaller shield cans can be nested around the transformer. Two or three Mumetal cans can

provide 60 dB and 90 dB of shielding respectively. In very strong fields, because high-permeability materials might saturate, an

iron or steel outer can is sometimes used.

Toroidal power transformers can have a weaker radiated magnetic field than other types. Using them can be an advantage if audio

transformers must be located near them. However, a toroidal transformer must be otherwise well designed to produce a low

external field. For example, every winding must completely cover the full periphery of the core. The attachment points of the

transformer lead wires are frequently a problem in this regard. To gain size and cost advantages, most commercial power

transformers of any kind are designed to operate on the verge of magnetic saturation of the core. When saturation occurs in any

transformer, magnetic field essentially squirts out of the core. Power transformers designed to operate at low flux density will

prevent this. Often a standard commercial transformer, when operated at reduced primary voltage, will have a very low external

field.

1.3 General Application Considerations

For any given application, a number of parameters must be considered when selecting or designing an appropriate audio

transformer. We will discuss how the performance of a transformer can be profoundly affected by its interaction with surrounding

circuitry.

1.3.1 Maximum Signal Level, Distortion, and Source Impedance

Because these parameters are inextricably inter-dependent, they must be discussed as a group. Although transformer operating

level is often specified in terms of power such as dBm or watts, the only thing that affects distortion is the equivalent driving

voltage. Distortion is caused by excitation current in the primary winding which is proportional to primary voltage, not power.

Referring to Figure 8, recall that RC represents the distortion producing mechanisms of the core material. Consider that, if both

RG (driving source impedance) and RP (internal winding resistance) were zero, the voltage source (by definition, zero impedance)

would effectively “short out” RC resulting in zero distortion! But in a real transformer design there is a fixed relationship between

signal level, distortion, and source impedance. Since distortion is also a function of magnetic flux density, which increases as

frequency decreases, a maximum operating level specification must also specify a frequency. The specified maximum operating

level, maximum distortion at a specified low frequency, and maximum allowable source impedance will usually dictate the type of

core material which must be used and its physical size. And, of course, cost plays a role, too.

The most commonly used audio transformer core materials

are M6 steel (a steel alloy containing 6% silicon) and 49%

nickel or 84% nickel (alloys containing 49% or 84% nickel

plus iron and molybdenum). Nickel alloys are substantially

more expensive than steel. Figure 17 shows how the choice

of core material affects low-frequency distortion as signal

level changes. The increased distortion at low levels is due

to magnetic hysteresis and at high levels is due to magnetic

saturation. Figure 18 shows how distortion decreases

rapidly with increasing frequency. Because of differences in

their hysteresis distortion, the fall-off is most rapid for the

84% nickel and least rapid for the steel. Figure 19 shows

how distortion is strongly affected by the impedance of the

driving source (the plots begin at 40 Ù because that is the

resistance of the primary winding). Therefore, maximum

operating levels predicated on higher frequencies, higher

distortion, and lower source impedance will always be

higher than those predicated on lower frequencies, lower

distortion, and lower source impedance.



Figure 18 - Measured THD at 0 dBu and 40 Ù Source vs

Frequency for the Cores of Figure 16

Figure 19 - Measured THD at 0 dBu and 20 Hz vs

Source Impedance for the Cores of Figures 16 and 17

As background, it should be said that THD or total harmonic

distortion is a remarkably inadequate way to describe the

perceived awfulness of distortion. Distortion consisting of

low-order harmonics, 2 or 3 for example, is dramaticallynd rd

less audible than that consisting of high-order harmonics, 7 th

or 13 for example. Consider that, at very low frequencies,th

even the finest loudspeakers routinely exhibit harmonic

distortion in the range of several percent at normal listening

levels. Simple distortion tests whose results correlate well

with the human auditory experience simply don’t exist.

Clearly, such perceptions are far too complex to quantify

with a single figure.

One type of distortion which is particularly audible is inter-

modulation or IM distortion. Tests frequently use a large

low-frequency signal and a smaller high-frequency signal

and measure how much the amplitude of the high frequency

is modulated by the lower frequency. Such inter-modulation

creates tones at new, non-harmonic frequencies. The classic

SMPTE (Society of Motion Picture and Television

Engineers) IM distortion test mixes 60 Hz and 7 kHz signals

in a 4:1 amplitude ratio. For virtually all electronic amplifier

circuits, there is an approximate relationship between

harmonic distortion and SMPTE IM distortion. For example,

if an amplifier measured 0.1% THD at 60 Hz at a given

operating level, its SMPTE IM distortion would measure

about three or four times that, or 0.3% to 0.4% at an

equivalent operating level. This correlation is due to the fact

that electronic non-linearities generally distort audio signals

without regard to frequency. Actually, because of negative

feedback and limited gain-bandwidth, most electronic

distortions become worse as frequency increases.

Distortion in audio transformers is different in a way which

makes it unusually benign. It is caused by the smooth

symmetrical curvature of the magnetic transfer characteristic

or B-H loop of the core material shown in Figure 9. The

non-linearity is related to flux density which, for a constant

voltage input, is inversely proportional to frequency. The

resulting harmonic distortion products are nearly pure third

harmonic. In Figure 18, note that distortion for 84% nickel

cores roughly quarters for every doubling of frequency, dropping to less than 0.001% above about 50 Hz. Unlike that in

amplifiers, the distortion mechanism in a transformer is frequency selective. This makes its IM distortion much less than might be

expected. For example, the Jensen JT-10KB-D line input transformer has a THD of about 0.03% for a +26 dBu input at 60 Hz.

But, at an equivalent level, its SMPTE IM distortion is only about 0.01% — about a tenth of what it would be for an amplifier

having the same THD.

1.3.2 Frequency Response

The simplified equivalent circuit of Figure 20 shows the high-pass RL filter formed by the circuit resistances and transformer

primary inductance LP. The effective source impedance is the parallel equivalent of RG + RP and RS + RL. When the inductive

reactance of LP equals the effective source impedance, low-frequency response will fall to 3 dB below its mid-band value. For

example, consider a transformer having an LP of 10 Henries and winding resistances RP and RS of 50 Ù each. The generator

impedance RG is 600 Ù and the load RL is 10 kÙ . The effective source impedance is then (600 Ù + 50 Ù ) in parallel with (10 kÙ

+ 50 Ù) which computes to about 610 Ù . A 10 Henry inductor will have 610 Ù of reactance at about 10 Hz, making response 3 dB

down at that frequency. If the generator impedance RG were made 50 Ù instead, response would be !3 dB at 1.6 Hz. Lower

source impedance will always extend low-frequency bandwidth. Since the filter is single-pole, response falls at 6 dB per octave.

As discussed earlier, the permeability of most core material steadily increases as frequency is lowered and typically reaches its



Figure 20 - Simplified Low Frequency

Transformer Equivalent Circuit

Figure 21 - Simplified High-Frequency

Transformer Equivalent Circuit

Figure 22 - Undamped Response Figure 23 - Proper Damping

maximum somewhere under 1 Hz. This results in an actual roll-off rate less than 6 dB per octave and a corresponding

improvement in phase distortion (deviation from linear phase). Although a transformer cannot have response to 0 Hz or dc, it can

have much less phase distortion than a coupling capacitor chosen for the same cutoff frequency. Or, as a salesperson might say

“it’s not a defect, it’s a feature.”

The simplified equivalent schematic of Figure 21 shows the parasitic elements which limit and control high-frequency response.

Except in bi-filar wound types discussed below, leakage inductance LL and load capacitance are the major limiting factors. This is

especially true when Faraday shields because of the increase in leakage inductance. Note that a low-pass filter is formed by series

leakage inductance LL with shunt winding capacitance CS plus external load capacitance CL. Since this filter has two reactive

elements, it is a two-pole filter subject to response variations caused by damping. Resistive elements in a filter provide damping,

Ddissipating energy when the inductive and capacitive elements resonate. As shown in the figure, if damping resistance R is too

high, response will rise before it falls and if damping resistance is too low, response falls too early. Optimum damping results in

the widest bandwidth with no response peak. It should

be noted that placing capacitive loads CL on

transformers with high leakage inductance not only

lowers their bandwidth but changes the resistance

required for optimum damping. For most transformers,

RL controls damping. In the time domain, under-

damping manifests itself as ringing on square-waves as

shown in Figure 22. When loaded by its specified load

resistance, the same transformer responds as shown in

Figure 23. In some transformers, source impedance

also provides significant damping.

In bi-filar wound transformers, leakage inductance LL

is very low but inter-winding capacitance CW and winding capacitances CP and CS are quite high. Leakage inductance must be

kept very small in applications such as line drivers because large cable capacitances CL would otherwise be disastrous to high-

frequency response. Also note that a low-pass filter is formed by series RG and shunt CP plus CS. Therefore, driving sources may

limit high-frequency response if their source impedance RG is too high. In normal 1:1 bi-filar output transformer designs, CW

actually works to capacitively couple very high frequencies between windings. Depending on the application, this can be either a

defect or a feature.

1.3.3 Insertion Loss

The power output from a transformer will always be slightly less than power input to it. As current flows in its windings, their dc

resistance causes additional voltage drops and power loss as heat. Broadly defined, insertion loss (or gain) is that caused by

inserting a device into the signal path. But, because even an ideal lossless transformer can increase or decrease signal level by

virtue of its turns ratio, the term insertion loss is usually defined as the difference in output signal level between the real

transformer and an ideal one with the same turns ratio.



Figure 24 - Insertion Loss Compares the Outputs of Real and Ideal Transformers

The circuit models,

Thevenin equivalent

circuits, and equations for

both ideal and real

transformers are shown in

Figure 24. For example,

consider an ideal 1:1 turns

ratio transformer and RG =

s pRL = 600 Ù . Since N /N is

1, the equivalent circuit

ibecomes simply E in series

with RG or 600 Ù . When

RL is connected, a simple

voltage divider is formed

o imaking E = 0.5 E (or a

6.02 dB loss). For a real

transformer having RP = RS

= 50 Ù , the equivalent

icircuit becomes E in series

with RG + RP + RS or 700 Ù .

o iNow, the output E = 0.462 E (or a 6.72 dB loss). Therefore, the insertion loss of the transformer is 0.70 dB.

Calculations are similar for transformers with turns ratios other than 1:1, except that voltage is multiplied by the turns ratio and

reflected impedances are multiplied by the turns ratio squared as shown in the equations. For example, consider a 2:1 turns ratio

itransformer, RG = 600 Ù , and RL = 150 Ù . The ideal transformer output appears as 0.5 E in series with RG/4 or 150 Ù . When RL

o iis connected, a simple voltage divider is formed making E = 0.25 E (or a 12.04 dB loss). For a real transformer having RP = 50

i oÙ and RS = 25 Ù , the equivalent circuit becomes 0.5 E in series with (RG + RP)/4 + RS or 187.5 Ù . Now, the output E = 0.222

iE (or a 13.07 dB loss). Therefore, the insertion loss of this transformer is 1.03 dB.

1.3.4 Sources with Zero Impedance

One effect of using negative feedback around a high-gain amplifier is to reduce output impedance. Output impedance is reduced

by the feedback factor which is open-loop gain in dB minus closed-loop gain in dB. A typical op-amp with an open-loop gain of

80 dB, set for closed-loop gain of 20 dB (feedback factor is 80 dB ! 20 dB = 60 dB or 1000) will have its open-loop output

impedance of 50 Ù reduced by the feedback factor to about 0.05 Ù . Within the limits of linear operation, i.e., no current limiting

or voltage clipping, the feedback around the amplifier forces the output to remain constant regardless of loading. For all practical

purposes this can be considered a true voltage source.

As seen in Figure 19, the distortion performance of ANY transformer is significantly improved when the driving source

impedance is less than the dc resistance of the primary. However, little is gained below about 10% of the winding dc resistance.

For example, consider a typical line output transformer with a primary dc resistance of 40 Ù . A driving source impedance well

under 4 Ù will result in lowest distortion. The line drivers shown in Figure 28 and Figure 29 use a paralleled inductor and resistor

to isolate or decouple the amplifier from the destabilizing effects of load (cable) capacitance at very high frequencies. Because its

impedance is well under an ohm at all audio frequencies, it is much preferred to the relatively large series or “build-out” resistor

often used for the purpose. It is even possible for an amplifier to generate negative output resistance to cancel the winding

resistance of the output transformer. Audio Precision uses such a patented circuit in their System 1 audio generator to reduce

transformer-related distortion to extremely low levels.

1.3.5 Bi-Directional Reflection of Impedances

The impedances associated with audio transformers seems to confuse many. Much of the confusion probably stems from the fact

that transformers can simultaneously reflect two different impedances. One is the impedance of the driving source, as seen from

the secondary, and the other is the impedance of the load, as seen from the primary. Transformers simply reflect impedances,

modified by the square of their turns ratio, from one winding to another. However, because of their internal parasitic elements,

transformers tend to produce optimum results when used within a specified range of external impedances.

There is essentially no intrinsic impedance associated with the transformer itself. With no load on its secondary, the primary of a

transformer is just an inductor and its impedance will vary linearly with frequency. For example, a 5 H primary winding would



Figure 25 - Impedance Reflection in a 1:1 Transformer

Figure 26 - Impedance Reflection in a 4:1 Transformer

Figure 27 - Multiple Loads

are Effectively Paralleled

have an input impedance of about 3 kÙ at 100 Hz, 30 kÙ at 1 kHz,

and 300 kÙ at 10 kHz. In a proper transformer design, this self-

impedance, as well as those of other internal parasitics, should have

negligible effects on circuit operation. The following applications

will illustrate the point.

A 1:1 output transformer application is shown in Figure 25. It has a

winding inductance of about 25 H and negligible leakage inductance.

The open circuit impedance, at 1 kHz, of either winding is about

150 kÙ . Since the DC resistance is about 40 Ù per winding, if the

primary is short circuited, the secondary impedance will be 80 Ù . If

we place the transformer between a zero-impedance amplifier (more

on that later) and a load, the amplifier will "see" the load through the

transformer and the load will "see" the amplifier through the

transformer. In our example, the amplifier would "look like" 80 Ù to

the output line or load and the 600 Ù load would "look like" 680 Ù

to the amplifier. If the load were 20 kÙ , it would "look like" slightly

less than 20 kÙ because the open circuit transformer impedance (150

kÙ at 1 kHz) is effectively in parallel with it. For most loads, this

effect is negligible.

A 4:1 input transformer example is shown in Figure 26. It has a

primary inductance of about 300 H and negligible winding

capacitance. The open circuit impedance, at 1 kHz, of the primary is

about 2 MÙ . Because this transformer has a 4:1 turns ratio, therefore

16:1 impedance ratio, the secondary open circuit impedance is about

125 kÙ . The DC resistances are about 2.5 kÙ for the primary and

92 Ù for the secondary. Since this is an input transformer, it must be

used with the specified secondary load resistance of 2.43 kÙ for

proper damping (flat frequency response). This load on the

secondary will be transformed by the turns ratio to "look like" about

42 kÙ at the primary. To minimize the noise contribution of the

amplifier stage, we need to know what the transformer secondary

"looks like," impedance wise, to the amplifier. If we assume that the

primary is driven from the line in our previous output transformer

example with its 80 Ù source impedance, we can calculate that the

secondary will "look like" about 225 Ù to the amplifier input.

Actually, any source impedance less than 1 kÙ would have little

effect on the impedance seen at the secondary.

Transformers are not "intelligent" — they can’t isolate, in the loading sense, outputs from one

another or magically couple signals in one direction only. Magnetic coupling is truly

bi-directional. For example, Figure 27 shows a three-winding 1:1:1 transformer connected to

drive two 600 Ù loads. The driver “sees” the loads in parallel or, neglecting winding

resistances, 300 Ù . Likewise, a short on either output will be reflected to the driver as a short.

Of course, turns ratios and winding resistances must be taken into account to calculate actual

driver loading. For the same reason, stereo L and R outputs driving two windings on the same

transformer are effectively driving each other, possibly causing distortion or damage.

1.3.6 Transformer Noise Figure

Although the step-up turns ratio of a transformer may provide “noise-free” voltage gain, some 20 dB for a 1:10 turns ratio, it’s

important to understand that improvements in signal-to-noise ratio are not solely due to this gain. Because most amplifying

devices generate current noise as well as voltage noise, their noise performance will suffer when turns ratio is above the optimum

(see Chapter 21 on mic preamps). Noise figure measures, in dB, how much the output signal-to-noise ratio of a system is degraded

by a given system component. All resistances, including the winding resistances of transformers, generate thermal noise.

Therefore, the noise figure of a transformer indicates the increase in thermal noise or hiss when it replaces an ideal noiseless

transformer having the same turns ratio, i.e., voltage gain. The noise figure of a transformer is calculated as follows:



1.3.7 Basic Classification by Application

Many aspects of transformer performance, such as level-handling, distortion, and bandwidth, depend critically on the impedance

of the driving source and, in some cases, the resistance and capacitance of the load. These impedances play such an important role

that they essentially classify audio transformers into two basic types. Most simply stated, output transformers are used when load

impedances are low, as in line drivers, while input transformers are used when load impedances are high, as in line receivers. The

conflicting technical requirements for output and input types make their design and physical construction very different. Of course,

some audio transformer applications need features of both input and output transformers and are not so easily classified.

Output transformers must have very low leakage inductance in order to maintain high-frequency bandwidth with capacitive loads.

Because of this, they rarely use Faraday shields and are often multi-filar wound. For low insertion loss, they use relatively few

turns of large wire to decrease winding resistances. Since they use fewer turns and operate at relatively high signal levels, output

transformers seldom use magnetic shielding. On the other hand, input transformers directly drive the usually high-resistance, low-

capacitance input of amplifier circuitry. Many input transformers operate at relatively low signal levels, frequently have a Faraday

shield, and are usually enclosed in at least one magnetic shield.

2 Audio Transformers for Specific Applications

Broadly speaking, audio transformers are used because they have two very useful properties. First, they can benefit circuit

performance by transforming circuit impedances, to optimize amplifier noise performance for example. Second, because there is

no direct electrical connection between its primary and secondary windings, a transformer provides electrical or galvanic isolation

between two circuits. As discussed in Chapter 37, isolation in signal circuits is a powerful technique to prevent or cure noise

problems caused by normal ground voltage differences in audio systems. To be truly useful, a transformer should take full

advantage of one or both of these properties but not compromise audio performance in terms of bandwidth, distortion, or noise.



Figure 29 - Low-Noise Unity-Gain Balanced Line Input Stage

Figure 28 - Microphone Pre-Amplifier with 40 dB Overall Gain

2.1 Equipment-Level Applications

2.1.1 Microphone Input

A microphone input transformer is driven by the

nominal 150 Ù (or 200 Ù in Europe) source

impedance of professional microphones. One of its

most important functions is to transform this

impedance to a generally higher one more suited to

optimum noise performance. As discussed in chapter

21, this optimum impedance may range from 500 Ù

to over 15 kÙ , depending on the amplifier. For this

reason, microphone input transformers are made with

turns ratios ranging from 1:2 to 1:10 or higher. The

circuit of Figure 28 uses a 1:5 turns ratio transformer, making the microphone appear as a 3.7 kÙ driving source to the IC

amplifier, which optimizes its noise. The input impedance of the transformer is about 1.5 kÙ . It is important that this impedance

remain reasonably flat with frequency to avoid altering microphone response by loading it excessively at frequency extremes.

In all balanced signal connections, common-mode noise can exist due to ground voltage differences or magnetic or electrostatic

fields acting on the inter-connecting cable. It is called common-mode noise because it appears equally on the two signal lines, at

least in theory. Perhaps the most important function of a balanced input is to reject (not respond to) this common-mode noise. A

figure comparing the ratio of its differential or normal signal response to its common-mode response is called common-mode

rejection ratio or CMRR. An input transformer must have two attributes to achieve high CMRR. First, the capacitances of its two

inputs (to ground) must be very well matched and as low as possible. Second, it must have minimal capacitance between its

primary and secondary windings. This is usually accomplished by precision winding of the primary to evenly distribute

capacitances and the incorporation of a Faraday shield between primary and secondary. Because the common-mode input

impedances of a transformer consist only of capacitances of about 50 pF, transformer CMRR is maintained in real-world systems

where the source impedances of devices driving the balanced line and the capacitances of the cable itself are not matched with

great precision [3].

Because tolerable common-mode voltage is limited only by winding insulation, transformers are well suited for phantom power

applications. The standard arrangement using precision resistors is shown in Figure 28. Resistors of lesser precision may degrade

CMRR. Feeding phantom power through a center tap on the primary requires that both the number of turns and the dc resistance

on either side of the tap be precisely matched to avoid small dc offset voltages across the primary. Normal tolerances on winding

radius and wire resistance make this a less precise method than the resistor pair in most practical transformer designs. Virtually all

microphone input transformers will require loading on the secondary to control high-frequency response. For the circuit in the

figure, network R1, R2, and C1 shape the high-frequency response to a Bessel roll-off curve. Because they operate at very low

signal levels, most microphone input transformers also have magnetic shielding.

2.1.2 Line Input

A line input transformer is driven by a balanced line and, most

often, drives a ground-referenced (unbalanced) amplifier

stage. As discussed in Chapter 37, modern voltage-matched

interconnections require that line inputs have impedances of

10 kÙ or more, traditionally called “bridging.” In the circuit of

Figure 29, a 4:1 step-down transformer is used which has an

input impedance of about 40 kÙ .

High common-mode noise rejection or CMRR is achieved in

line input transformers using the same techniques as those for

microphones. Again, because its common-mode input

impedances consist of small capacitances, a good input

transformer will exhibit high CMRR even when signal sources

are real-world equipment. Electronically-balanced stages, especially simple differential amplifiers, are very susceptible to tiny

impedance imbalances in driving sources. However, they usually have impressive CMRR figures when the signal source is a

laboratory generator. The pitfalls of measurement techniques will be discussed in section 3.1.



Figure 32 - Universal Isolated Output Application

Figure 30 - Preamp for 25 Ù to 40 Ù Moving-Coil Pickups

Figure 31 - Typical Line Output Application Circuit

As with any transformer having a Faraday shield, line input transformers have significant leakage inductance and their secondary

load effectively controls their frequency response and roll-off characteristics. The load resistance or network recommended by the

manufacturer should be used to achieve specified bandwidth and transient response. Input transformers are intended to

immediately precede an amplifier stage with minimal input capacitance. Additional capacitive loading of the secondary should be

avoided because of its adverse effect on frequency and phase response. For example, the capacitance of two feet of ordinary

shielded cable, about 100 pF, is enough to significantly degrade performance of many input transformers.

2.1.3 Moving-Coil Phono Input

Moving-coil phonograph pickups are very low-impedance, very

low-output devices. Some of them have source impedances as low

as 3 Ù , making it nearly impossible to achieve optimum noise

performance in an amplifier. The transformer shown in Figure 30

has a three-section primary that can be series-connected as a 1:4

step-up for 25 Ù to 40 Ù devices and parallel-connected as a 1:12

step-up for 3 Ù to 5 Ù devices. In either case, the amplifier sees a

600 Ù source impedance that enables low-noise operation. The

transformer is packaged in double magnetic shield cans and has a

Faraday shield. The loading network R1, R2, and C1 tailor the

high-frequency response to a Bessel curve.

2.1.4 Line Output

A line-level output transformer is driven by an amplifier

and typically loaded by several thousand pF of cable

capacitance plus the 20 kÙ input impedance of a

balanced "bridging" line receiver. At high frequencies,

most driver output current is actually used driving the

cable capacitance. Sometimes, terminated 150 Ù or

600 Ù lines must be driven, requiring even more driver

output current. Therefore, a line output transformer must

have a low output impedance that stays low at high

frequencies. This requires both low resistance windings

and very low leakage inductance, since they are

effectively in series between amplifier and load. To

maintain impedance balance of the output line, both

driving impedances and inter-winding capacitances must

be well matched at each end of the windings. A typical

bifilar-wound design has winding resistances of 40 Ù

each, leakage inductance of a few micro-henries, and a

total inter-winding capacitance of about 20 nF matched to within 2% across the windings.

The high-performance circuit of Figure 31 uses op-amp A1 and current booster A2 in a feedback loop setting overall gain at 12

dB. A3 provides the high gain for a dc servo feedback loop used to keep dc offset at the output of A2 under 100 ìV. This prevents

any significant dc flow in the primary of transformer T1. X1 provides capacitive load isolation for the amplifier and X2 serves as a

tracking impedance to maintain high-frequency impedance balance of the output. High-conductance diodes D1 and D2 clamp

inductive kick to protect A2 in case an unloaded output is driven

into hard clipping.

The circuit of Figure 32 is well suited to the lower signal levels

generally used in consumer systems. Because its output floats, it

can drive either balanced or unbalanced outputs, but not at the

same time. Floating the unbalanced output avoids ground loop

problems that are inherent to unbalanced interconnections.

In both previous circuits, because the primary drive of T1 is

single-ended, the voltages at the secondary will not be

symmetrical, especially at high frequencies. THIS IS NOT A



Figure 34 - Push-Pull Vacuum-Tube Power Amplifier

Figure 33 - Double Cathode-Follower Line Driver

PROBLEM. Contrary to widespread myth and as explained in Chapter 37, signal symmetry has absolutely nothing to do with

noise rejection in a balanced interface! Signal symmetry in this, or any other floating output, will depend on the magnitude and

matching of cable and load impedances to ground. If there is a requirement for signal symmetry, the transformer should be driven

by dual, phase-inverted drivers.

The circuit of Figure 33 uses a cathode follower circuit which

replaces the usual resistor load in the cathode with an active current

sink. The circuit operates at quiescent plate currents of about 10 mA

and drives the transformer with a source impedance of about 60 Ù ,

which is less than 10% of its primary dc resistance. C2 is used to

prevent dc flow in the primary. Since the transformer has a 4:1 turns

ratio (or 16:1 impedance ratio), a 600 Ù output load is reflected to the

driver circuit as about 10 kÙ . Since the signal swings on the primary

are four times as large as those on the secondary, high-frequency

capacitive coupling is prevented by a Faraday shield. The secondary

windings may be parallel connected to drive a 150Ù load. Because of

the Faraday shield, output winding capacitances are low and the

output signal symmetry will be determined largely by the balance of

line and load impedances.

2.1.5 Inter-Stage and Power Output

Inter-stage coupling transformers are seldom seen in contemporary equipment but were once quite popular in vacuum-tube

amplifier designs. They typically use turns ratios in the 1:1 to 1:3 range and, as shown in Figure 34, may use a center-tapped

secondary producing phase-inverted signals to drive a push-pull output stage. Because both plate and grid circuits are relatively

high impedance, windings are sometimes section-wound to reduce capacitances. Resistive loading of the secondary is usually

necessary both to provide damping and to present a uniform load impedance to the driving stage. Although uncommon, inter-stage

transformers for solid-state circuitry are frequently bifilar wound units similar to line output designs.

The classic push-pull power output stage, with

many variations over the years, has been used in

hi-fi gear, PA systems, and guitar amplifiers.

The turns ratio of the output transformer is

generally chosen for a reflected load at the tubes

of several thousand ohms plate-to-plate. A

typical 30:1 turns ratio may require many

interleaved sections to achieve bandwidth

extending well beyond 20 kHz.

If the quiescent plate currents and the number of

turns in each half of the primary winding are

matched, magnetic flux in the core will cancel at

dc. Since any current-balancing is temporary at

best, these transformers nearly always use steel

cores. The relatively high driving impedance of

the tube plates results in considerable transformer related distortion. To reduce distortion, feedback around the transformer is often

employed. But to achieve stability (freedom from oscillation), very wide bandwidth (actually low phase shift) is required of the

transformer when a feedback loop is closed around it. As a result, some of these output transformer designs are very sophisticated.

Some legendary wisdom suggests “as a rough guide” that a good-fidelity output transformer should have a core weight and volume

of at least 0.34 pounds and 1.4 cubic inches respectively per watt of rated power [4].

A “single-ended” power amplifier is created by removing the lower tube and the lower half of the transformer primary from the

circuit of Figure 34. Now plate current will create a strong dc field in the core. As discussed in section 1.2.1, the core will likely

require an air gap to avoid saturation. This reduces inductance (limiting low-frequency response) and increases even-order

distortion products. Such a single-ended pentode power amplifier was widely used in 5-tube table radios of the fifties and sixties.



Figure 35 - Condenser Microphone

Output Transformer

Figure 36 - A 3-Way Microphone “Splitter” Box

Figure 37 - A Transformer-Isolated “Direct Box”

2.1.6 Microphone Output

There are two basic types of output transformers used in microphones, step-up and

step-down. In a ribbon microphone, the ribbon element may have an impedance of

well under 1 Ù , requiring a step-up transformer with a turns ratio of 1:12 or more to

make its nominal output impedance around 150 Ù . Typical dynamic elements have

impedances from 10 Ù to 30 Ù , which require step-up turns ratios from 1:2 to 1:4.

These step-up designs are similar to line output transformers in that they have no

Faraday or magnetic shields, but are smaller because of lower signal levels.

A condenser microphone has integral circuitry to buffer and/or amplify the signal

from its extremely high-impedance transducer. Since this low-power circuitry

operates from the “phantom” supply, it may be unable to directly drive the 1.5 kÙ

input impedance of a typical microphone preamp. The output transformer shown in Figure 35, which has an 8:1 step-down ratio,

will increase the impedance seen by Q1 to about 100 kÙ . Due to its high turns ratio, a Faraday shield is used to prevent capacitive

coupling of primary signal to the output.

2.2 System-Level Applications

2.2.1 Microphone Isolation or “Splitter”

The primary of a transformer with a 1:1 turns ratio can

“bridge” the output of a 150 Ù to 200 Ù microphone

feeding one pre-amp and the secondary of the

transformer can feed a duplicate of the microphone

signal to another pre-amp. Of course, a simple “Y” cable

could do this but there are potential problems. There are

often large and noisy voltages between the grounds of

two pre-amplifiers. The isolation provided by the

transformer prevents the noise from coupling to the

balanced signal line. To reduce capacitive noise

coupling, Faraday shields are included in better designs

and double Faraday shields in the best. As discussed in section 11.1.3.5, the input impedances of all the pre-amps, as well as all

the cable capacitances, will be seen in parallel by the microphone. This places a practical upper limit on how many “ways” the

signal can be split. Transformers are commercially available in 2, 3, and 4-winding versions. A 3-way splitter box schematic is

shown in Figure 36. Since the microphone is directly connected only to the “direct” output, it is the only one that can pass

phantom power to the microphone. To each preamp, each isolated output "looks like" a normal floating (ungrounded) microphone.

The ground lift switches are normally left open to prevent possible high ground current flow in the cable shields.

2.2.2 Microphone Impedance Conversion

There are some legacy dynamic microphones which are high-impedance (about 50 kÙ) and have two-conductor cable and

connector (unbalanced). When such a microphone must be connected to a standard balanced low-impedance microphone pre-amp,

a transformer with a turns ratio of about 15:1 is necessary. Similar transformers can be used to adapt a low-impedance microphone

to the unbalanced high-impedance input of a legacy pre-amplifier.

2.2.3 Line to Microphone Input or “Direct Box”

Because its high-impedance input accepts line-level signals and its output

drives the low-level, low-impedance microphone input of a mixing

console, the device shown in Figure 37 is called a “direct box.” It is most

often driven by an electric guitar, synthesizer, or other stage instrument.

Because it uses a transformer, it provides ground isolation as well. In this

typical circuit, since the transformer has a 12:1 turns ratio, the impedance

ratio is 144:1. When the microphone input has a typical 1.5 kÙ input

impedance, the input impedance of the direct box is about 200 kÙ . The

transformer shown has a separate Faraday shield for each winding to

minimize capacitively coupled ground noise.



Figure 39 - Unbalanced Output to Unbalanced Input

Top = None, Center = Output, Bottom = Input

Figure 40 - Unbalanced Output to Balanced Input


Figure 38 - Balanced Output to Balanced Input


2.2.4 Line Isolation or “Hum Eliminators”

There are a remarkable number of “black boxes” on the market

intended to solve “ground loop” problems. This includes quite a

number of transformer-based boxes. With rare exception, those

boxes contain output transformers. Tests were performed to

compare noise rejection of the original interface to one with an

added output transformer and to one with an added input

transformer. The tests accurately simulated typical real-world

equipment (see the definitions at the end of this section).

Figure 38 shows results of CMRR tests on a balanced interface

using the IEC 60268-3 test procedure (discussed in section 3.1.2).

This test recognizes that the impedances of real-world balanced

outputs are not matched with the precision of laboratory

equipment. While the output transformer reduces 60 Hz “hum” by

over 20 dB, it has little effect on “buzz” artifacts over about 1

kHz. The input transformer increases rejection to over 120 dB at

60 Hz and to almost 90 dB at 3 kHz, where the human ear is most

sensitive to faint sounds.

Figure 39 shows results of ground noise rejection tests on an

unbalanced interface. By definition, there is 0 dB of inherent

rejection in an unbalanced interface (see Chapter 37).While the

output transformer reduces 60 Hz “hum” by about 70 dB, it

reduces “buzz” artifacts around 3 kHz by only 35 dB. The input

transformer increases rejection to over 100 dB at 60 Hz and to

over 65 dB at 3 kHz.

Figure 40 shows results of CMRR tests when an unbalanced

output drives a balanced input. A two-wire connection of this

interface will result in zero rejection (see Chapter 37). Assuming a

three-wire connection, the !30 dB plot shows how CMRR of

typical electronically-balanced input stages is degraded by the 600

Ù source imbalance. Again, the output transformer improves 60

Hz “hum” by over 20 dB, it has little effect on “buzz” artifacts

over about 1 kHz. The input transformer increases rejection to

over 120 dB at 60 Hz and to almost 90 dB at 3 kHz.

Figure 41 shows results of ground noise rejection tests when a

balanced output drives an unbalanced input. Because our

balanced output does not float, the direct connection becomes an

unbalanced interface having, by definition, 0 dB of rejection.

While the output transformer reduces 60 Hz “hum” by about 50

dB, it reduces “buzz” artifacts around 3 kHz by less than 20 dB.

The input transformer increases rejection to over 105 dB at 60 Hz

and to almost 75 dB at 3 kHz. In this application it is usually

desirable to attenuate the signal by about 12 dB (from +4 dBu or

1.228 volts to !10 dBV or 0.316 volts) as well as provide ground

isolation. This can be conveniently done by using a 4:1 step-down

input transformer such as the one in Figure 29, which will

produce rejection comparable to that shown here.

One might fairly ask “Why not use a 1:4 step-up transformer when

an unbalanced output drives a balanced input to get 12 dB of

signal gain?” Because of the circuit impedances involved, the

answer is because it doesn’t work very well. Recall that a 1:4

turns ratio has an impedance ratio of 1:16. This means that the



Figure 41 - Balanced Output to Unbalanced Input


input impedance of the “pro” balanced input we drive will be

reflected back to the “consumer” output at one-sixteenth that.

Since the source impedance (usually unspecified, but not the same

as load impedance) of a consumer outputs is commonly 1 kÙ or

more, the reflected loading losses are high. A 1:4 step-up

transformer would have its own insertion losses, which we will

rather optimistically assume at 1 dB. The table below shows

actual gain using this transformer with some typical equipment

output and input impedances (Z is impedance).

Consumer Pro Balanced Input Z

Output Z 10 kÙ 20 kÙ 40 kÙ

� (625 Ù) (1.25 kÙ) (2.5 kÙ)

200 Ù 8.6 dB 9.7 dB 10.3 dB

500 Ù 5.9 dB 8.1 dB 9.4 dB

1 kÙ 2.7 dB 5.9 dB 8.1 dB

Not only will gain usually be much less than 12 dB, the load

reflected to the consumer output (shown in parentheses) may

cause headroom loss, increased distortion, and poor low-

frequency response. Often the only semi-technical description of a consumer output is “10 kÙ minimum load.” It is futile to

increase the turns ratio of the transformer in an attempt to overcome the gain problem — it only makes the reflected loading losses

worse! In most situations, a 1:1 transformer can be used because the “pro” equipment can easily provide the required gain. Of

course, a 1:1 input transformer will provide far superior noise immunity from ground loops.

The point here is that the noise rejection provided by an input transformer with a Faraday-shield is far superior to that

provided by an output type. But the input transformer must be used at the receiver or destination end of an interface cable. In

general, input transformers can drive no more than two feet of typical shielded cable — the capacitance of longer cables will erode

their high-frequency bandwidth. Although output type (no Faraday shield) transformers are not as good at reducing noise, their

advantage is that they can be placed anywhere along an interface cable, at the driver end, at a patch-bay, or at the destination end,

and work equally well. In all the test cases discussed in this section, results of using both an output and an input transformer

produced results identical to those using only an input transformer. For example, an unbalanced output does not need to be

“balanced” by a transformer before transmission through a cable (this is a corollary of the balance versus symmetry myth), it needs

only an input transformer at the receiver. There is rarely a need to use both types on the same line.

Definitions:

“Balanced Output” means a normal, non-floating source having a differential output impedance of 600 Ù and common-mode

output impedances of 300 Ù , matched to within ± 0.1%;

“Balanced Input” means a typical electronically-balanced stage (an “instrumentation” circuit using 3 op-amps) having a

differential input impedance of 40 kÙ and common-mode input impedances of 20 kÙ , trimmed for a CMRR over 90 dB when

directly driven by the above “Balanced Output”;

“Unbalanced Output” means a ground-referenced output having an output impedance of 600 Ù . This is representative of typical

consumer equipment;

“Unbalanced Input” means a ground-referenced input having an input impedance of 50 kÙ . This is representative of typical

consumer equipment;

“No Transformer” means a direct wired connection;

“Output Transformer” means a Jensen JT-11-EMCF (a popular line output transformer); and

“Input Transformer” means a Jensen JT-11P-1 (our most popular line input transformer).



Figure 42 - Step-Up

Auto-Transformer

2.2.5 Speaker Distribution or “Constant Voltage”

When a number of low-impedance speakers are located far from a power amplifier, there are no good methods to interconnect

them in a way that properly loads the amplifier. The problem is compounded by the fact that power losses due to the resistance of

the inter-connecting wiring can be substantial. The wire gauge required is largely determined by the current it must carry and its

length. Borrowing a technique from power utility companies, boosting the distribution voltage reduces the current for a given

amount of power and allows smaller wire to be used in the distribution system. Step-down “matching” transformers, most often

having taps to select power level and/or speaker impedance, are used at each location. This scheme not only reduces the cost of

wiring but allows system designers the freedom to choose how power is allocated among the speakers. These so-called “constant-

voltage” speaker distribution systems are widely used in public address, paging, and background music systems. Although the

most popular is 70-volt, others include 25-volt, 100-volt, and 140-volt. Because the higher voltage systems offer the lowest

distribution losses for a given wire size, they are more common in very large systems. It should also be noted that only the 25-volt

system is considered “low-voltage” by most regulatory agencies and the wiring in higher voltage systems may need to conform to

ac power wiring practices.

It is important to understand that these nominal voltages exist on the distribution line only when the driving amplifier is operating

at full rated power. Many specialty power amplifiers have outputs rated to drive these lines directly but ordinary power amplifiers

rated to drive speakers can also drive such lines, according to the following table:

Amplifier Rated Output, Watts Output

at 8 Ù at 4 Ù at 2 Ù Voltage

1,250 2,500 5,000 100

625 1,250 2,500 70.7

312 625 1,250 50

156 312 625 35.3

78 156 312 25

For example, an amplifier rated to deliver 1,250 watts of continuous average power into an 8 Ù load

could drive a 70-volt distribution line directly as long as the sum of the power delivered to all the

speakers doesn’t exceed 1,250 watts. Although widely used, the term “rms watts” is technically

ambiguous [5]. In many cases, the benefits of constant-voltage distribution are desired, but the total

power required is much less. In that case a step-up transformer can be used to increase the output voltage of an amplifier with less

output. This is often called “matching” it to the line because such a transformer is actually transforming the equivalent line

impedance down to the rated load impedance for the amplifier. Most of these step-up transformers will have a low turns ratio. For

example, a 1:1.4 turns ratio would increase the 50-volt output to 70 volts for an amplifier rated at 300 watts into 8 Ù . In such low-

ratio applications, the auto-transformer discussed in section 11.1.2.2 has cost and size advantages. Figure 42 is a schematic of an

auto-transformer with taps for turns ratios of 1:1.4 or 1:2 which could be used to drive a 70-volt line from amplifiers rated for

either 300 or 150 watts respectively at 8 Ù . Several power amplifier manufacturers offer such transformers as options or

accessories.

A “line to voice-coil” transformer is usually necessary to step-down the line voltage and produce the desired speaker power:

Speaker Power in Watts Speaker Transformer Step-Down Turns Ratio Required

16 Ù 8 Ù 4 Ù Volts 100 V 70 V 35 V 25 V

32 64 128 22.63 4.42 3.12 1.56 1.10

16 32 64 16 6.25 4.42 2.21 1.56

8 16 32 11.31 8.84 6.25 3.12 2.21

4 8 16 8 12.5 8.84 4.42 3.12

2 4 8 5.66 17.7 12.5 6.25 4.42

1 2 4 4 25 17.7 8.84 6.25

0.5 1 2 2.83 35.3 25 12.5 8.84

0.25 0.5 1 2 50 35.3 17.7 12.5

0.125 0.25 0.5 1.41 71 50 25 17.7



Figure 43 - Transformer

with Secondary Taps for

Power Selection

Figure 45 - “Repeat Coil” Ground Isolation for 600 Ù Lines

Figure 44 - Transformer

with Primary Taps for

Power Selection

These step-down transformers can be designed several ways.

Figure 43 shows a design where the line voltage is selected at the

primary side and the power level is selected at the secondary while

Figure 44 shows a design where power level is selected on the

primary side and speaker impedance is selected at the secondary.

As clear from the repeating patterns in the table above, there are

many combinations of line voltage, speaker impedance, and power

level that result in the same required turns ratio in the matching

transformer.

Since the constant-voltage line has a very low source impedance,

and the transformer is loaded by a low-impedance loudspeaker,

transformer high-frequency response is usually not a design issue.

As in any transformer, low-frequency response is determined by primary inductance and total source impedance, which is

dominated by the primary winding resistance since the driving source impedance is very low. Winding resistances of both primary

and secondary contribute to insertion loss. In efforts to reduce size and cost, the fewest turns of the smallest wire possible are often

used, which raises insertion loss and degrades low-frequency response. Generally, an insertion loss of 1 dB or less is considered

good and 2 dB is marginally acceptable for these applications.

It is very important to understand that, while the frequency response of a transformer may be rated as !1 dB at 40 Hz, its rated

power does NOT apply at that frequency. Rated power, or maximum signal level is discussed in section 1.3.1. In general, level

handling is increased by more primary turns and more core material and it takes more of both to handle more power at lower

frequencies. This ultimately results in physically larger, heavier, and more expensive transformers. When any transformer is driven

at its rated level at a lower frequency than its design will support, core saturation is the result. The sudden drop in permeability of

the core effectively reduces primary inductance to zero. The transformer primary now appears to have only the dc resistance of its

winding, which may be only a few ohms. In the best scenario, some ugly-sounding distortion will occur and the line amplifier will

simply current limit. In the worst scenario, the amplifier will not survive the inductive energy or “kick” fed back to it as the

transformer comes out of saturation. This can be especially dangerous if large numbers of transformers saturate simultaneously.

In 1953, the power ratings of “speaker matching transformers” were based on 2% distortion at 100 Hz [6]. Traditionally, the

normal application of these transformers has been speech systems and this power rating standard assumes very little energy will

exist under 100 Hz. The same reference recommends that transformers used in systems with “emphasized bass” should have

ratings higher than this 100-Hz “nominal power” rating and those used “to handle organ music” should have ratings of at least four

times nominal. Since the power ratings for these transformers is rarely qualified by a specification stating the applicable frequency,

it seems safe to assume that the historical 100 Hz power rating applies to most commercial transformers available today.

If a background music system, for example, requires good bass response, it is wise to use over-rated transformers. Reducing the

voltage on the primary side of the transformer will extends its low-frequency power handling. Its possible, using the table above,

to use different taps to achieve the same ratio while driving less than nominal voltage into the transformer primary. For example, a

70-volt line could be connected to the 100-volt input of the transformer in Figure 33 and, for example, the 10-watt secondary tap

used to actually deliver 5 watts. In any constant-voltage system, saturation problems can be reduced by appropriate high-pass

filtering. Simply attenuate low-frequency signals before they can reach the transformers. In voice-only systems, problems that arise

from breath pops, dropped microphones, or signal switching transients can be effectively eliminated by a 100-Hz high-pass filter

ahead of the power amplifier. In music systems, attenuating frequencies too low for the speakers to reproduce can be similarly

helpful.

2.2.6 Telephone Isolation or “Repeat Coil”

In telephone systems it was sometimes necessary to

isolate a circuit which was grounded at both ends. This

“metallic circuit” problem was corrected with a “repeat

coil” to improve “longitudinal balance.” Translating

from telephone lingo, this balanced line had poor

common-mode noise rejection which was corrected with

a 1:1 audio isolation transformer. The Western Electric

111C repeat coil was widely used by radio networks and

others for high-quality audio transmission over 600 Ù

phone lines. It has split primary and secondary windings



Figure 46 - Two-Transformer “Hybrid”

Figure 47 - Single-Transformer “Hybrid”

Figure 48 - Step-Up Transformer for

Moving Coil Phono Pickup

and a Faraday shield. Its frequency response was 30 Hz to 15 kHz and it had less than 0.5 dB insertion loss. Split windings allow

them to be parallel connected for 150 Ù use.

Figure 45 shows a modern version of this transformer as a general purpose isolator for low-impedance circuits, such as in a

recording studio patch-bay. Optional components can be useful in some applications. For example, network R1 and C1 will flatten

the input impedance over frequency, R2 will trim the input impedance to exactly 600 Ù , and R3 can be used to properly load the

transformer when the external load is high-impedance or “bridging.”

2.2.7 Telephone Directional Coupling or “Hybrid”

Telephone “hybrid” circuits use bridge nulling principles to separate signals

which may be transmitted and received simultaneously on a 2-wire line. This

nulling depends critically on well-controlled impedances in all branches of

the circuits. This nulling is what suppresses the transmit signal (your own

voice) in the receiver of your phone while allowing you to hear the receive

signal (the other party).

A two-transformer hybrid network is shown in Figure 46. The arrows and

dashed lines show the current flow for a signal from the transmitter TX.

Remember that the dots on the transformers show points having the same

instantaneous polarity. The transformer turns ratios are assumed to be 1:1:1.

NWhen “balancing network” Z has an impedance that matches the line

L Limpedance Z at all significant frequencies, the currents in the Z loop

N(upper) and Z loop (lower) will be equal. Since they flow in opposite

directions in the RX transformer (right), there is cancellation and the TX

signal does not appear at RX. A signal originating from the line rather than TX is

not suppressed and is heard in RX. A common problem with hybrids of any kind is

Nadjusting network Z to match the telephone line, which may vary considerably in

impedance even over relatively short time spans.

If the transmitter and receiver are electrically connected, the single transformer

method, shown in Figure 47, can be used. Any well-designed transformers with

accurate turns ratio can be used in hybrid applications.

2.2.8 Moving-Coil Phono Step-Up

Outboard boxes are sometimes used to adapt the output of low-output,

low-impedance moving-coil phono pickups to pre-amplifier inputs

intended for the more ordinary high-impedance moving-magnet pickups.

These pre-amplifiers have a standard input impedance of 47 kÙ . Figure 48

shows a 1:37 step-up transformer used for this purpose. It has a voltage

gain of 31 dB and reflects its 47 kÙ pre-amplifier load to the pickup as

about 35 Ù . This keeps loading loss on the pickup to about 1 dB. The

series RC network on the secondary provides proper damping for smooth

frequency response. Double magnetic shield cans are used because of the

very low signal levels involved and the low-frequency gain inherent in the

RIAA playback equalization. In these applications, it is extremely

important to keep all leads to the pickup tightly twisted to avoid hum from

ambient magnetic fields.



Figure 49 - Transmission Tests for Output Types Figure 50 - Transmission Tests for Input Types

Figure 51 - Common-Mode Test for Output Types Figure 52 - IEC Common-Mode Test for Input Types

3 Measurements and Data Sheets

3.1 Testing and M easurements

3.1.1 Transmission Characteristics

The test circuits below are the basic setups to determine the signal transmission characteristics of output and input type

transformers, respectively, shown in the diagrams as DUT for “device under test.” In each case, the driving source impedance

must be specified and is split into two equal parts for transformers specified for use in balanced systems. For example, if a 600 Ù

balanced source is specified, the resistors Rs/2 become 300 Ù each. The generator indicated in both diagrams is understood to

have symmetrical voltage outputs. The buffer amplifiers shown are used to provide a zero source impedance, which is not

available from most commercial signal sources. The generator could be used in an unbalanced mode by simply connecting the

lower end of the DUT primary to ground. The specified load impedance must also be placed on the secondary. For output

transformers, the load and meter are often floating as shown in Figure 49. For input transformers, a specified end the secondary is

generally grounded as shown in Figure 50.

LThese test circuits can be used to determine voltage gain or loss (turns ratio when R is infinite), frequency response, and phase

response. If the meter is replaced with a distortion analyzer, distortion and maximum operating level may be characterized. Multi-

purpose equipment such as the Audio Precision System 1 or System 2 can make such tests convenient. Testing of high-power

transformers usually requires an external power amplifier to boost the generator output as well as some hefty power resistors to

serve as loads.

3.1.2 Balance Characteristics

Tests for common-mode rejection are intended to apply a common-mode voltage through some specified resistances to the

transformer under test. Any differential voltage developed then represents undesired conversion of common-mode voltage to

differential mode by the transformer. In general terms, CMRR or common-mode rejection ratio, is the ratio of the response of a

circuit to a voltage applied normally (differentially) to that same voltage applied in common-mode through specified impedances.

This conversion is generally the result of mismatched internal capacitances in the balanced winding. For output transformers, the

most common test arrangement is shown in Figure 51. Common values are 300 Ù for RG and values from zero to 300 Ù for Rs/2.

Resistor pairs must be very well matched.

Traditionally, CMRR tests of balanced input stages involved applying the common-mode voltage through a pair of very tightly-

matched resistors. As a result, such traditional tests were not accurate predictors of real-world noise rejection in some very widely

used electronically-balanced inputs. The IEC recognized this a number of years ago and solicited help to revise the test. The

problem arises from the fact that the common-mode output impedances of “balanced” sources in typical commercial equipment

are not matched with laboratory precision. Imbalances of 10 Ù are quite common. This author, through an educational process

about balanced interfaces in general, suggested a more realistic test which was adopted by the IEC in their document 60268-3

“Testing of Amplifiers” in August, 2000. The “Informative Annex” of this document is a concise short-course explaining the

nature of a balanced interface. The method of the new test, as shown in Figure 52, is simply to introduce a 10 Ù imbalance, first in

one line and then in the other. The CMRR is then computed based on the highest differential reading observed.



Figure 53 - Impedance Tests

3.1.3 Resistances, Capacitances, and Other Data

Other data which can be very helpful to an equipment or system designer includes resistances

of each winding and capacitances from winding to winding or winding to Faraday shield or

transformer frame. Do not use an ohmmeter to check winding resistances unless you are able

to later demagnetize the part. Ordinary ohmmeters, especially on low-ohm ranges, can

weakly magnetize the core. If an ohmmeter simply must be used, use the highest ohm range

(where the test current is least).

Capacitances are usually measured on impedance bridges and, to eliminate the effects of

winding inductances, with all windings shorted. Total capacitances can be measured this way, but balance of capacitances across a

winding must be measured indirectly. CMRR tests are effectively measuring capacitance imbalances.

As shown in Figure 53, sometimes the input impedance of a winding is measured with specified load on other windings. This test

includes the effects of primary resistance, secondary resistance, and the parallel loss resistance RC shown in Figure 8 and Figure

13. If specified over a wide frequency range, it also includes the effects of primary inductance and winding capacitances.

Breakdown voltages are sometimes listed as measures of insulation integrity. This is normally done with special equipment,

sometimes called a “hi-pot” tester, which applies a high voltage while limiting current to a very low value.

3.2 Data Sheets

3.2.1 Data to Impress or to Inform?

As with other products, many data sheets and other product specifications are designed to impress rather than inform.

Specifications offered with unstated measurement conditions are essentially meaningless, so a degree of skepticism is always

appropriate before comparisons are made. A few examples:

“Hum Eliminator” and “Line Level Shifter” products with no noise rejection specs of any kind!

“Line Level Shifter” products with no gain spec at all! Section 2.2.4 explained why you likely never see one.

“Maximum Power” or “Maximum Level” listed with no frequency and no source impedance specified!

Other specifications, while true, may mislead those not wise in the ways of transformers.

“Maximum Level” and “Distortion” are commonly specified at 50 Hz, 40 Hz, 30 Hz, or 20 Hz. Be careful, the 50 Hz specs will

always be much more impressive than those at 20 Hz! There is an approximate 6 dB per octave relationship at work here. A

transformer specified for level or distortion at 40 Hz for example, will handle about 6 dB less level at 20 Hz and have at least

twice the distortion!

Seen in advertising copy: “Frequency response 10 Hz to 40 kHz ±1 dB into 10 kÙ load” and “Distortion less than 0.002% at

1 kHz.” What about the source impedance? Response at 10 Hz and low distortion is a lot easier from a 0 Ù source than from a

1 kÙ source — and 1 kHz is not a very revealing frequency for distortion tests. Section 1.3.1 explains.

3.2.2 Comprehensive Data Sheet Example

For reference, the following is offered as a sample of a data sheet that has been called truly useful and brutally honest.

LINE INPUT TRANSFORMER1:1 FOR "BALANCED BRIDGING" INPUTS

M Ideal for balancing any high-impedance unbalanced inputM Wide bandwidth: !3 dB at 0.25 Hz and 100 kHzM Recommended for levels up to +20 dBu at 20 HzM High input impedance: 13 kÙ with 10 kÙ loadM High common-mode rejection: 107 dB at 60 Hz

This transformer is designed for use in wideband line input stages. Distortionremains very low and CMRR remains high, even when driven by high sourceimpedances. The primary is fully balanced and its leads may be reversed toinvert polarity, if required. A 30 dB magnetic shield package is standard.

TYPICAL APPLICATION

JT-11P-1

PARAMETER CONDITIONS MINIMUM TYPICAL MAXIMUM

Input impedance, Zi 1 kHz, +4 dBu, test circuit 1 12.3 kÙ 13.0 kÙ 13.7 kÙ

Voltage gain 1 kHz, +4 dBu, test circuit 1 !2.6 dB !2.3 dB !2.0 dB

Magnitude response,ref 1 kHz

20 Hz, +4 dBu, test circuit 1, Rs=600 Ù !0.15 dB !0.04 dB 0.0 dB

20 kHz, +4 dBu, test circuit 1, Rs=600 Ù !0.15 dB !0.05 dB 0.0 dB

Deviation from linear phase (DLP) 20 Hz to 20 kHz, +4 dBu, test circuit 1, Rs=600 Ù +0.6E ±2.0E

Distortion (THD)1 kHz, +4 dBu, test circuit 1, Rs=600 Ù <0.001%

20 Hz, +4 dBu, test circuit 1, Rs=600 Ù 0.025% 0.10%

Maximum 20 Hz input level 1% THD, test circuit 1, Rs=600 Ù +18 dBu +20 dBu

Common-mode rejection ratio (CMRR)50 Ù balanced source

60 Hz, test circuit 2 107 dB

3 kHz, test circuit 2 65 dB 73 dB

Common-mode rejection ratio (CMRR)600 Ù unbalanced source

60 Hz, test circuit 3 100 dB

3 kHz, test circuit 3 68 dB

Output impedance, Zo 1 kHz, test circuit 1, Rs=50 Ù 2.34 kÙ

DC resistancesprimary (RED to BRN) 1.45 kÙ

secondary (YEL to ORG) 1.55 kÙ

Capacitances @ 1 kHzprimary to shield and case 98 pF

secondary to shield and case 110 pF

Turns ratio 0.999:1 1.000:1 1.001:1

Temperature range operation or storage 0E C 70E C

Breakdown voltage(see IMPORTANT NOTE below)

primary or secondary to shield and case, 60 Hz,1 minute test duration

250 V RMS

JT-11P-1 SPECIFICATIONS (all levels are input unless noted)

All minimum and maximum specifications are guaranteed. Unless noted otherwise, all specifications apply at 25EC. Specifications subject to changewithout notice. All information herein is believed to be accurate and reliable, however no responsibility is assumed for its use nor for any infringements ofpatents which may result from its use. No license is granted by implication or otherwise under any patent or patent rights of Jensen Transformers, Inc.IMPORTANT NOTE: This device is NOT intended for use in life support systems or any application where its failure could cause injury or death. Thebreakdown voltage specification is intended to insure integrity of internal insulation systems; continuous operation at these voltages is NOT recommended.Consult our applications engineering department if you have special requirements.

JENSEN TRANSFORMERS, INC., 7135 Hayvenhurst Avenue, Van Nuys, CA 91406-3807, USA

1/01 (818) 374-5857 • FAX (818) 374-5856 • www.jensen-transformers.com



4 Installation and Maintenance

4.1 A Few Installation Tips

! Remember that there are very tiny wires inside an audio transformer. Its wire leads should never be used like a handle to pick

it up. The internal bonds are strong, but one strong tug might result in an open winding.

! Be careful with sharp tools. A gouge through the outer wrapper of an output transformer can nick or cut an internal winding.

! Use either the supplied screws or ones no longer than recommended to mount transformers in shield cans. If the screws are

too long, they’ll bore right into the windings — big problem!

! Be careful about using magnetized tools. If a screwdriver will pick up a paper clip, it shouldn’t be used to install an audio

transformer.

! Don’t drop a transformer. It can distort the fit of the laminations in output transformers and affect their low-frequency

response. Mechanical stress (as in denting) of the magnetic shield can of an input transformer will reduce its effectiveness as a

shield. For the same reason, don’t over-tighten the clamp on transformers mounted with them.

! Twisting helps avoid hum pickup from ambient ac magnetic fields. This is especially true for mic level lines in splitters, for

example. Separately twist the leads from each winding — twisting the leads from all windings together can reduce noise

rejection or CMRR.

4.2 De-Magnetization

Some subtle problems are created when transformer cores and/or their shield cans become magnetized. Generally, cores become

magnetized by having dc flow in a winding, even for a fraction of a second. It can leave the core weakly magnetized. Steel cores,

because of their wider hysteresis loops, are generally the most prone to such magnetization. The only way to know if the core has

some permanent magnetization is to perform distortion measurements. A transformer with an un-magnetized core will exhibit

nearly pure third harmonic distortion, with virtually no even order harmonic distortion while magnetized ones will show

significant even order distortion, possibly with 2nd harmonic even exceeding 3rd. A test signal at a level about 30 or 40 dB below

rated maximum operating level at 20 or 30 Hz is typically the most revealing because it maximizes the contribution of hysteresis

distortion.

Microphone input transformers used with phantom power are exposed to this possibility whenever a microphone is connected or

disconnected from a powered input. However, distortion tests before and after exposure to the worst-case 7 mA current pulses

have shown that the effects are indeed subtle. Third harmonic distortion, which normally dominates transformer distortions, is

unaffected. Second harmonic, which normally is near the measurement threshold, is typically increased by about 20 dB but is still

some 15 dB lower than the third harmonic. Is it audible? Some say yes. But even this distortion disappears into the noise floor

above a few hundred Hz. In any case, it can be prevented by connecting and disconnecting microphones only when phantom

power is off. However, such magnetized transformers can be de-magnetized.

Demagnetizing of low level transformers can generally be done with any audio generator having a continuously variable output It

may take a booster of some sort to get enough level for output transformers (be sure there’s no dc offset at its output!). The idea is

to drive the transformer deeply into saturation (5% THD or more) and slowly bring the level down to zero. Saturation will, of

course, be easiest at a very low frequency. How much level it takes will depend on the transformer. If you’re lucky, the level

required may not be hazardous to the surrounding electronics and the de-magnetizing can be accomplished without disconnecting

the transformer. Start with the generator set to 20 Hz and its minimum output level, connect it to the transformer, then slowly (over

a period of a few seconds) increase the level into saturation — maintain it for a few seconds — then slowly turn it back down to

minimum. For the vast majority of transformers, this process will leave them in a demagnetized state.

Shield cans are usually magnetized by having a brief encounter with a strongly magnetized tool. Sometimes, transformers are

unknowingly mounted on a magnetized chassis. When the shield can of an input transformer becomes magnetized, the result is

microphonic behavior of the transformer. Even though quality input transformers are "potted" with a semi-rigid epoxy compound

to prevent breakage of very fine wires, tiny movements between core and can activate what is essentially a variable reluctance

microphone. In this case, a good strong tape head de-magnetizer can be used to de-magnetize the can. At the end of the production

line, most transformers are routinely demagnetized with a very strong de-magnetizer just prior to shipment. Although I haven't

tried it, I would expect that something like a degausser for 2" video tape (remember that!) would also de-magnetize even a large



steel-core output transformer.

References

[1] Magnetic Shield Corporation, Frequently Asked Questions, www.magnetic-shield.com/faq.html.

[2] Sowter, G.A.V., Soft Magnetic Materials for Audio Transformers: History, Production, and Applications, Journal of the

Audio Engineering Society, October 1987, www.sowter.co.uk/pdf/GAVS.pdf.

[3] Whitlock, Bill, Balanced Lines in Audio: Fact, Fiction, and Transformers, Journal of the Audio Engineering Society, June

1995, pp 454-464.

[4] Smith, F. Langford, Radiotron Designer’s Handbook, Wireless Press, Sydney, 4 Edition, 1953, p 208.th

[5] Woolf, Lawrence, RMS Watt, or Not?, Electronics World, December 1998, pp 1043-1045.

[6] Smith, F. Langford, op. cit., p 227.

Co-Netic® is a registered trademark of Magnetic Shield Corp.

HyMu® is a registered trademark of Carpenter Technology Corp.

Mumetal® is a registered trademark of Telcon Metals, Ltd.

Permalloy® is a registered trademark of B & D Industrial & Mining Services, Inc.

http://www.magnetic-shield.com/faq.html.

Soft Magnetic Materials for Audio Transformers:History, Production, and Applications*

G. A. V. SOWTER

Sowter Audio Transformers, Ipswich IP1 2EL, Suffolk, UK

The history of soft magnetic materials is traced from 1000 B.C. to the present time. Thisincludes a description of the work of Oersted and Faraday who invented the firsttransformer, and the gradual improvements in core material over the last 150 years. Thesecover soft iron, silicon iron, grain orientation, Hi-B steels, domain control by lasers, andspark ablation. Amorphous metallic glasses are also detailed. Finally the design andcharacteristics of a wide range of audio transformers and magnetic shields are discussed, inparticular with regard to Mumetal, which with other nickel-iron alloys has been the author'slifetime occupation.

0 INTRODUCTION

The term "soft" relates to that class of metals or alloyswhich can be easily magnetized and demagnetized asopposed to "hard" magnetic materials used for permanentmagnets. This paper deals exclusively with soft materials,particularly for audio applications.

As far back as 1000 B.C. certain iron ores were found,mainly in Magnesia, a district of Macedonia, pieces ofwhich attracted and repelled each other. These containedFe304 (magnetite) and became known as lodestone, fromthe Saxon "loeden," to lead or direct. Lodestones as foundwere permanently magnetized and their power was named"magnetism." Around 55 B.C. Lucretius wrote "I haveseen Samothracean iron rings even jump up, and at thesame time filings of iron rave within brass basins whenthe magnet stone has been placed under." Later Plinyobserved that iron which has been well touched andrubbed with lodestone is able to take hold of other piecesof iron.

The first use of lodestone as a mariner's compass isattributed to the Chinese. Even before then, it was knownthat a piece of lodestone freely suspended always turns tothe North. The first compasses were magnetized ironneedles on floating straws, but pivoted devices weredeveloped. While visiting the Chinese National Museumin Peking some 20 years ago, the author was shown thewhole range of early Chinese compasses.

The first authentic treatise on the science of magnetismwas written in Latin by William Gilbert of Colchester

* Presented at the 82nd Convention of the Audio EngineeringSociety, London, 1987 March 10-13.

760

PAPERS

who had also studied electrostatics, a science dating backto about 600 B.C., when Thales, by rubbing amber withfur, gave the amber the power of picking up certainobjects. The Greek word for amber was "elektron," andfrom this our word "electricity" is derived. Over thecenturies considerable experimentation and the productionof friction machines to generate electrostatic charges werecompleted and included capacitors and spark dischargedevices. It was not until 1796 that Volta evolved thevoltaic pile to generate a continuous flow of electricity.This consisted of copper and zinc disks placed alternatelyin column form but prevented from touching each otherby means of pieces of moist cloth. This was later replacedby the voltaic cell which consisted of a copper and a zincstrip placed in dilute sulfuric acid and capable of beingjoined externally by copper wires to feed a load. As isknown, hydrogen gathers on the surface of the copperstrip and polarization takes place, limiting the currentoutput.

1 OERSTED'S DISCOVERYOF ELECTROMAGNETISM

Before we consider transformers, the production of amagnetic field by the presence of current is fundamental.In early 1820 the Danish physicist Oersted gave a series oflectures on magnetism and electricity. He made thecurrent from a galvanic trough (voltaic cells in series) passthrough a platinum wire to illustrate the heating effect(forerunner of modern electric heaters). Adjacent was acompass covered with glass, and in the course of thedemonstration, on making the circuit, in the presence ofthe audience, a slight flick of

J. Audio Eng. Soc., Vol. 35, No. 10, 1987 October

PAPERS

the compass needle was noticed. It was not considered tobe of very great significance, but months later, in 1820July, he resumed the research and confirmed that theneedle did actually move. By putting the compass, above,below, and on the sides of the wire carrying current heestablished that the wire was surrounded by a magneticfield. He immediately published a fourpage quartodocument in Latin, describing this epochmaking discovery,and sent it to all learned bodies and distinguishedscientists.

When several turns of wire were wound on amagnetizable core, such as iron, the field was greatlyenhanced, and in 1825 Sturgeon produced the firstelectromagnet. A typical example is the Royal Institution'sgreat electromagnet illustrated in Fig. 1. Electromagnetswere constructed by Franklin in the United States and G. I.Moll of Utrecht, Holland. Magnetizable materials knownat that time were various steels, wrought iron, nickel, andcobalt. (It is interesting to note that the nickel-cobalt alloyPermendur, a 20th-century U.S. discovery now inproduction, has the highest saturation induction of allwell-used commercial alloys, particularly for pole pieces.Some rare earth alloys with even higher saturations existbut are too expensive to come into general use.

Fig. 1. The Royal Institution's great electromagnet.J. Audio Eng. Soc., Vol. 35, No. 10, 1987 October

SOFT MAGNETIC MATERIALS FOR AUDIO TRANSFORMERS

2 FARADAY'S DISCOVERYOF ELECTROMAGNETIC INDUCTION

In the years 1821-1831 Michael Faraday became deeplyinterested in experimentation with electrically producedmagnetic fields and in November 1825 came very close todiscovering electromagnetic induction. He had fiveseparate wires, each 5 ft long, adjacent to each other, andhe passed a current through one of them trying to detectany effect on any of the neighboring wires. Unfortunatelyhis galvanometer was not a delicate one and no effect wasobservable. At that time a galvanometer, orcurrent-measuring device, was no more than a crudecompass near a coil of wire.

On 1828 February 15, at the usual Friday eveninggathering at the Royal Institution in London, there washeld what could have been the first meeting of our AudioEngineering Society. The subject of the lecture was"Resonance or the Reciprocation of Sound." Music wasdemonstrated on instruments from Java, the jew's harp, andwhistles, and a second meeting included sirens andstringed instruments. At the first lecture resonances wereproduced by the then well-known method of strewing sandon a circular disk and drawing a violin bow across theedge. The Chladni (1785) figures showed the naturalresonances of the disk. A second disk of similardimensions was placed under the energized one, which wassimilarly lightly covered with sand. It was then shown thatthe sand on the unenergized disk exhibited the samepattern of Chladni figures.

Michael Faraday was present at these demonstrationsand he perceived that the mechanical work of bowing hadbeen converted into sound energy and then reconvertedinto work on the second disk. This gave him a germ ofinspiration to determine whether electrical energy mightbe converted into magnetism and then reconverted intoelectricity.

Incidentally N. W. McLachlan and the author, in 1930,made Chladni figures with sand and lycopodium powderon disks and wide-angled metal and paper cones todiscover the natural resonances of loudspeakers bybowing. Subsequently energization of cones of many sizesand materials was made by passing audio frequencycurrent through the moving coil attached to the cone andthe sand studied. The frequencies at which these occurredwere confirmed by bridge measurement of the variationsof impedance and radiation resistance at each resonance[1]. It is worth recording that even the resonances of theactual moving coils were found to be audible by bowing,and the frequencies were measured.

Faraday, in 1831 August, did confirm that electricenergy could be converted to magnetism and back toelectricity by the following entry in his diary [2]:Have had an iron ring made (soft iron) round and '1/8 inchesthick and ring 6 inches in external diameter. Wound manycoils of copper wire round one half, the coils beingseparated by twine and calico-there were 3 lengths of wireabout 24 feet long and they could be connected as onelength or used as separate lengths. By trial with a trough,each was insulated from the other. Will call this side of the

761

SOWTER

side of the ring, but separated by an interval, was wound in twopieces together amounting to about 60 feet in length, the directionbeing as with the former coils; this side call B. Since the coil on Aintensified the effect of the current it was logical to presume thatcoil B would intensify the effect of the forces in the ring. Coil Awas capable of being connected to a trough and coil B wasconnected to a Galvanometer. When all was ready, connected theends of one of the pieces on A side with battery; immediately asensible effect on the needle. It oscillated and settled at last in theoriginal position. On breaking connection of A side with batteryagain a disturbance of the needle wave apparently short and sudden.

This is exactly what Faraday wrote in his diary.The discovery of electromagnetic induction resulted

from many months of experimental research which hecontinued for almost 30 years.

Faraday's induction ring was the first transformer evermade, and his description of the toroidal core andwindings does not differ greatly from that of a moderntoroidal mains transformer now so extensively used inaudio equipment (Fig. 2). He even had some idea of theeffect of the turns ratio but suffered from the fact thatcovered insulated wire was not then available.

During the nineteenth century wire coverings of silk orcotton in single or double layers, impregnated papers,Gutta Percha for submarine cables, and rubber wereutilized, to be followed eventually by enamel coatings.

It is worth recording that Faraday also invented the firstdynamo, which gave a supply of direct current from arotating disk (Fig. 3). This greatly enhanced the use ofdirect current for experimental and other purposes andbasically led to the manufacture of highpower commercialgenerators.

Toward the end of that century considerable researchwas undertaken on soft magnetic materials for generatorsand power transformers. The latter, in some instances,consisted of toroidal copper windings with as manysmall-diameter iron wires as possible, forced through thecentral aperture and bent back on themselves to completea magnetic core. Similar construction was used for smallcommunication transformers for telephones. For powertransformers an alternative construction was the use ofsoft iron plates bolted together, but these had appreciablelosses and suffered from deterioration due to aging.

3 PRODUCTION OF NICKEL-IRON ALLOYSIn about 1890 J. A. Ewing had published a book entitled

Magnetic Induction in Iron and Other Metals [3]. This is amost comprehensive study covering various magneticmeasurements, including Weber's ballistic method,magnetization of iron rings and long wires, steel, cast iron,nickel, cobalt, and wrought iron wires. A chapter dealswith hysteresis and the effects of vibration, together withmagnetizing in weak and strong fields. He also studiedeffects of temperature and stress, torsion and twisting, witha final chapter on practical magnetic testing. Consideringthat the period was 1890-1900, it is astonishing that suchcomprehensive

762

PAPERS

research was being carried out on so many magneticmaterials.

Another reason for mentioning this treatise is thatEwing, Hopkinson, and others almost anticipated thediscovery of modern high-permeability alloys such asMumetal and Permalloy, so widely used in audiotransformers. At that time tests were made on nickel-ironalloys containing 4.7% Ni, 25% Ni, 30% Ni, and 33% Ni.

Even the effects of annealing were observed, and hadthe nickel contents been increased further up to 80%, therewould have been created elementary forms of Invar (35%NiFe), Radiometal (50% NiFe), and Mumetal andPermalloy (73-80% NiFe).

After leaving the university in 1922, the author's firstlaboratory work was to measure the magnetic properties ofnickel-iron rods about 5 ft long and 0.25

Fig. 2. Page from Faraday's diary describing experiment andshowing his induction ring the first toroidal transformer.


PAPERS

in diameter, containing 10%, 20%, 30%, 40%, 50%, 60%,70%, 80%, and 90% Ni. These had previously been heatedto about 1000°C and slow cooled. The test equipmentemployed, invented by Weber, is illustrated in Fig. 4,where a. ballistic galvanometer is used to measure flux.

These tests quickly indicated that as the nickel contentwas increased, there was an enormous improvement inmagnetic properties around about 78% Ni content, whichgave the optimum permeabilities.

High permeability is closely allied with lowmagnetostriction, and some years later the author mademagnetostriction tests on a Mumetal rod, again using theWeber ballistic equipment. A 6-ft length of annealed thinMumetal rod was inserted in the magnetizing solenoid, andone end was securely fixed in a large lead block. At theother end a 2000x linear magnification Reichert measuringmicroscope with oil immersion was focused on the grainboundary of a crystal exposed by etching with nitric acid.The first observations showed that the whole laboratory,situated within the works, was in a state of vibration due tothe operation of hot and cold rolling mills and particularlya steam hammer. The result was that the measurements hadto be made in the middle of the night when all was quiet.The magnetostriction movement on the grain boundarywas on the order of one-millionth of its length for thatparticular specimen (Fig. 5). It is interesting to observe thathad magnetostriction measurements been made on theaforementioned series of rods, the optimum compositionfor high permeability might have been confirmed.

Another test carried out by the author was to measurethe permeability of a vertically suspended annealedMumetal wire when various loads were applied to thelower end. This clearly showed that as loads were in

Fig. 3. Page from Faraday's diary showing sketch of firstdynamo.



creased, there was first an improvement in permeabilityand then a decline. It is interesting to note that on modernHi-B transformer steel a small tensile stress is obtained byusing a glass surface coating applied at high temperatureand then cooling. This reduces the losses and raises thepermeability.4 DISCOVERY OF IRON ALLOYEDWITH SILICON FOR TRANSFORMER CONES

During the latter half of the nineteenth centuryconsiderable research on magnetic materials had beencarried out by such persons as Ewing, Rowlands, S. P.Thomson, Steinmetz, and many others, and measurementtechniques became well established. Many properties ofwrought iron, steels, nickel, cobalt, and even somenickel-iron alloys were determined, and it is to beregretted that the full import of the results was notrealized.

In the early 1900s that first major improvement inmaterials for transformers took place when Sir RobertHadfield introduced iron alloyed with silicon which gavehigher permeabilities and appreciably less loss than earliersteels. Various percentages of silicon were utilized and thealloys were sold under a variety of trade names. Thesewere produced from hot-rolled sheets and hadomnidirectional properties. Strain-relieving annealing wassometimes employed, and various coatings were used toreduce eddy current loss.

These sheets were used in the form of butt lapped stripsfor the magnetic cores or power transformers and had onlyabout half of the previous iron losses. The most popularalloy was 3-4% SiFe. Larger values of silicon contentwere investigated even up to 7%, which was found tohave superior magnetic properties, but the material wasbrittle and not easily machinable or stamped.

As an indication of the quality of silicon iron availablein 1915, reference is made to an IEE paper by N. W.McLachlan on Stalloy plates 0.5 mm thick for instrumenttransformers [4]. He found by measurement at 50 Hz thatat 0.01 T the complex permeability was 780 and at 0.1 T,2760. At 0.5 T the value was only 3000.

Fig. 4. Connections for testing iron rods by search coilmethod due to Weber.

763

SOWTER

5 GRAIN-ORIENTED SILICON IRON

Silicon irons with a number of improvements wereutilized for transformers until the late 1930s when abreakthrough occurred due to the introduction ofgrainoriented silicon iron. This was the importantinvention of N. P. Goss, who termed the product Gossiron [5].

This was achieved by altering the silicon content in thesteel, cold rolling the strip to the desired thickness,followed by high-temperature annealing at 1200°C toevolve secondary recrystallization. Large grains wereproduced, oriented in the rolling direction and resulting ingreatly improved magnetic properties along the strip. Moreloss, however, arose across the strip, and this led toconsiderable research on mitered joints, butt joints, andmethods of utilizing as far as possible constructions wherethe flux went along the grains. Obviously toroids here hada big advantage, and subsequently C cores and E coreswere introduced, particularly for small transformers.

While improvements were taking place prior to the1960s such as making thinner Goss material to reduceeddy currents, research was continuing to produce bettersteels. Japan came to the fore and patented their Hi-BSteel which is extensively used today. Here larger grainsare evolved and a small tensile stress is imparted to thesteel by using a glass surface coating applied at hightemperature and resulting in reduced electrical loss.

6 RECENT DEVELOPMENTS

Even in the last few years significant improvements inelectrical steel production have been obtained. As is wellknown, magnetic losses in a core consist partly ofhysteresis, which varies linearly with frequency, and eddycurrent loss, which is proportional to the squares of sheetthickness, frequency, and induction, but inverselyproportional to resistivity. There is however a third loss,mentioned by the author in 1941 [6], which was termeddisaccommodation loss or Nachwirkung loss. It has beenfound that this loss depends on the distance betweendomain walls, and recently by a process of scribing andlaser treating the surface of the strip, losses can be reducedby as much as 10%. Richardson [5] gives further details ofthese treatments and states that electrical steels developedtoday give a 40% improvement on the Goss 0.35-mm strip.British steels are now using spark ablation to give the sameresults as laser scribing. So far, for use in audiotransformer cores, several grades of oriented strip areavailable, termed M grades, from M2 to M7, and these areutilized for the production of small toroids andlaminations. For these purposes the aforementioned veryhigh grade materials are not available yet, possibly foreconomic reasons.

It is noted that all efforts to improve steel materials areconcerned with reducing losses. Fortunately low lossusually means higher magnetic permeability, which in thecase of audio transformers is a most desirable

764

PAPERS

feature to obtain high inductance with the smallest numberof turns. The latter is required to minimize the capacityeffects, as described later.

7 METALLIC GLASS OR AMORPHOUS SOFTMAGNETIC ALLOYS

One final development in magnetic materials over tilelast 20 years has been the production of metallic glass, oramorphous soft magnetic alloys: These are like glass andhave no crystalline structure. They are produced bycontinuous casting and rapid continuous quenching, whichresults in a quick transition from the fluid to the solidphase. The virtue of these materials is that thin strips, suchas 0.05 mm thick (and up to 1 m wide in one instance), canbe made directly from the casting line, thus avoiding theusual hot rolling, cold rolling, and intermediate annealingprocesses. Unfortunately, like glass, they are hard and verybrittle which makes handling and cutting uneconomic.

The composition of metallic glasses may consist ofsome of the following: iron (Fe), boron (B), phosphorus(P), nickel (Ni), carbon (C), copper (Cu), and molybdenum(Mo), a few of these elements constituting a particularbrand. Table 1 gives the properties of metallic glasses thatexisted a few years ago, but research continues [7].

Amorphous metal has been employed in smalldistibution transformers, and a 16-kVA unit which hasonly 20% of the loss of normal oriented silicon steels hasbeen constructed. Amorphous metal is unlikely to be usedin large power transformers owing to its low saturationinduction, but in due course there is a possibility for itsuse in audio transformers if it can be considerablyreduced in price as compared even with Mumetal.

8 THE ORIGIN OF MUMETAL

In the early 1920s Mumetal was developed to act as aloading material for submarine telegraph cables. It wasproduced in high-frequency induction furnaces (theoriginal microwave oven principle), and the 20-lb ingotswere used to make wire 0.010 in diameter. In 1926 for thePacific submarine cable between Bamfield and Fanning,3370 nautical miles in length, thousands of miles of thisMumetal wire were drawn for wrapping around the centralcopper conductor to increase its inductance. This involvedsubsequent annealing to develop the high permeabilityrequired. The effect of the Mumetal wire was to reducegreatly the attenuation of the signals and increase theword-handling capacity. By passing the loaded copperthrough a continuous furnace at about 900°C in a nitrogenatmosphere it also meant continuous measurements ofinductance by the author and others on a definite length ofconductor after passage through the furnace.

1 Mumetal is a registered trademark of Telcon MetalsLtd., Crawley, Sussex, UK.


PAPERS

It is worthy of mention that since so much Mumetalwire was required, quite a number of firms were engagedin its production. It was found that the wire from one firmalways had higher permeabilities than any other, and ittranspired that they used fewer passes betweenintermediate softenings. It was thus proved that coldworking or work hardening produced better magneticqualities, the forerunner of grain orientation.

To cover the improvement in permeability a worldpatent was taken out (British Patent 366523, Smith,


Garnett, and Randall, 1930) and subsequently sold to theU.S. company engaged in the production of orientedsilicon iron.

In the early 1930s the demand for loaded submarinecables slackened and fresh fields for the utilization ofMumetal were explored. Magnetic shields began to berequired and Mumetal toroids for precision instrumenttransformers soon became the fashion. In addition ademand arose for shielding cathode-ray tubes, particularlyfor oscilloscopes and eventually radar equipment.

Fig. 5. Longitudinal magnetostriction effect in nickel-iron alloys.

J. Audio Eng. Soc., Vol. 35, No. 10, 1987 October 765

PAPERS

It is worthy of mention that since so much Mumetalwire was required, quite a number of firms were engagedin its production. It was found that the wire from one firmalways had higher permeabilities than any other, and ittranspired that they used fewer passes betweenintermediate softenings. It was thus proved that coldworking or work hardening produced better magneticqualities, the forerunner of grain orientation.

To cover the improvement in permeability a worldpatent was taken out (British Patent 366523, Smith,


Garnett, and Randall, 1930) and subsequently sold to theU. S. company engaged in the production of orientedsilicon iron.

In the early 1930s the demand for loaded submarinecables slackened and fresh fields for the utilization ofMumetal were explored. Magnetic shields began to berequired and Mumetal toroids for precision instrumenttransformers soon became the fashion. In addition ademand arose for shielding cathode-ray tubes, particularlyfor oscilloscopes and eventually radar equipment.

Table 1. Properties of metal glasses.


Fig. 5. Longitudinal magneto striction effect in nickel-iron alloys.

SOWTER

9 HISTORY OF AUDIO FREQUENCYTRANSFORMERS

The author's first experience with audio transformerswas in 1919 when he examined a war surplus audioamplifier which used "R" bright emitter valves andcontained three kinds of audio transformers. These wereinlet, intervalve, and output types and were all about 2 in 3

with Stalloy (4% silicon iron) cores. It is of interest to notethat a small brass plate on the amplifier case stated "madeby Captain Mullard" with a South London, Streatham,address. It is believed that subsequently he was the founderof the firm of Mullard.

In the 1920s Ferranti Ltd. produced much improvedtypes of intervalve and output transformers, termed the"AF series," which persisted for many years. Thetransformer cores were Armco iron, where the initialpermeability was about 600 and the maximum less than4000. The advantage of these transformers was that thecore section was generous, the windings sandwiched togive good magnetic coupling and spaced for minimumcapacitance. This led to a respectable frequency responsefrom 50 Hz to 8 kHz or slightly above, which wasadequate for the various types of loudspeakers then beingmanufactured.

In the early 1930s N. W. McLachlan and the authorwere engaged in research for a proposed transatlantictelephone cable, to be 2300 nautical miles in length. Thefrequency range was 250 Hz to 2500 Hz, with the receivedsignal strength at the highest frequency only about 1 uVThe transmitter with a shaped frequency characteristic,emphasizing highest frequency, had an input of 200 W.Due to cable attenuation the signal strength on the cablefalls as the operating frequency is increased, and the ratiobetween 250 and 2500 Hz was as indicated in Table 2.The frequency characteristic of the receiving amplifierwas designed to offset this by the aid of Mumetal-coredresonant transformers fed from the then newly inventedscreened grid valves (Tetrodes). British Patent 304710gives the circuit of this amplifier, as shown in Fig. 6.Elaborate precautions had to be taken to decouple thefeeding supplies and screen the transformers. In its finalform the amplifier had six stages, and the followingamplification figures (which include amplification due tothe input and output

PAPERS

transformers) were obtained by measurements made by theauthor and given in British Patent 304710.

The input transformer with a Mumetal core operated atsuch a low level that "noise" had to be minimized and thewinding resistances were reduced by immersing thistransformer in liquid air. Due to the Wall Street crash, thetransatlantic telephone cable could not be financed, but theproject worked well in the laboratory using artificial lineswhich corresponded to the proposed cable.

10 TONE COMPENSATOR FOR PHONOGRAPHRECORDS

Based on the experience gained with resonanttransformers, it was decided to design a tone compensatorfor use with a pickup for phonograph records. This becameknown as the Novotone, which correctly compensated forthe low-note loss due to groove limitations by the use of acarefully designed transformer resonating at about 30 Hz.A second transformer resonating at 4000 Hz and having atertiary winding loaded with a variable resistor permittedvariable high-note compensation as shown in Fig. 7. Thisinstrument was patented and became a commercialsuccess. It is interesting to note that when the twotransformers in the Novotone were first connected up sothat the primaries and the secondaries were in series, theauthor's measurements showed an unexpected reduction ofvoltage in the midfrequency band. After pondering aboutthis for some time, it was realized that the output voltagefrom the 30-Hz transformer, being above the resonantfrequency, was capacitive and that from the 4-kHztransformer, being below the resonant frequency, wasinductive. When the two secondaries were connected inopposition,

Table 2. Voltage amplification.

Frequency (Hz) Voltage amplification250500

1000150020002500

2.4 x 1022.3 x 1035.8 x 1046.8 x 1055.4 x 1061.1 x 10'

Fig. 6. Amplifier with "resonant intervalve transformers" to give voltage amplification of 240 at 250 Hz and 11 million at 2500 Hz (BritishPatent 304710).

766 J. Audio Eng. Soc., Vol. 35, No. 10, 1987 October

PAPE

the performance curve in Fig. 7 was achieved. It shouldperhaps be stated that to get the exact frequencycharacteristics of the resonant transformers, a considerableamount of research was required.

By the early 1930s, nickel-iron alloys such as Mumetal,Permalloy C, Radiometal, and others were firmly

SOFT MAGNETIC MATERIALS FOR AUDIO

established as materials for audio frequency transformers.A very large variety of sizes of laminations becameavailable and eventually led to the formation of acommittee which produced a document giving thepreferred types, particularly for government departments.Many of these sizes are still being manufactured.

Fig. 7. McLachlan Novotone compensator for electrical reproduction of disk records using resonant audio transformers

J. Audio Eng. Soc., Vol. 35, No. 10,

SOWT

During World War II, munitions and communicationsmade great demands on these high-permeabilitylaminations, but a few unusual types of transformers areworth mentioning. Thus the Royal Aircraft Establishmentat Farnborough was interested in determining the vibrationfrequencies and their amplitudes on the mainly woodenaeroplane, the Mosquito, specially designed to avoid radar.In conjunction with the De Havilland Company, whichbuilt Mosquitos, the author designed and manufacturedMumetal-cored transformers operating over the range 4 Hzto 1 kHz, which suited the R. A. E. program. Thetransformers handled the small voltages set up bynickel-chromium wire transducers glued to the vibratingparts and a six-channel amplifier-recorder was built.

After this, a similar demand arose from the Wellcomeresearch laboratories for transformers to operate over 4 Hzto 1 kHz or above for the encephalograph. This is a devicefor measuring the tiny voltages set up between electrodesgummed to the patient's head for the study of brain tumors.Today as many as nine electrodes can be utilized.

It is amusing to recall that the author was invited to thelaboratories to witness the first demonstration of theequipment utilizing an anesthetized dog, on the head ofwhich had been fixed electrodes feeding the amplifier anda recording oscilloscope. Rhythmic signals at lowfrequencies were being observed when the doctor in chargefacetiously asked the author, "Would you like to have yourbrains tested?" Fearing the worst, the author agreed andwas asked to observe the pattern on the screen when heworked out an elementary mathematical calculation. To hisastonishment he found that the record showed a burst ofvoltages during the calculation, which was immediatelyfollowed by a second similar burst. He was told thatsubconsciously he checked his calculations although hewas unaware of this.

Another outstanding device considered during the warwas modified transformers for the Asdic antisubmarineequipment. Toward 1945 the author was also asked toredesign the normal transformers so that, without loss ofperformance, they could be appreciably reduced in size.

After the war ended in 1945 there began improvementsin recording on disks and tapes and a frequency rangespectrum of 40 Hz to 16 kHz became common, althoughsome recording companies specified 20 Hz to 20 kHz,which is normal for many transformers today.

Harmonic generation, today called distortion, then hadbecome important, and the author made a detailed studyof the properties of the nickel-iron alloys from this aspect.For this he was awarded an external Ph.D. by LondonUniversity, his thesis being entitled "Harmonic Distortionin Transformers and Chokes with Nickel Iron Cores" [8].The superiority of Mumetal over other alloys with respectto low distortion was studied and distortion coefficientswere evolved. These enabled designers to predicttransformer distortion on finished transformers, providedthe associated circuit

768

PAPE

parameters were disclosed. In the thesis the importance ofuniform flux density throughout the magnetic circuit wasshown to be essential if distortion is to be minimized. Fig.8 illustrates the distortion coefficients of various magneticmaterials in the form of interleaved assemblies oflaminations. These coefficients are directly proportional tothe resulting distortion, and the superiority of Mumetal isapparent. Fig. 9 gives the distortion coefficients forMumetal in forms other than laminations and emphasizesthe low distortion of highpermeability spiral cores.11 MAGNETIC CORES FOR AUDIOFREQUENCY TRANSFORMERS

The desirable properties for audio frequency magneticcores varies somewhat according to the type oftransformer. For those handling voltages over a widefrequency band, particularly starting at 20 Hz,highpermeability cores are essential to restrict the numberof turns and keep the leakage inductance down. Highresistivity of the magnetic material and low hysteresis andeddy current loss are desirable so that overall core lossesare minimized. Where actual power handling is small andlow cost is desirable, Mumetal 0.38 mm thick is mostlyemployed, although thinner laminations can offer certainadvantages, especially as regards permeability. This isparticularly the case for very small transformers requiredfor printed circuit board mounting. For this the range oflaminations available is somewhat limited but can vary insize from about 10 mm 2 up to a few square centimeters.On the Continent DIN standard sizes exist. Fig. 10 gives afew of the lamination sizes in general use, although forhigh per

Fig. 8. Distortion coefficients of various magnetic materials in theform of interleaved assemblies of laminations.

J. Audio Eng. Soc., Vol. 35, No. 10,

PAPERS

performance or appreciable audio output larger sizes areavailable.

Laminations are generally in the form of Es and Is (tomake up a rectangular form) or Ts and Us, and care mustbe taken in assembly to avoid excess compression orbending since the permeability is easily reduced thereby.There are advantages in higher permeability by usingsingle E laminations (Fig. 11). Impregnation can haveundesired effects, and immersion in incorrect grades ofwax can lead to microphony, that is, minor voltages set upin the windings generally due to relative movements-ofwindings and cores.

It will be noticed that generally the core sectional areasare such that uniform flux density throughout the magneticcircuit is obtained. This is most important to minimizeharmonic generation, and the presence of holes for boltsfor fixing purposes does cause nonuniform fluxconcentration. Laminations are normally interleaved toform a stack, but when this is the case, a strikingphenomenon is observed.

Referring to Table 3, which gives the latest properties ofthe various grades of Mumetal, it will be observed that thesaturation flux density is 0.77 T. Now by examining themanufacturers' curves for Mumetal laminations type 187(Fig. 12) it will be seen that the permeability falls rapidlybeyond 0.3 T, or less than half of the ferric inductionsaturation. This obviously limits the practical maximuminduction at which the transformer can operate since highdistortion starts at this density. Fig. 13 shows the widerange of permeabilities


found over a number of batches of laminations, as sold,which shows initial permeability varying from 16 000 to27 000 and a comparable divergence over the whole usefulrange of flux density. This must be taken into accountwhen designing. The reason for the limitation of maximumworking induction is given in detail in [9], from whichFigs. 14 and 15 are taken. Basically it is due to thecrowding of flux at the imbricated joints in the laminationassembly, which is discussed in detail in this paper.

12 DISTORTION IN AUDIO TRANSFORMERS

Referring to Figs. 8 and 9 it will be noticed thatdistortion increases as the flux density is raised so thatwherever there are flux concentrations, additionaldistortion is produced. With an audio transformer thehighest operating flux levels are at the lowest frequency,and here the maximum distortion occurs. As frequency israised, for a definite operating voltage, the flux isprogressively reduced so that harmonic generation falls.For the higher frequencies in the audio range there is the"skin effect," that is, the flux tends to concentrate on theouter surface of the laminations, which accounts for thefall in effective permeability, as shown in Fig. 16.Obviously for best operation at high audio frequenciesthin material, such as 0.1 mm thick, has advantages asregards both inductance and distortion, but it is expensive.

13 INCREMENTAL OPERATING CONDITIONS

The passage of direct current through winding carryingaudio, frequency currents causes magnetization whichresults in a severe diminution in permeability and limitsthe audio output. This is of particular importance withtransistor amplifiers where heavy direct current can beavailable in the output.

In one case encountered by the author a 300-VA

Fig. 9. Distortion coefficients of Mumetal. O-cut ringstampings with interleaved joints; x-ring-stampings; 0- Fig. 10. Sizes and types of some of the laminations used forspiral core; A-high-permeability spiral core. audio frequency transformers.


SOWTER

push-pull audio output transformer designed by him andfed from two transistor amplifiers was said to have beenbalanced to eliminate any direct current. In fact it wasfound to be unbalanced and passing 2 A direct currentthrough the primary winding, and the effect on the qualityof the speech and music, to put it mildly, was mostpronounced. When the unbalance was found, there hadnever before been such a rush to put two 10 000-RFcapacitors in the lines. Here was a case of enormousdistortion, and the author was reminded of a wartimerequirement of a large number of transformers he designedespecially to create maximum distortion for radiotransmitters to jam unwanted radio reception.

PAPERS

14 DEMAGNETIZATION OF SOFT MAGNETICMATERIALS

When a transformer has been subjected to large valuesof direct current or has been in the vicinity of a strongpermanent magnet, such as too near a loudspeaker, it willassume a polarized state or become magnetized.Fortunately this seldom has a permanently harmful effecton the core material, although in its magnetized state itwill have higher distortion and reduced audio output. Theprocess of demagnetizing is quite simple and consists ofapplying to one winding an alternating current (forexample, at 50 Hz) of value appreciably exceeding

770 J. Audio Eng. Soc., Vol. 35, No. 10, 1987 October

Fig. 11. Higher permeability of E laminations as compared with E and I of same material.Table 3. Summary of typical characteristics of modern Mumetal.

Standard Mumetal Mumetal Plus SupermumetalMagnetic properties

Initial permeability* do 114 60 000 80 000 140 000Maximum permeability do 240 000 300 000 350 000Saturation ferric induction Bsat (Tesla) 0.77 0.77 0.77Remanence, Brem from saturation (Tesla) 0.45 0.45 0.5Coercivity, H° do (A/m) 1.0 0.8 0.55Hysteresis loss at Bsat (J/m3 cycle) 3.2 1.3 0.9Total loss at 0.1 Tesla 50 Hz, 0.1 mm spirated cores (mW/kg) 0.7 0.55 0.35Curie temperature (°C) 350 350 350

Physical properties (similar for all grades)Coefficient of linear expansion, per °C 13 x 10-6Resistivity, uOhm . m 0.6Specific gravity 8.8Thermal conductivity, W/m . °C 33Specific heat, J/kg . °C 440* u4 is measured at 0.4 A/m.

PAPERS SOFT MAGNETIC MATERIALS FOR AUDIO TRANSFORMERS

Fig. 14. Variation of permeability of interleaved laminations with flux density as compared with butted and gapped assemblies. 0-interleave0-butted; O-with 0.001-in gap in one outer limb and butt joint in the other; x -with 0.001-in gap in middle and in both outer limbs. Notechange of scale of ordinate at [L5-5000.


SOWTER

that required for saturation. This current should then bereduced smoothly and gradually over a period of time,such as one-half to one minute. In its most elementaryform this can be done with a suitable variable resistance,such as a potential divider capable of handling the largecurrent. It is important to spend the bulk of the time on thelow values of demagnetizing force, which should bereduced to absolute zero. Another demagnetizing methodis to reverse direct current continuously while reducing itsvalue to zero. It should be stressed that Mumetal does noteasily acquire unidirectional magnetization under normaloperating conditions. Silicon iron, however, can becomepolarized to a small extent if low-frequency ac signals orpulses of values approaching saturation are encountered,and this increases distortion.

15 HISTORY OF MAGNETIC SHIELDING AS USEDFOR AUDIO TRANSFORMERS

Nowadays a good proportion of audio transformers arecontained in high-permeability (usually Mumetal)magnetic shielding cans where 50-Hz "hum" may exist,and these are most effective.

The first evidence of magnetic shielding was in the

Fig. 15. Variation of permeability with flux density of Mumetalspiral cores showing higher possible working densities.x-high-permeability specimen; O-medium-permeability specimen;A-low-permeability specimen.

PAPERS

early 1820s, when it was demonstrated that a horseshoepermanent magnet freely suspended and rotated over acopper disk caused the latter to rotate. (This is theprinciple of the induction motor.) By interfacing variousnonferrous disks between the magnet and the copper diskthere was little effect. When, however, an iron disk wasinterfaced, the copper disk did not move, and this was thefirst evidence of magnetic shielding. Other metals thancopper were tried instead for the rotatable disk, and it wasfound that silver was comparable but bismuth reacted veryweakly. It is now known that electric conductivity is thedesideratum for motion of the disk.

On 1883 November 8 Willoughby Smith, as presidentof the Society of Telegraph Engineers and Electricians,read a paper entitled "Volta-Electrical Induction" [10].This society became the present Institution of ElectricalEngineers (IEE) some years later in 1888.

For demonstration purposes he utilized two woodenframes about 36 in 2 in which were supported flat helicesof insulated copper wire, as indicated in Fig. 17. The coilswere placed some distance apart, and switches D and Ewere mechanically controlled and could be synchronized.

Faraday had also experimented with similar coils withhand-operated static switches as in Fig. 17(b) and foundthat when the space between the coils was filled withinsulating bodies such as sulfur and shellac, there was noeffect on the galvanometer deflection when the circuitwas made or broken. Copper and other nonmagneticmaterials also had no observable effect, and WilloughbySmith wrote:

It is to be regretted that so sound a reasoner and so carefulan experimenter had not the great advantage of the assistance ofsuch suitable instruments for this class of research as theMirror-Galvanometer and the Telephone.

It is noteworthy that both these instruments wereavailable in 1879, Sir William Thomson's mirror re-flecting galvanometer being described in detail by Wil-loughby Smith in his paper read before the Society ofTelegraph Engineers on 1879 February 12 [11].

In his presidential paper Willoughby Smith was able tosend interrupted current through coil A and measure theinduced currents in coil B at various frequencies.

Fig. 16. Curves stressing fall in permeability with increasing frequency and showing advantage of thin laminations.

772 J. Audio Eng. SOC., Vol. 35, No. 10, 1987 October

PAPERS

The results of his tests by placing between the coils sheetsof copper, zinc, tin, iron, and lead show wide differencesin shielding effects and particularly variations as thefrequency increased. (This author found all this equipmentsome 40 years later in his company's stores.)

A more scientific series of shielding tests was carriedout by Constable and Aston at the National PhysicalLaboratory some 40 years ago, and results are given inTable 4. It will be noted that at 50 Hz Mumetal is easilythe most effective, but while copper 1/32 in thick has a valueof 4 dB at 50 Hz it becomes 26 dB at 3200 Hz.

With the passage of time and research on the effects ofimpurities in Mumetal, the permeabilities have greatlyimproved, and comparatively thinner thicknesses forshields are employed. It is even possible to obtain


some degree of shielding by wrapping a transformer invery thin Mumetal tape, 0.05 or 0.1 mm thick, but the usesare limited.

For shielding audio frequency transformers the author,in the 1930s, devised about 10 deep-drawn cylindricalMumetal cans, and many of these sizes have persisted upto the present time. The normal reduction in hum by theuse of these cans is 30-40 dB, but where 50-Hz fields areintense, it is customary to use double shields or even aMumetal shield encased in a second shield having a highersaturation induction, such as Radiometal 50% Ni-50% Fealloy.

When low- and high-frequency audio fields are to beminimized it is possible to copper plate Mumetal shields,although an inner lining of copper foil is preferable.

The usual method of measuring hum reduction is to

Fig. 17. Magnetic shielding measurements by Willoughby Smith in 1882.

Table 4. Screening effects of various materials compared with Mumetal.


SOWT

create a uniform 50-Hz magnetic field by the use ofHelmholtz coils and to measure the voltage setup in atransformer first without a shield, then when encased in ashield.

For some applications, particularly for electromedicalpurposes or electron-microscope observations or where theearth's field is to be minimized or eliminated, it is possibleto line a room with a Mumetal sheet, or even to make aMumetal cabin capable of housing an operator.

16 SOME NOTES ON FREQUENCY RESPONSEOF AUDIO TRANSFORMERS

The lower register of the frequency response is obtainedby the correct selection of the grade and size of themagnetic core and by the number of primary turns. Themaximum operating flux density must be chosen to keepwithin the acceptable distortion range, and this generally isquite a straightforward procedure.

The middle register up to about 10 kHz will generallybe acceptable with the above provisos for the bass,provided that the copper and iron losses are not excessive.

The top register from, say, 10 kHz to 20 kHzdepends

PAPE

to a great extent upon the number of sections in thewindings and the minimization of capacitance betweencoils and other parts of the transformer. The magneticcoupling between primary and secondary windings is ofconsiderable importance since this controls the leakageinductance, and high permeability of the core material ishelpful in effecting this. At high frequencies it can beassumed that the self- and mutual coil capacities, and thosebetween the windings and the core, and electrostatic andmagnetic shields, can be regarded as lumped together toform capacitance C20 in the equivalent circuit (Fig. 18a).If L' represents the leakage inductance and R' is thegenerator resistance plus the equivalent resistance of thetransformer, this forms a series resonant circuit whosefrequency is

This will have an amplification factor Q whose value willbe controlled by the resistance R'.

An alternative method of controlling the Q, whichcauses the rise in voltage output at the resonant frequency,is by loading the transformer with a resistor

Fig. 18. (a) Equivalent circuits of audio transformer at high audio frequencies. (b) Simplest form of equivalent circuit of audio transformer athigh audio frequencies.

77 J. Audio Eng. Soc., Vol. 35, No. 10,

PAPERS

or increasing the secondary winding resistance. Fig. 19 is aseries of curves showing the frequency response of a veryold (1944) fairly bad input transformer for different loadingresistances. With modern transformers the leakageresonance is usually above 20 kHz, but loading thesecondary gives similar effects. The type of loading on a veryhigh grade input transformer can have astonishing effectswhere the frequency response can be made to extend from20 Hz up to beyond 100 kHz, as indicated by the curves inFig. 20. This transformer was designed by the author forDolby Laboratories and was extensively used by them fortheir broadcast equipment.17 PERFORMANCE REQUIREMENTSFOR AUDIO FREQUENCY TRANSFORMERS

A specification could include all or a great deal of thefollowing, much of which depends on the magnetic core.

1) Frequency response, for example, 20 Hz to 20 kHzgenerally, although 2 Hz to 10 kHz has been specified forvibration study transformers and 320 Hz to 320 kHz forhigh-speed cassette-copying equipment

2) Maximum operating level at the lowest operatingfrequency

3) Turns ratio4) Copper resistance

5) Inductance of primary or secondary at a specifiedfrequency and value of excitation and sometimes leakageinductance

6) Permissible transmission loss and whether correctionis to be made by turns adjustment

7) Source impedance8) Load or loads

9) Insulation test on winding to winding and to magneticcore and housing


10) Flash test at prescribed voltage11) Permissible distortion or harmonic generation12) Balance or common-mode rejection13) Mechanical size with any limitations for insertion in a

module14) Type of mounting, such as printed circuit board,

one-hole fixing, or grommet15) Color-coded leads or terminal blocks16) Electrostatic interwinding shields if required

17) Magnetic shielding, usually by Mumetal can andbeing of the order of 30-40 dB at 50 Hz

18) Freedom from microphony19) For large transformers, such as 300 VA, freedom

from acoustic noise generated by core and winding20) For telecommunication transformers, return loss

21) Isolation test (such as 1500 V rms at 50 Hz for 1minute), between windings and metal parts withsubsequent insulation test

22) Maximum voltage permissible if core accidentlysaturated.

A few of the specialized requirements encountered are:1) To operate at 90°C at bottom of oil well drilling2) To operate in liquid nitrogen at 77 K

3) To operate inside diver's helmet with compensation forloss of bass

4) Very low power bridge input transformers5) Audiometric transformers to medical specifications

6) High-ratio transformer in liquid helium for noiseresearch.

18 RANGE OF AUDIO FREQUENCYTRANSFORMERS

It is not normally realized that there is a very largenumber of different audio frequency transformers; thefollowing list covers some of those in general demand:

Frequency, cycles per second

Fig. 19. Frequency and phase characteristics of a circuit containing an input or interstage transformer, showing the effects of varying thesecondary resistance.


SOWTER PAPERS

Fig. 20. Typical performance of wide-band audio transformers showing effects of different secondary loadings.

1) Microphone transformers (all types), including 26) Transformers for electrostatic speakers up to 2000

those for phantom powering2) Transformers for dynamic and moving-coil pickups3) Input transformers (all types)4) Output transformers for mixers5) Multisecondary output transformers6) Bridging transformers7) Line transformers8) Line transformers to isolating test specifications9) Impedance matching transformers, including thosefor high-power loudspeaker distribution10) Balanced transformers, input and output11) Double-screened transformers12) Experimental transformers (all types) for researchprojects13) Audio output transformers for power amplifiersup to 1 kW14) 100-V line transformers for audio amplifiers upto 1kW15) Output transformers for valve amplifiers up to500 W16) Hi-fi loudspeaker transformers for all ratings17) Column loudspeaker transformers for plain andfocused outputs18) Tapped autotransformers for volume control onloudspeakers19) Printed circuit board mounting transformers formixing and recording desks20) Miniature audio transformers for most modules21) Microphone split!er/combiner transformers22) Antimicrophonic transformers23) Low-frequency pulse transformers

24) Vibration study transformers (2 Hz upward)25) Direct injection transformers (for guitars etc )

776

V do27) Hi-fi output transformers for Compact Disc

reproduction28) Induction loop transformers (all ratings).

19 REFERENCES

[1] N. W. McLachlan and G. A. V. Sowter, Philo. Mag.,vol. 11, p. 15 (1931 Jan.)

[2] L. P. Williams, Michael Faraday (Chapman & Hall,London, 1965).

[3] J. A. Ewing, Magnetic Induction in Iron, etc.(Electrician Printing and Publishing Co., 1890).

[4] N. W. McLachlan, J. IEE, vol. 53 (1915 Mar.).[5] B. Richardson, "Transformer Core Losses,"

Electronics c& Power (IEE) (1986 May).[6] G. A. V. Sowter, "Magnetic Properties of Nickel Iron

Alloys," J. Brit. IRE (1941 Aug.).[7] R. Boll, Soft Magnetic Materials (Heyden & Sons,

1979).[8] G. A. V. Sowter, "Harmonic Distortion in Transformers

and Chokes with Nickel Iron Cores," Ph.D. thesis, LondonUniversity, 1944.

[9] G. A. V. Sowter, "Characteristics of Soft MagneticMaterials for Instruments," Proc. IEE, vol. 98, p. 11 (1951Dec.).

[10] W. Smith, "Volta-Electric Induction," presidentialaddress, Society of Telegraph Engineers and Electricians(1883 Nov.).

[11] W. Smith, "Working of Long Submarine Cables,"Society of Telegraph Engineers (1879 Nov. 8).

[12] MIT., Magnetic Circuits and Transformers(Chapman & Hall, London, 1944).


PAPERS SOFT MAGNETIC MATERIALS FOR AUDIO TRANSFORMERS

THE AUTHOR

G. A. V. Sowter was born in London, U.K., and educatedat London University, where he was awarded his B.Sc. inengineering in 1922. He then joined the TelegraphConstruction and Maintenance Co., which made and laid thefirst Atlantic submarine cable, and he was engaged inresearch on magnetic materials. This included the early workand evolution on Mumetal and kindred alloys by the team. Inthe early 1930s, he worked with N. W. McLachlan, thecelebrated pioneer of moving-coil loudspeaker research, anddeveloped a transmitter and receiver for a projectedtransatlantic telephone submarine cable. The economicworld depression of the 1930s terminated this project andlaboratory work on moving coil loudspeakers wasundertaken.

A number of technical papers were published by Dr.Sowter and Dr. McLachlan on loudspeaker articles in thePhilosophical Magazine. During this period the standardtextbook Loudspeakers by N. W. McLachlan was published,based on the extensive measurements carried on by Dr.Sowter. Dr. Sowter became chairman of the measurementsdivision of the Institution of Electrical Engineers, was onthe council for a number of years, and served on many oftheir committees. He received a fellowship of the I.E.E.Previously he had been Chairman of the Council of theBritish Institution of Radio Engineers (now I.E.E.I.E.) andstill enjoys his ham radio-his callsign G20S being allocatedin the early 20s. He has chaired a number of BritishStandard Committees on magnetic materials and has writtenseveral technical papers on this subject. He became GroupCommercial Director of several Telcon Metals companieswhich included factories producing Mumetal

and other alloys.Prior to World War II Dr. Sowter spent several years

teaching at Higher National Certificate level as well asmanaging Telcon Metals Ltd. This resulted in the award ofan external Ph.D. degree at London University, where histhesis was on "Harmonic Distortion in Transformers andChokes with Nickel Iron Cores."

In the late 1930s he became consultant to SowterTransformers, Ipswich, U.K., which produces every type ofaudio frequency transformer. He is still active in this capacityand has made hundreds of designs resulting in the sale ofthousands of transformers handling from a few microwattsup to a kilowatt. His expertise in the properties andapplications of high-permeability magnetic alloys led to thedesign of transformers for the Royal Shakespeare Co.Barbican, Royal Opera House Covent Garden, BBC Studios,Dolby Laboratories, plus many others.

Dr. Sowter has traveled extensively, including a visit toChina 25 years ago. He is treasurer and member of the"Dynamicables" Club, which has celebrated its Centenaryrecently and consists of the 100 outstanding British ElectricalEngineers.

He is also a registered chartered engineer by the BritishEngineering Council. At the 73rd Convention of the AudioEngineering Society in Eindhoven, Dr. Sowter was awardeda fellowship in recognition of his achievements in the audiofield. The citation, approved by Ray Dolby, chairman of theAES Awards Committee, reads: "For contributions to audiotransformer and loudspeaker design, particularly the optimalemployment of magnetic materials."


Documents

Audio Transformers Chapter