The Audio Note Transformer Design Philosophy
Article By Andy Grove And Peter Qvortrup

The Theory

  The electronic valve is a high voltage, low current device that is incapable of driving a low impedance loudspeaker directly. Although output-transformer-less (OTL) designs have appeared from time to time, with these types many devices are connected in parallel and a large amount of negative feedback is used to achieve a workable, but not necessary satisfactory result. The efficiency of the output transformer less circuit is also always very, very low due to the severe impedance mismatch, so a large amount of power must be dissipated within the output valves to achieve a tiny output into the loudspeaker. The only way to correctly match a valve output stage to a low impedance loudspeaker is via a step-down output transformer.

The transformer is sometimes seen as a barrier to amplifier performance, and whilst on a theoretical level this is to an extent is true. A transformer does have a finite bandwidth, but as will be shown and discussed later in this article, when properly researched, designed and made the limits achievable in practice are more than wide enough for what is required by the harmonic envelope of a musical signal. Most of the problems normally referred to in this regard relate to problems in amplifiers utilizing negative feedback. The limited bandwidth (which may still exceed that of the human ear) and associated phase shifts can make the amplifier unstable, this situation is made considerably worse if there is a strong high frequency resonance present in the transformer itself.

The Design
At Audio Note we have spent many hundreds of hours involved in a combination of theoretical research and experimental work to develop and combine proprietary interleaving methods and winding techniques to extend the bandwidth of our transformers to the point at which they could be considered not only excellent components from the technical standpoint, but virtually invisible from a sonic perspective. In some of our designs as many as five wires are wound onto the bobbin at the same time, using these methods a bandwidth of 5Hz to 200kHz is achievable with a transformer for single ended operation of a 300B triode. This extended bandwidth presents the valve with a constant impedance load across the audio range thereby minimizing distortion, and allows all of the harmonic overtones and transient events of the music to be accurately reproduced within the harmonic envelope.

Perhaps not surprisingly, we have found that the materials used within the transformer greatly affect the both the sound quality and measured performance. This is an area largely overlooked both now and in the past cost and ease of use was and still is the primary considerations.

Theoretically speaking, the Interleave Insulation and Primary to Secondary Insulation acts as the dielectric in a distributed capacitor, therefore it can be seen that the properties of the dielectric material will affect the electrical and sonic performance of the transformer. Electrical quantities to be considered are dielectric constant, which affects the magnitude of the resultant distributed capacitor and dielectric absorption, which causes distortion by hysteresis. A vacuum is off course the ideal choice as it has a low dielectric constant and no dielectric absorption, but a vacuum is as impractical as it is unrealizable in anything but a laboratory. We have therefore experimented with every man-made plastic insulating material available, but in the end we found that the best sounding material is a special type of paper. Paper is a natural material, and although subject to variations as are all such natural materials, it is more conducive to creating a natural sound. As with all Audio Note™ products the ear was the final arbiter as to which material was to be used.

The Wires
The wire used to wind the transformer is also critical and in this area as in many others Audio Note was the company that pioneered the finest, Silver and it was therefore natural to put silver to good use in our best output transformers. Why silver sounds so superior is still not fully understood, but it is unlikely to be simply a function of conductivity.

One theory puts forward the notion that the intense AC electrical and magnetic fields within the transformer interact in some way with the wire material. Another theory considers the crystalline structure of each material copper is very sensitive to impurities, in particular oxygen, it is also possible that the differences are caused by effects that occur on the surface of the material. Surface chemistry is different to that of the bulk material, the atoms at the surface are exposed, rather than being enclosed within the crystal lattice. When the metal is drawn into wire the surface will quickly adsorb components of the air, particularly oxygen and nitrogen as they are most prevalent and despite our best efforts (we coat our immediately it leaves the die), some contamination still takes place. After a while a bulk reaction takes place producing a layer of oxide and sulphide. Silver and copper compounds are similar chemically but not identical. Copper oxide is a rather poor semiconductor compound capable of producing rectification effects whereas silver oxide is a good conductor and is used in switch contacts and batteries. It may be possible to draw wire in an inert atmosphere such as argon and then cover the wire before it reaches the air or to chemically treat the surface before coating to further improve the wires.

The Cores
The core of the transformer is vital for it's operation. In our standard transformers we use good quality silicon steels but in our finest specialist transformers we make no compromises and use the very best and very expensive nickel irons such as Radiometal. 3% silicon steel is widely used around the world and is produced in vast quantities. China, America, Japan, Russia and the UK are amongst the countries where this material is manufactured. For our economy transformers we use a material known as M6, in laminations of 0.35mm thickness. The material is first cold rolled, to align the grain structure, into a tape then it is punched into laminations. The problem with this is that the flux runs anti-parallel to the preferred direction at the back of the "E". This means that at that point the materials full potential is not realized at that point increasing losses and decreasing effective permeability. M6 steel has reasonably low hysteresis, good permeability (approximately 10,000) and high saturation flux density (approximately 2T or 20,000 Gauss). The problem of poor grain orientation is alleviated if we move from I-E laminations to a C-Core. Here the metal tape, after being cold rolled, is wound into a loop and then cut, now the magnetic flux always travels in the preferred direction in the steel, this alone gives a significant increase in performance. When we move up to a C-Core we change the material's specifications to M0 or HiB silicon steel a material that has slightly lower losses and higher permeability than M6, the permeability of HiB can be 40,000 or more. HiB is processed in a different way to M6 giving it a different grain structure this special material is manufactured in Japan and America only. Our finest transformers use two versions of Radiometal core in the form of a C-Core. Radiometal is a 36% Nickel iron and Superradiometal a 48% Nickel iron alloy of excellent magnetic properties the permeability is similar to that of HiB but it's saturation flux density is lower at 1.6T or 16000 Gauss. Radiometal has a much lower hysteresis loss than silicon steel and is far more sensitive to small signals. If one is to firstly listen to a transformer with the best silicon steel core and then change to one with the Radiometal core, one experiences more colours and texture in the performance and more low level details are present. The high frequencies are so much clearer. It is like the difference between an artificial light and sunlight.

A Little History
Traditional transformer designers still use winding calculations and technology that were established in the 1940's, just after WWII, these calculations are designed to yield the best results from a standard M6 type C-core and companies like Partridge, Savage, Parmeco and Gardners made excellent examples of these types of conventional Push Pull output transformers in the 1950's and 1960's, however, when the new highly permeable magnetic materials such as the Radiometals emerged in the early 1950's no-one realized that they require a quite different approach to winding technique to get the best magnetic coupling between the windings and the core possible and thereby utilizing the capabilities of these fine magnetic materials fully.

Over 40 years later Audio Note™ is so far the only company in the world to conduct such work. Work, which is further enhanced by the advantage of having both in-house transformer and circuit design capabilities side by side, something which allows Audio Note™ to design our transformers specifically for a specific circuit thus maximizing the harmonic envelope and dynamic transfer and utilizing the best combination of both, because we can always check the sonic properties of any given combination during the prototype stages.

No other audio manufacturer have this in-house facility, and they therefore have to source standard designs from transformer manufacturers who do not have the ability or necessary understanding of electronic circuitry to test and design the best possible transformer for each specific application, but will always supply a compromise.

In contrast Audio Note designs its best transformers practically without cost restraints a fact which has resulted in a transformer quality undreamed of even 20 years ago, the completely "invisible" transformer is a goal so far unattainable, the Audio Note™ silver wired Super Perma 55% nickel C-core transformer is the closest alternative!

The Single-Ended Transformer
One final point of interest with a S.E. transformer is the air gap. This is necessary in order to bring the operating point of the core to the correct region on its B-H curve. It does not seem that anyone has ever experimented with anything other material than paper or plastic for use as a spacer between the core limbs. At Audio Note we have discovered that the use of a metallic spacer reduces the distortion produced by the transformer and the improvement in the sound of the transformer is considerable provided the correct material is used and it is applied in the correct way.

Overall a transformer could be described in a similar way to a culinary dish. To get the best flavor one must use the best ingredients and cook them in the correct way and as new ingredients emerge and are developed, be sure that we at Audio Note™ will be the first cooks to write the new recipes...

The Audio Note Group C
Double C-Core Output Transformers
Shortform Introductory Article 

Why Does Audio Note Choose Double C-Cores Rather Than I-E Core?
This is an area of very little understanding in the modern audio industry and as a result much controversy, so Andy Grove and I would like to give at least some background as to why Audio Note's best transformers use C-cores and always will do, despite cost and availability problems.

Basic (very) domain theory tells us that magnetic steels function by two main processes, domain growth and domain rotation.

Under low magnetization, the field domains, which are oriented in the direction of the applied field, grow at the expense of their non-oriented and anti-parallel neighbors, this low field domain growth is generally reversible if the field is removed.

Under a medium applied field again domain growth is the predominant factor. However there will be some non-reversible growth of the domains, and a reverse field is required to return them to their original state.

Under a large magnetization field those domains, which were not oriented in the direction of the applied field start to rotate towards that direction. Eventually all of the domains are pointing in the direction of the applied field and saturation is reached.

This neatly explains the familiar shape of the B-H curve and hysteresis loop.

Essentially iron crystallizes in a body cubic form and the domains are oriented parallel to the edges of the crystal, therefore an iron crystal will be easier to magnetize if the applied field is parallel to an edge, and will be most difficult to magnetize in a diagonal direction across the cube.

In a non-oriented material the crystals and the domains are oriented randomly, therefore it will magnetize much the same in any direction. However no direction is aligned with the preferred direction of all of the crystals, and a lot of the crystals will be oriented in the worst direction.

Therefore permeability is low and losses are high. The hysteresis loop will be wide and rounded. In a singly oriented steel (M4 etc, there are cubic oriented types which we use as well) the crystals are oriented so that two of the faces are perpendicular to the strip rolling direction, two of the edges are parallel to it and the other edges are at 45 degrees to the strip surface. In other words the plane of the strip cuts the diagonal of the faces, which are perpendicular to it. This means that the material is very easy to magnetize by a field parallel to the strip rolling direction as the domains are facing in that direction.

This makes for a material with a high permeability, low losses and a narrow rectangular hysteresis loop when the field is in the strip direction. But it also means that a field in any other direction in the plane of the strip will be trying to magnetize the crystal in it's worst possible mode. The highest losses always occur at approximately 45 degrees to the rolling direction, in the plane of the strip.

This diagram shows the magnetization behavior of GOSS and non-GOSS strip relative to rolling direction

The I-E Core
Now, laminations are punched out of a steel strip such that the "arms" of the E point in the strip direction, however the back of the E is perpendicular to the rolling direction. The I is punched so that it's longest side is parallel to the rolling direction. This of course means that with an I-E laminated transformer the flux has to curve round, across the grain at the corners, both at the junction of E and I and at the back of the E, and travel perpendicular to it across the back of the E.

In addition I-E laminations generally have whopping great holes punched just where you don't want them. This means that a stack of grain oriented laminations ends up with better properties than a non-oriented stack but not by much. Quite serious curvature of the B-H curve starts to appear at around 1.2T to 1.3 T (Tesla) even though true saturation doesn't occur until about 1.6T to 1.8T.

The C-Core
Enter the C-Core, as the C-Core is wound out of the strip the flux always traverses the preferred direction. This means that a C-Core remains linear almost to saturation and then hits a brick wall around 1.8T maybe a bit more. The losses are much lower as well, and that translates into lower distortion as the hysteresis loop is narrow and straight sided. Let's say that 1.3T peak is as far as one would like to go on lams, and 1.7T peak for a C-Core that is a ratio of 1:1.3. So for a given number of turns on a core of equal dimension the C-Core could sustain a 30% higher voltage across that winding.

That is an increase in power of 70% for a given level of core distortion.

Or translated into the realm of mains power transformers it explains why strip wound cores, especially toroidal transformers are so small for their power rating. Of course increasing the cross sectional area of a stack of lams can equalize the power rating but that brings about an increase in winding length and hence an increase in leakage inductance and capacitance.

In the Audio Note™ high quality output transformers we use 50% or 55% nickel iron alloys, both oriented and non-oriented through a carefully developed, customized and proprietary heat treatment processes depending upon the application. These materials offer greatly reduced distortion at low signal levels, due to the very narrow hysteresis loop of these materials. The downside is their expense and lower saturation flux density, which in single-ended low power amplifier applications is not an issue. A small word about winding technology.

Traditionally, transformer design has focused on achieving the widest possible frequency response and this has been the main tenet of design priority for the past 80 or so years, in our research into the behavior and interaction between coil and core we have discovered that when dealing with the highly permeable nickel cores, purely looking at frequency as the main arbiter of transformer quality is woefully inadequate and as a result we have spent years developing and refining the best way of winding the coil with special focus on improving the low level linearity and bandwidth, the end results speak for themselves.

From a purely practical standpoint, thin materials in the 0.1mm range are impossible to handle as large laminations, especially in the very mechanically soft nickel irons, and the C-Core format allows their use. Very thin laminations or strips are more important to very high permeability materials because eddy currents in thicker material, greatly reduces the effective permeability.

But remember that flux density is inversely proportional to frequency so at 1kHz the flux density in an output transformer will only be 2% of that at 20Hz, and 0.1% of it at 20kHz. Assuming 1.3T peak at 20Hz (for lams of M6) that gives 26mT peak at 1kHz and 1.3mT peak at 20kHz.

A high frequency power transformer such as used in a switch mode supply transformer would run the core at maybe 0.5T or more peak at 20kHz and then losses would become very significant. This is where ferrites with their very high intrinsic resistance become important, but useless for wideband audio applications.

Cobalt irons offer high saturation flux densities but they have a very wide hysteresis loop, not far from a semi-hard material and they don't lend themselves to audio output transformer work where low level resolution is paramount.

This fact does not relegate cobalt based materials from other audio applications, for example permendur (49% cobalt) has uses in pole pieces for magnets in phono cartridges, loudspeakers and for electromagnets as the high saturation flux density allows for a greater density in the gap.

Here the material is generally driven into saturation by a DC polarizing field.

Company Information
Audio Note (UK) Ltd.
25 Montefiore Road,
Hove,
East Sussex
BN3 1RD
United Kingdom

Voice: +44 (0)1273 220 511
Fax: +44 (0)1273 731 498

E-mail: info@audionote.co.uk
website: www.audionote.co.uk