For the PDF version of this article, click here.
Thirty years ago, when the premier issue of Solid-State Power Conversion magazine rolled off the presses — it was the father of the Power Electronics Technology you're reading now — power supplies that provided dc voltages to all kinds of electronic equipment were a much different technology than they are today. And so was power magnetics, the design and manufacture of transformers and inductors that are the most difficult and costly aspect of power-supply design. With its mathematically heavy theory, magnetics has always been a subject designers approach with apprehension. Some industry observers claim that, because of its complexity, even now, it is the most neglected aspect of analog design. Nevertheless, magnetics is at the core of every power supply that connects to the ac line to produce dc voltages. To understand where magnetics technology was three decades ago, you have to go back to the types of power supplies that were being designed for electronic equipment of that time.
In 1977, a Motorola Semiconductor application note, “Switched Mode Power Supplies — Highlighting a 5-V, 40-A Inverter Design,” divided the dc power-supply universe in 1974 into four categories: controlled ferroresonant transformer, SCR phase control, linear regulators (series and shunt) and switched-mode regulators (switching regulators and pulse-width modulated [PWM] inverters). The table lists the features — the advantages, disadvantages, cost, efficiency and size — of each type given in the note. The note went on to state that the first three types “all use bulky 60-Hz transformers for isolation between line and load, so they suffer the common disadvantage of large size. Switched-mode regulators on the other hand, which operate above audio frequencies, can use small 20-kHz power transformers.”
Since the emphasis at that time was on energy conservation, efficiency and small size, because of the emergence of computers on a chip (i.e., 4- and 8-bit microprocessors), the age of the high-frequency switching regulator was born, and with it new challenges for power magnetics. Note, however, that some types of switching supplies were developed in the 1960s for military and aerospace systems, but it was not until the 1970s that improvements in the technology and lower-cost components made them practical for commercial applications.
The trend away from linear supplies and toward switching-regulator types was evident in the lead design article of that first 1975 edition of Solid-State Power Conversion, “Optimizing the High Current Transistor Converter,” by Wally Hersom of LH Research. While acknowledging the excellent regulation characteristics of the then-popular series regulator, the author recognized the drawbacks of its magnetics by stating, “The major material costs in the economical series regulator are associated with basic commodities, such as iron and copper, which (unpleasantly) creep up inexorably.” His choice for the next generation of dc power supplies was the off-line converter, which is similar to a series switching regulator that is operated directly from a rectified ac input and whose rectified and filtered dc output is controlled through pulse-width modulation of a high-frequency symmetrical wave. One of the benefits of an off-line converter from both cost and size standpoints is that it eliminates the need for a large 60-Hz power transformer. Instead, the converter uses a more economical high-frequency (20-kHz) power transformer, similar to that described in the Motorola application note.
Hersom's article describes the design of a 5-V, 100-A converter with an extensive description of the 20-kHz transformer calculations needed for the power stage. The analysis begins with equations to determine the transformer volumetric efficiencies of the different types of bridge circuits that could be used in the output stage (full wave, half wave, etc.). Next come suggestions for the kind of transformer core to be used. At that time, metal tape-wound cores were the standard approach in transformers, but Hersom notes the “recent availability of high-flux density ferrites, which are provided in a variety of economical molded shapes, making this material ideal for converter applications.”
Ferrites were becoming a viable alternative because of their low losses at high frequencies and lower cost compared to tape-wound cores. The section concludes with equations for manually calculating the transformer's primary and secondary voltages and the number of primary and secondary turns required for the core.
Comparing Design Techniques
Some striking comparisons can be made between Wally Hersom's 1975 design techniques and the way things are done today. First, in the absence of personal computers and the modeling software available now, all design computations back then had to be carried out by hand using some fairly unwieldy equations. In fact, the mathematics involved with magnetics theory and design is conceded by many in the transformer and inductor business to be the most difficult aspect of power-supply design.
Moreover, no guidance is given for dealing with the more complex transformer parameters, such as core dissipation consisting of eddy current and hysteretic losses, and how to compare flux density versus core losses for different types of core materials. In order to evaluate such parameters, designers had to obtain data sheets from manufacturers containing a complicated set of electrical, mechanical, winding and physical parameters for various types of core configurations. Before an actual transformer could be built, much more information had to be provided to the transformer builder. With today's software and Internet accessibility, obtaining such data from magnetics manufacturers and plugging it into the programs can make complex transformer and inductor calculations much easier than in the past.
Today's computer-aided engineering (CAE) software for magnetics design falls into two basic categories. One type performs the analysis and synthesis of transformers and inductors based on design parameters the user enters into the program. These could include data such as wire size and type, insulation thickness and inductance, among others. The program can use this data to synthesize a core type and winding arrangement from among many different cores.
The second type uses finite element analysis (FEA) techniques, which chop a structure such as a transformer into a large number of tiny elements for analysis. The latest FEA programs (see “Modeling Integrated Magnetic Components,” Power Electronics Technology, March 2005, pp. 14-22) can define the core shape and material and the winding arrangement. Both kinds of programs enable a designer to not only synthesize a magnetic component, but also to analyze how that component performs electrically in the specific topology (forward, push-pull, etc.) of the power supply.
The great advantages of CAE are that it allows designers to quickly perform many “what-if” scenarios to evaluate design tradeoffs much faster than using manual calculations. Moreover, the programs perform as many as thousands of tedious, complex calculations involved in the design, again saving time and eliminating computation errors. After all the design information is finalized, a SPICE model of that data must be sent to a manufacturer to build a prototype.
Frequencies Move Higher
In the mid-1970s, most switching-power-supply operating frequencies topped out at a high limit of about 50 kHz. But the benefits of higher-frequency operation were apparent to researchers and designers, because the two largest elements in a dc-dc converter, the power transformer and output filter capacitor, could become much smaller at higher operating frequencies. In a dc-dc square-wave converter design, the maximum available load power from a given transformer is proportional to its operating frequency. For a given load power, doubling the operating frequency will roughly halve the volume of the transformer.
Another development in the late 1970s led to the practical development of higher-frequency switching supplies. In a 1980 Intersil application note, “The Design of Switchmode Converters Above 100 kHz,” author Rudy Severns wrote, “With the introduction of power MOS switches, and the general improvement in other components, it is now possible to design switchmode converters with switching rates in the range of 100 kHz to 5 MHz, and from both off-line and low-voltage dc bases.” Since high-frequency switchers were in the earliest stages of development, much of the application note's analysis dealt with the possible converter topologies (voltage fed, current fed and resonant), control options and the operation of converter components at frequencies of 100 kHz and above.
In a section on high-frequency magnetics, Severns said, “The high-frequency power transformer is the most difficult component in a high-frequency switcher. While a number of satisfactory solutions have been found, this is an area where a great deal more work is necessary.” One area that received extensive attention was transformer winding, which required, according to Severns, “great ingenuity…to reduce the leakage inductance, and to produce symmetrical and predictable voltages in the ouput windings.”
Several suggestions for achieving this were to interleave layers of the primary and secondary, and better coupling is possible if the primary and secondary are wound multifilar in each layer. Even better coupling is possible if primary and secondary are wound of twisted coaxial transmission lines (Fig. 1). Several ideas were presented for constructing high-frequency transformers, such as the use of Litz wire instead of single copper wire to reduce ac resistance and equalize current distribution, and the use of ferrites in place of tape cores.
The Road Ahead
Ferrite cores have ascended to the top of the pecking order in today's hierarchy of magnetics materials for power-supply applications. Ferrites are easy to manufacture, can be fabricated in many sizes and shapes, and are lower in cost than other magnetics cores. They are metal-oxide ceramic materials (about 50% iron oxide) that can be magnetized to a high degree. The basic component is iron oxide combined with binder materials such as nickel, manganese, zinc and magnesium. The two major types are manganese zinc and nickel zinc. Applications for ferrites fall in the frequency range below 500 kHz and operating temperatures between -80°C and 100°C. The manganese-zinc type has high permeability but low bulk-resistivity, and is suitable for low-frequency applications. Nickel-zinc has lower permeability and high bulk-resistivity, and is used in high-frequency magnetics.
Today's challenge for magnetics designers is to find ferrite material combinations that can meet the power demands of computer and communications converters that must run at high frequencies (500 kHz to 1 MHz) to reduce package size and supply high currents to the low-voltage ICs contained in such systems. A high-saturation flux density is one of the requirements in the material, meaning that the permeability of the material must remain high at elevated frequencies since permeability predicts the flux density for a given amount of external magnetic force. But permeability tends to roll off at high frequency, indicating that the flux density will also be lower (Fig. 2).
Other important parameters to increase a transformer's power throughput are the core size, the number of winding turns and reducing core loss at high frequencies. Currently, magnetics designers are looking for materials that will provide high-flux density while exhibiting low core losses at temperatures at or above 100°C.
Ferrites play a key role in what is likely the most important advance in magnetics technology in the past 10 years: the field of planar magnetics, which uses coils encapsulated within the multiple layers of printed circuit (pc) boards. The coils are sandwiched between low-profile ferrite cores to make small, flat transformers and chokes ideally suited for pc-board mounting. Planar magnetics fits perfectly with the trend towards smaller, low-profile power supplies needed for communications and computer equipment. Components fabricated with planar technology can have height dimensions of 0.5 in. or less.
A planar magnetics transformer can be constructed as a standalone component or integrated directly onto a pc board together with the circuitry it powers (Fig. 3). The major benefits of planar, and the reasons it will probably replace bobbin-wound magnetics in the coming years are the following:
Much lower profile than conventional wirewound devices, resulting in a height that is 50% or more lower.
High surface-to-volume ratio and larger surface area than wirewound components make for better heat conduction in today's high power-density converters.
Unlike wirewound transformers, which can have varying parasitics that cause voltage and current spiking, planar parasitics are very constant from lot to lot.
A high degree of repeatability is achieved with planar because the wiring is etched within the pc-board layers.
Planar designs can be developed to fit a specifically designed space and pin layout, making them more flexible than the fixed layout of bobbin transformers.
Magnetic coupling is better in planar, making for easier supply design.
Transformers and inductors can be integrated into a single planar structure, saving space and cost.
The road ahead for magnetics technology depends on how power-distribution architectures will evolve to support the integrated circuits and systems of future generations of electronic equipment. As power supplies change to adapt to new architectures, magnetics will change to meet these new challenges.