A Systematic Approach Optimizes Magnetics

For the PDF version of this article, click here.

Power electronic system designers and equipment manufacturers face significant challenges in the specification and design of magnetic components. Devices such as inductors and transformers represent a significant portion of the bill-of-materials (BOM) cost for power electronic systems. And with the recent increases in copper and steel prices (Fig. 1), the need to optimize these components for cost and performance is greater than ever.

A systematic approach to the optimization of magnetic components proves useful for all stakeholders, including systems engineers, magnetics design engineers and manufacturers. The magnetics design process has two main steps: the development of the component specification and the detailed design of the component itself. Each step requires careful consideration and attention to detail to make sure the resulting component is optimized for the particular application. If this type of disciplined process is followed, the end result will be a component that helps to maximize the value of the end product and the profitability of the OEM.

The first step, the specification development process, involves synthesizing both the market-driven requirements and the system-level requirements into the specification for the particular component. For example, market-driven requirements may include efficiency and cost, while system-level requirements may include the specific inductance and cooling conditions. The choices made by the systems engineer in developing the component specification will have a profound impact on the resulting cost and performance of the component.

The next step, the detailed design process, involves meeting the specification requirements at the lowest possible cost. Here the magnetic-component designer has several choices to make, including core material, winding material and configuration, and physical topology among many other options.

During both the specification development and the detailed design phases, open lines of communication must exist between the two parties to ensure the final result will be optimized. For example, the magnetics designer needs to let the systems engineer know the specific impact of the choices the engineer has made, such as the impact of efficiency on cost, so that the whole process may be optimized in a closed-loop manner. And finally, the entire process must produce the detailed manufacturing documentation required to build the optimized components.

Component Specification

The specification of a magnetic component has an enormous impact on the component's realized performance and cost, and therefore requires careful optimization from a system level. The following application example will be used throughout this section and will provide a baseline for how the various specification choices impact the component cost and performance (Table 1). Each of the design options in the specification represents levers the systems engineer has available to pull to optimize the component.

The inductor design example referred to in Table 1 is a power-converter type designed for use in a three-phase inverter with 35-kW, 480-V_RMS line-to-line output. The inductor must be rated to handle 42 A_RMS at 60 Hz with a peak current of 200% rated current plus peak ripple current. In this example, the harmonic voltage is equivalent to 200 V_RMS at a 15-kHz switching frequency. In addition, maximum ambient temperature is 55°C.

Inductance

The primary function of an output inductor is to control the output voltage and current harmonics. In the reference design option listed in Table 1, the inductance is set to 500 µH, and the resulting equivalent harmonic current at 15 kHz is 4.24 A_RMS — approximately 10% of the rated fundamental current. In reality, however, engineers select a wide range of inductance for different reasons.

If the inductance is doubled, for example, the ripple current will be reduced by half; Table 1 shows that the cost will increase more than 60%. However, a filter capacitor with a larger capacitance value may be able to achieve the same goal with a much lower cost impact. This is an example of a system-level tradeoff that must be made.

Switching frequency (ƒ_SW)

A high switching frequency may significantly reduce the inductor power loss and cost, as well as reduce ripple-current amplitude. The downside is almost always higher power dissipation in the switching devices. As illustrated in Table 1, by comparing the reference design versus the low ƒ_SW design, if the switching frequency is reduced from 15 kHz to 7.5 kHz, increased harmonic current requires an increase of inductor size to maintain the same temperature rise. As a result, inductor cost increases by 25%.

Peak current (I_PK)

The peak current that occurs under transient conditions is related to the inductor's magnetic saturation capability and should be specified as low as possible to minimize cost. Under saturation, the effective inductance is decreased, which can affect system stability. The system control must be designed to tolerate a certain degree of “soft” saturation. Table 1 shows that if the inductor peak current requirement is increased by an additional amount equal to the rated current amplitude — in other words, the peak current requirement is approximately three times the rated current — the cost of the high I_PK inductor will increase 39% over the reference design.

Insulation class

High-temperature insulation classes, such as Class R (220°C), enable magnetic components to operate at higher temperatures and with increased electrical and magnetic loading, without decreasing their operating life. To reduce the rising cost of core and conductor materials, it is practical to push for higher operating temperatures. In this example, if the Class F (155°C) insulation system of the reference design is raised to Class R (Class 220), Table 1 shows a cost reduction of 22%.

Cooling

Another practical approach to reduce cost from a thermal perspective is to improve cooling. Enhanced cooling allows for an increase of electrical and magnetic loading, which leads to less material usage. The forced air design option in Table 1 has a 28% cost reduction compared with reference. Normally, the cost of adding a blower is small in comparison to the component cost reduction, especially if the blower may be used to cool other components such as the power semiconductors.

Efficiency

Efficiency conflicts directly with cost in most cases. High-efficiency magnetic components usually require the use of high-performance materials (at high cost) or larger quantities of materials. From a system level, a high efficiency also may require a lower switching frequency in the power semiconductor devices, which will directly lead to cost and mass increase of magnetic components.

The requirements for magnetic-component efficiency are application and market dependent. In the photovoltaic (PV) industry, efficiency is critical, and many governmental rebate programs take into account inverter efficiency to determine the dollar amount of the rebate. For a more conventional power supply (for example, a PC power supply), however, low cost is more important than high efficiency. Compared with the reference design, the low-cost option cuts the material cost by 65%, by changing the magnetic core material, adding forced-air cooling and raising the insulation class from Class F (155°C) to Class H (180°C).

Magnetic-Component Design

Once the component specification has been optimized from the system perspective, a detailed design optimization of the component itself must be performed to achieve the maximum benefit from the entire process. Due to a great variety of available materials, configurations and second-order effects, magnetics optimization is not an easy task. The main goal of the detailed design process is to achieve the specification at the lowest possible cost.

Core material

The most common core materials are shown in Table 2, along with the relative performance and cost. Note that each kind of material has a wide range of performance and cost. For example, silicon steel has a thickness from 0.001 in. to 0.025 in., which is associated with a correspondingly wide range of power loss and cost.

While no best magnetic core material exists for all applications, there is the best or the most suitable material for each particular application. The process to select a core material for a given application is both a science and an art that requires a thorough understanding of the available materials, configuration, interaction with windings, thermal capabilities, manufacturing issues and economics.

Normally cores with low power loss are selected for high-efficiency applications or applications with a large amount of high-frequency voltage or current harmonics. Nanocrystalline materials are especially suitable for common-mode choke application.

Winding material

The most common winding materials are shown in Table 3, along with a relative comparison of their performance under different operating conditions. For example, a strand of AWG#20 round wire can be used to conduct 1 A_RMS at 10 kHz for a transformer if the number of turns is less than 10 and the skin and proximity effects are not severe.

Compared with the Litz wire, round wire has advantages of high-winding fill factor and low cost. However, for the same current of 1 A_RMS at 10 kHz, if the number of turns is more than 100, and the skin and proximity effects are strong, Litz wire may be needed to limit winding power loss and winding temperature.

Second-order effects

It is not a trivial job to optimize magnetic components for power electronics applications. Due to high frequencies and the existence of air gaps, the second-order issues such as leakage and fringing flux, eddy currents and proximity effects may be significant or even overwhelm the design considerations in many cases.

If these issues are not well understood and carefully considered in the design, unnecessarily high costs will result. For example, not understanding the true nature of the losses may result in a magnetics designer adding more material when it is not really necessary.

Ignoring these important effects can result in a long and iterative design cycle. Because the magnetics design and manufacturing cycle can be long and costly, it is imperative to get the design right on the first or second iteration.

Modern computer analytical approaches such as finite element analysis (FEA) are important to design optimization and may provide significant insight into the second-order effects. Fig. 2 shows an example of the FEA applied to a transformer design. In this example, the power rating is specified for single-phase, bidirectional flux for 1.6 kW at 10 kHz. Primary voltage is 32 V_RMS and primary current is 50 A_RMS. The turns ratio on the transformer will be 3-to-2 with 6 turns of 0.5-mm × 36-mm copper foil on the primary and 4 turns of 0.75-mm × 36-mm copper foil on the secondary. The core is a Ferroxcube E71/33/32, which is one set with two pieces.

This example illustrates that even when using the same materials, the difference in winding arrangement can have a significant performance impact. Fig. 2a shows the winding configuration where the primary winding and the secondary winding are separate. Fig. 2b plots the current distribution in the windings with the peak value of 8.4 A/mm². A finite element study calculates the total winding loss of 12.2 W. The improved design makes the windings interleaved as shown in Fig. 2c. The current distribution plotted in Fig. 2d becomes more uniform and the total winding loss reduces to 6.5 W.

Design for manufacturing

One of the most important — but too often overlooked — aspects of magnetic-component design optimization is making sure the component can be easily manufactured. The first guideline is to know the selected manufacturer's processes and capabilities and make sure the component is designed accordingly. Using standard materials such as commonly available bobbins, wire, insulation and varnishes may greatly aid the manufacturability of the components — particularly in low-volume production.

It is also critical to minimize labor costs, even if the components are to be manufactured in overseas factories. This involves making sure the design is easily assembled and the “packing” factors (the amount of wire and insulation to be installed into the core window) are not excessive. Reducing the labor content and complexity will reduce cost and improve the manufacturing yield and reliability.

Worth the Effort

Since magnetic components often represent a significant — and increasing — portion of the BOM cost for power electronic products, it can be easily justified to spend the appropriate amount of time and resources to optimize them.

Through careful work in the specification and design phases of the magnetic-component development process, breakthrough results in terms of cost and performance can be achieved. Design-cycle times also can be reduced significantly by meeting all of the requirements on the first or, if necessary, the second iteration.

Table 1. Tradeoff of inductor specifications.Design option Reference High inductance Low f_SW High I_PK Class 220 Air cooled Low cost Inductance (µH) 500 1000 500 500 500 500 500 I_PK (A) 128 123 136 187 128 128 128 f_SW (kHz) 15 15 7.5 15 15 15 15 Cooling Convection Convection Convection Convection Convection Forced air Forced air Core material Amorphous Amorphous Amorphous Amorphous Amorphous Amorphous M5 Power loss (W) 181 243 215 223 244 273 365 E_FF inductor (%) 99.49 99.31 99.39 99.37 99.31 99.23 98.97 Temp. rise (°C) 90 90 90 90 145 90 115 Active material cost ($) 154 248 193 215 120 111 55 Notes: Peak current (I_PK) includes both the fundamental current and harmonic current; power loss is the inductor power loss for three phases; the temperature rise (°C) is for hot spot in the windings; and active material cost is the best estimate for three phases and for core and conductor only. Table 2. Core materials. Power loss Saturation Cost Silicon steel High High Low Metglas Low Medium High JFE super cores Low Medium Medium Nickle alloy Low Medium High MPP Low Low High Sendust Medium Low Medium Iron powder High Medium Low Ferrite Low Low Medium Nanocrystalline Low Medium High Table 3. Winding materials. Fill factor Power loss* Cost Round wire Medium High Low Rectangular wire High High Low Foil Medium Medium Medium Litz wire Low Low High *Power loss due to high-frequency current harmonics.