IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 531 Comparison of Realization Techniques for PFC Inductor Operating in Discontinuous Conduction Mode Vytenis Leonavičius, Maeve Duffy, Associate Member, IEEE, Ulrich Boeke, and Seán Cian Ó Mathúna Abstract—Several design techniques are compared for producing a power factor correction (PFC) boost inductor operating in discontinuous conduction mode. In this application, fringing fields associated with high frequency current pulses cause problems of winding loss in the vicinity of an air gap, in particular with planar core shapes. For this reason, a planar inductor in which a lumped gap is replaced by a distributed air gap material is described and investigated. The consequences of lumped versus distributed air gap for the losses of the boost inductor are investigated. A significant reduction in ac winding loss of the planar structure with the composite core is demonstrated. However, the trade-off between reduced winding loss and increased core loss for this technique has to be considered along with the selection of proper winding technology. Four boost inductor design realizations are built and compared. Presented winding loss models are verified with measurements on prototypes operating in an 80 W PFC converter. have disadvantages in modern power converters, considering mechanical integration and manufacturability of the magnetic component. Low profile (planar) magnetic cores may incur even higher ac winding losses due to the fringing fields in the vicinity of the air gap. It is important to choose the proper magnetic core structure, material, air gap strategy and winding technology to achieve optimum performance of the magnetic component. However, the design procedure is somewhat complicated where topologies involve variable switching frequency and variable current amplitude, which is considered in this application. These issues are addressed in the paper, giving the comparison of different techniques to achieve the optimum PFC inductor design. The performance of several boost inductor realizations is compared in an 80 W PFC converter. Index Terms—Air gaps, chokes, composite magnetic core, discontinuous conduction mode, inductors, loss measurement, magnetic devices, magnetic materials, planar magnetics, power factor correction, windings. II. ISSUES FOR BOOST INDUCTOR IN DCM OPERATION I. INTRODUCTION T HE BOOST inductor in power factor correction (PFC) circuits operating under discontinuous conduction mode (DCM) requires careful design. Conventional PFC inductors operating in continuous conduction mode (CCM) with constant switching frequency have a straightforward design procedure, which allows accurate selection of the inductor realization, core size, and winding type and size. Here, low ac ripple often has negligible impact on the winding and core design strategy, which mainly depends on dc current levels in the component. In DCM operation, significant core losses may be incurred due to corresponding high levels of magnetic flux swing. Large current ripple causes problems of ac winding losses, which requires special attention to the winding design. Conventional wire-wound component designs with high profile cores using litz wire can help in reducing ac effects. Such realizations often Manuscript received November 21, 2002; revised August 21, 2003. This work was supported in part by PEI Technologies. Recommended by Associate Editor J. A. Ferreira. V. Leonavičius and S. C. Ó Mathúna are with the Energy Processing for ICT, National Microelectronics Research Centre, Cork, Ireland (e-mail: vytenis.leonavicius@nmrc.ie; cian.omathuna@nmrc.ie). M. Duffy is with the Department of Electronic Engineering, Nun’s Island, National University of Ireland, Galway, Ireland (e-mail: maeve. duffy@nuigalway.ie). U. Boeke is with Philips Research Laboratories, Group Electronic Modules, Aachen 52066, Germany (e-mail: ulrich.boeke@philips.com). Digital Object Identifier 10.1109/TPEL.2003.823249 PFC boost converters in DCM operation have several interesting features [1]. A sophisticated control technique is used in this particular application to minimize conducted differential mode interference [2]. The switching losses in the boost diode are low because of a reduced reverse recovery current. The power metal oxide semiconductor field effect transistors (MOSFET) can operate with zero voltage switching for input voltages less than 50% of the output voltage. Also, less energy must be buffered in the boost inductor compared with converter designs operating in continuous conduction mode (CCM) [3]. However, the resulting complex current waveform in Fig. 1 has three variable parameters: current amplitude, frequency and duty cycle. High switching frequency and duty cycle are modulated over a period of 10 ms (half the mains period). Such operation complicates the conventional design procedure for the magnetic component [4]–[6], as the evaluation of core and winding losses is not straightforward due to variability of electrical parameters over the wide range. Inductors designed to operate in DCM have lower inductance values than those operating in CCM, so that a smaller physical inductor size may therefore be expected. However, due to large value of ripple current in DCM, ac winding losses may be larger than in CCM. Similarly, large flux swings have the potential to incur higher core losses than with CCM. As the design of high frequency magnetic components is driven mainly by the thermal performance, the net reduction in size of a DCM inductor is therefore limited by the need to reduce the thermal resistance of the structure to allow higher ac power losses. Existing wire-wound and planar inductor structures with two different winding technologies—litz wire and PCB—will be compared. Emphasis will be given to 0885-8993/04$20.00 © 2004 IEEE 532 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 Fig. 1. Boost inductor current waveform (a) over rectified mains period and (b) waveform over 1 ms. the investigation of planar magnetic structures as they offer improved manufacturability (integration and packaging of the converter) and thermal performance (larger surface to volume ratio). The distributed air gap technique, which involves the combination of the low permeability and high permeability magnetic materials, will be presented as a way of reducing ac effects in inductor winding. III. PFC INDUCTOR TOPOLOGIES In general, the most important limiting factors in PFC inductor design are core saturation and temperature rise arising from winding and core losses. The conventional approach is a wire wound magnetic component with a high permeability gapped ferrite core or the toroidal core with low permeability distributed gap material as shown in Fig. 2. Litz wire is generally used to control ac winding loss due to skin and proximity effects caused by the large ripple current. Gapped core structures usually have a significant fringing field around the lumped air gap because of the large flux swing in DCM operation. Winding turns positioned close to these fringing fields will generate high eddy current losses. High frequency electromagnetic fields in the vicinity of the gap may cause problems of interference with neighboring circuitry [7]. These issues may decrease the performance of the inductor as well as the overall circuit. Toroidal cores with powdered metal material (low permeability) have inherently distributed air gaps. Having the winding distributed uniformly around the entire core will keep stray magnetic flux and EMI propagation very low. A less desirable feature of distributed air gap materials is their large levels of specific core loss compared with high performance ferrites. Most popular distributed air gap materials currently available for power applications—Permalloy powder (MPP), Kool Fig. 2. Wire-wound magnetic components: (a) gapped ferrite U core, (b) distributed gap toroidal core. and powdered iron—have specific core loss 5 to 20 times greater than ferrites. DCM operation at switching frequencies above 50 kHz will result in a core size which is not fully utilized. High performance MPP toroidal cores usually are more expensive than ferrite cores. Considering the additional costs for manufacturing such components, toroidal cores are not an attractive solution for this application and will not be discussed in this paper. While litz wire can reduce high frequency losses, this advantage has limited frequency range and is not a universal solution [8]. The need to remove turns from the vicinity of the air gap (with large fringing fields) requires additional efforts during manufacturing. An alternative approach is a planar design as shown in Fig. 3. Planar cores have a higher surface area to volume ratio compared with conventional cores, so that structures with lower thermal resistance and consequently lower temperature rise may be designed for the same power losses [9]. Moreover, planar windings with thin tracks offer a convenient solution for reducing ac winding losses due to skin and proximity effects. This technology is compatible with multilayer PCB manufacturing which provides improved manufacturability of planar magnetic components. One of the main drawbacks of PCB windings is the reduction in window copper utilization when compared with wire-wound design [10]. This also impacts on the issue of fringing fields [11], where due to limited window space, it is not possible to remove windings from the vicinity of the gap. The resulting effects of increased winding losses may therefore inhibit the use of planar structures in DCM applications. LEONAVIČIUS et al.: COMPARISON OF REALIZATION TECHNIQUES FOR PFC INDUCTOR 533 Fig. 3. Planar magnetic component, lumped gap. Fig. 4. Planar magnetic component, distributed air gap. By replacing the lumped gap with a low permeability magnetic material which has a distributed air gap, a method for overcoming the problems described is provided. The use of distributed air gap materials has been described previously both for wire-wound high profile and planar magnetic structures, including number of examples of quasidistributed gap techniques [12]–[20]. The resulting improvement in ac winding losses has been attributed to the redirection of the magnetic leakage fields parallel to the length of the winding structure in a 1-D field pattern [13]. The planar structure presented in Fig. 4 is proposed as being compatible with planar windings (PCB or similar technology). Here the I-part of a standard planar ferrite EI core set is replaced with a distributed air gap material. A similar structure has been described in [21] for application in a combined inductor-transformer device. Composite core with 2 different magnetic materials is an attractive solution, which allows exploitation of inherent advantages of the materials while minimizing each material’s shortcomings [22]. Analysis of winding and core loss is necessary to demonstrate the trade-off between reduced winding loss and increased core loss in the distributed air gap solution using composite planar core. To illustrate the level of improvement in winding loss, 2-D finite element analysis (FEA) simulation of a planar winding structure was carried out for 3 different versions of a gapped E-PLT 22 core set [23]. These included a lumped gap in the central core leg, a lumped gap distributed between central and outer core legs (spacer gap) and a composite core with distributed air gap material for I-plate. During simulation, a pure sinusoidal ac current with frequency of 200 kHz was used to produce the magnetic excitation. The four-layer PCB winding had 20 turns with copper track thickness of 0.105 mm (3 oz. copper), which is less than the skin depth at the operating frequency. The gap size in the lumped gap designs (centre gap and spacer gap) is normalized to the core window height. For the distributed air gap design, the permeability of the I-plate in a composite core is adjusted so that it produces the same total reluctance as of the lumped gap design; from here the equivalent air gap size is calculated assuming high permeability of the ferrite in both Fig. 5. (a) AC winding resistance versus effective air gap and (b) total winding resistance versus ripple. core parts E and I. The factor of increase in ac winding resistance over dc resistance is plotted as a function of the effective gap length in Fig. 5(a). As expected, ac winding resistance in the distributed air gap solution is smaller than that in either of is only 1.5 for the disthe lumped gap approaches; tributed air gap approach, as compared with factors of 4 and 14 for the lumped gaps. However, in a typical DCM inductor application, the total current includes components of ac (ripple) and dc current so that the level of improvement in winding loss provided by the distributed gap approach is generally less than given in Fig. 5(a). The total winding resistance over dc resistance as a function of percentage ripple current in Fig. 5(b) illustrates the level of improvement in winding loss to be expected for different planar inductor realizations. For example, 200% ripple means DCM operation, taking into account ac current ripple (sinusoidal wave shape for simplified analysis) with a dc component present. Total winding resistance in the distributed for I-plate) design increases very little over air gap ( . Equivalent spacer gap its dc resistance (0.25 mm spacer) design results in a doubled value of winding resistance. However, centre gap (0.5 mm) design yields over five times higher total winding resistance over its dc resistance. The improvement of the ac winding effects in the planar structure with the composite core is evident. However, composite cores with low permeability magnetic part will have higher core 534 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 TABLE I BOOST INDUCTOR SPECIFICATIONS Design no. 1: wire-wound design, gapped ferrite core, litz wire winding [Fig. 2(a)]. Design no. 2: planar design with gapped ferrite core, litz wire winding (Fig. 3). Design no. 3: planar design with gapped ferrite core, PCB winding (Fig. 3). Design no. 4: planar design with distributed air gap, composite core, PCB winding (Fig. 4). A. Application Specifications Specifications for the boost inductor were derived from circuit simulation of the PFC converter described in [1], [2]. Both switching frequency and duty cycle are modulated over a period of 10 ms (half the mains period) producing inductor current waveform as shown in Fig. 1. Corresponding specifications for the boost inductor are given in Table I. B. Inductor Design Procedure Fig. 6. (a) Core loss and (b) total inductor losses for different plate materials. losses when compared with cores made of ferrite material only. and powPopular distributed air gap materials MPP, Kool dered iron were compared and values of core loss corresponding to the conditions described above are presented in Fig. 6(a). Results for the planar ferrite gapped core (0.25 mm spacer) are also included for comparison. Calculations were carried out for a dc current of 1.5 A and a variable sinusoidal ripple current at 200 kHz. As shown, 3F3 ferrite material from Ferroxcube [23] has the lowest level of the specific core loss. MPP and Kool materials [24] have higher levels, while powdered iron from Micrometals [25] has the highest core loss. On the other hand, total inductor losses (winding plus core losses) compared in Fig. 6(b), show that design realizations with MPP and Kool material outperform gapped ferrite core design, i.e. the reduction in ac winding losses provided by presence of the distributed air gap material is larger than the corresponding increase in core losses. This illustrates that planar design realization with composite core is a viable alternative to the conventional lumped gap design. It also lends itself an attractive solution with less manufacturing effort when compared with the quasidistributed gap technique [15]. IV. PFC INDUCTOR DESIGN Following the discussion in the previous section, four inductor design structures were selected for further technical analysis and practical measurements. The standard procedure in selecting the optimum magnetic core size is the area product approach [4], which relates electrical and magnetic parameters with the core size and geometry. High frequency effects can be included to get a more accurate method [6]. Due to the continuously varying waveform over time and high peak current to rms ratio, optimum design of the boost inductor cannot be achieved in a single attempt. A procedure that gives the initial design for the maximum thermal stress of the component is described in Appendix I. The assumption of the maximum operating frequency, flux density and corresponding peak current yields a selection of the core size larger than optimum. Since the average frequency and flux density values over the mains period are lower than the chosen maximum boundary conditions, design optimization is necessary to produce the smallest inductor size for the maximum allowed thermal stress. The optimization takes into account component losses in the core and the selected winding structure. Estimations of winding and core losses must account for the time-varying nature of the applied current waveform in Fig. 1. For the purpose of this work, each half of the mains period was divided into 10 equal time intervals and one representative pulse was used to describe the current waveform over each time interval. Each point presented in Table II has fixed current am, frequency and duty . plitude Core losses of the magnetic component can be calculated using modified Steinmetz equation for each representative current pulse, with an equivalent switching frequency to account for nonsinusoidal waveform [28], [29] LEONAVIČIUS et al.: COMPARISON OF REALIZATION TECHNIQUES FOR PFC INDUCTOR 535 TABLE II EXTRACTED TIME POINTS FROM THE CURRENT ENVELOPE (1) Here, —core loss density , f—switching fre—equivalent frequency, —peak magnetic flux quency, , x, y—empirical parameters [36], which can be density, obtained from the magnetic core manufacturer. Total core losses are calculated as the average power loss density of all ten time intervals from Table II multiplied by the effective core volume (2) The winding design can be optimized using analytical methods [5], [8], [26] or FEA methods to account for 2-D field effects [27]. Skin and proximity losses in wire wound designs with litz wire can be calculated using 2-D analytical calculation method [30], which accounts for the H-field distribution in the winding area including fringing effects. FEA simulation tool [31] will be used to calculate winding losses for the planar gapped core with multilayer PCB winding. Such a structure presents a 2-D problem as shown by the H-field distribution in Fig. 7(a), which is difficult to evaluate accurately with analytical tools. For the distributed gap design in Fig. 7(b), magnetic field in the winding area is almost 1-D and the Dowell’s analytical approach [33] can be used to estimate ac effects in the PCB winding as described in [34] (3) where —ac resistance factor, p—number of parallel layers (=PCB layers), h—copper track thickness, —skin depth at given frequency, —layer packing factor, —turns per layer, wc—copper track width, ww—core window length. The ac effect correction factor in (3) is for one fundamental harmonic frequency. To account for the nonsinusoidal nature of the inductor current waveform, the losses can be calculated using Fourier series expansion for each of the representative time point i over the mains period H Fig. 7. -field distribution for different air gap techniques: (a) planar EI core with spacer gap, (b) planar EI core with distributed air gap in the I part. ac current harmonic, —ac resistance factor for jth harmonic and frequency. For the triangular current waveform have the following expressions: (4) —winding losses for the ith triangular pulse extracted where from the inductor current envelope, —dc winding resistance, —dc current component, —amplitude of the jth (5) of triangular waveform and duty D are deThe peak value fined in Table II. Fourier series expansion up to the 7th harmonic can be taken to give the accurate result. This 536 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 TABLE III COMPARISON OF INDUCTOR DESIGNS, L = 120 H method proves to be useful to calculate winding losses in combination with FEA simulations tool [31]. Alternative calculation procedure for arbitrary current waveforms is described in [26], which presents a straightforward formula to calculate effective winding resistance without performing Fourier analysis. Having calculated winding losses for each time interval, the total losses are equal to the average value (6) As the last step temperature rise of the magnetic compoof the nent can be evaluated taking the thermal resistance magnetic component (7) Previous work in [9], [35], [36] describes an empirical forof a magmula which relates the value of thermal resistance netic component directly to the value of the effective magnetic volume . As mentioned earlier, the optimization involves the winding design for the selected core size. A further optimization can be done for the core size and shape considering the component temperature rise. This complete design procedure is valid for the boost inductor operating in DCM, where both core and winding losses must be taken into account. V. COMPARISON OF DIFFERENT REALIZATIONS Four different inductor designs mentioned in the beginning of Section IV are summarized in Table III. Design no. 1, made with standard U15/11/6 core of 3C94 ferrite [23] was chosen for the initial wire-wound solution. In order to control eddy current losses, 63 0.071 mm litz wire was used in the winding. The air gap was realized inserting spacers between the U-core halves on each leg. Design no. 2 is a low profile solution with reduced thermal resistance (almost by 33% if compared with design no. 1). It has the planar E-PLT 22 core of 3F3 ferrite material [23]. The same litz wire 63 0.071 mm was used to construct the winding. Such a winding is very difficult to manufacture for a planar core construction. An alternative cost effective solution with improved manufacturability is the design no. 3. The winding is made with four layer 3 oz. copper PCB. Each layer in the winding has 5 turns with track pitch of 1.04 mm (0.300 mm spacing). Both design no. 2 and no. 3 have the same air gap size, which is realized with the spacer between the E part and the I part of the planar core. Design no. 4 represents distributed gap approach. It has a composite planar E-PLT 22 and I core, where E part is made of 3F3 ferrite material with permeplate was cut from the slab of Kool [24]. As in the design no. 3 the winding is made ability with four-layer PCB. The number of turns for this design was deliberately increased in order to reduce the peak flux density in the core. A. Comparison of AC Winding Resistance All four designs are compared in terms of the winding resistance under ac current excitation for the frequency range up to 400 kHz. Calculations for wire wound designs no. 1 and no. 2 (with litz wire) were done using 2-D analytical model described in [30]. Design no. 3 and no. 4 (with PCB winding) analysis was done with 2-D FEA [31]. The simulation procedure is based on the division of the winding of the magnetic component into two parts, which produce field distribution in different planes. The planar conductor inside the magnetic core window is modeled as the first part, and the end-turns outside core are modeled as the second part. Planes which represent H-field distribution of these two segments are perpendicular to each other. In this way it is possible to perform an accurate studies of a 3-D planar E-PLT structure using 2-D FEA tool. LEONAVIČIUS et al.: COMPARISON OF REALIZATION TECHNIQUES FOR PFC INDUCTOR 537 TABLE IV WINDING RESISTANCE This so-called “Double 2-D” method is described in [32]. Table IV shows calculated values of the winding resistance for several operating frequencies. i.e. ac winding resistance The factor over dc resistance is calculated to compare all designs. According to the calculations in Fig. 8, ac effects are lowest for the design with distributed gap approach (design no. 4), over the because of the smallest values of the ratio frequency range. This can be attributed to the 1-D magnetic field distribution in the winding area (H-field has dominant horizontal component, minimum eddy currents). On the other hand, design no. 2 (planar core, litz wire) shows the highest ac effects. The winding in the planar structure is closer to the air gap (lower window height) with high fringing field perpendicular to the winding. Litz wire does not help to reduce ac effects. Design no. 1 has a similar trend to design no. 2, except that the first has U core with larger window to accommodate the winding further away from the air gap. Design no. 3 (planar core, PCB winding) has better performance than design no. 1 and no. 2. PCB winding is shifted away from the air gap, although 2-D field distribution within the winding area has larger vertical component than for the design no. 4 and this causes significant ac winding losses. The 2-D FEA calculation for the planar gapped core with multilayer PCB winding (design no. 3) gives a reasonably accurate representation of ac effects in the winding, while 1-D analysis based on Dowell’s approach gives false result as shown in Fig. 9(a). For the distributed gap design, 1-D analysis shows comparable results with 2-D FEA simulations as demonstrated in Fig. 9(b). Therefore, it validates the (3) to estimate ac winding effects for the design no. 4. Fig. 8. AC winding resistance versus frequency. B. Comparison of Inductor Power Losses In this section high frequency power losses of the inductor operating in PFC converter are examined. Calculation results require experimental validation under actual operating conditions. Performance of the inductor design realizations was input and studied in a 80 W PFC converter with 85 output. The measurement setup for total inductor 385 power losses is described in Appendix II and the summary of these results is presented in Table V. The lowest power losses have been measured for design no. 1 with a U-core [Fig. 10(a)]. The planar design no. 3 has twice the power losses compared with design no. 1. However, these higher losses produce only about 30% higher temperature rise [Fig. 10(b)] since the planar core has a lower thermal resistance to ambient. The higher power losses in this case are due to higher winding losses. PCB winding in the design no. 3 has 3 times lower copper cross section (see Table III) when compared Fig. 9. Comparison of ac resistance calculation methods: (a) design no. 3 and (b) design no. 4. with litz wire winding in design no. 1, resulting in higher rms winding losses. Calculations show that this is even worse for the design no. 4, which has rms losses similar in the magnitude of the total losses for the designs with litz wire. Gapped core designs, however, show significant ac winding losses accounting around 40% for the design no. 1 and no. 3 to 57% for the design no. 2 of the total winding losses. A possible solution to reduce eddy currents in PCB windings of a planar magnetic component is illustrated with planar inductor design no. 4. Skin and proximity winding losses for 538 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 TABLE V INDUCTOR PERFORMANCE IN PFC CONVERTER; COMPARISON OF CALCULATED (LEFT) AND MEASURED DATA (RIGHT), (N.V=NO VALUE) plate. higher core losses due to the properties of the Kool As a result, design no. 4 has the highest total losses of all inductors investigated. Nonetheless, planar inductor design no. 2 with litz wire shows by how much the winding rms loss of design no. 3 and no. 4 could be reduced for the case of increased copper fill factor. This design also has the lowest temperature rise as demonstrated in Fig. 10(b). The higher core loss in this case is due to lower ferrite core temperature and the temperature characteristic of 3F3 ferrite material, where the minimum core loss is observed at 80 [23]. The accuracy of the models applied for predicting winding and core losses for the complex current waveform associated with DCM has been verified by measurements for the design no. 1 (no error) and the design no. 4 (2.5% error). Somewhat larger errors between calculations and measurements are observed for the design no. 2 (12% error) and the design no. 3 (8.75% error). In design no. 2 the presence of the lumped air gap and the leakage fields that link with the winding paths outside the core were not accurately evaluated. Possible error may come from the winding termination [37] in the measurements of prototypes, which was not addressed in the technical analysis. VI. CONCLUSION Fig. 10. Inductor performance in PFC converter: (a) power losses and (b) measured peak temperature rise. the current waveform of Fig. 1 are at least half those in designs with litz wire (design no. 1 and no. 2) and more than 5 times smaller than those produced in the design no. 3 with PCB winding [Fig. 10(a)]. Unfortunately, the improvement in ac winding losses has been achieved with much higher rms losses in the winding due to the lowest copper fill factor, and also The design of planar inductor with a distributed air gap instead of a lumped gap is described for application in a PFC boost converter operating under DCM. The motivation for removing the lumped gap in planar core structure is the potential for reducing ac winding losses caused by high ripple current levels in this application. However, as shown, additional core losses are incurred in distributed air gap materials. Analysis of winding and core losses is presented to illustrate the overall performance and MPP achieved with different materials available. Kool materials are predicted to provide improved performance over an equivalent lumped gap design. inductor realizations were built to compare Several 120 the performance of design strategies, involving different winding technologies (litz wire versus PCB), core shapes (high profile versus planar) and gapping techniques (lumped air gap LEONAVIČIUS et al.: COMPARISON OF REALIZATION TECHNIQUES FOR PFC INDUCTOR versus distributed). The performance was investigated analytically and confirmed with operation in an 80 W PFC converter. A significant improvement in ac winding losses of planar core structures was observed with distributed air gap technique. However, large ac flux density levels in DCM operation result material. This might have in high core losses for Kool been improved by applying materials with lower specific loss (e.g. MPP), if the required shapes were available. Also, poor copper utilization in PCB windings resulted in significant rms winding losses. These are the two major reasons why overall losses for the distributed gap approach were larger than for the lumped gap designs. Also, it has been shown that winding loss for planar structure may be reduced by increased copper utilization. This may be achieved by reducing the thickness of dielectric layers e.g. by using FLEX PCB or stamped copper windings. From the technical analysis, it appears that 1-D approach is in a good agreement with 2-D FEA simulations for the distributed gap design, which simplifies the design procedure for boost inductor with the complex current waveform. From the practical analysis some discrepancies were observed between the calculations and measurements. This may be due to the fact that winding termination was not addressed properly in the technical analysis. However, total power loss measurements show good agreement with the calculated results. Although wire wound component with conventional high profile gapped core outperformed other design realizations, all structures including the distributed air gap inductor meet performance specifications. It is worth noting that elimination of the lumped air gap may provide a solution with reduced EMI. APPENDIX I INITIAL INDUCTOR DESIGN PROCEDURE The initial design assumes fixed values of the frequency and the flux density, which produces the maximum thermal stress of the magnetic component. The standard procedure using area product approach [4], allows the selection of the core size for the given specification parameters from Table I. Maximum operating frequency of 220 kHz and maximum ac flux density (for the ferrite material) can be assumed as a boundary condition. Further calculation of physical winding and air gap is governed by the well known inductance expressions (8) , —corresponding peak where: L—inductance value, values of the current and flux density, N—number of turns, —cross section area of the core, —air gap length, —gap . In calculating the air gap required area, to achieve accurate inductance value, the effective area of the gap is increased by adding the gap length to both the width and breadth of the core section to correct for fringing fields. This iterative calculation method is described in [12]. Gap length can be calculated iterating (9) This procedure is common for the design realizations with gapped cores (design no. 1, 2, and 3). For the design with com- 539 posite core using two different magnetic materials initial design procedure is based on the reluctance model shown in Fig. 11. Reluctance of the particular core part is described by (10) where: —relative permeability, —length and —area of the section . E part of the core is made of a standard high permeability fer) and I part of low permerite material (typically ability distributed air gap material. Material MPP has the lowest specific losses among other distributed gap materials [24]. However, as this material is only available in the form of ring cores, with similar material properties has been chosen as Kool the next best option. E cores (from which I plate can be cut) with permeability values of 26, 40, 60, and 90 are commercially available [24]. The dimensions of the magnetic I plate assumed equal to those of the standard planar EI core set in order to keep the constant flux density across the entire structure. Using the permeof the different Kool material and the ability value , the number of turns effective plate length in the model for is calculated N which provides the total inductance for each material (11) Flux density culated with in the corresponding core part can be cal(12) Core loss for the boundary conditions are calculated separately for the E part and the I plate. Specific loss data for the and ferrite materials are provided by the manufacKool turer in [23], [24], and [36]. Suitable composite core design can be selected based on the minimum total component losses, adding up winding and core losses for each design. Following this methodology, it can be shown that the lowest losses in the composite core are predicted for the plate with the lowest perin this case). meability value ( APPENDIX II INDUCTOR POWER LOSS MEASUREMENTS The direct electrical measurement of inductor power losses under the large signal operation, illustrated with Fig. 1, is not possible since the reactive power flow already claims most of the resolution of measurement equipment. Additionally, the complex current time function with triangular waveform modulated in amplitude and frequency with twice of the mains frequency would require an enormous resolution in time. Thus a thermal technique [38], [39] has been used to identify inductor power losses using component temperature rise as an intermediate result. During the measurements, the inductor has been mounted in a plastic test box filled with quartz sand illustrated with Fig. 12(a). Due to the quartz sand, the heat transportation is limited to conduction and both convection and radiation inside the box are not present. The absolute thermal resistance of this test box is less K/W . However, it is very important that important 540 Fig. 11. Reluctance model of planar design no. 4 with composite core. IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 2, MARCH 2004 measured with 3 PT100 resistors mounted on component surface positions illustrated in Fig. 12(b). In the third test the inductor has been mounted on its own 100 mm size. Other heat printed circuit board of 160 mm generating power components were removed from the board. In this test the inductor has been stressed again with the boost converter current like in the first test. The measured temperature rise of the winding and the core is given separately in Table V. Finally, the last row of Table V gives the ratio of peak temperature rise divided by the measured power losses as an indication for thermal resistance of such a component mounted on a printed circuit board. The range of measured thermal resistances for E-PLT 22 core set is 24 K/W 29 K/W. These values are somewhat lower than the values (31 K/W 42 K/W) presented in [39] for the same E-PLT 22 core set. This mainly resulted from different test conditions. In [39], heat transportation was mainly due to convection, while in our test additional heat conduction was present from the planar core to the ambient via the PCB. Errors of the power loss measurements have the following sources: 2% accuracy of the boost converter current; — 2% accuracy of the dc power load; — 2 to 6% accuracy of each temperature rise measure— ment for a constant absolute error of 1 ; slightly different heat distribution around the inductor — for the operation with boost converter and dc current. ACKNOWLEDGMENT The authors wish to thank Dr. P. Luerkens, Philips Research Laboratory Aachen, for discussions of the boost converter design, and G. Kowalin, Philips Research Laboratory Aachen, for the preparation of magnetic material plates. REFERENCES Fig. 12. Thermal measurement setup: (a) test box and (b) position of temperature sensors. thermal resistance does not change between two tests. In the first test, the inductor is stressed with current of the boost converter. The task here was to measure the steady state value of the inductor peak temperature rise due to power losses and the thermal resistance of the box. This temperature has been reached typically after 4 h. In a second test, the inductor has been stressed with a constant dc current generating winding losses only. The value of this dc current has been adjusted so, that temperature rise of the winding is equal to the peak temperature rise of the first test. We assume that the dc power losses of the second test are equal to the unknown power losses of the first test since both tests generate the same temperature rise for the same thermal resistance. The temperatures during these two tests have been [1] P. Luerkens, “Step-Up Converter,” European Patent EP 1 083 648 A2, 2001. [2] M. Albach and F. A. Wegener, “AC-DC Converter Triggered by Variable Frequency Pulses,” U.S. Patent 4 719 552, Jan. 12, 1988. [3] J. Zhang, M. Jovanovic, and F. C. Lee, “Comparison between CCM single-stage and two-stage boost PFC converters,” in Proc. Appl. Power Electron. Conf. (APEC’99), Dallas, TX, 1999, pp. 335–341. [4] W. T. McLyman, Transformer and Inductor Design Handbook. New York: Marcel Dekker, 1978. [5] J. P. Vandelac and P. Ziogas, “A novel approach for minimizing high frequency transformer copper losses,” IEEE Trans. Power Electron., vol. 3, pp. 266–276, July 1988. [6] W. G. Hurley, W. H. Woelfle, and J. G. Breslin, “Optimized transformer design: inclusive of high-frequency effects,” IEEE Trans. Power Electron., vol. 13, pp. 651–659, July 1998. [7] P. Evans and W. Heffernan, “Electromagnetic considerations in power electronic converters,” Proc. IEEE, vol. 89, pp. 864–875, June 2001. [8] B. Carsten, “High frequency conductor losses in switchmode magnetics,” in Proc. Intertec High Freq. Power Conv. Conf., 1986, pp. 155–176. [9] Design of Planar Power Transformers, Ferroxcube. (1997, May). [Online]. Available: http://www.ferroxcube.com/appl/info/plandesi.htm [10] C. Quinn, K. Rinne, T. O’Donnell, M. Duffy, and S. C. O’Mathuna, “A review of planar magnetic techniques and technologies,” in Proc. Appl. Power Electron. Conf. (APEC’01), Anaheim, CA, 2001, pp. 1175–1183. [11] K. D. T. Ngo and M. H. Kuo, “Effects of air gaps on winding loss in high-frequency planar magnetics,” in Proc. 19th Annu. Power Electron. Spec. Conf., Apr. 1988, pp. 1112–1119. [12] Magnetics Design Handbook MAG100A [Online]. Available: http://focus.ti.com/download/zip/sem_magnetics_design.zip [13] W. M. Chew and P. D. Evans, “High frequency inductor design concepts,” in Proc. 22nd Annu. Power Electron. Spec. Conf., June 1991, pp. 673–678. LEONAVIČIUS et al.: COMPARISON OF REALIZATION TECHNIQUES FOR PFC INDUCTOR [14] P. D. Evans and W. M. Chew, “Reduction of proximity losses in coupled inductors,” Proc. Inst. Elect. Eng. B, vol. 138, no. 2, pp. 51–58, Mar. 1991. [15] J. Hu and C. R. Sullivan, “AC resistance of planar power inductors and the quasidistributed gap technique,” IEEE Trans. Power Electron., vol. 16, pp. 558–567, July 2001. [16] , “Quasidistributed gap technique for planar inductors: design guidelines,” in Proc. Annu. Meeting IEEE Ind. Applicat. Soc., vol. 2, 1997, pp. 1147–1152. [17] N. H. Kutkut and D. M. Divan, “Optimal air-gap design in high-frequency foil windings,” IEEE Trans. Power Electron., vol. 13, pp. 942–949, Sept. 1998. [18] U. Kirchenberger, M. Marx, and D. Schroder, “A contribution to the design optimization of resonant inductors in high power resonant converters,” in Proc. IEEE Ind. Applicat. Soc. Annu. Meeting, vol. 1, Oct. 1992, pp. 994–1001. [19] W. A. Roshen, R. L. Steigerwald, R. J. Charles, W. G. Earls, G. S. Claydon, and C. F. Saj, “High-efficiency, high-density MHz magnetic components for low profile converters,” IEEE Trans. Ind. Applicat., vol. 31, pp. 869–878, July/Aug. 1995. [20] M. A. Preston, R. W. DeDoncker, R. C. Oney, C. M. Stephens, and M. Hernes, “High-current high-frequency inductors for resonant converters,” in Proc. Euro. Conf. Power Electron. Applicat., Florence, Italy, Sept. 1991, pp. 242–246. [21] M. Meinhardt and M. Duffy et al., “New method for integration of resonant inductor and transformer—design, realization, measurements,” in Proc. Appl. Power Electron. Conf. (APEC’99), Dallas, TX, 1999, pp. 1168–1174. [22] G. Orenchak, “Composite cores offer the best of all worlds,” in Proc. PCIM/HFPC Conf. Power Electron., 2000, http://www.tscinternational.com/techmain.html. [23] Data Handbook Soft Ferrites and Accessories (2002). [Online]. Available: http://www.ferroxcube.com/appl/info/HB2002.pdf [24] Powder Cores [Online]. Available: http://www.mag-inc.com/library.asp [25] Iron Powder Cores for Power Conversion and Line Filter Applications [Online]. Available: http://www.micrometals.com/materials_index.html [26] W. G. Hurley, E. Gath, and J. G. Breslin, “Optimizing the ac resistance of multilayer transformer windings with arbitrary current waveforms,” IEEE Trans. Power Electron., vol. 15, pp. 369–376, Mar. 2000. [27] L. Ye, G. R. Skutt, R. Wolf, and F. C. Lee, “Improved winding design for planar inductors,” in Proc. 28th Annu. IEEE Power Electron. Spec. Conf. (PESC’97), vol. 2, 1997, pp. 1561–1567. [28] M. Albach, T. Duerbaum, and A. Brockmeyer, “Calculating core losses in transformers for arbitrary magnetizing currents: a comparison of different approaches,” in Proc. IEEE Power Electron. Spec. Conf. (PESC’96), 1996, pp. 1463–1468. [29] M. Albach, “Design of magnetic components,” in Proc. Sem. 25, Power Conv. Intell. Motion Conf. (PCIM), 2000. , “Two-dimensional calculation of winding losses in transformers,” [30] in Proc. IEEE Power Electron. Spec. Conf. PESC’00, vol. 3, 2000, pp. 1639–1644. [31] Maxwell EM 2D Field Simulator Manual, Ansoft Corp., Pittsburgh, PA. [32] R. Prieto, J. A. Cobos, O. Garcia, P. Alou, and J. Uceda, “Model of integrated magnetics by means of ‘double 2-D’ finite element analysis techniques,” in Proc. IEEE Power Electron. Spec. Conf. (PESC’99), vol. 1, 1999, pp. 598–603. [33] P. L. Dowell, “Effects of eddy currents in transformer windings,” Proc. Inst. Elect. Eng., vol. 113, no. 8, pp. 1387–1394, 1966. [34] M. C. Smit and J. A. Ferreira et al., “Technology for manufacture of integrated planar LC structures for power electronic applications,” in Proc. Eur. Conf. Power Electron. Applicat., Brighton, U.K., Sept. 1993, pp. 173–178. 541 [35] S. A. Mulder, “On the design of low profile high frequency transformers,” Application Note, Philips Components, Amsterdam, The Netherlands, 1990. [36] , “Loss formulas for power ferrites and their use in transformer design,” Applicat. Note, Philips Components, Amsterdam, The Netherlands, 1994. [37] G. Skutt, F. C. Lee, R. Ridley, and D. Nicol, “Leakage inductance and termination effects in a high-power planar magnetic structure,” in Proc. Applicat. Power Electron. Conf. (APEC’94), vol. 1, 1994, pp. 295–301. [38] T. G. Imre, W. A. Cronje, J. D. van Wyk, and J. A. Ferreira, “Experimental validation of loss calculations for a planar inductor,” in Proc. IEEE Annu. Power Electron. Spec. Conf., vol. 1, June–July 1999, pp. 586–591. [39] A. Lewalter and B. Ackermann, “A thermal model for planar transformers,” in Proc. 4th IEEE Int. Conf. Power Electron. Drive Syst. (PEDS), Indonesia, 2001, pp. 669–673. Vytenis Leonavičius was born in Kalvarija, Lithuania, in 1974. He received the M.Sc. degree in electrical engineering from Kaunas University of Technology, Lithuania, in 1998 and is currently pursuing the Ph.D. degree in microelectronics at the Energy Processing for ICT Team, National Microelectronics Research Centre, University College, Cork, Ireland. His main research area is in the design of planar magnetics for switch mode power supplies, with the focus on functional integration and packaging. Maeve Duffy (S’95–A’96) was born in Monaghan, Ireland. She received the B.E. degree (with honors) in electronic engineering and the Ph.D. degree in planar magnetics from the National University of Ireland, Galway (NUIG), Ireland, in 1992 and 1997, respectively. She was a Research Officer with PEI Technologies, National Microelectronics Research Centre, Cork, Ireland, from 1997 to 2001. She is currently a Lecturer with the Department of Electronic Engineering, NUIG. Her main research interests are in modeling and design of magnetic components, including planar magnetics and magnetic sensors. Ulrich Boeke received the M.Sc. degree from the Darmstadt University of Technology, Germany, in 1991. He has joined the Electronic Modules Group, Philips Research Laboratories, Aachen, Germany. He is a Senior Scientist in the field of power electronics for consumer electronics such as displays and lighting systems. Seán Cian Ó Mathúna received the B.E., M.Eng.Sc., and Ph.D. degrees from the National University of Ireland, Cork, in 1981, 1984, and 1994, respectively. From 1982 to 1993, he was instrumental in establishing the Interconnection and Packaging Group, National Microelectronics Research Centre (NMRC), University College Cork, Ireland, where he held the position of Senior Research Scientist. In 1993, he joined PEI Technologies, NMRC, as Technical/Commercial Director, where he was responsible for power packaging, planar/integrated magnetics, and product qualification. In 1997, he rejoined NMRC as Group Director with responsibility for Microsystems. In 1999, he was appointed as Assistant Director for NMRC with responsibility for microelectronics integration with research themes in ambient electronics, biomedical microsystems, and energy processing for ICT.