Sunteți pe pagina 1din 10

www.ietdl.

org
Published in IET Power Electronics Received on 31st January 2010 Revised on 27th July 2010 doi: 10.1049/iet-pel.2010.0039

ISSN 1755-4535

Design methodology for universal line input boost power factor correction magnetics
ndido2 M.L. da S. Martins3 H.L. Hey2 J.R.R. Zientarski1 R.C. Beltrame2 D.B. Ca
1 2

Federal Institute of Education, Science and Technology Catarinense, IFC, 89650-000 Videira, SC, Brazil Federal University of Santa Maria, UFSM, 97105-900 Santa Maria, RS, Brazil 3 Federal University of Technology Parana, UTFPR, 85503-390 Pato Branco, PR, Brazil E-mail: mlucio@utfpr.edu.br

Abstract: This study presents a design methodology for inductors employed in single-phase power factor correction (PFC) boost converters operating in continuous conduction mode (CCM). A simulation tool based on the proposed design methodology performs an automatic selection of magnetic core taking into account the standard limits for conducted electromagnetic interference (EMI), single- or multi-layer winding pattern and the limit of temperature rise on the inductor core. Through a sweep simulation, the algorithm chooses the switching frequency and the input current ripple that minimise the magnetic volume of boost inductor considering the characteristics of some powder cores (sendust, molypermalloy and high-ux). The design methodology is conrmed by experimental results obtained from a 650 W universal line input PFC prototype, considering a switching frequency of 70 kHz and an input current ripple of 40%.

Introduction

Electronic equipment is part of everyday human life, providing communication, innumerous forms of entertainment and other facilities including transportation, appliances, utilities and leisure. Most electronic equipment requires direct current (DC) for its proper operation, and the rectication process from the utility alternating current (AC) is usually provided just by using diode bridges with capacitive lters. In spite of cheap and simple, diode bridges with large output capacitors deteriorate the line current producing electromagnetic interference and leading to a poor utilisation of the power source capability [1]. International standards are growing up regarding these drawbacks, as the IEC 61000-3-2, which dene limits of the harmonic content injected into the utility. In order to comply with the standard requirements, many converters with power factor correction (PFC) circuits were proposed [2]. In regard to information technology equipment, the frontend rectier implemented by a boost PFC circuit operating in continuous conduction mode (CCM), as shown in Fig. 1, is widely adopted for medium power applications (400 W up to a few kilowatts), because of the continuous nature of the input current, its high efciency, high-power density, simplicity and low cost. Additionally, information technology equipment must comply with electromagnetic interference (EMI) standards, such as CISPR-22, which determines limits for EMI noise generated at frequencies from 150 kHz up to 30 MHz [3]. The noise generated in this frequency range can cause undesirable effects in other equipments in the converters vicinity, making the employment of EMI lters necessary in order to suppress
IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715724 doi: 10.1049/iet-pel.2010.0039

them. As the boost PFC converter is the rst stage of static power conversion in information technology equipments, it is essential to determine its impact in conducted EMI emissions. Compliance with EMI standards is commonly the last part of a power supply design and, in most cases it involves trial and error methods [4]. Predicting EMI behaviour even before implementing the circuit would be advantageous because it would reduce the development time of the power converter. Therefore the EMI lter could be estimated as well as the nal volume of the converter. The differential mode (DM) noise has a main inuence in the lower frequency range and frequently determines the size of the lter [5]. Unlike the common mode noise, the DM noise is strongly affected by the switching frequency, input current ripple and the parasitic elements of the boost inductor. Wherefore, the EMI prediction performed in this work considers only the DM noise in order to evaluate the effects of the EMI noise in the volume and in the converter performance. Power electronics engineers and researchers have concentrated efforts to design a robust, reliable and low-cost PFC converters. Many papers have been published suggesting modications in the boost PFC converters to reduce their main disadvantages, such as losses in the semiconductors (conduction and switching) and EMI noise generated [6, 7]. These characteristics will determine the volume of heat sink and lters, which together with the boost inductor determine the total volume of the converter and its nal cost. In this scenario, the boost inductor plays a fundamental role in the converter design, because its volume depends mainly of the switching frequency and ripple current, which is also closely related to the volume of
715

& The Institution of Engineering and Technology 2011

www.ietdl.org
EMI lters by calculating the attenuation required in critical harmonic frequency (rst harmonic inside the standard range), does not take into account the effects of these parasitic elements in the EMI noise generated by the converter, and it may lead to a wrong design of the lter. To avoid this problem, many papers have been published suggesting techniques to reduce the parasitic elements in boost inductors and in the components of EMI lter [5]. A simple way to reduce these parasitic elements is the use of single-layer winding for boost inductor. Owing to parasitic capacitances and resistances of the winding and the magnetic core, the inductor acts as a high-order resistorinductor-capacitor (RLC) network, whose behaviour can be modelled by using distributed circuit models [9, 10]. The total parasitic capacitances, named stray capacitance, consist of the following components: the capacitances between turns of the same layer and between turns of adjacent layers (turnto-turn capacitances), and the capacitances between turns and the magnetic core (turn-to-core capacitances), [11, 12]. The parasitic capacitance between adjacent layers have largest magnitude, followed by the capacitance between the turns of the rst layer and the magnetic core (turn-to-core), whereas the capacitance between turns of the same layer (turn-to-turn) is very small and can be neglected [9]. This means that in single-layer inductors the stray capacitance can be signicantly reduced, as there is no capacitance between layers. As an example, two inductors were assembled to make an experimental comparison between single- and multi-layer inductors applied to a boost PFC converter. The compared inductors have the same specications, but the rst is designed to have a single-layer winding and the second uses a standard multi-layer winding. The impedance curve of the compared inductors is presented in Fig. 2a, where it

Fig. 1 Boost PFC converter with a PI-type EMI lter

the other components. The increase of the input current ripple results in a reduction in the boost inductor volume. On the other hand, this leads to an increase in semiconductor losses and in the EMI lter volume necessary to suppress the DM noise. Therefore the choice of appropriate values for both current ripple and switching frequency can yield in a reduction of the total magnetic volume. In relation to semiconductor losses, which determine the heat sink size, the use of recent semiconductor technologies, such as CoolMOSw Power MOSFETs and silicon carbide diodes, can drastically reduce switching losses [8]. Thus, the inuence of the switching frequency and current ripple on the semiconductor losses is alleviated, and the heat sink size becomes almost unaffected by the operation point of the converter. Therefore the heat sink volume will be disregarded in the following analyses. Another important characteristic in the PFC converter design is the magnetic material employed to manufacture the boost inductor. The powder cores with a single-layer winding layout are being adopted in order to reduce the leakage capacitances and the irradiated EMI noise caused by the discrete gap [9 13], which is present in some materials, such as ferrite. Moreover, in a single-layer winding, the magnetic core must t the required number of turns in a single-layer winding. This can affect the minimum core size, and must be considered in its selection. A similar approach for the boost inductor was adopted in [14], which investigated the operation points that reduce the magnetic volume of boost PFC converters. In that study, soft saturation effects, temperature limit and conducted EMI were considered. However, a simple rst-order boost inductor model was used for simulation analysis, which is unable to predict high-frequency EMI noise. In addition, the magnetic cores resulting from the analysis may not have commercial sizes, making the prediction of the inductor volume occasionally be different from that being used in practice. Therefore this paper proposes to expand upon this approach, presenting an algorithm able to simulate the boost converter and to choose the best core array, including inductors with commercial dimensions of the materials sendust, molypermalloy (MPP) and high-ux. The proposed algorithm considers a third-order inductor model for the input current simulation, enabling a good precision of the DM EMI prediction for frequencies up to 30 MHz, which is the upper limit of the range limited by the standard for this application.

Boost inductor modelling

Owing to the existence of parasitic elements in the boost inductor and in the EMI lter, ltering high-frequency components of DM EMI noise has been a difcult task to the engineers, mainly because these parasitic capacitances and resistances are distributed parameters and hence, difcult to model. As presented in some papers [15, 16], the design of
716

Fig. 2 Inuence of the inductors parasitic elements on the conducted EMI


a Comparison between single- and multi-layer boost inductor impedance magnitudes b Measured input EMI noise for converter employing the multi-layer inductor c Measured input EMI noise for converter employing the single-layer inductor IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715 724 doi: 10.1049/iet-pel.2010.0039

& The Institution of Engineering and Technology 2011

www.ietdl.org
is possible to note that the impedance magnitude at the valley for the single-layer inductor is almost nine times larger than in the multi-layer inductor. This magnitude difference affects the measured conducted EMI, as shown in Fig. 2b for the multilayer inductor and in Fig. 2c for the single-layer one. It is evident that by using single-layer winding on the inductor can reduce signicantly the EMI noise at high frequencies. On the other hand, the use of single-layer windings can cause a signicant increase on magnetic volume of the inductor in some values of switching frequency and input current ripple. Therefore this work investigates this inuence and helps the designer of boost PFC converters to determine how far the use of single-layer windings is advantageous. In order to evaluate the benets of the single-layer winding, the boost inductor model must comply with the real inductor behaviour at frequencies up to 30 MHz. Nevertheless, the behaviour of inductors at high frequencies is quite different from their low-frequency behaviour, which can be simply represented by a single self-inductance (L1) as shown in Fig. 3a. The parasitic capacitances and resistances are distributed parameters that are negligible at low frequencies but play a role of increasing signicance as the operating frequency increases, as was addressed in [9 12]. Normally, the inductor windings with distributed parasitic parameters are modelled in a simple way by a lumped parameter equivalent circuit that consists of an inductance L1 , a series AC resistance of the coil R1 , and an overall stray capacitance C1 , as shown in Fig. 3b. These lumped parameters are calculated using some ready-to-use expressions [17]. Regrettably, to represent more accurately the behaviour of the inductor in the entire range of frequencies under the EMC standards, a third-order equivalent circuit model has been used, Fig. 3c, in order to include the self-resonance frequencies of the real inductor prototype. A comparative analysis among the real inductor and its rst-, second- and third-order models are depicted in Fig. 3d. As can be seen, the third-order model permits to represent accurately the valley of impedance magnitude at high frequencies, which impacts on the EMI noise as previously demonstrated in Fig. 2. This way, the third-order equivalent circuit model of Fig. 3c is used in the input current simulation routine which is described in the following section. As an example, the resulting current waveform is presented in Fig. 3e. It is important to highlight that the ringing on the current waveform, detailed in Fig. 3e, is also responsible for the high-frequency EMI noise. The third-order model of Fig. 3c presents the transfer function dened in (1), where the constants aj and bj ( j 0, . . . , 3) are dened and calculated in the Appendix ZL (s) = a3 s3 + a2 s2 + a1 s + a0 b3 s3 + b2 s2 + b1 s + b0 (1)

Design methodology

Fig. 3 Boost inductor models


a First-order equivalent model b Second-order equivalent model c Third-order equivalent model d Comparison among the inductor impedances of real and equivalent models e Comparison between the input currents for rst- and third-order equivalent models IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715724 doi: 10.1049/iet-pel.2010.0039

The choice of the magnetic materials for the inductor cores can signicantly inuence the losses, size and cost of the boost PFC converter. The size of the inductor is directly related to the maximum ux density (Bmax) supported by the magnetic material before appearing the saturation effect. It means that small size inductor can be achieved by high ux density materials. Nevertheless, the losses because of hysteresis and eddy currents usually increase with the ux density. Thus, there is a trade-off between the converter size and efciency that must be taken into account during the choice of the magnetic material. Two different types of core material are commonly used for the inductor in a boost PFC converter: powder cores and ferrite cores. Powder cores have a distributed air gap structure giving a soft saturation characteristic that presents many design benets including an overall smaller core size and also alleviates the fringing ux difculties that occur if a discrete gap design is used. Owing to constructive restrictions, powder cores are commonly available as toroids and are composed by ve different materials: (a) kool Mm, (b) MPP, (c) high-ux, (d) iron powder and (e) XFlux. Iron powder and XFlux materials operate in low-frequency range, which make them unsuitable for the analyses carried out in this paper. Aiming to reduce the magnetic volume in the converter, an algorithm (see Fig. 4) implemented in the software MATLABw performs an automatic design of the boost PFC inductor for a wide range of switching frequencies and input current ripples. For each operating point, the program selects the best magnetic arrangement regarding a maximum temperature rise and the DM EMI generated by the converter. The algorithm is able to stack cores to obtain the best arrangement and also can consider two different congurations for the winding used in the boost inductor, which can be single- or multi-layer. In the case of singlelayer winding, the core selection must be done to t all the required turns in a single-layer of winding.
717

& The Institution of Engineering and Technology 2011

www.ietdl.org
magnetisation are taken into account to accurately estimate the core losses and DM noise [13]. For each inductor current simulation step, the algorithm calculates the effective inductance and the magnetic losses of the inductor. The total losses in the inductor (in the copper and in the magnetic core) allow a prediction of the temperature rise using empirical models [9]. Thus, the maximum temperature rise dened by designer becomes a restrictive factor for the minimum volume of the boost inductor. 3.1 Saturation criteria

The soft saturation is characterised by a continuous variation of the permeability with the DC bias current. As in boost PFC converters the low-frequency current is sinusoidal, there will be a quite variation of the effective inductance during an input voltage period. Thus, to take it into account, the effective permeability (meff ) curves for each material must be computed. The effective permeability can be modelled by (2), where mi is the initial permeability of the core, H is the magnetic eld and the coefcients cj ( j 1, . . . , 4) are obtained from the manufacturer catalogue [18]

meff =

3 4 2 m2 i c1 mi H + c2 mi H 2 1 + c3 mi H + c4 m2 iH

(2)

It is important to note that the B H curve of soft saturation materials does not present a well-dened knee. Hence, in the following analyses, a reduction up to 90% in meff is assumed as the saturation point. 3.2 Single-layer criteria

To ensure that single-layer winding can be implemented, the restriction dened in (3) must be observed, where dci is the internal diameter of the core, N is the number of turns of the inductor winding and dw is the diameter of the selected wire, which is given by (4). Additionally, iLrms is the RMS component of the inductor current and J is the current density

pdci Ndw
dw = 3.3 iLrms J

(3) (4)

Temperature rise prediction

The inductor temperature rise (DT ) is estimated by the empirical equation (5) [18]. Equation (5) represents a simplied model based on the total surface area ST for the heat transfer by natural cooling and on the total losses for the wound core. Additionally, Pcopper is the copper losses and Pcore is the core losses DT = Pcopper + Pcore ST
0.833

(5)

Fig. 4 Design methodology owchart

After the magnetic core selection and the design of the boost inductor, the input current of the boost PFC converter is simulated. As a particularity of the soft saturation materials, inductance variations as a function of core
718 & The Institution of Engineering and Technology 2011

It is important to mention that it is not an easy task to dene the maximum allowable temperature rise, as it has a strong dependence upon certain operating conditions, such as airow and temperatures outside and inside of the equipment case. In the case of powder cores, they can operate up to 2008C, but such magnitude can be harmful to the board, to nearby components and can reduce the core lifetime [19]. Selecting 1008C as a reasonable maximum
IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715 724 doi: 10.1049/iet-pel.2010.0039

www.ietdl.org
core temperature, regarding the mentioned restrictions, it is possible to allow a temperature rise of 508C, considering that the temperature inside the equipment case can be as high as 508C. It is clear that other values for the maximum temperature rise could be used and this is a task that should be based on the designers experience for each specic case. 3.3.1 Copper losses: The copper losses are calculated considering copper resistivity, the skin effect and the proximity effect for each harmonic frequency of the inductor current as in (6) and (7) [20] Rec (f ) = 4 p
(1/4)

where T a R1 Ta 0 1 L L1 1 T 0 1 a A= L2 Ta Ta Ta R2 + R3 1 C1 C1 C1 R2 R3 0 Ta 1 and B= L2 , C = 0 1 R 3 Ta R3 C1 ,

D=

N lturn

rcu mcu p f 2(Nl 1) 1+ dw td 3


1

1 R3

(6)

Pcopper =

Rdc i2 L rms

+
k =1

Rec (kf1 ) irms (kf1 )

(7)

Ta is the sampling period, and the voltage across the inductor (VL[k]) is given by (11) V L [k ] = Vin [k ], Vin [k ] Vout , for kTa , DTS for DTS , kTa , TS (11)

where Rec is the eddy current resistance, f is the frequency, lturn is the average length of a turn, rcu is the copper resistivity, mcu is the copper permeability, td is the distance between the centers of two adjacent turns, Nl is the number of layers of winding, Rdc is the DC resistance, k is the harmonic current order, f1 is the fundamental frequency, iLrms is the RMS inductor current and irms(kf1) is the RMS inductor current of the kth harmonic. 3.3.2 Core losses: The core losses are calculated using the method derived in [21], where the duty cycle variations are automatically taken into account, which is an important consideration in PFC applications. In this method, the losses are calculated at each commutation period by (8), where Vc is the core volume, a, m, n are the Steinmetz equation coefcients, D is the duty cycle of the switch, Ts is the switching period and DB is magnetic ux variation in the switching period Pcore = Vc a DBm (D1n + (1 D)1n ) (2Ts )n (8)

The output voltage of the boost PFC converter is assumed to be constant and to ensure that the current tracks its reference in the CCM operation mode, a PI controller [23] is used. The inductor current is calculated for each simulation step kTa , which must observe the Nyquist Theorem. Thus, in order to reproduce the current harmonic at 30 MHz, the simulation step should be less

The recommended Steinmetz coefcients are those taken from a square wave excitation of the core. However, in general, the manufacturers provide just the coefcients resulting from sine wave excitations. Fortunately, the losses with sine wave excitation are just slightly higher than those from square wave excitation in powder cores [22]. Therefore the sine wave coefcients are used in (8), which can be considered as a conservative approach. 3.3.3 Boost inductor current simulation: To make possible to compute several variables of the design methodology algorithm, it is required to simulate the current through the boost inductor. In order to accomplish this task, the system dened by (9) and (10) is solved for a half period of the AC line voltage. The state vector of system (9) and (10) is dened as X [k] [iL1[k] iL2[k] vC1[k]]T, where iL , iL1 , iL2 and vC1 are depicted in Fig. 2c X [k + 1] = A X [k ] + B VL [k ] iL [k ] = C X [k ] + D VL [k ]
IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715724 doi: 10.1049/iet-pel.2010.0039

Fig. 5 DM EMI noise


a Simplied equivalent circuit for LISN b Predicted DM EMI noise and CISPR 22 limits

Table 1
Parameter

Boost PFC converter specications Value 90 240 Vrms 60 Hz 400 V 650 W 20 160 kHz 20 80% 508C

(9) (10)

input AC line voltage (Vin) input AC line frequency ( f1) output voltage (Vout) output power (Pout) switching frequency range ( fs) maximum current ripple range (DiL) maximum temperature rise (DT(max))

719

& The Institution of Engineering and Technology 2011

www.ietdl.org
than 16.67 ns, which gives 500 000 points in half period of a 60 Hz AC line. As the fast Fourier transform demands a number of points multiple by two, the number of simulation points selected is 524 288 or 219. 3.4 DM EMI prediction the AC line. Additionally, the EMI analyser determines the EMI by measuring the voltage on an internal resistor (50 V) of the LISN. Since EMI noise is estimated, the next step is to determine the cutoff frequency of the DM EMI lter. As discussed in the following analysis, the cutoff frequency can be determined by evaluating the necessary attenuation of the critical harmonic (rst harmonic inside the frequency range determined by the EMI standard). 3.4.1 LISN model: A model that represents the LISN circuit in the frequency range of the EMI standard is proposed by Mainali and Oruganti [4] and presented in

Aiming to dene the cutoff frequency of the DM EMI lter, the EMI noise injected into the grid by the boost PFC converter must be estimated. The DM EMI noise is measured by means of a line impedance stabilisation network (LISN), which aims to standardise the AC line impedance, as well as, to block the noise originated from

Fig. 6 Inductors volume behaviour in function of the switching frequency and input current ripple
a b c d e f Manufactures chart core selection volume for sendust material Proposed core selection volume for sendust material Manufactures chart core selection volume for MPP material Proposed core selection volume for MPP material Manufactures chart core selection volume for high-ux material Proposed core selection volume for high-ux material IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715 724 doi: 10.1049/iet-pel.2010.0039

720 & The Institution of Engineering and Technology 2011

www.ietdl.org

Fig. 7 Comparison of the use of single- or multi-layer winding for the boost inductor
a For the sendust material b For the MPP material c For the high-ux material

Fig. 5a with an additional capacitor included into the model to further improve its accuracy. It must be highlighted that iin is the current through the input of the boost PFC converter simulated for a complete AC line period and the Uint is a voltage level that represents the DM EMI noise, which is expressed in dB mV. Therefore it can be concluded that the EMI prediction accuracy is proportional to the simulated input current, which is ensured by employing the thirdorder model discussed in Section 2. The current iin can be obtained by (12), where f1 is the AC line frequency and iL is given by (10). Additionally, the EMI noise in dB mV can be calculated by (13) iL [k ], iL [k ], 1 2f1 1 for kTa . 2f1 for kTa

complies with the standard, the higher order harmonics will also comply. Fig. 5b shows that the harmonics above 210 kHz are within the frequency range of the CISPR 22 standard (150 kHz 30 MHz). In this case, the harmonic situated in 210 kHz is named critical harmonic. This way, the cut-off frequency of the lter is calculated to meet this critical frequency, as in [14].

Inductor volume analysis

iin [k ] =

(12)

This section presents the simulation results for a single-phase boost PFC converter with a universal input voltage, as summarised in Table 1. It must be highlighted that the inductor design is made for an input voltage of 90 Vrms because this condition results in the highest current level through the converter devices.

Uint(dBmV) = 20 log(106 |Uint (s)|)

(13)

3.4.2 DM EMI lter cutoff frequency: As the CISPR 22 standard limits the noise magnitude only above 150 kHz, any harmonic component below of this value is ignored. On the other hand, in most cases the rst harmonic of the spectra above 150 kHz denes the required EMI lter attenuation [24]. It means that whether the lowest critical harmonic
Table 2
Parameter input AC line voltage (Vin), Vrms input AC line frequency (f1), Hz output voltage (Vout), V output power (Pout), W switching frequency (fs), kHz maximum current ripple (DiL(max)), % core part number core volume, cm3 inductor total volume, cm3 initial inductance, mH effective inductance, mH

Boost PFC prototype specications Sendust 90 60 400 650 70 40 2 77716 31.86 38.22 451 279 MPP 90 60 400 650 70 40 2 55716 31.86 37.4 343 279 High-ux 90 60 400 650 70 40 2 8716 31.86 37.86 419 279

Fig. 8 Selection of the input current ripple for experimental evaluation (at 70 kHz)
a DM EMI lters cutoff frequency in function of the current ripple b Inductor volume in function of the current ripple 721

Standard cores from the manufactures catalog [18]

IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715724 doi: 10.1049/iet-pel.2010.0039

& The Institution of Engineering and Technology 2011

www.ietdl.org
It is important to note that the volumes achieved in this simulation and presented in Fig. 6 are strongly dependent of the maximum temperature rise dened by the designer, Table 1. The inuence of the use of single- or multi-layer winding on the volume of the boost inductor can be evaluated by the simulation results illustrated in Fig. 7, which shows the total inductor volume (core and winding) as function of the input current ripple for a selected switching frequency of 70 kHz. The use of low values of input current ripple brings a high increase on the volume for the single-layer inductor, but, in high values of ripple, this difference is not large. In many cases, the volume of the core is not limited by the layout criteria, but by the limit of temperature rise (usually in larger values of current ripple). In these cases, there are no differences between the volumes on the two types of winding, the chosen core is the same and the number of turns forces the winding to have a single-layer layout. As can be observed in Fig. 7b, the material with less magnetic losses (MPP) results in major differences in the volume. It occurs because the minimum volume is not limited by the temperature rise. In the case of the material high-ux, Fig. 7c, this behaviour does not happen, and at low values of current ripple, the difference in the volume is large because of its high ux density characteristic. At high values of current ripple these differences does not appear because of the volume is not limited by the winding layout, but by the temperature limit, caused by the high magnetic losses present in this material. As can be observed, the points that minimise the volume of the boost inductor occur generally in medium values of current ripple, depending on the material. In these points, the proposed magnetic core selection method presents small differences in the volume between single- or multi-layer winding. As a result, it is possible to infer that for the design specications adopted in this paper, the use of single-layer windings is advantageous, since it can further reduce the high-frequency conducted EMI without compromising signicantly the volume of the inductor.

Fig. 9 Measured and simulated input current for the sendust material

Using the proposed magnetic selection method, the algorithm makes a simulation sweeping for a wide range of values of input current ripple and switching frequencies, as dened in Table 1, for the materials sendust, MPP and high ux. It must be noticed that all the cores presented in the manufactures catalogue [18] are take into account in the analyses. The results are the graphs shown in Fig. 6. As can be observed, the magnetic volume behaviour in function of the input current ripple and switching frequency in the proposed selection method (Figs. 6b, d and f ) is quite different to that using the manufacturers Li2 method (Figs. 6a, c and e) [18]. This occurs due to the restrictions of temperature rise and layout (single- or multi-layer, as can be seen in the owchart of Fig. 4) that enlarge the magnetic volume aiming to assure the mentioned limits for all operations points. This leads to great differences in the volume behaviour among the three analysed materials. Using the proposed method for MPP material, the points that minimise the inductor volume occur at high frequencies (from 100 to 140 kHz), and medium values of current ripple (around 60%), as can be seen in Fig. 6d. It occurs because the magnetic losses are smaller, allowing a large variation in ux density (dB/dt ), which is directly related to the inductance and, consequently, the value of maximum allowed current ripple. In the case of material high-ux, the point that minimises the inductor volume also occurs in high switching frequency, but low input current ripples (less than 40%), as shown in Fig. 6f. This result occurs because of the losses in the magnetic material are higher, and the use of high values of current ripple requires the use of larger volume cores to limit the temperature rise. Otherwise, the sendust material presented an intermediate result to the two other materials, as can be seen in Fig. 6b, because of their moderated characteristics of losses and ux density.

Experimental results

In order to evaluate the core temperature rise and the DM EMI estimation used in the minimum magnetic volume simulation algorithm, a boost PFC converter was assembled with specications summarised in Table 2, which also includes the specication of the analysed cores. As can be seen in Table 2, the prototype was implemented considering a switching frequency of 70 kHz and an input current ripple of 40%. This switching frequency was chosen

Fig. 10 Experimental inductor thermal images (at 70 kHz, 650 W)


a For the sendust material b For the MPP material c For the high-ux material 722 & The Institution of Engineering and Technology 2011 IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715 724 doi: 10.1049/iet-pel.2010.0039

www.ietdl.org
because it presents a good relationship between the magnetic volume and the conducted EMI generated using single-layer winding. Fig. 8a presents the cutoff frequency of the DM EMI lter for this switching frequency. The input current ripple of 40% was chosen because in this operation point, the magnetic volume for the three materials is the same, as shown in Fig. 8b, as well as, the cutoff frequency of the DM EMI lter is similar, Fig. 8a, allowing a fair comparison among them. Aiming to demonstrate that the implemented converter operates in CCM, presents a sinusoidal input current, as well as the predicted current ripple is about 40%, in Fig. 9 are shown the measured and simulated input currents for the converter employing an inductor implemented with sendust material. Fig. 10 presents the experimental results obtained for the temperature rise in the core, measured using the Flukew Ti20 thermal imager. As expected, the material high-ux presented a higher temperature rise caused by the higher losses. The measured temperatures are listed in Table 3. As can be seen, the predicted temperature rise is very much similar to the measured value for the three materials. Fig. 11 presents the experimental results for the conducted DM EMI noise, obtained with the spectrum analyser Agilentw E7402A against the prediction performed by the algorithm for the materials sendust, MPP and high-ux, respectively. As can be observed, the measured and simulated values are very close for frequencies below 1 MHz. It evidences a good precision of the cutoff frequency estimation for the EMI lter. It is important to highlight that no EMI lter was used in the measurements because the objective of this analysis was just to compare the predicted noise and the measured one. Additionally, the prediction results are close to the experimental values for frequencies above 1 MHz, showing the accuracy of the model at high frequencies and giving to the designer an outline of the behaviour of generated EMI at this frequency range. Furthermore, the measured conducted EMI noise on critical harmonic (210 kHz) was slightly larger than the predicted one (see Table 3), leading to a conservative design of the DM EMI lter.

Fig. 11 Predicted EMI against experimental measurements


a For the sendust material b For the MPP material c For the high-ux material

Conclusions

In this paper an alternative design methodology was presented for an inductor of single-phase PFC boost converters. An investigation of the dependence of volume on the switching
Table 3
Parameter measured input current ripple, % ambient temperature, 8C measured core temperature, 8C simulated total inductor losses, W simulated temperature rise, 8C experimental temperature rise, 8C temperature prediction error, % simulated conducted EMI on critical harmonic (210 kHz), dB mV measured conducted EMI on critical harmonic (210 kHz), dB mV conducted EMI prediction error, dB mV Boost PFC prototype measurements Sendust 39.5 28.0 62.4 7.08 36.7 34.4 2 3.7 126.8 125.4 1.4 MPP 37.9 29.5 59.0 4.8 26.6 29.5 + 4.9 127.6 126 1.6 High-ux 36.3 28.5 68.1 7.53 38.7 39.6 + 1.32 127.2 125.9 1.3

frequency and the input current ripple amplitude was performed, ensuring the electromagnetic compatibility with the standards IEC 61000-3-2 and CISPR 22. Through an input current simulation that takes into account several converter characteristics, a method is proposed that selects the smallest magnetic arrangement for each operating point of the converter. This selection method considers the design restrictions of saturation, single-layer winding pattern and maximum temperature rise. A sweep of the input current ripple and switching frequency in the proposed selection method results in a graph that allows the choice of an optimised switching frequency and input current ripple amplitude that minimises the magnetic volume of the inductor for each material. It was veried that the points that minimise the volume of the boost inductor occurs generally in medium values of current ripple, depending on the material. In these points, the proposed magnetic core selection method presents small differences in the volume between the use of single- or multi-layer winding. As a result, it is possible to infer that for the design specications adopted in this paper, the use of single-layer windings is advantageous, since it propitiates good results regarding the high power density, low-cost, reduction of leakage capacitances and EMI noise without compromising signicantly the inductor volume. The experimental results conrm the results obtained by the algorithm. Even using empirical models to estimate the temperature rise of the boost inductor, the temperature error in the experimental results is less than 5%, which can be considered a good approximation to prevent the overheating of the core. In addition, the conducted EMI prediction model presented a good response with an error of 1.6 dB/ mV at critical frequency, conrming that the used model is enough to estimate the EMI lter required.
723

IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715724 doi: 10.1049/iet-pel.2010.0039

& The Institution of Engineering and Technology 2011

www.ietdl.org
7 Acknowledgments
o The authors would like to express their gratitude to Coordenac a de Aperfeic oamento de Pessoal de Ensino Superior CAPES and Conselho Nacional de Desenvolvimento Cient co e Tecnolo gico CNPq (proc. 307798/2009-7, proc. 478154/2009-7 and proc. 554103/2010-9) for nancial support, Icotron an EPCOS Company, Thornton Inpec Electro nica Ltda and Xilinx for material support.
23 Silva, E.T.: Ana lise e Projeto de Compensadores para o Conversor o de Mestrado, UFSC, Floriano Boost. Dissertac a polis-SC/Brasil, 1994, p. 107 24 Shih, F.-Y., Chen, D.Y., Wu, Y.-P., Chen, Y.-T.: A procedure for designing EMI lters for AC line applications, IEEE Trans. Power Electron., 1996, 11, (1), pp. 170181

Appendix

References

1 Wichakool, W., Avestruz, A.T., Cox, R.W., Leeb, S.B.: Modeling and estimating current harmonics of variable electronic loads, IEEE Trans. Power Electron., 2009, 24, (12), pp. 28032811 2 Agamy, M.S., Jain, P.: Performance comparison of single-stage threelevel resonant AC/DC converter topologies, IEEE Trans. Power Electron., 2009, 24, (4), pp. 10231031 3 Nussbaumer, T., Heldwein, M.L., Kolar, J.W.: Differential mode input lter design for a three-phase buck-type PWM rectier based on modeling of the EMC test receiver, IEEE Trans. Ind. Electron., 2006, 53, (6), pp. 16491661 4 Mainali, K., Oruganti, R.: Simple analytical models to predict conducted EMI Noise in a power electronic converter. Proc. Industrial Electronics Society, IECON 2007, 2007, pp. 1930 1936 5 Wang, S., Lee, F.C., Odendaal, W.: Improving the performance of boost PFC EMI lters. Proc. Applied Power Electronics Conf. and Exposition, APEC 03, 2003, pp. 368 374 6 Huber, L., Jang, Y., Jovanovic, M.M.: Performance evaluation of bridgeless PFC boost rectiers, IEEE Trans. Power Electron., 2008, 23, (3), pp. 13811390 7 Jang, Y., Jovanovic, M.M., Fang, K.H., Chang, Y.M.: High-powerfactor soft-switched boost converter, IEEE Trans. Power Electron., 2006, 21, (1), pp. 98104 8 Lu, B., Dong, W., Zhao, Q., Lee, F.C.: Performance evaluation of CoolMOSTM and SiC diode for single-phase power factor correction applications. Proc. Applied Power Electronics Conf. and Exposition, APEC 03, 2003, pp. 651 657 9 Wang, S., Lee, F.C., Odendaal, W.: Single layer iron powder core inductor model and its effect on boost PFC EMI noise. Proc. Power Electronics Specialist Conf., PESC 03, 2003, pp. 847 852 10 Hole, M.J., Appel, L.C.: Stray capacitance of a two-layer air-cored inductor, IEE Proc. Circuits Devices Syst., 2005, 152, (6), pp. 565 572 11 Grandi, G., Kazimierczuk, M.K., Massarini, A., Reggiani, U.: Stray capacitances of single-layer solenoid air-core inductors, IEEE Trans. Ind. Appl., 1999, 35, (5), pp. 1162 1168 12 Massarini, A., Kazimierczuk, M.K.: Self-capacitance of inductors, IEEE Trans. Power Electron., 1997, 12, (4), pp. 671676 13 Yang, L., Lu, B., Dong, W., et al.: Modeling and characterization of a 1 KW CCM PFC converter for conducted EMI prediction. Proc. Applied Power Electronics Conf. and Exposition, APEC 04, 2004, pp. 763 769 14 Pieniz, M., Pinheiro, J.R., Hey, H.L.: An investigation of the boost inductor volume applied to PFC Converters. Proc. Power Electronics Specialists Conf., PESC 06, 2006, pp. 1 7 15 Rossetto, L., Buso, S., Spiazzi, G.: Conducted EMI issues in a 600-W single-phase boost PFC design, IEEE Trans. Ind. Appl., 2000, 36, (2), pp. 578585 16 Lu, B., Dong, W., Wang, S., Lee, F.C.: High frequency investigation of single-switch CCM power factor correction converter. Proc. Applied Power Electronics Conf. and Exposition, APEC 04, 2004, pp. 1481148 17 Massarini, A., Kazimierczuk, M.K., Grandi, G.: Lumped parameter models for single and multiple-layer inductors. Proc. Power Electronics Specialists Conf., PESC 96, 1996, pp. 295301 18 Magneticsw: Powder cores design manual and catalog. http://www. mag-inc.com, accessed November 2008 19 Busquets-Monge, S., Crebier, J.-C., Ragon, S., et al.: Design of a boost power factor correction converter using optimization techniques, IEEE Trans. Power Electron., 2004, 19, (6), pp. 13881396 20 Bartoli, M., Reatti, A., Kazimierczuk, M.: Modelling iron-powder inductors at high frequencies. Proc. Industry Applications Society Annual Meeting, IAS94, 1994, pp. 12251232 21 Liu, J., Wilson, T., Wong, R., Wunderlich, R., Lee, F.C.: A method for inductor core loss estimation in power factor correction applications. Proc. Applied Power Electronics Conf. and Exposition, APEC 2002, 2002, pp. 439 445 22 Thottuvelil, V., Wilson, T., Owen, H.: High-frequency measurement techniques for magnetic cores, IEEE Trans. Power Electron., 1990, 5, (1), pp. 4153 724

The coefcients employed in the inductor transfer function (1) are those dened in (14). As can be seen, the coefcients are function of the effective inductance L1 , as well as the stray parameters that represent the inductor behavior at high frequency. In order to obtain the values of stray parameters, three inductors for each evaluated material (sendust, MPP and high-ux) were designed to operate in frequencies from 46.6 to 170 kHz and 650 W, and their impedance magnitudes were experimentally measured, as shown in Fig. 12. Therefore the stray parameters can be determined by tting the experimental impedance curves with the theoretical one, given by (1) a0 = R1 R2 R3 a1 = L2 R3 (R1 + R2 ) + R2 (L2 R1 + L1 R3 ) a2 = L1 L2 (R2 + R3 ) + L2 CR1 R2 R3 a3 = L1 L2 CR2 R3 b 0 = R 3 (R 1 + R 2 ) b1 = L2 (R1 + R2 ) + R3 (L1 + CR1 R2 ) b2 = CR2 (L1 R3 + L2 R1 ) + L1 L2 b3 = L1 L2 CR2 Since the algorithm varies the operation frequency, the stray parameters employed during the simulation were approximated by making an arithmetic mean of those obtained from the curve tting. This way, the stray parameters for the inductor third-order model of Fig. 3c employed in the simulation are those summarised in Table 4. (14)

Fig. 12 Measured boost inductor impedance magnitudes Table 4


Parameter R1 , mV R2 , kV R3 , kV C, pF L2 , mH Parameters for the inductor third-order model Sendust 40 53.00 2.40 14.33 6.00 MPP 40 41.33 2.27 18.33 7.17 High-ux 40 26.00 3.57 19.67 7.50

& The Institution of Engineering and Technology 2011

IET Power Electron., 2011, Vol. 4, Iss. 6, pp. 715 724 doi: 10.1049/iet-pel.2010.0039

S-ar putea să vă placă și