Calculation of Leakage Inductance in Transformer Windings William G. Hurley, Senior Member, IEEE, and David J. Wileox, Member, IEEE Absiract—A formula is presented to calculate mutual imped ance between transformer windings on ferromagnetic cores The formula i based on the solution of Maxwells Equations for foils on ferromagnetic cores and a such offers the ulate in accuracy. The formula I frequency dependent, taking into sccount the effect of eddy currents in the core on the fhux iribution as well as representing the eddy current core lass fs an equivalent resistance. Experimental results are presented for leakage Inductance and an Mlustrative example fs presented Showing how leakage inductance affects the operation of ypleal Sitching mode power supply. Approximations for the formula fare also presented to simplify the calulations under certain ‘operating conditions. NoMeNcLATURE In general, subscript I refers to coil | and subscript 2 refers, to coll 2. A Average col radius Ae Core cross-sectional area ‘25,42 Inside and outside radi of a col 5 Core radius GMD Geometric mean distance between cols. k Radial height of coil Iq(2)-K#) Modified Bessel functions of the fist and second kind, respectively: Lh Self inductance associated with primary (air field. Lu Self inductance of a col In Component of Ly: deo Aux linking coll 2 L Sel inductance associated with reflected i th Leakage inductance t Sruve Function. 1 “Magnetic pathlength of core (measured on center ax). M, Mutual inductance associated with primary (ai) it. m Gena) N umber of tens in col R ‘Average cil radius iota side and outside rai of a coi ° Width of cot z + jf, Mutual impedance z Ro toNe, Mutual impedance associated ‘with reflected field. Mansi csv July 30, 1950 eid Septet 8,193. Tis wok ‘nu pully fol by EOLAS, the rsh Scents ad Tooley Ase ‘Mite porum of avn ecology ia Power Eicon a Internal impedance of a conductor. : Distance between filaments or coil centers, ‘measured along the core axis r VP + jemepotee) én Flux linking coils T/and 2 due to current in coil 2 bs “Magnetic permeability of free space. rx 10-7 Him, Relative permeability of core Conductivity of the core material Conductivity of the winding conductor. Angular frequency. 1. Ivrropuction TIS A wellknown fact that successful miniaturization of power supplies requires that magnetic components, such ‘as transformers and inductors, operate at higher frequencies, Manufacturers depend on empirical models to predict high frequency operation of ferromagnetic cores. Power densities ‘of 100 Win" are only possible, with current materials, by ‘Switching at frequencies approaching 10 MHz. The inductance fof a coil on a ferrite core at 10 MHz is quite different from the inductance of the same coil at 100 kHz, Clearly precise frequency dependent mode! is required to investigate the limits of current materials and to characterize those properties which mike high frequency operation possible. The issues involved in ransformer design at 10 MHz are clearly enunciated i (1. Previous work [2] has shown that itis possible 10 take proper account of frequency and eddy current effects in transformer and inductor cores by means of a new frequency fore shed addtional light on the leakage inductance problem, paricularly at high frequencies. Tl, MUTUAL IMPEDANCE IN A ToRoIAt Cont “The detailed derivation of the formula is given in (2) and js summarised here for completeness. The closed magnetic path of the toroidal core, as shown in Fig. (a), is modeled by imagining the actual core to have been cut open at A 1 and straightened out (0 its Fength . The magnetic circuit, is then reclosed by providing a retum flux path of zero reluctance between the two ends, This is done by placing the straightened out core between two infinite plates of ideal magnetic material (je = co, = 0) a5 shown in Fig. 1(b), Maxwell's Equations are solved for the structure of Fig. 1(b) using integral transform techniques, as described in [2] Experimental results [3} have show that the stracture of Fi. 1b) is a very good approximation to the closed core. Since, in practice, the length of the core is much greater than its diameter, the field in the vicinity of the cols is not unduly affected. Furthermore, it can be shown that the magnetic endplates in Fig. 1b) can be removed and replaced by images of coil | at distances = = +h, k= 1,2, 9... on a core of infinite length shown in Fig. 1(c), This isa mathematical device to satisfy the field boundary conditions in the ease of a closed core (2) TREE TRANSACTIONS ON FOWER FLECTRONIS, YOK 9, NO. JANUARY 14 “The mutual impedance between the two coils is given by o ywahatab® (ad (Cob) {eer Gatien) 4295 Puts ts) Paras Pern) Q( 01 Betea) 2 jul +2, toxin SiMe 20 = HOR oyhata ~onjetina} ® Ao(B0) J A neh 2) and P= VOR + jupetiodeo) i = KO fl Po = Venet0%) (as. Bu) = (Be) ~ BIN Pls) = x=[R2)lo(2) + fale) Kol Q(B. By) = 2feos|s(x ~ 9/21 ~ coslCe + 9)/2))/5 [Mf isthe mutual inductance between coil 2 and. coil 1 and its images as shown in Fig. 1), without the core. These mutual inductances can be calculated from tables given by Grover 17), Altematively the coils may be replaced by equivalent filaments separated by an axial distance equal tothe geometric mean distance (GMD) between the coils or in the case of self inductance, the geometric mean distance ofthe coil crss- ‘section, ln this approximation, My is given by: M anf R onmsante( E +235 1IHAR)R(HAPoa(HsGmD)} ‘Ais the mean rau of coil 1 and Fis the mean radius of Sait 2 The romaining dimension are given in Fig. MZ. represents the rected il! due to the presence of the core Z has both real and imaginary components (2. M$ jo). fe represents the eddy cue loss in the core td AF. represeny the aktiona coi inductance de to the ore presence Th are ied Bess fantions ofthe fist and Second kin respectively Li the motfed Stave fton [8 Note ta ther are there are two component in 2 and 3). ‘The int component is independent of the ditance = between the coils and Coesponds fo the nraight ines of ox which Tink all the coils om the core, The summation ters represent the effects of Ban which outside the core, incline Aux stich scapes fom the core. Inshore summation ferns Tennesent leakage fon A. Self Impedance Tests Self impedance tests were caried out on a Micrometls, 7400-26 10 cm O.D. powdered core with a distributed wap.28 0 Full details ofthe test core are given in the Appendix, and the coil dimensions are given in Fig. 2. Measurements were taken sing a Hewlet Packard impedance analyzer, from 100 Hi. to 1 MHz. Calculation of self impedance is based on (1)-(), (0), and both measured and calculated results are shown in Fig 3. Note that measurements beyond about | MHz ae not practical asa consequence of the inter-tum capacitance effects inthe coil, The value of core relative permeability given by the manufacturer is 75 which is low on account ofthe distibuted gap. The value of effective core resistivity for the bes i is (0.01 2m, which is much higher than the intrinsic resistivity fof the iron on account of eddy current suppression inherent in powdered iron design, It is interesting to compare the above results with the classical formula for inductance of toroidal cores. At low Frequencies Py = V/jutso 0) 0 and from [8] Pobho(Cot ey Tab) and if 1. > 1 then the inductance due t0 the constant core term becomes, for & simple coil of NV turns, epoN?r8 _ sno! T T 6 whieh is recognized as the classical formula for the induc tance of a toroid. Further insight into the nature of leakage inductance can be obtained by examining the conditions whieh reduce the proposed formula to the familiar, and widely used, Tow frequency Formula: = Negleet air term. — Core taken to be lossless with high magnetic pesme- ability Flux density uniform over the core cross-section Tenore terms corresponding to leakage Mux “The self inductance of the test coil, calculated from the classical formula (5). is 85 Hl (manufacturer's data gives Bly). The calculated value using the new formula (1), whieh atinuctance Ge) ie ee eee Fe 3. (a ndctve component ofa impetancs. (Resse component fl npedanc. ‘agrees with the measured value, is 144 HH at 100 Ha. This is made up as follows constant core term 85 jill summation core term 34 ‘constant aterm 8 pH summation air term 17 ph “The classical formula predicts the constant core term as ex- pected, but the other components ae almost equally important, making up 41% ofthe toa In the case of a coil which is wound on the coe so that», the coil width, is equal to the core length J, then Q(B, y) in the summation tenm of (2) becomes Asin®(Hul/2) _ 4sin®(2xkl/20) eae ‘casing the summation tem to disappear. In this ease there is ho Teakage fx and the inductance reduces to that given by the classical toroidal formula, This is what the manofacturer’s data is based on, Clealy, in a practical situation where there ate many windings on a eore the simplified formula is quite inadequate “The calculations for Fig. 3 include a conection for the {nernal impedance of the coil conductor, due tothe skin effect, aiven by (9) rurlolmr) h(n) m= Vas00.Where dy is the conductivity ofthe copper conductor and r is the conductor radius. In Fig. 3(b) the value of resistance at low frequency is that ofthe coil and test leads (0.19 9). At the lower frequencies, say up 10 50 kHz, self inductance is high and self resistance is ow. However atthe higher frequencies there is a substantial reduction in inductance and a substantial increase in losses. The deterioration in performance accurately predicted by the formula, explains why the manufacturers ‘deem the core unsuitable for operation above SO KH, B. Approximations forthe Impedance Formula tis unfortunate thatthe proposed formula isso complicated and it is natural to seek simplified working approximations {0 it, There are two regions of the self inductance curve of Fig. 31a) which show asymptotic behaviour, one at low frequency and the other at high frequency. The formulas for T and T show that frequency and condactivty appear together asthe product wr. In this context low frequency and low conductivity ae interchangeable. A core exhibiting low ‘conductivity such asa ferrite, would have a lat self inductance characteristic up to about | MHz. This justifies a simplifestion (of (3). Noting that in practice f/b > 25 for toroidal cores and eh 1o(5b) Kola) ‘The constant core term is given by the classical toroidal formula, (5), since Ty 0. As expected the result gives inductance only and the reflected component 1. of the self inductance of a coil of N ums is given by 43) X25 Ps has }Q(He, Hem) (6) ‘The high frequency asymptote of the self inductance curve in Fig. Xa) corresponds to very high frequency or very high ‘conductivity and represents the limiting case of zero flux pperation into the core (as a result of skin effect caused by eddy currents). In this ease I’ — 00 and _hi(Pob) rie Publ by ‘The self inductance of coil with AV turs is then ny HONS wD eaeau a cate (hu AT 2D PP Aaa Peer YOQCAaw. H20)(%.) (1) and _ hob, Ki} ‘Tre inductance in tis cas is actually reduced by the presence of the core. Evidently itis independent of the magnetic Permeability of the core. a) = Fes Leakage stuctace IIL LEAKAGE INDUCTANCE Leakage inductance is the property of one winding with respect to another. Consider two air coils shown in Fig. 4, If the instantaneous current in coil 1 is i; and coil 2 i open circuited, the general character of the magnetic field produced by 1 Is shown in the field map of Fig. (a). The dete lines show the fux liking both coils and the solid ines indicate flux which links coil 1 only, this is leakage Mux of coil 1. Fig. 4b) shows the corresponding situation when coil 2 is carrying current ig and coil | is open citcuited, The mature ofthe 10 leakage fields is quite different. With both coils energized the flux in each coil as three components, on = du +a + O12 an = da + Oa + ba ‘where diy is the leakage flux due 101, 1 i dhe ux liking both cols due to 1, isthe flux linking both coils due to fa and 6 the leakage ux due to ig. In terms of indvetance ‘we can write the voltages atthe terminals of each coil iy Vi = [a + all + a i Yaw Mas 4 et bul The self inductance of coil Lis Lin. Lay = bn + La which can be measured. We can then extract the letkage term by noting that: ay = (Na/Na)May daa = (Nai Ns)Miz MF is the mutual inductance between the two Mu = Ma coils Ly (Mi /No)ME 1M in this case is simply the inductive component of Z in (1) and Ly is found from (1) with = = 0.9 “s 32 Ved | § ool i © wens | ie ot : i liciemaes ie to a 10 (e) °o 9 a 80 8000 aT rt ary , Bp eee eaar eae © vont ty 2, ' @® ‘ ine : .tesage nt Tes io ‘ Leatagelnacanes le wee cae ct onthe 10 om ls Sac teaitatdeg oer cischameiny ts —mpepin TV, LEAKAGE INDUCTANCE IN A PWM CONVERTER eee eee eee eee ab Pee cesta cuits ale coe as ‘When the switch @ is closed, the input DC voltage is applied 8'¥!"® diode D is forward biased and diode Dz is reverse biased When switch @ is opened, the ux in the core is reset through diode Ds. Under ideal conditions, where there is no leakage inductance in the tansformer, diode D, is now reverse bised and diode Dp is forward biased. The detailed operation ofthis circuit is piven in (10). Ideally, current should commutate instantaneously from D. to Dz and vice versa. When leakage inductance is present in the transformer secondary, instantaneous commutation is not possible. When Q is turned on the current in Dy builds up through the leakage inductance Fy and diode Dz remains fon until the load current is established in Dy. Since both diodes ar on the secondary voltage appears across the leakage inductance ‘We are assuming thatthe ripple inthe output current i small 48 it should be in «well designed converter, ‘Wheo the load current is established in Ds, Da is tuened off. ‘When @ is turned of, the current in Dy decreases linearly to zero and the secondary voltage across the leakage inductance {1N;)ar so thatthe tum-of commutation periods Since the input voltage is not presented to the output during the turn on commutation period, the average output is reduced from its ideal value = 1D =r /T I c duty cycle and T the petiod of the converter.Another consequence of the leakage inductance is thatthe maximum frequency of operation is limited bY tx + "yo ‘This has serious implications for PWM contro, because higher Frequencies lead to smaller magnetic components and higher power densities, V. Coxcuusion A formulas been presented which allows accurate calcu lation of leakage inductance in a wasformer, The formula has been experimentally validated for a powdered iron core. It has been shown that the classical formula ean be grossly inade {quate for seif inductance calculations. It as also been shown that neglecting the core in leakage inductance calculations can lead to inaceurate results, Teas been shown that leakage inductance limits the upper frequency of operation ina typical PWM. converter and reduces the output voltage, The toroidal core with a homogeneous magnetic material js the closest physical representation t0 the ideal model of 8 straight core between magnetic endplates. Since leakage effects are concentrated around the windings, the authors feel that the model would be suitible for any core with a closed magnetic path, such as a pot cor. APPENDIX Core radius 10.8 mm Number of turns 28 Coil inside radius 15 mm Coil outside radius 25 mm Coil width 10 mm Magnetic path length 280 mm Relative permeability 5 Core effective resisivity 0.01 4m RerERENCES IN} A. Golders 1 6. Kassin snd ML PSs Tes eae 0 2) BPs! Wor Nt Cankn an W. GH, "Calon of sf and na ings on eat TEE Pe Pa vol 38 pp. 70-76 Sep 8 toy Bi wide, We tiny ant M Conon, “Calculation fs and tu impedances tween seco of tanfomet Nidps” IEE Pr: Part Cv. 1p M831 Sep 188 (8) VA Niele. Sha A ME Using ¥-. Chang. Gog A. Owen Ie and: Wong, “kato th sh Su indents ‘ailing tansfomet ts emery.“ TEEE Poner econ Spc Co Red SRDS. p. 788, 15 PIR Mem A eee a Conco Gaeted for shr cet impalaers EEE Power Elcom Spe a Con Recor SOCHORTS pp. 823 0. 10 (6) Pol Dowel tee o fy cet i tnfomer windings” Pr: 1 the eh, wl TT pp 378-1398, Aug 1368, In] EW, Grover Indo Calan: Woking Formulas and Tales, Now You Boner, Is] NW MeL, Bete! Futons or Elers, Ch 1 and VI 2nd {8 Oxon Clon Prev 1988 (01 8 Ramo. JR Whienery and T” Yan Duss, Fede and Wines a Commons Elcs Rew Yo: Wey 968 10) 174mm, MOF” Shs an C-Veree, Pine of Power Etec Ch 7 MAS Adit Wey, 19 lam Gerard Harkey (M7789) 9 8 ia'Cad, isn! io 1952 fe nese the BE Aegina emcee fo te Neos ney tea NU Co 1 he ev nite Tuy sd ttn (NU Gay in 98 2 pods engineer fm 1977 199 al ar vy on 79 Teen rie lecture in Electrical Engineering at Universty Calls, Gav. clin He [the decir ofthe Per Elenicr Resch Coste David shin Wile (BERD) was or in se rg in 198% He ended fo ne Une fof Manis stn of Secs aed Tool THe has ol cadre sppsment the Fos st Poecinge de Mone (905-72 he ‘esi of Sng (97270 ra Une Uiste-s, nd Natoel Uses cand {Uom-pesea, He cuentas he Chai of conic Fini at Unive Coleg, Caen. elm ote Wikow fa Fel othe incon o Eggers of lad,» Mente af the nan of tical Exner (Lopon it Chered