Free positioning for inductive wireless power system Eberhard Waffenschmidt Philips Research Europe Eindhoven, The Netherlands eberhard.waffenschmidt@philips.com Abstract— Wireless power transmission suggest the freedom of placement for power transmission. However, efficiency and emitted magnetic fields limit the inductive power transfer to close to a surface. But even there a lateral displacement of the receiver coil to the transmitter coil may lead to a change of the coupling factor and thus an unwanted variation of the power transfer. Here, an algorithm to determine the turn distribution to achieve homogeneous coupling between coils of different diameter is described. As long as the coils overlap, the variation of the coupling factor is very low. To achieve a lateral displacement over an even larger area, an array of transmitter coils can be used. The size of the receiver coil is selected such that it always covers a complete transmitter coil. If only the covered transmitter coil is activated by a suitable detection circuit, the power transmission area can be arbitrary large with homogeneous magnetic coupling. I. can be arbitrary large. This way, homogenous magnetic coupling is achieved on top of one of the transmitter coils and over the whole array. II. TURN DISTRIBUTION The idea to find a suitable turn distribution is: First find a current density distribution, which generates a homogeneous field above the coil. Then find in a next step a turn distribution, which narrows this current density distribution, with equal current in each turn. In a third step, an approximate analytical expression is derived. This is illustrated in Figure 1. H Specified magnetic field INTRODUCTION Wireless power transmission suggest the freedom of placement for power transmission. However, efficiency and emitted magnetic fields limit the inductive power transfer to close to a surface [1]. In inductive wireless power transmission system a lateral displacement of the receiver (Rx) coil to the transmitter (Tx) coil leads to a change of the coupling factor and thus an unwanted variation of the power transfer. For single coil systems literature describes solutions with larger transmitter coils, which have a non-uniform distribution of winding turns [2] [3]. Both sources use empirical approaches, which lead to non-optimal designs. Another approach is to use an array of transmitter coils, where at least one transmitter coil is always covered by the receiver, as e.g. published by the Wireless Power Consortium [4]. But this approach may still lead to a variation of the coupling, if the receiver’s position is varied on top of one transmitter coil. This paper aims in merging and improving these two approaches. First, a procedure to calculate a winding turn distribution to achieve a homogeneous magnetic coupling for different sized coils is introduced. Contrary to existing single coil solutions, this winding design is applied to the receiver coil, which overlaps the transmitter coil. The transmitter coils are arranged to an array, such that always one transmitter is completely covered. If only the covered coil is activated by a detection circuit, the power transmission area 1. step Current density in a disk 2. step Turn distribution Figure 1: Illustration of the approach to calculate the turn density A. Solving the inverse field problem The task of finding a suitable current density distribution relates to the task of solving the inverse magnetic field problem. Here, a disk shaped coil with a limited outer radius is used as example to derive and explain the algorithm. H1 H2 H3 H4 a11 a12 a13 a14 J1 J2 J3 J4 HN a1N JN Figure 2: Illustration of the inter-dependence of currents and magnetic fields. A discrete approach is used. The winding width is divided into equally spaced current traces. Each current trace contributes to the magnetic field at each magnetic field position, as illustrated in Figure 2. The dependence of the magnetic field on one of the current traces can be expressed with a coefficient. The current values can be combined to a vector J , the values for the magnetic field can be combined to a vector H and the coefficients can be combined to a matrix A. Then, Figure 2 can be expressed as equation system: (1) H1 a11 a12 a13 ... H2 a21 a22 a23 ... H3 = a31 a32 a33 ... : : : : HN aN1 aN2 aN3 ... H = a1N a2N a3N . : aNN J1 J2 J3 : JN . J A (2) 1 A H Thus, the unknown current distribution can be calculated from the inverted coefficient matrix multiplied with the vector of required magnetic field values. Current distribution 0.6 2 Current Magnetic field 0.4 0.3 1 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Spec. magnetic field H / Ho Trace current I / Io 0.5 An example for the resulting current distribution among the traces is shown in Figure 3 (red curve). It also shows the radial positions, at which the magnetic field is specified. The figure is calculated for a planar coreless inductor. The radial position is scaled to the outer radius Rout of the coil. It is calculated for a number of magnetic field points and current traces NI = 10. The magnetic field is specified at a height above the coil of 5% of the coil diameter. To avoid instabilities in the calculation, the magnetic field is not specified until exactly the edge of the coil, but only to 90% of the outer radius. This method as described here is applicable to problems with one-dimensional symmetry. Here, it will be applied to circular inductors. In a similar manner, it can be applied to a linear (or stadium shaped) inductor, if the round ends can be neglected compared to the linear length. It’s even simpler, because the magnetic field generated by a linear track is simply calculated. To obtain the current distribution, this equation is inverted to: J coefficients can be calculated using Finite Element Method (FEM) simulations. 0 1 Radial position r / Rout Figure 3: Calculation of the current density distribution in the winding. Red: Current density. Blue: Specified magnetic field. To obtain one of the coefficients of matrix A, the magnetic field at one position must be calculated from the current in one of the traces for one arbitrary current value. The coefficients of a circular coil in air can be calculated from loops [5] [1] (see Appendix). If one side of the coil is shielded with a soft-magnetic material, the algorithm described in [6] can be used. For a general case, the However, the general idea can also be applied to problems, where there is no symmetry. This has not been derived in detail in the course of this work, but I suggest using finite sized current loops representing magnetic dipoles distributed over the area. The resulting two-dimensional current density distribution could then be converted into a non-symmetric turn distribution in a similar manner as in the following section. As a further modification of the algorithm the number of magnetic field points could be overdetermined. Then the current vector is to be optimized by a least square error fit. This may lead to a better stability of the algorithm and might avoid possible oscillations between the data points. B. Distribution of turns To match the current distribution to a turn distribution, the following algorithm is used: At first, the current per turn Iturn is calculated from the sum of all currents I0. To determine the width of the track w, the current density times a small x is summed up as well as the x to a width. If the sum of all the small current equals the required current for one turn Iturn, the necessary width of the track w is reached. Then, the same current, summed up over the width w, flows in the track of the distributed turn coil and in the traces of the equally spaced structure. As a further approximation, the resulting turns can be assumed as infinite thin with a position in the centre of the planar track. A listing of the algorithm will be presented in the final paper. Summarizing the algorithm: Distribute the currents to the tracks with variable width until one track has the right amount of current. The result is shown in Figure 4. It is calculated for the previous example with a number of turns Nw = 10. The red curve shows the current density distribution based on the results shown in the previous Figure 3, but distributed over the width of each current trace. The blue curve shows the current density distribution for the distributed turns, where each turn has a different width, but the current density adds up to a constant track current. It is clearly seen that both curves match as much as the discrete tracks allows. Current density distribution D. Resulting magnetic field The resulting magnetic field is compared in Figure 6. The gray curve shows the resulting magnetic field for the case of equally spaced current traces with variable current with NI = 10 current traces. The curve matches exactly the specified points. The blue curve corresponds to the structure with distributed turns with number of turns Nw = 10 turns. Current variation Trace position variation NI 10 4 Nw 10 3 2 1 0 0 0.2 0.4 0.6 0.8 Radial position r / Rout Figure 4: Calculation of the current density distribution in the winding. Red: Constant trace width and distance. Blue: Variable trace width with constant trace current. C. Fit function To ease the calculation of a turn distribution, a fit function for a circular coil is derived. Figure 5 shows the turn distribution as calculated in the previous section. N is the number of turns and r(i) is the trace position of the turn with index i. From the shape of this curve, an equation is guessed: (3) w Here, the ripple is higher as for the current distribution, because the track density is rather low at the inside of the coil. The magnetic field for the fit function with w = 0.2 and Nw = 10 is shown in red. The ripple is comparable to the distribution calculated by solving with the inverse magnetic problem. Resulting magnetic field 1.5 Magnetic field H / [A/m] Current density J / Jo 6 5 The parameter w is a fit parameter. It determines the “bending” of the curve”. The term in the denominator scales the curve to 1 at the outer turn. As Figure 5 shows w = 0.2 matches the optimal turn distribution for this particular case. For a different height, the parameter w varies slightly. 1 0.5 0 i 1 N 1e r( i) 1e w w 0.8 0.2 0.7 0.3 0.6 10 0.5 0.8 1 1.2 1 0.5 0 Specification Nw 50 Variable current density z H Variable turn width 0.1 Rout Approx. turn width 0.5 0.4 0.3 1 0.2 0 0.2 0.4 0.6 0.8 1 1.2 Radial position r / Rout 0.1 0 0.6 Resulting magnetic field w= 0.05 0.1 0.4 0.1 1.5 Magnetic field H / [A/m] Turn position r / Rout 0.9 0.2 zH R out Figure 6: Calculation of the resulting magnetic field for different algorithms at height above the coil of zH = 0.1.Rout and NW = 10 turns. Turn distribution 1 0 N w 10 Radial position r / Rout 1 Rout Specification Variable current density Variable turn width Approx. turn width 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Turn number i / N Figure 5: Trace positions. Dots: Results from solving the inverse magnetic field problem at a height of zH = 0.1.Rout. Lines: Fits with different fit parameter w. Figure 7: Calculation of the resulting magnetic field for different algorithms at height above the coil of zH = 0.1.Rout and NW = 50 turns. The ripple reduces, if the number of turns is increased as Figure 7 shows. It was calculated for a winding distribution with Nw = 50 turns. However, the equation system was solved for only 10 current traces because of un-stabilities at a higher number of traces. The figure clearly shows that the “original” winding distribution has a ripple which is rarely visible. The approximation function, however, has a slight deviation in the centre of the winding, but is low with less than 10%. A remaining ripple further reduces, if the finite extension of a coupled coil is taken into account. This effect becomes visible a the measurements in chapter 0. Resistance R / Rref 100 Concluding, the fit function according to equation (3) with a fit parameter of w = 0.2 gives sufficient good results for a homogeneous magnetic field. 10 100 2 f L 0 0.1 wmin 0.2 0.3 0.5 10 , where f is the operating frequency. To see, whether the increase of the resistance R or of the inductance L is more dominant, both and the quality factor Q are calculated for exemplary structures with different turn distributions, expressed by a variation of the fit parameter w. The results are scaled to reference (index “ref”) values of an inductor with the same copper thickness, but equally distributed turns. The inductance values are calculated like in the previous section and the resistance is calculated according to Ohm’s law. The results turned out to be independent of the number of turns in a wide range. To reduce discretization errors, the calculations are performed for a “high” number of turns N = 50. The results are shown in Figure 8 as the red thick curves. As expected, for low values of w, where the turns are concentrated at the outer edge, the resistance, but also the inductance increase. But overall, the resistance increase exceeds the inductance increase such that the quality factor decreases for inductors with distributed turns at the outer edge. For the typical design case of w, = 0.2, the resistance increases to about 10 times compared to the reference coil. The inductance increases only by factor of 3. Therefore, the quality factor decreases by 1/3 compared to the reference coil. 1 0.1 1 10 Fit parameter w b) 10 Quality factor Q / Q ref R 1 Fit parameter w a) (4) Q 10 1 0.1 Inductivity L / Lref E. Quality factor of printed circuit board coils If the coil with distributed turns is realized in printed circuit board (PCB) technology, the width of the tracks is usually adapted such that a maximum amount of the copper layer is used. However, since the turns of an optimal distribution are concentrated at the outer edge, these tracks are significantly thinner than the average width. Therefore, the coil with optimized turn distribution has a significant higher resistance as a reference coil with an equal distribution of the turns with the same number of turns. However, the concentration of the turns at the outer edge also increases the inductance compared to the reference coil. Important for an application is the ratio of the inductance L to its resistance R, expressed as the quality factor Q: 0 0.1 wmin 0.2 0.3 0.5 1 0 0.3 0.01 0.1 c) 0.1 wmin 0.2 0.1 0.5 1 10 Fit parameter w Figure 8: Change of a) resistance, b) inductivity and c) quality factor of a coil with distributed turns scaled to equally spaced turns (reference) as a function of the fit parameter w. Contrary, for a fit parameter > 0.5 the quality factor does no longer degrade significantly and is better than 90% of the reference coil. However, a design with those fit parameters lead to an inhomogeneous magnetic field distribution. To avoid the tracks becoming too thin, a minimum track width wmin is introduced: no track must be smaller than this value. To make the parameter independent of the particular structure, the minimum track width wmin is related to the track width of a reference structure with equal turn distributions. To achieve a turn distribution satisfying this criterion, the turns are first distributed from the outer edge to the inside as close as possible. As soon as it is possible to place a track on the optimal position (according to equation (3) )without violating the width condition, the track is place there. Thus, at the inside of the coil the turns are on the same position as in the optimal distribution. The effect of the introduction of a minimum track width on resistance, inductance and quality factor is shown in Figure 8. Especially for small values of w, where the turns are concentrated at the outer edge, a significant improvement with larger minimum track widths can be seen. Using a minimum track width of wmin = 0.2 at an optimal fit parameter of = 0.2, the quality factor improves from 30% of the reference value (red) to 60% (green) of the reference value. A minimum width of wmin = 0.5 (blue) even improves the quality factor to 90% of the reference value. Concluding, the homogeneous magnetic field must be paid with reduction of the quality factor, if the coil is manufactured in PCB technology. However, the quality factor reduction can be minimized by introducing a minimum track width. For wire wound coils, this effect is less obvious, because of the fixed wire diameter and wires can overlap. It should be mentioned that these investigations consider only the DC (direct current) resistance. The AC (alternating current) resistance is usually higher and dependent on the track width. Therefore, for higher operating frequencies the result may look different and are worth further investigations. III. EXAMPLE AND MEASUREMENTS Based on these considerations a transmitter and a receiver coil are manufactured in wire-wound technology. The purpose is to prove the calculations on the homogeneous field distribution, not on quality factor. The geometric dimensions of the coils are listed in Table 1. 2 1.5 Figure 10: Tx coil placed on Rx coils, both realized with litz wire. 1 0 0.5 0 -0.5 -1 0 Coupling measurements 0.1 w = 0.2 wmin 0.2 0.3 0.35 Nw = 10 0.3 0.5 0.2 Z = 0.1 .R out 0.4 0.6 0.8 Radial position r / Rout 1 1.2 Figure 9: Magnetic field of distributed turns with minimum track width. The influence on the magnetic field is shown in Figure 9. As can be seen in the figure, the magnetic field in the centre part of the coil is hardly affected by modifying the turn distribution. Increasing the minimum track width only leads to a less steep “edge” of the magnetic field at the outer edge of the coil. As can be seen, a minimum track width of wmin = 0.5 (blue) leads to a wide area of decay at the outer edge, which is not desired. However, at a minimum track width of wmin = 0.2 (green) the magnetic field shows hardly a difference to the magnetic field of the optimal distribution (red). Coupling Coupling factor factor kk Magnetic field H / [A/m] 2.5 0.25 0.2 0.15 z=0mm, meas. 0.1 z=5mm, meas. max. overlap 0.05 0 0 5 10 15 20 Radial displacement r / mm 25 30 Figure 11: Measured coupling factor of the litz wire coils for a radial displacement with vertical distance as parameter. The turn distributions are calculated using the fit function. Different to the exemplary simulations in the previous chapter, the larger coil has only 2x8 turns, which resulted from inductivity requirements of the intended application (not further discussed in this paper). The receiver is larger than the transmitter and selected such that it always completely covers a transmitter coil in a hexagonal array. Figure 10 shows the smaller transmitter coil on top of the receiver coil. In the final system, the arrangement would be reversed. Figure 11 shows coupling measurement results at different vertical distances z. The measurements are done with an impedance analyzer HP4194 and manual placement. The red curve is measured at a vertical distance of 5 mm (= 0.1.Rout) and can be compared to the simulation shown in Figure 6. Clearly, a good lateral homogeneity is visible. The ripple is lower despite the less turn positions in the previous chapter, because the extension of the smaller coil averages out the ripple. The coupling factor is very constant up to a displacement of 28 mm, which marks the relevant range of operation. TABLE 1: GEOMETRIC PROPERTIES OF THE LITZ-WIRE COILS: Rx coil Design Name Number of turns per layer Number of serial layers Wire type Outer coil diameter / mm Radial turn positions / mm Tx coil Design Name Number of turns per layer Number of serial layers Wire type Outer coil diameter / mm Radial turn positions / mm Hexagonal Quadratic Hex triple Hex overlap Hex triple overlap Figure 12: Coil arrangement options: Red: Receiver, Green: Activated cells, Blue: Non-active cells. TABLE 2: TRANSMITTER COIL ARRANGEMENT OPTIONS N_tpl N_lay d_out r N_tpl N_lay d_out r #1 8 2 180 x 0.03 100 23.4 #1 4 2 180 x 0.03 44 15.8 35.9 42.6 45 46.25 47.5 48.75 50 Diameter / m #of cells per area /[1/m²] # of layers per coil System quality factor Coupling factor FOM Qk Hex Hex Hex triple Receiver Quadratic Hexagonal overlap triple overlap 100E-3 41.4E-3 46.4E-3 63.4E-3 33.3E-3 50.0E-3 583 536 1149 1039 1848 2 3 3 1 3 1 144.3 152.8 103.1 123.0 106.0 0.201 0.237 0.374 0.262 0.426 29.1 36.3 38.6 32.3 45.2 Coupling factor conditions: 5mm distance, Tx 5 turns, Rx non-equal turn distance 10 turns 19.8 20.9 22 At the vertical distance of z = 0 the coupling variation is remarkably larger, because this is not the distance for which the system is designed. However, still a deviation of +/-10% of the average coupling factor is achieved, which is still good compared to conventionally designed coil combinations. IV. coupling between the transmitter cells is considered and an effective coupling of the three combined coils to the receiver coil is calculated. The Rx diameter is 10 cm, copper thickness is 70 µm, frequency is 420 kHz. SELECTION OF TRANSMITTER COIL ARRANGEMENT In this chapter, the most suitable options for transmitter arrays are analyzed and weighted. Figure 12 shows the most reasonable arrangements of transmitter coils. For each arrangement, the size of the transmitter coils is adapted such that always at least one Tx coil is covered by the receiver. The first row shows arrangements, where only one Tx cell activated at the same time. In further arrangements, not one, but three transmitter coils are activated simultaneously (proposed by [4]). For comparison, the product of the quality factor Q and the coupling factor k is calculated as a figure of merit (FoM), which should be as large as possible. It is assumed that all are made from the same copper thickness, e.g. printed circuit board (PCB) coils. As a further decision factor the required number of Tx cells per area is calculated. Inductivity and resistance are calculated according to [7] neglecting AC effects and the coupling calculations are based on [1]. For the triple cell arrangements, the mutual The calculation results are listed in Table 2. The overlapping arrangements allow larger transmitter coils with better coupling. However, the thickness can be only 1/3. Therefore, the quality factor of the coils is lower. The triple hex overlap arrangement achieves the best FOM because of the good coupling of the triple cell to the receiver. However, it requires by far the most number of cells. The least number of transmitter cells are needed for the hexagonal arrangement. It has a reasonable efficiency similar to the hex overlap structure and the hex triple structure. Concluding, the hexagonal arrangement in one layer is the preferred solution for a low cost inductive power transmission system with free positioning. V. TRANSMITTER ARRAY As shown in the previous chapter, a regular hexagonal arrangement is preferred. To give an impression of the coupling homogeneity of a larger area consisting of 19 hexagonal arranged transmitter coils, its coupling is simulated. The layout used for the simulation is shown in Figure 13. The larger receiver coil (blue) is specified to 10 cm diameter and consists of 6 turns. The smaller transmitter coils (red) are designed to 4.4 cm diameter and have 4 turns. The turns at the outer edge may be difficult to distinguish in the plot due to the printing resolution. In the simulation, always only one transmitter coil is activated. For this purpose, the first coil within a “detection radius” of 1.2 of the transmitter coil outer radius is used in the final plot. 0.1 Transmitter coils Receiver coil 0.08 y-Position / m 0.06 0.04 The resulting coupling position dependence is shown Figure 14 in a three dimensional plot. The z-axis of his plot corresponds to the induced magnetic flux in the receiver coil as a measure for the coupling. In plot a) showing the full scale, no structure of a coupling variation is visible. Therefore, the display range is limited to a range from 80% to 100% of the maximum flux value in the bottom figure b). Only by reducing the display range the slight variations of the magnetic coupling become visible. VI. 0.02 Free positioning of a wireless power receiver over a large area can be achieved with a suitable winding design, selecting a larger receiver coil which always covers one transmitter and by using a hexagonal array of transmitter coils. 0 -0.02 -0.04 -0.06 -0.08 -0.1 -0.12 -0.08 -0.04 0 0.04 x-Position / m SUMMARY ACKNOWLEDGMENT 0.08 0.12 Figure 13: Coil layout for the simulation of a larger coil array. Thanks to my colleague Reinhold Elferich, who triggered the idea to the method, and to Michael Deckers, at that time at Philips Research Aachen, for the design of related hardware. REFERENCES [1] 2.74 µVs 2.19 µVs 1.50 µVs [2] [3] 0.82 µVs [4] a) 0 µVs [5] 2.74 µVs [6] 2.63 µVs 2.49 µVs 2.35 µVs b) 2.19 µVs Figure 14: 3D visualization of the coupling homogeneity simulation. Detection radius is 27.6 mm from Tx cell centre, maximal one Tx cell is activated. a) Linear scale from 0 to maximum. b) Reduced scale from 80% to 100% of maximum! [7] Eberhard Waffenschmidt and Toine Staring, "Limitation of inductive power transfer for consumer applications", 13th European Conference on Power Electronics and Applications (EPE 2009), Barcelona, Spain, 8.-10.Sept. 2009, paper #0607. Xun Liu and S.Y.(Ron) Hui, "Optimal design of a hybrid winding structure for planar contactless battery charging platform", IAS 2006. Joaquin J. Casanova, Zhen Ning Low, Jenshan Lin and Ryan Tseng, "Transmitting coil achieving uniform magnetic field distribution for planar wireless power transfer system", Proceedings of IEEE Radio and Wireless Symposium 2009, Jan. 18-22, 2009, p.530, paper #TU4B-5. System Description Wireless Power Transfer, Volume I: Low Power, Part 1: Interface Definition. Available at http://www.wirelesspowerconsortium.com/blog/2010/08/31/qispecification-availble-for-download/ Kathleen O’Brian, “Inductively coupled radio frequency power transmission system for wireless systems and devices”, PhD Thesis, TU Dresden, 5.12.2005, Shaker Verlag Aachen 2007, ISBN 978-38322-5775-0. Eberhard Waffenschmidt, "Shielding properties of soft-magnetic layers for planar inductors", 14th International Power Electronics and Motion Control Conference - EPE-PEMC 2010, Ohrid, Republic of Macedonia on September 6-8, 2010. E. Waffenschmidt, B. Ackermann, "Size advantage of coreless transformers in the MHz range", EPE 2001 (European Power Electronic Conference), Graz, Austria, 27.- 29. Aug. 2001. APPENDIX A. Magnetic field of a coreless loop For a single circular loop centered on the z-axis, the axial magnetic field intensity Hz is (see [5], p. 34): (5) a 2 r2 z2 1 Hz( r z a I) I E( k ( a r z) ) K ( k ( a r z) ) 2 2 2 2 2 ( a r) z ( a r) z with I = current in the loop a = radius of the loop r = radial position of the point z = axial position of the point K(k) is the Elliptic Integral of the first kind: (6) 2 K( k ) 1 1 k sin d 2 0 E(k) is the Elliptic Integral of the second kind (7) 2 E( k ) 2 1 k sin d 0 And the auxiliary function k is defined as (8) k ( a r z) 2 ar 2 2 ( a r) z