Skin effect and Proximity Losses in High Speed Brushless Permanent Magnet Motors Mircea Popescu David G. Dorrell Motor Design Ltd Elsmere, UK mircea.popescu@motor-design.com University of Technology Sydney Sydney, Australia david.dorrell@uts.edu.au Abstract—This paper investigates eddy current and proximity losses in windings of different high speed brushless permanent magnet machines. Several papers have already addressed this but in this paper different topologies and winding connections are investigated to give a comparative study. This is done different operating speeds and loadings. This is important because more machnes are being developed to address the requirements of applications such as automotive electric drives. These often operate at the thermal limits of the machine at high current density during the transient torque cycling. winding will be tight so that strands will not change position between coil sides. This can generate higher losses. I. INTRODUCTION There is an increasing use of variable high-speed brushless permanent magnet machines. Traditionally it was considered that machines with air-gap windings were most liable to eddy current losses, hence the use of Litz wire in these machines, which is expensive. However, with the development of high performance rare-earth magnet machines, which have very wide speed ranges, then even windings in slots can be prone to additional ac copper losses. These machine are typically found in applications such as hybrid and electric vehicles. Parallel strands are often used in addition to Litz wire [1] but this can lead to circulating currents in some circumstances [2], and if many strands are used then this can reduce the fill factor although the cooling efficiency can be improved. In addition, when the parallel paths of a coil turns are located in different parts of the slot they will experience a proximity effect with circulating current between strands and also eddy currents in individual strands since there may be alternative flux linkage between them from the rotor magnets and electric loading leakage and mutual coupling. This should be differentiated from skin effect in windings as described in [3]. The differences are noted in [4] and [5]; Fig. 1 illustrates these in a similar manner where individual effects on strands (skin effect) and bundles of parallel strands (proximity) are considered. Even with thin wire to eliminate skin effect, proximity effects will exist. In [3], it was illustrated that the turns can be divided into bundles of parallel wires (strands-in-hand) where the series turns are stacked in the slot and the coils spans only one slot; the end Figure 1. Skin and proximity effects on coils. There are several papers that address losses in motor windings. Several use analytical calculations [6]-[12] while others use hybrid finite element analysis (FEA) and analytical methods or do comparisons of the two [13]-[19]. Some of these papers illustrated that more recently it has been possible to use time-stepped FEA to calculate the losses in individual conductors. There are now publications that only use time stepped FEA in the analysis since there is now a confidence in the modeling solution [2][3][20][21]. There is a drive to use an integrated design solution to motor design as illustrated in [22]. Increasing computational power means that individual conductors can be simulated and calculated. This can also include shell modelling of the conductor bundling to assess heat transfer [23]. The work in this paper assumes constant temperature but it can be included in an iterative simulation scheme when coupled with a thermal simulation. In this paper the variation of losses is investigated in different machine topologies with different connections and positioning of conductors in slots. This is for different loadings and speeds. This is relevant to many modern applications. The modeling use Cedrat Flux 2D [24] although similar packages could be utilized. II. MOTORS DESIGNS The analysis is carried out on two different machines. The cross sections of these are given in Fig. 2 They represent a high speed gas turbine generator and a hybrid electric vehicle drive similar to a Toyota Prius machine. conductors are the bottom of the slot and one has the conductors at the top. This machine will have substantial resin content for insulation and thermal conduction of the copper loss. The winding arrangement for the gas turbine machine in the Flux 2D model is shown in Fig. 3. Two phases are kept as simple bulk conducting regions with the resistance set to the DC value. One phase has the coil broken down into individual conductors as illustrated. This allows the ac losses to be assessed and calculated and direct comparison made between the ac and dc losses. This is a valid simulation method because the machine is current fed and it can be seen that the machine is fed by two current sources in the three phase connection. A third source is not needed. Magnet eddy current loss is not considered because this is not the focus of this paper. “Go” conductors in one slot “Return” conductors in one slot Current sources Phase B winding Phase C winding Figure 3. Flux 2D connection for gas turbine machine Figure 2. Different motor topologies The gas turbine generator design is shown in Fig. 2(a). This runs up to a speed of 150000 rpm. This is essentially a generator but to help starting it can be used in motoring mode. There are 18 conductors in each slot as illustrated. These have a diameter of 2.34 mm so that the cross section is 4.3 mm2. For this machine the conductors are series and with forced gas or liquid cooling so that the rated conductor density could be up to about 15 A/mm2 rms with a conductor current is up to 64.5 A rms. This gives a peak current of 91.2 A for sinusoidal operation. Therefore three peak current levels are considered which are within the rated values: 0, 50 and 100 A, with one transient over-current at 150 A and one severe fault current of 200A. Two winding scenarios are considered for the winding in the slots as shown. One has the The Hybrid vehicle machine is shown in Fig 2(b). This is similar to the Toyota Prius Hybrid machine although it is not the actual machine. The 2nd Generation Prius machine operated up to 6000 rpm and has been commonly used in hybrid drive motor benchmarking comparisons [25]. The later 3rd Generation machine reduced the magnet content and increased the speed to 13900 rpm [26]. This machine lies somewhere between the two. In the design here each slot contains 8 conductors. It can be connected in different configurations. As more parallel connections are made, the voltage is reduced and the current increased. The conductors are 3.76 mm in diameter. This is a high value and more likely to be split up into several strands. However, we are attempting to illustrate eddy current and proximity issues. Thick conductors can be used if there is a need to run fluid ducts around them for cooling. In these types of machines the current can run very high, and up to 25 A/mm2 rms in a transient acceleration is quite possible (with series conductors), but this is only up to the base speed. With the conductors used (11 mm2) then the peak transient rms current can be set to 250 A. Typically this will be at about 1/4th or 1/5th of the maximum speed because they have an extended torque/speed range [25]. Therefore this will be at about 3000 rpm. At maximum speed that current will be about 70 A rms since there is a constant power range from 3000 rpm to 12000 rpm. In steady-state the current will be substantially lower than 250 A even at low speed. 150000 rpm at 0, 50 and 150 A peak with the conductors in the slot bottoms. It can be seen that there is little proximity current when there is no phase current. When there is 50 and 150 A peak in the phase winding then it can be seen that there are eddy and proximity currents in the top layer of turns. The scales give the level and the top and bottom and edges are high. (a) No phase current (b) Phase current = 50 A peak Figure 4. Flux 2D connection for vehicle drive motor. III. ANALYSIS In this section the two designs are analyzed, the two pole machine in terms of the conductor positioning over a range of speeds and currents and the hybrid vehicle addresses parallel connections. A. Two pole turbine generator. This machine runs up to 150000 rpm. The current density can also be high since this is a high performance gas turbine machine that may have forced fluid cooling. The two different positionings of the discretized coil are shown in Fig. 2(a). Fig. 5 shows the current density distributions at (c) Phase current = 150 A peak Figure 5. Winding current density at 150000 rpm for different currents – conductors in slot bottom. In Fig. 6 the exercise is repeated with the conductors at the top of the slots. It can be seen that now there are proximity currents in the conductors at the top of the slot (a) No phase current (b) Phase current = 50 A peak (c) Phase current = 150 A peak Figure 6. Winding current density at 150000 rpm for different currents – conductors in slot top. The instantaneous Joule loss at 150000 rpm and phase currents of 50 and 100 A peak are shown in Fig. 7. These are compared to the dc copper loss. Obviously the dc loss will cycle between a maximum and zero twice per current cycle as illustrated (180 mech. deg. is half a current cycle). “dc loss” is a misnomer to some extent. It implies there are no eddy or proximity losses rather than carrying dc current. It can be used to compare to the “ac loss” in the phase where there are eddy and proximity losses. These show that there is continuous loss in the conductors since the eddy and proximity EMFs that drive in the currents are not necessarily in-phase with the main back-EMF, nor do they necessarily have the same frequency. Bear in mind that these are the summed proximity losses from all the individual conductors where the eddy and proximity loss currents in individual conductors will have different phase, magnitude and possibly frequency. Joule loss (instantaneous) per phase [W] even when there is no current. When there is 50 and 150 A peak phase current then inspection of the scales relative to Fig. 5 illustrates there is somewhat more eddy and proximity current and loss. These would lead to conductor hot spots particularly in the top center conductor. I = 50 Apk, N = 150 K rpm, Bottom; 50 W I = 50 Apk, N = 150 K rpm, Top; 88 W I = 100 Apk, N = 150 K rpm, Bottom; 236 W I = 100 Apk, N = 150 K rpm, Top; 343 W I = 50 Apk, DC loss; 36 W I = 100 Apk, DC loss; 145 W 500 450 400 350 300 250 200 150 100 50 0 0 45 90 135 Rotor rotation [mech deg] 180 Figure 7. Instantaneous joule, or copper, losses under DC condistiosn and at 150000 rpm for condustros in slot bottom na slot top. This is per phase. In Fig. 8 the Joule losses per phase are expressed for the different values of current over the full speed range from 1000 rpm up to 150000 rpm. What should be remembered about these losses is that they are only a small percentage of the output power since this is a high speed and high efficiency machine, but they do represent a substantial amount of loss for the size of machine (it has an outer diameter of 64 mm and a core length of 60 mm). In this case it is not the percentage loss that is relevant, rather the actual loss that the cooling system is able to cope with and the individual hot spots that occur. In the figure, it can be seen that when the turns are at the slot top there is a substantial increase in losses and these are mostly in the conductors at the top of the slot. This can create hot spots. In Fig. 9 the instantaneous Joule loss is shown for center conductors at the top and the bottom of the slot. It can be seen that in the center top location with the coil bundle located at the top of the slot, the conductor has very high losses and this would create a hot spot and a weakness. Litz wire or use of thinner wire with several strands in hand would help reduce this. However, the exercise here is to illustrate that there can be high eddy and proximity losses in a high speed machines even when the slots have narrow openings as is the case here. I = 50 A pk, Bottom I = 50 A pk, Top I = 100 A pk, Bottom I = 100 A pk, Top I = 150 A pk, Bottom I = 150 A pk, Top 1400 I = 200 A pk, Bottom I = 200 A pk, Top Pmax (IPM) = Pmax (Prius) 2 ⎛ Volrotor (IPM) ⎞ ×⎜ ⎟ ⎝ Volrotor (Prius) ⎠ Lstk =constant; Slot area increases with square of radius Joule loss per phase [W] 1200 ( Magnet weight/Vol )rotor (IPM) × ( Magnet weight/Vol )rotor (Prius) Same 1000 magnet 800 Base speed (IPM) × Base speed (Prius) 600 400 200 0 0 50000 100000 Speed [rpm] 150000 Figure 8. Variation of Joule loss at different speeds and current. Conductor 16 (Top) 100 A pk, 150000 rpm; 27 W Conductor 16 (Bottom) 100 A pk, 150000 rpm; 8 W 70 Conductor 4 (bottom) 100 A pk, 150000 rpm; 2 W and the parameters for the machines are given in Table I. This shows that the rating of the machine is about 215 KW. The approximate operating envelope is shown in Fig. 10 and the base speed is about 3000 rpm. The scaling exercise carried out earlier is very much an approximation. When the current is set to 100 A peak (70.7 A rms) the torque is about 160 Nm which gives 200 KW of power at 12000 rpm. This is for a series connected machine. When parallel paths are used the current is multiplied by the number of parallel paths. TABLE I. IPM MOTOR PARAMENTERS. 60 Parameter 50 Stator OD [mm] Core length [mm] Rotor OD [mm] Rotor volume [mm2] Magnet weight [Kg] Magnet weight per rotor vol. [Kg/mm3] Speed range [rpm] Base speed [rpm] Maximum power rating [KW] 40 30 20 10 0 0 45 90 135 180 Rotor rotation [mech deg] Figure 9. Instantaneous Joule losses for the centre conductor 4(bottom of slot) and 16 (top of slot). B. 8 Pole Hybrid Electric Vehicle Drive Motor The Hybrid electric vehicle was simulated to address the effects of parallel paths. As already discussed previously the machine has torque speed curve that has a narrow maximum torque range and an extended maximum power range. Up to the base speed these machines can operate at high current density (possibly up to 25 A/mm2 across the bulk conductor material) but only in the transient acceleration state for a few seconds and they have fluid cooling. The example machine here is similar to the Toyota Prius although not identical. The diameter is 282 mm while the axial length is 84 mm. This is a different size. In order to rate the machine then a comparison was carried out. The maximum power was estimated using Torque [Nm] or Mech Power [kW] Conductor Joule loss (instantaneous) [W] (1) IPM motor studied here 282 84 184 2.23 × 106 1.14 6.28 × 10-6 12000 3000 215 2nd Generation Toyota Prius machine 269 84 160.4 1.70 × 106 1.05 6.29 × 10-6 6000 1200 50 800 700 Power 600 Torque 500 400 300 200 100 0 0 3000 6000 9000 12000 Speed [rpm] Figure 10. IPM motor approximate operating envelope. The most relevant operating point is at high speed and simulations were carried out at 12000 rpm and 100 A peak phase current (for series connection). Fig. 11 shows the instantaneous Joule loss in the coil represented by discrete turns. The mean power loss is given in the legend. It can be seen that as parallel connections are increasingly used, the Joule loss increases. For comparison, the dc Joule loss is also give. And it can be seen that there is a substantial increase in copper loss whatever the connection is. When there are 8 parallel paths the loss increases by nearly ten times compared to the dc value. DC loss - 24 W Series discrete conductors - 134 W Disrete conductors - 2 parallel paths - 164 W Disrete conductors - 4 parallel paths - 185 W Disrete conductors - 8 parallel paths - 205 W 500 To examine the lower speed operation then Fig. 13 illustrates that the series connection loss is 40 W when the speed is reduced to 3000 rpm. This is still a AC/DC loss ratio of 1.67 compared to 5.58 at 12000 rpm. Parallel connection gives a ratio of 3.58. When the current is increased to give maximum torque at 3000 rpm the ratio is now 1.42 for series connection. These machines operate at margins of thermal stability at high torque and this could be very significant; hence, thinner conductors should be used. 450 400 Instantaneous coil loss [W] The current density distribution is examined in Fig. 12. Only the series and 8 parallel path connections are examined. The legend shows that the parallel path connection has much higher current density in the top conductor compared to the series connection due to circulating currents between the parallel legs. Hence the loss increases from 134 W to 205 W. 350 300 250 DC loss - 24 W 12000 rpm Series discrete conductors - 134 W 3000 rpm Series discrete conductors - 40 W 3000 rpm Parallel discrete conductors - 86 W 200 150 350 100 300 0 0 30 60 Rotor position [mech deg] 90 Figure 11. Instantaneous coil loss at 12000 rpm for different connections – current = N × 100 A peak where N is the number of parallel paths. dc loss included for reference. Instantaneous coil loss [W] 50 250 200 150 100 50 0 0 30 60 Rotor position [mech deg] 90 Figure 13. Instantaneous coil loss at 3000 rpm with seres and parallel connections compared to 12000 rpm amd DC losses. DC loss at 250 A rms - 299 W 3000 rpm Series discrete conductors - 250 A peak - 426 W 800 (a) Current density on load with series connection 700 Instantaneous coil loss [W] 600 (b) Current density on load with 8 parallel connections Figure 12. Current dersnity distribution at 12000 rpm with series and parallel turns and 100 A peak load. 500 400 300 200 100 0 0 30 60 Rotor position [mech deg] 90 Figure 14. Instantaneous coil loss at 3000 rpm and high torquwe (I = 250 A rms). The current density is examined in Fig. 15 at 3000 rpm. The legend illustrates that the current is more evenly distributed at the lower speed when series connected compared to the high speed operation. This is to be expected. Series discrete conductors - open circuit - 3.46 W 16 Parallel discrete conductors - open circuit - 5.45 W 14 Instantaneous coil loss [W] 12 (a) Current density at 3000 rpm – Series connection 10 8 6 4 2 0 0 30 60 90 Rotor position [mech deg] Figure 16. Instantaneous coil power loss on open circuit at 12000 rpm with series and parallel coil turns. (b) Current density at 3000 rpm – 8 parallel paths (a) Current density on open circuit with series connection (c) Current density at 3000 rpm – Series connection and high 250 A high torque current Figure 15. Current density at low speed and high torque with series and parallel connections. It is also worth examining the eddy and proximity losses in the machine when running at 12000 rpm and on open circuit. The results of a comparison between series connection and 8 parallel path connection is shown in Fig. 16. The loss is relatively low, even for parallel connection. This suggests the effects of the magnets on the parallel connections is relatively low and it is mostly eddy current and proximity effects rather than circulating currents between parallel paths that causes the increase in copper loss. This is illustrated in Fig. 17 where there is a difference but not a large difference in current distribution. The current density is still highest in the top turns. (b) Current density on open circuit with 8 parallel connections Figure 17. Current density distribution on open circuit at 12000 rpm. C. Comments Both of these machine designs exhibit high levels of eddy and proximity losses and they was a deliberate design flaw. This is because they contain thick single conductors and the aim of the paper is to illustrate these are susceptible to these effects. In reality the conductor bundles in both machines would be broken down into multiple strands-in-hand and parallel paths. The 2 pole generator illustrates that even with semi-closed slots, and copper losses that appear to be low compared to the output power, the conductors can still have hot spots that are potential failure points. The hybrid machine likewise illustrates that that the thick conductors lead to high eddy and proximity losses. It also illustrates that parallel strands can lead to circulating currents. IV. CONCLUSIONS With the developments in wide speed-range drive motors for automotive applications that are often very thermally critical the losses in the windings of such machines has to be carefully considered. Thermal analysis is now being broken down into conductor-level simulation and this will lead to greater understanding of the heat loss distribution. This will allow for possible hot-spots to be identified and better design to be carried out. REFERENCES [1] X. Tang and C. R. Sullivan, "Stranded Wire With Uninsulated Strands as a Low-Cost Alternative to Litz Wire," PESC, 2003, pp 1-7. [2] M. Popescu and D. G. Dorrell, "Proximity Losses in the Windings of High Speed Brushless Permanent Magnet AC Motors with Single Tooth Windings and Parallel Paths," IEEE Transactions on Magnetics, vol. 49, no. 7, July 2013 [3] M. Klauz and D. G. Dorrell, “Eddy Current Effects in a Switched Reluctance Motor”, IEEE Trans. on Magn., Vol. 42, Iss. 10, 2006, pp 3437 - 3439. [4] C.R. Sullivan, "Optimal choice for number of strands in a litz-wire transformer winding," IEEE Trans. on Power Electr., Vol. 14, Iss. 2, 1999, pp 283 - 291. [5] J. McCabe and M. Young, “Beating The Heat: Temperature Control Of A High-Performance Spindle," The Technical Writer’s Handbook, Mill Valley, CA: University Science, 1989. [6] H. Hämäläinen, J. Pyrhönen and J. Nerg, “AC Resistance Factor in One-Layer Form-Wound Winding Used in Rotating Electrical Machines,” IEEE Trans. on Magnetics, vol. 49, no. 6, 2913, pp 2967 – 2973. [7] R. P. Wojda and M. K. Kazimierczuk, “Analytical Optimization of Solid–Round-Wire Windings,” IEEE Trans, on Ind. Electr. vol. 60, no. 3, 2013, pp 1033 – 1041. [8] A.T. Phung, G. Meunier, O. Chadebec, X. Margueron, and J.-P. Keradec, "High-Frequency Proximity Losses Determination for Rectangular Cross-Section Conductors," IEEE Trans. on Magn., Vol. 43, No. 4, 2007, pp1213-1216 [9] X. Nan and C. R. Sullivan, "An Improved Calculation of ProximityEffect Loss in High-Frequency Windings of Round Conductors," IEEE Power Electronics specialists Conf., June 2003, pp 853-860. [10] C. R. Sullivan, "Computationally Efficient Winding Loss Calculation with Multiple Windings, Arbitrary Waveforms, and TwoDimensional or Three-Dimensional Field Geometry," IEEE Trans. on Power Electronics, Vol. 16, No. 1, 2001, pp 142-150. [11] C. R. Sullivan, " Winding Loss Calculation with Multiple Windings, Arbitrary Waveforms and Two-Dimensional Field Geometry," IEEE Ind. Appl. Soc. Annual Meeting, Oct 1999, pp 2093-2099. [12] X. Nan and C. R. Sullivan, "Simplified High-Accuracy Calculation of Eddy-Current Losses in Round-Wire Windings," IEEE Power Electronics specialists Conf., June 2004, pp873-879. [13] R. Wrobel, A. Mlot and P. H. Mellor, "Investigation of End-Winding Proximity Losses in electromagnetic devices," XIX Int. Conf. on Elect. Machines (ICEM), Rome, 2010, pp 1-6. [14] R. Wrobel, A. Mlot and P. H. Mellor, "Contribution of End-Winding Proximity Losses to Temperature Variation in Electromagnetic Devices,” IEEE Trans. on Ind. Electr. vol. 59, no. 2, 2012, pp 848 – 857. [15] R. Wrobel, P. H. Mellor and N. McNeil,, "Analysis of Proximity Losses in a Brushless Permanent Magnet Motor," XII Int. Symp. on Electromagnetic Fields in Mechatronics, Electrical and Electronic Eng., Baiona, Spain, 15-17 Sept 2005. [16] P. B. Reddy, T. M. Jahns and T. P. Bohn, "Transposition Effects on Bundle Proximity Losses in High-Speed PM Machines," IEEE Ind. Appl. Soc Annual Meeting, 2009, pp 1919-1926. [17] P. B. Reddy, Z. Q. Zhu, Seok-Hee Han and T. M. Jahns, "StrandLevel Proximity Losses in PM Machines Designed for High-Speed Operation," Int. Conf. on Elec. Machines (ICEM), 2008, pp 1-6. [18] A. S. Thomas, Z. Q. Zhu and G. W. Jewell, " Proximity Loss Study In High Speed Flux-Switching Permanent Magnet Machine," IEEE Trans. on Magn., Vol. 45, No. 10, 2009, pp 4748-4751. [19] S. Iwasaki, R. P. Deodhar, Y. Liu, A. Pride, Z. Q. Zhu and J. J. Bremner, "Influence of PWM on the Proximity Loss in PermanentMagnet Brushless AC Machines," IEEE Trans. on Ind. Appl., vol. 45, No. 4, 2009, pp 1359 - 1367. [20] P. B. Reddy, T. M. Jahns and T. P. Bohn, “Modeling and Analysis of Proximity Losses in High-Speed Surface Permanent Magnet Machines with Concentrated Windings,” IEEE Energy Conversion Congress and Exposition, ECCE, 2010, pp 996 – 1003. [21] P. B. Reddy and T. M. Jahns, “Scalability Investigation of Proximity Losses in Fractional-Slot Concentrated Winding Surface PM Machines during High-Speed Operation,” IEEE Energy Conversion Congress and Exposition, ECCE, 2011, pp 1670 – 1675. [22] M. Popescu, D. A. Staton, D. G. Dorrell, F. Marignetti and D. Hawkins, “Study of the thermal aspects in brushless permanent magnet machines performance,” IEEE Workshop on Electrical Machines Design Control and Diagnosis, WEMDCD, 2013, pp 60 – 69. [23] D. G. Dorrell, “Combined Thermal and Electromagnetic Analysis of Permanent Magnet and Induction Machines to Aid Calculation”, IEEE Trans. on Ind. Electr., vol 55, no 10, Oct 2008 pp 3566-3574. [24] Cedrat Ltd, France, http://www.cedrat.com. [25] D. G. Dorrell, A. M. Knight, M. Popescu, L. Evans and D. A. Staton, “Comparison of different motor design drives for hybrid electric vehicles,” Record of IEEE Energy Conversion Congress and Exposition, ECCE, 2010, pp. 3352-3359. [26] T. A. Burress, S. L. Campbell, C. L. Coomer, C. W. Ayers, A. A. Wereszczak, J. P. Cunningham, L. D. Marlino, L. E. Seiber and H. T. Lin, Evaluation of the 2010 Toyota Prius Hybrid Synergy Drive System, Oak Ridge National laboratory Report ORNL/TM-2010/253, USA, 2010.