Electromagnetic Modelling Process to Improve Cabling of Power Electronic Structures J. Aimé (1, 2), E. Clavel (1), J. Roudet (1), G. Meunier (1), P. Loizelet (2) (1) G2Elab, Electrical Engineering laboratory of Grenoble BP 46 Domaine Universitaire 38402 Saint Martin d Hères, France, jeremie.aime@g2elab.grenoble-inp.fr (2) STIE, Schneider Toshiba Inverter Europe, Rue André Blanchet 27120 Pacy-sur-Eure, France Abstract Uncontrolled high costs which can be generated by the compliance with EMC standards become a real problem for industrials. Anticipation and better control of EMC performances associated with low cost solutions to reduce the disturbances are required. That is why the modelling of electromagnetic behaviour of power electronics structures is investigated. Several modelling methods are needed to establish accurate EMC models of complex industrial products. This paper deals with an electromagnetic modelling process which is applied to the computation of a commercialized variable speed drive. Moreover, routing rules are deduced and cabling improvements are developed and experimentally verified. Introduction Due to the frequency and/or power rising, EMC constraints become more and more important in industry. In consequence, modelling tools are required to predict EMC performances of industrial power electronic structures. Nevertheless, modelling must take into account environment heterogeneity (magnetic components, dielectric), complexity of multilayer cabling geometries and near and far field computations. An accurate electromagnetic modelling process is presented in this paper. Moreover, the process can be used to model EMC behaviour of power electronic devices but also to improve them. The modelling process is applied to an industrial variable speed drive. The EMC behaviour of the initial product can be computed and using the same process, the cabling geometry can be optimized. The following part is a presentation of the modelling process. Then, the case study is described. Finally, results illustrate the effectiveness of the approach. Electromagnetic modelling process Introduction An electromagnetic modelling process has been developed (fig. 1). Mechanical part and cabling Back to conception Equivalent impedance computation Field computation Electrical simulator Equivalent impedance Excitation sources Fig. 1. Electromagnetic modelling process
Prediction of optimized cabling for improvement of EMC performances can be performed on power electronic structures by computing the emitted field. EMC perturbations of a static converter are generated by the differential and common mode currents as in [1]. Both currents must be taken into account in order to compute near and far fields in a large frequency domain and in the three spatial dimensions. The equivalent impedance of the converter layout is determined by considering the geometry like an equivalent electrical circuit using localized elements. The differential mode is associated with resistive, inductive and mutual terms [2]. A capacitive model is added in order to model the common mode [3]. The modeling methods used to compute the model are presented in the next part. Modeling methods Due to meshing considerations, multilayer topologies are difficult to describe using classical variational methods. Integral methods are more adapted to such problems [4]. An equivalent electrical circuit is extracted from geometries (fig. 2). Zeq = ([ R][ L][ C] ) Fig. 2. Equivalent impedance of the static converter (layout and mechanical part) In other words, an equivalent impedance of the mechanical part and cabling is computed. The impedance depends on frequency with resistive, inductive (with mutual), and capacitive terms. PEEC formulations implemented in InCa3D, an inductance calculation dedicated tool, are used to determinate resistive, inductive and mutual parts [5-6]. The capacitive model is determined using an Adaptative Multi-Level Fast Multipole Method (AMLFMM). This algorithm accelerates the computation of interaction coefficients and is low-memory consuming by using truncated multipole decomposition of interactions [7-9]. Excitation sources The complete power electronic structure is then simulated by taking into account the geometry which can be complex and not negligible [10]. The equivalent impedance is imported in an electrical simulator like PSpice or Saber. A Fast Fourier Transform of the time domain simulated signals is done in order to extract the excitation sources (differential and common modes) (fig. 3). Static converter model Elements of the power electronic structure: semi-conductors, passive components + Geometry Z eq =([R] [L] [C]) Time domain simulation FFT Excitation sources (differential and common modes) Field computation Geometry Z eq =([R] [L] [C]) Fig. 2. Electromagnetic modelling process
By this way, the non linearity induced by the switching is modeled and the geometry, through the equivalent impedance, is taken into account. The amplitude and phase of the sources for each frequency are extracted from frequency domain signals. Then, the sources are connected to the geometry model in order to achieve a field computation frequency by frequency because the equivalent scheme is frequency dependent. The emitted field is deduced from the computation of the differential [2] and common mode currents. The field computation is done by using Biot and Savart law (1). r B P µ 0 4π r j MP ( ) = * dv 3 Point M is an element of the volume where the current density is known, here the tracks, and P is the field point. Due to standards, it is also necessary to compute the electric field. This entity is deduced from the magnetic field using the wave impedance given by (2). ( ) Z d,f M V MP ( + jkd) ( dk) 2 (1) jωµ 0d 1 = (2) 1+ jkd Where d is the distance between the source and the field point, k is the propagation constant defined in the air by (3). ω k = (3) c By this way, optimization and virtual prototyping of complex industrial structures can be realized. Case study Using the electromagnetic modelling process, a virtual prototyping is carried on a variable speed drive commercialized by Schneider Toshiba Inverter Europe (S.T.I.E.) [11]. This power electronic device is complex. Due to thermal constraints, the layout is composed of four layers. A three phase's common mode filter is connected to a power module including a rectifier and an inverter (fig. 4). C3 C2 C1 Common mode filter C5 C6 C4 SW100 Earth C8 C7 C9 Rectifier Power module PA L opt PC Inverter U V W 1 2 3 4 W2 W1 W3 PA PC Power module U V W Electrical scheme Fig. 4. Variable speed drive structure Geometry The plus of the DC bus is named PA. A cable connects PA to the DC bus capacitors. An inductance Lopt can be added in order to filter parasitic currents. PC is the minus of the DC bus. The floating potentials (U, V and W) are common mode sources. The equivalent impedance is computed thanks to two modelling environments (fig. 5).
PEEC environment (InCa3D) Capacitive meshes (Flux3D) Fig. 5. Variable speed drive models As previously mentioned, the inductive and resistive parts are computed by using PEEC method and the capacitive meshes are done by using the Finite Element environment with Flux3D [6]. The meshes are used to compute the capacitances by using the AMLFMM. The equivalent circuit has been imported in Saber (Fig. 6). (R, L, M) model Capacitive model Power electronic devices Fig. 6. Motor drive model in Saber The differential mode sources are deduced from the simulation of the currents of line, and of the DC bus. The common mode sources are extracted from the values of the floating potentials U, V, and W. In a first time, the magnetic field is computed (fig. 7). Measured Modeled Fig. 7. Radiated magnetic field at 32 khz
The measured and modeled magnetic fields are in good agreements. The equivalent circuit of such application is complex. In consequence, we have verified that the model is in agreements with the impedance of the variable speed drive. To do that, the resonances in the conducted frequency range have been measured and modeled (Fig. 8 and Table 1). F0=1.314MHz F1=5.166MHz F2=19.16MHz F3=25.83MHz Fig. 8. Measured resonance in the conducted frequency range TABLE I RESONANCES Frequencies (in MHz) F0 F1 F2 F3 Measured 1.314 5.166 19.16 25.83 Simulated 1.221 5.198 none 24.76 The model has been validated by a harmonic response. The simulated and measured resonances are closed. The resonance F2 which is linked to the cable connecting the converter to the motor does not appear because the cable is not modeled in the simulation. The modelling process is validated on such industrial power electronic structure. It is interesting to note that this is a first step in the development of a simulation platform able to predict the EMC behaviour of industrial power electronic devices by taking into account the real complexity of such structures. Moreover, it is interesting to use the models in order to improve the initial product by proposing optimized geometries. Cabling optimization Power part The cabling is modified with the aim of reducing the radiated perturbations. The connecting technology is modified between the common mode filter and the rectifier. The cables W1, W2 and W3 are removed. The connection is performed by using copper tracks. The modelling process is used to validate the new cabling geometry. For example, considering the filtering effectiveness, it is essential to avoid the couplings between the tracks, and connecting the entry to the inductance with the "new" tracks ', ' and ' connecting the inductance with the rectifier (fig. 9).
Floating potentials U PA Cp V Cable PC W Cm C1 C2 C3 Shielding plane The shielding plane adding Fig. 9. Prototype 1 2 3 4 PA 5 6 PC Geometry U V W Vias Plan Shielding écran plane Moreover, a shielding plane is added between the floating potentials U, V and W to recycle the common mode currents inside the structure [12-13]. Due to the fact that the cable, because of the common mode excitation, is the most important radiating source, the aim is to create a preferential path inside the converter for the common mode currents recycling. Capacitors Cp and Cm are added in order to fix the potential of the shielding plane. Their values are determined thanks to the modelling process too. Moreover, the shielding plane is designed by taking into account several constraints: the couplings between the common mode filter and the power module which must be imperatively avoided; mechanical constraint with the placement of the connectors, and the placement of the electronic components. The magnetic field Hz is measured above the variable speed drive at 32 khz (fig. 10). Measured near magnetic field (dbµa/m) Initial product Prototype Fig. 10. Experimental validation of the emitted field reduction The magnetic field density is reduced significantly. More measurements have been carried on several frequencies and spatial components. The tendency is always the same, with a reduction of the radiated perturbations. The improvements carried on the power part of the industrial variable speed drive show good results. The modelling process is useful for designing and optimizing the cabling. Electronic part Most of the perturbations are issued from the power part but the electronic one can also generate problems for compliance with standards. The quartz used in the variable speed drive is cadenced at 40MHz. High field levels appearing at 120MHz and 160MHz are a real problem for homologation. Using routing rules deduced from modelling, cabling modifications are carried on this part too (fig. 11).
Shielding plane 2 SAFE T1 P P7F T4 Pass T6 Layer 4 Shielding plane 1 Layer 2 Layer 3 Quartz Shielding plane 2 Quartz area Shielding of the tracks connected to the quartz New layout implantation Fig. 11. Modifications around the quartz The tracks connected to the quartz are placed between two shielding planes. Moreover, the conductors routed around the "quartz area" (defined as the placement of the quartz with the associated tracks), are rerouted. By this way, couplings are reduced and limited. Results show an important reduction of the far field emission levels (fig. 12). Fig. 12. Reduction of the quartz influence The emission levels are less important for all the incriminated harmonics. Our cabling modifications are validated by measurements on the electronic part. Conclusion An electromagnetic modelling process is presented. By mixing PEEC and FMM methods, all the parasitic elements of modelled power electronic devices are taken into account. The process has been applied to an industrial variable speed drive. Its EMC behaviour has been modelled and simulations have shown good agreements with measurements. Moreover, the process has been used to optimize the cabling. Significant improvements have been carried on and validated thanks to measurements. Anticipation and optimization, key words for the cost reduction of EMC for industrials, are possible with the presented tools. This approach is a first step to the development of an industrialized simulation platform able to characterize accurately EMC behaviour of industrial power electronic systems. This will be carried on in future works.
Acknowledgement We would like to thank Ouafae Aouine and Cécile Labarre for the experimental validation of the near magnetic field reduction carried on at the ENSM of Douai. References [1] C.R. Paul, A Comparison of the Contributions of Common Mode and Differential Mode Currents in Radiated Emissions, IEEE Transactions on Electromagnetic Compatibility, Vol. 31, Issue 2, pp. 189-193, May 1989. [2] J. Aimé, J. Roudet, E. Clavel, O. Aouine, C. Labarre, F. Costa, J. Ecrabey, Prediction and measurement of the magnetic near field of a static converter, IEEE International Symposium on Industrial Electronics, ISIE 2007, pp. 2550-2555, June 2007. [3] V. Ardon, J. Aimé, O. Chadebec, E. Clavel, E. Vialardi, MoM and PEEC Method to Reach a Complete Equivalent Circuit of a Static Converter, The 20 th International Symposium on Electromagnetic Compatibility, 2009. [4] J. Aimé, T-S. Tran, E. Clavel, G. Meunier, Y. Le Floch, Ph. Baudesson, Magnetic Field Computation of a Common Mode Filter using Finite Element, PEEC methods and their coupling, IEEE International Symposium on Industrial Electronics, IEEE-ISIE08, 30 June 2 July 2008, Cambridge UK. [5] A. E Ruehli, Inductance calculations in a complex integrated circuit environment, IBM Journal on R&D, September 1972. [6] http://www.cedrat.com [7] J. Carrier, L. Greengard, and V. Rokhlin, A Fast Adaptive Multipole Algorithm for Particle Simulations, SIAM J. Sci. Statist. Comput., 1988. [8] K. Nabors, J. Withe, Multipole-Accelerated Capacitance for 3-D Structures with Multiple Dielectrics, IEEE Transactions on Circuits and Systems, vol. 39, N 11, Nov. 1992. [9] A. Buchau, W. M. Rucker, Capacitance Computation of Thin Conductors with the Fast Multipole Method, International Journal of Applied Electromagnetics and Mechanics, vol.17, 75 89, 2003. [10] J. Aimé, E. Clavel, J. Roudet, P. Baudesson, Determination of the Layout Influence on the Effectiveness of a Three-Phase Common Mode Filter by using Equivalent Circuits and PSpice, IEEE International Symposium on Industrial Electronics, ISIE 2008, pp. 1-6, July 2008. [11] http://www.schneider-electric.com/sites/corporate/en/home.page [12] N. Mutoh, J. Nakashima, M. Kanesaki, Multilayer Power Printed Structures Suitable for Controlling EMI Noises Generated in Power Converters, IEEE Transactions on Industrial Electronics, Vol. 50, N. 6, pp. 1085-1094, December 2003. [13] J. Aimé, E. Clavel, J. Roudet, P. Baudesson, Determination of the Layout Influence on the Effectiveness of a Three-Phase Common Mode Filter by using Equivalent Circuits and PSpice, IEEE International Symposium on Industrial Electronics, ISIE 2008, pp. 1-6, July 2008.