A Novel Approach for EMI Design of Power Electronics

Transcription

1 A Novel Approach for EMI Design of Power Electronics Bernd Stube 1 Bernd Schroeder 1 Eckart Hoene 2 Andre Lissner 2 1 Mentor Graphics Corporation, System Design Division, Berlin, Germany {Bernd_Stube, Bernd_Schroeder}@mentor.com 2 Fraunhofer Institute for Reliability and Microintegration, System Design & Integration, Berlin, Germany {Eckart.Hoene, Andre.Lissner}@izm.fraunhofer.de Abstract The placement of passive components significantly influences the EMI behavior of power electronic systems. Particularly filter components are affected by magnetic field coupling reducing filter performance. In this paper we introduce a novel approach for a methodical EMI design of power electronic circuits. Based on the results of EMI prediction design rules for component placement are derived. To meet the design rules a prototype of a dedicated placement tool was developed. This tool has much interactive and automatic placement functionality to solve the very complex design task efficiently. Using the proposed approach in the design stage allows both a statement on achievable performance with the given components and the minimization of the system volume. Development costs can be relevantly reduced. 1. Introduction The progress in switching efficiency allows a significant reduction in cost and size of power electronic devices. Nevertheless, EMC requirements are working in the opposite direction. The smaller the components for switching, energy storage and cooling become, the higher percentage especially of device volume is needed for filtering electromagnetic interference. Filtering electromagnetic interference in lines is a task mostly solved with passive components. In practice, EMI filter design is mainly carried out by trial and error due to the complexity of the topic. For a filter performance prediction many parasitic properties of components and setups have to be considered, which are often not specified or completely unknown. Additionally, coupling effects between the passive components gather influence when placed close to each other, as mentioned in a recent research paper [1]. Some effort was made to find out more about the coupling effects and how to calculate them [2-5]. Focus of this paper is to demonstrate the influence of placement of passive components on EMC behavior, to introduce a method of predicting these effects, to derive design rules for placement and to present a prototype of a dedicated placement tool for power electronic systems. The urgent demand for a solution is demonstrated on an example from practice, a buck converter for automotive applications. Depending on the component placement the board shows severely different EMI behavior by using the same components, circuit topology and placement area (Figure 1 and 2). 8 1 Figure 1: Conducted noise of a buck converter with unfavorable component placement, measurement according to CISPR 25 The system performance is affected by component interactions, which cannot be considered in circuit simulation up to now, leading to several redesigns without a specific approach. In the considered frequency range the cause for these interactions are mainly magnetic coupling effects, nevertheless capacitive coupling gain more influence at higher frequencies. The proposed approach comprises a circuit simulation on system level, magnetic field computation of different components, sensitivity analysis to reduce complexity by /DATE8 8 EDAA

2 detecting the major magnetic couplings. The results show the potential to significantly accelerate product development. 8 1 Figure 2: Optimized placement reduces emissions up to db conducting structures are modeled using a simplified winding setup (segmented rings in Figure 11). The calculated values for inductance and mutual inductance of this simplified structure are adapted by the effective permeability for the influence of the ferrite. This procedure neglects the redirection of field lines by the ferromagnetic materials, but as the focus of this investigations are stray fields, magnetic field lines propagate most of their way through non ferromagnetic material. Therefore the redirection of field lines is limited we observe a calculation mistake for practical setups in the range of 15%, which is acceptable for EMI simulation. 2. Prediction method The evaluation of magnetic field coupling in electronic circuits requires an examination in two domains, field and circuit simulation. To calculate the 3D magnetic coupling the Partial Element Equivalent Circuit (PEEC) method [3] was chosen, which only requires the discretization of sources of electromagnetic field leading to reasonable calculation time. Results obtained in terms of equivalent circuits can be added in a circuit simulation environment and a system simulation can be performed. By using the PEEC method easy to use models can be created by simplifying the complex structure of passive components. These are combined to a full 3D model of the investigated circuit, comprising lines, passive and active components and mechanical structures like housing or heat sinks. Figure 3 gives an example for the field generating structure of a tantalum capacitor and its model in the PEEC simulating environment. Figure 3: X-ray photograph and PEEC model of SMD tantalum electrolytic capacitor A drawback of PEEC method is the missing capability to model the effect of inhomogeneous permeability. This affects especially the modeling of chokes. A suggestion for a work around overcoming this restriction was introduced in [4]. By adapting the simulation results with a factor called effective permeability the influence of ferrite cores is considered. Under certain circumstances this enables the modeling of inductive components. Figure 4 shows the magnetic flux lines of two coupling inductors with bobbin cores, in this case simulated by a finite element tool. In PEEC environment only the Figure 4: Magnetic field coupling between two inductors with bobbin cores, result of a finite element calculation The work flow for interference prediction starts with a circuit simulation of the device including the active and passive components, parasitic properties like equivalent series inductance (ESL) of capacitors or inductances of lines [5]. The function of the circuit is simulated either in time or frequency domain. In the next step a sensitivity analysis is carried out to trace those parts of the circuit which are sensitive to magnetic coupling. Therefore magnetic coupling factors between inductances are inserted and their influence on emitted interference of the whole circuit characterized. Especially components on positions with low interference levels are affected by magnetic stray fields of components with high interference levels. The sensitivity analysis generates a ranking list of the most influencing coupling factors. In the next step - the calculation of the magnetic coupling coefficients between the different parts of the circuit - only the relevant ones have to be simulated in the field simulating environment. By using this simplification the electromagnetic calculation of a whole circuit becomes feasible. By including these results in the circuit simulation a full prediction of interference can be carried out. 3. Design Rules Understanding the magnetic fields generated by the components allows to derive design rules for component placement, helping to minimize interactions and finally to

3 avoid non optimal designs. Rules comprise statements about minimum distance between components and optimum opposite orientation. The minimum distance between two capacitors avoiding interaction depends on various factors, for example on their function in the circuit, the presence of shielding planes like ground planes and of course their geometric setup. A general statement can be made on the dependency of magnetic coupling and distance between capacitors, as shown in Figure 5. The coupling factor is reduced proportional to the distance. Coupling factor Center-to-center distance of chokes [mm] coupling factor C 2 C 1 Φ22 i IN i 2 i 1 ΦM distance [mm] i OUT Figure 7: Coupling factor of two bobbin coils of different size A component with higher complexity is the current compensated choke, which is used for filtering power lines with two or three windings. The two winding design offers preferred placements for capacitors as shown in Figure 8, while the three winding design generates almost rotating stray fields and therefore no decoupled position for adjacent components can be found. Figure 5: Distance dependency of the magnetic coupling factor of two 1.5 µf x-capacitors with parallel magnetic axes As a coupling factor with an amount of.1 already severely influences the behavior of for example a π filter circuit other possibilities to decouple components must be found. This can be done by rotating one component by 9, thereby the equivalent current paths in the capacitors are in perpendicular position. Figure 6 visualizes the correlation of two capacitors, depicted by rectangles with two pins for each capacitor. Orthogonality: Minimum distance Reduced distance due to rotation Parallelism: Maximum distance Figure 6: Placement rules for arrangement of two capacitors Similar relations can be found for bobbin coils, as shown in Figure 7. The exact values for the coupling factors vary with the size of the components and have to be recalculated for every component combination. Minimum distance Maximum distance Minimum distance Maximum distance Figure 8: Placement of capacitors next to common mode chokes 4. Placement tool In order to reduce the design time of power electronic circuits significantly we developed a prototype of a dedicated placement tool (Figure 9). This tool provides much interactive and automatic design functionality to solve the complex 3D layout task. Particular intention was paid to handle the pair wise given minimum distance among the component system to meet EMC requirements. Supposed the circuit consists of n components then (n*(n- 1)/2) minimum distances can be defined for placement. To meet these constraints and to place the components in the given areas successfully, a suited rotation angle for each component has to be computed. Since this layout task is very complex the placement tool gives needed assistance. It is well known that layout problems are NP hard concerning their algorithmic complexity [6]. This means there are no exact algorithms which can solve the layout problem in polynomial time. Therefore, it is necessary to decompose the placement problems in sub-tasks and to solve them with efficient heuristic methods.

4 PEMD ij Figure 9: Graphical user interface of the placement tool For using the tool all placement relevant circuit data (e.g. 3D description of the components, net list) and given design rules are read in using an ASCII-file interface. To meet the design requirements of complex power electronics, the tool can handle the following design rules. These rules can result from the board characteristic, the modeling of the housing, manufacturing issues and the EMC requirements. Geometrical and technological constraints 1 or 2 rigid connected boards can be given for placement Different arbitrary shaped placement areas, keepins and 3D keepouts with/without z-offset Preplaced components Allowed and preferred placement areas and rotation angles for each component Clearances EMC constraints Groups of components To handle the EMC problem functional groups can be defined which have to be placed separately in coherent areas Maximum total length of electrical nets Minimal distance rules for component pairs The minimum distance rules (PEMD ij ) are result of the EMI prediction process and are defined by parallel magnetic axes, i.e. the angle between the magnetic axes is equal degree. Figure 1 shows two chokes with their magnetic fields and axes. Figure 1: Minimum distance between two chokes This minimum distance is changed by rotation of the components proportional to the cosine function. So, the really effective value of the electrical minimum distance that has to be considered during placement is computed by EMD ij = PEMD ij * cosine(alpha ij ). In the case of 9 degree between the magnetic axes the electrical minimum distance is equal and the components can be placed close to each other without any electromagnetic coupling effects. Automatic placement method We developed an automatic method to place the devices by taking all constraints into account. This method internally works in three steps: 1) Optimal rotation We compute optimal component angles to minimize the total sum of minimum distances. 2) Partitioning (optional) In the case of two boards for placement the circuit can be partitioned. The resulting partitions are assigned to board sides for placement. 3) Placement of components Computation of legal component locations fulfilling all design rules and by taking optimization criteria into account. Based on a design rule depending prioritization of the components, they are placed on board sequentially. To have the ability to compute tight 3D layouts the developed placement technique uses the continuous plane (no grid placement). Therefore, to compute free areas for placement of a specified component and to handle collision issues all placement relevant objects on board (components, keepouts) are rectilinear approximated by rectangles or cuboids. The automatic placement method was successfully applied to complex power electronic boards. The result of such a board computed by the automatic placement method is shown in Figure 9. The task for the method was to place 29 devices on a specified area by taking 1 minimum distances into account. Three functional groups were defined. The result is a legal component arrangement and was computed by the method in seconds. With this performance the method is well suited to assist interactive design work.

5 Interactive placement adviser functionality During interactive movement/rotation of a selected component the user can utilize different placement adviser functionality to meet the requirements. Online design rule checks visualize design rule violations immediately by changing the colors. By using this functionality a minimization of the system volume is possible since relevant constraints are controlled simultaneously. 5. Results The developed approach is demonstrated by examining and improving a buck converter, equipped with an input and output EMI filter, as a typical power device (Figure 11). Modeling the device with included parasitic effects of components but neglected magnetic couplings leads to severe differences between measurement and simulation results (Figure 12 and 13). Good coincidence is achieved only by including magnetic couplings (Figure 14). Two layouts presented in Figure 1 and 2 use the same topology and same components and obey all commonly known EMC design rules. Only the placement of passive components leads to a severely different EMI behavior. In case of suboptimal design the conducted emission limits are exceeded. This effect can be forecasted by modeling the passive components and connecting structures using the presented PEEC models, which makes measurements replaceable Figure 13: Simulated interference neglecting magnetic coupling Figure 14: Prediction of EMI behavior by including magnetic couplings, good correlation with measurements By using the developed interface of the placement tool we load in both the original buck converter layout (Figure 1) and the needed minimum distances. Figure 15 shows the visualization of these data in the placement tool. The user can immediately see the magnetic coupling violating the design rules (indicated by red circles) and which components are the sources of violations. Figure 11: A buck converter as test object and PEEC model of used components, traces, vias and GND Figure 12: Measured conducted noise of a buck converter, no correlation to prediction (Figure 11) due to neglected magnetic couplings Figure 15: Visualization of magnetic coupling problems of the original buck converter layout Now, a layout for the buck converter is computed by application of the automatic placement function (Figure 16). As Figure 17 shows all specified minimum distance rules are met (indicated by green circles) and therefore no magnetic coupling problems occur. Since 3 functional groups of components were specified the user can see that these groups are placed in separate coherent areas (Figure 18). The computation time for this placement result was less than 1 second. Based on this legal layout the user can

6 try to minimize the system volume using the provided interactive functionality. Since every design rule violation during interactive component movement is visualized the adherence of the constraints is ensured. Figure 16: Result of the automatic placement function Figure 18: Functional groups displayed 6. Conclusion This paper introduces methods to predict magnetic fields generated by passive components and their effect on circuit behavior using a coupled field and circuit analysis. To handle the complexity of this kind of investigation several simplifications are used, e.g. reduced itemization of components and the calculation of coupling factors with PEEC method. The proposed sensitivity analysis furthermore reduced the number of necessary field simulation results. The derived minimum distance rules between components are input for the developed placement tool. This tool assists the design of power electronic systems efficiently. Based on both the high performance of the automatic placement method and the availability of much dedicated interactive placement and advisory functionality this prototype can be used for a specific minimization of the system volume. The feasibility of the methods is shown on a DC/DC converter for automotive application. References Figure 17: Visualization that all distance rules are met [1] S. Wang, F.C. Lee, Dan. Y. Chen, W. G. Odendaal, Effects of Parasitic Parameters on EMI Filter Performance, IEEE Transactions on Power Electronics, Vol. 19, No. 3, May 4 [2] S. Wang, R. Chen, F.C. Lee, J. D. van Wyk, Improved passive filter configurations for high-frequency conducted EMI in power electronics, EPE 5, Dresden, Germany [3] A.E. Ruehli, Equivalent Circuit Models for Three- Dimensional Multiconductor Systems, IEEE Transaction on Microwave Theory and Techniques, 1974 [4] E. Hoene, A. Lissner, S. Weber, S. Guttowski, W. John, H. Reichl, Simulating Electromagnetic Interactions in High Power Density Converters, PESC 5, Recife, Brazil [5] A. Lissner, E. Hoene, S. Guttowski, H. Reichl, Predicting the Influence of Placement of Passive Components on EMI Behaviour, European Power Electronic Conference, 7, Aalborg, Denmark [6] T.Lengauer, Combinatorial algorithms for integrated circuit layout, John Wiley, 199