Optimized design of high power density EMI filters for power electronic converters

Nowadays, power density of power converters and related EMI filters is gaining more and more attention. This severely impacts on the design constraints in several application domains. A conventional design of the EMI filter does not guarantee the selection of components/configuration leading to the best power density. For this reason, an optimized design procedure of discrete EMI filters in terms of power density is proposed in this paper. It is based on a previously developed rule-based design procedure, introducing here additional features to obtain a more effective optimization of EMI filter's power density. The proposed approach has been validated in terms of EMI filter's performance and size reduction, by a comparison between an EMI filter designed using the optimized procedure and a conventionally designed one, for two different case studies.


I. INTRODUCTION
Power electronic converters are used in a broad variety of industrial applications. In many of these (e.g., aeronautic, automotive, etc.) a very compact system is strongly required. Therefore, power density is becoming an increasingly stringent design constraint.
The use of power devices with high speed commutation allows to reduce the size of the energy storage passive elements (i.e., inductors and capacitors). On the other hand, high-speed switching operation of power electronic converters usually causes relevant electromagnetic interference (EMI), potentially affecting other devices placed in their proximity. For this reason, EMI is one of the major issues in power electronic systems design, particularly when dealing with stringent standard limits [1]- [3].
The size of the EMI filter can contribute up to 30% of the total size and weight of the converter system. Therefore, a filter design matching the maximum power density is strongly desired, especially for the applications in which compactness and low weight are the primary constraints [4]- [7]. It is worth noting that, once the filter topology has been chosen and the values of its components (C ommon ModelD ifferential Mode inductors/capacitors) have been defmed, there is a huge amount of possibilities for practical configurations. Moreover, the identification of the configuration leading to the optimal power density is a challenging task. The conventional design of EMI filters requires a significant computational effort and it does not guarantee the optimal choice of filter components in order to obtain the maximum power density. Therefore, in order to reduce the effort and the time needed for properly selecting the filter components, an automatic rule based EMI filter design procedure that allows obtaining optimal power density has been previously proposed [8]. In this paper, the algorithm underlying the filter design procedure of [8] has been suitably improved and refined to obtain a more effective optimization of EMI filter's power density; on the other hand, easy implementation features and low computational demand have been maintained. Starting from the EMI noise to be attenuated, the proposed procedure allows a fast selection of EMI filter components, circuit configuration and number of stages, leading to the minimum overall filter size. The effectiveness of the proposed design approach has been validated using a suitably devised experimental setup and considering two different case studies.

II. GENERAL STEPS OF EMI FILTER DESIGN
The generation and coupling mechanisms, as well as the Common Mode (CM) and Differential Mode (DM) EMI paths, are different; thus, separated filter sections are needed in order to obtain a suitable attenuation for EMC compliance. Therefore, the initial step of an effective EM! filter design is the separate evaluation of the two noise modes.
Firstly, it is necessary to identify the crucial point on the experimental curve of the EM! emission, generally the emission peak at the lowest frequency within the EMI range (150 kHz -30 MHz).
The required attenuations Attreq_CM, Attreq_DM for the CM and DM noise are expressed as: (1) (2) where Ah_CM' Ah_DM are the amplitudes of the harmonic to be attenuated, Limit is the maximum amplitude allowed by the reference standard at the frequency of interest and SM denotes an additional safety margin, usually set to 6 dBfl V.
The cutoff frequency of the CM or DM filter is then expressed by (3), where !aU_h is the harmonic frequency to be attenuated, Attfllter is the filter inherent attenuation, which is related to the filter circuit configuration, and n is the number of filter stages.

fa CMIDM
(3) When both comer frequencies are known, the inductance and capacitance values of the CMlDM sections of the EMI filter can be determined. The CM capacitance is constrained by standards to comply with safety issues (ground current), whereas DM parameters can be selected exploiting some degrees of freedom, considering that the value of the inductance LDM will be reduced if the value of the capacitance COM is increased and vice-versa. For this reason, the DM section design can consider either the case of a LOM obtained by the leakage inductance of the CM choke (Lleakage=0.2-.;-2% LcM, depending on the core material) or the case of a LOM realized by means of additional inductors [7], [8 ]. The equations that are used to determine the inductors and capacitors values are (4) and (5), respectively.
Aft er determining the values of the CM and DM filter components, it is critical to choose the inductors' cores with suitable geometrical dimensions and magnetic properties so to prevent magnetic saturation, considering the maximum values of the CM/DM noise currents measured in time domain [9] [20].
The conventional design of EM! filters requires a significant computational effort and it does not guarantee the optimal choice of filter components in order to obtain the maximum power density. For this reason, an optimized design procedure of discrete EMI filters in terms of power density is described in the following section.

III. EMI FILTER OPTIMIZED DESIGN PROCEDURE
The optimized design procedure presented in this paper is based on a rule-based algorithm previously developed by the same authors in [8] and suitably improved to obtain a more effective design. The design procedure starts from the basic steps illustrated in Section II and introduces the additional objective of pursuing the best power density for the EMI filter.
The optimized design procedure takes into account the main characteristics of the filter application: the power electronic circuits under study, the filter design constraints and some databases with parameters extracted by datasheets of commercial components for the realization of EM! filters. It repeats the steps of the conventional design procedure for different configurations (e.g., varying core material and model, number of stages, etc.) and chooses the configuration exhibiting the best power density. Since multi-stage filter can occupy a smaller volume than single stage one, depending on the components which are used, the optimized design procedure considers the possibility to span a number of filter stages ranging from 1 to 5.
To avoid the degrading of the desired EMI filter performance, the algorithm verifies the fulfillment of the condition Bmax<Bsat for the considered magnetic materials (nano-crystalline or ferrite). Saturation of the magnetic core is verified in both the CM and DM section design procedures. Finally, the algorithm calculates the EMI filter volume and weight of all possible configurations and select the one with the best power density. More details about the databases and the rule-based algorithm have been described in [8]. With respect to [8] additional features have been added to obtain a more effective design. In particular, besides the core database, the database of commercial Cx and Cy capacitors has been extended and a new database of conducting wires has been included. In this way, the volume/weight contribution given by the inductor wires (non-negligible when dealing with rated power of hundreds of watts and beyond) is included in the EMI filter calculations. Furthermore, the CM capacitor can now be chosen either on the basis of SAE AES 1831 standard requirements [22] or by explicitly specifying the maximum ground leakage current.
The cores' dimensions have been chosen so as to design EMI filters for applications able to manage powers up to some kW.

IV. THE CASE STUDIES
The improved optimized EM! filter design procedure has been validated on two suitable experimental rigs. Moreover, the optimized EMI filter performance has been verified against both military and civilian technical standards.
The first case study (case study 1) consists on a PWM IGBT Voltage Source Inverter (VSI) that supplies a three phase induction motor with the following characteristics: rated voltage UN = 48 V, rated power PN = 220 W, maximum current Imax = 5 A. The VSI is based on a STGIPSI0K60A power module and on an Altera Cyclone III FPGA board implementing the PWM modulator.
The VSI switching frequency is equal to 20 kHz and the output voltage is 48 V. This is a typical case of induction motor drive supplied by a DC power grid; it is very common, for example, either in vehicle applications (road vehicles, marine vehicles, aircrafis), as well as in some residential/commercial smart buildings [25], [26].
In the second case study (case study 2), the device under study (D UT) is a DC motor drive with rated voltage of 12 V. It is composed of a voltage regulator based on a DCIDC boost converter and a DC motor with a rated power of 30 W.
The boost converter is based on the following devices: MURB820: Ultrafast Rectifier, IRFPI50N: Power MOSFET, Inductor 320 f.,lH, output capacitance 220 f.,lF, switching frequency 20 kHz. This is a typical application for automotive environment in which the presence of low-power loads supplied with different voltage levels requires the use of DC/DC converters [27]. A dual DC Line Impedance Stabilization Network (USN) with a voltage capability up to 600 V has been set-up and used to measure the conducted EM! [21]. The experimental rigs have been equipped with a R&S FSH4 (100 kHz -3.6 GHz) spectrum analyzer and a RF current probe R&S EZ-17 for measurements in the frequency range 20 Hz -100 MHz with a maximum DC current of 300 A.

V. EXPERIMENTAL V ALlDATlON AND DISCUSSION
Measured conducted disturbances have been compared with the limits reported in CISPR 25 [23] and MIL-STD-461G [24] standards to verify the compliance with both civilian and military standards. As for CISPR 25 standard, it deals with limits and methods of measurement of radio disturbance for the protection of receivers used on vehicles. Its limits are referred to five different classes, in rising order of required reduction of the maximum electromagnetic disturbance level that the devices can produce on board. Two CISPR 25 limit curves are reported in Figs. 3 and 4: the Class 5 limit, which is the most stringent, and the Class 4 limit.
In particular, the emission of case study 1 has been compared with the limits of both standards, i.e., CISPR 25 and MIL-STD-461G; on the other hand, only the CISPR 25 has been considered for case study 2, since it is the only one that can be applied for the given DOT.
As shown in Figs. 3 and 4, the emission profiles in both the considered case studies exceed the limits of the standard that has been chosen as a reference. This calls for a suitable input EMI filtering.
On the basis of the conventional and optimized design procedures described in sections II and III, EM! filters have been set up and the following required attenuations have been used for the design procedures:  Tables I and II have been used to run the optimized design algorithm. It is worth noting that SAE AS 1831 standard has been used for C y selection in case study 1; on the contrary, the maximum ground leakage curr ent has been considered in case study 2.
For both case studies, the conventional design procedure led to a single stage configuration, whereas the optimized procedure selected a double stage configuration. A comparison of the optimized filters with the conventionally designed ones has been carried out verifying their size and performance. Tables III and IV summarize the results. Evaluating the volume and weight of the two filters, it is possible to observe that the optimized design leads to a reduction in volume and weight in both case studies. In particular, a reduction of 44% in volume and of 53% in weight is obtained in case study 1. On the other hand, a reduction of 38% in volume and of 41 % in weight is obtained in case study 2.   In order to visually perceive the difference between the two approaches, a picture of the optimized EMI filter compared to the conventionally designed one in the case study 1 is shown in Figure 3: the higher compactness of the optimized filter is evident. Fig. 3. Picture of conventionally designed EMl filter (on the right) and optimized EMI filter (on the left), in case study 1.
Morever, the EM! filters, designed according to the data of Tables III and IV, have been experimentally realized.
In order to evaluate the EMI filters mitigation performance, EMI measurements have been carried out in case study 1 and 2 without any filter, with the conventionally designed filter and with the optimized filter.
It is possible to observe in Figs. 4 and 5 that the optimized and the conventionally designed filters exhibit a satisfactory behavior. In particular in both case studies, with regard to CISPR 25, the filters allow to obtain a fully compliant behavior with the Class 4 limit and an acceptable behavior for the Class 5 limit. Therefore, despite the higher compactness and power density achieved, the optimized EM! filter still allows to obtain the compliance of the power electronic systems under study with the reference standards.

CONCLUSIONS
This work deals with the power density issue in EMI filters used for mitigating EMI in power electronic systems. In detail, this paper proposes an improved rule-based optimized design procedure for the optimal and fast selection of discrete EMI filter components and configuration, aiming at obtaining the best power density. In order to validate the method, two case studies have been considered, devising suitable experimental test rigs. In particular, an inverter-fed induction motor drive and a DC motor drive supplied by a DC/DC boost converter have been used, respectively. The size and the perfonnance of the EMI filters designed by the proposed procedure have been compared with those obtained by conventionally designed ones. The comparison, based on experimental tests, demonstrates the effectiveness of the proposed approach. In particular, a volume reduction of 44% and a weight reduction of 53% is obtained in the first case study and a reduction of 38% in volume and of 41% in weight is obtained in the second case study.