Accurate modeling of layout parasitic to forecast EMI emitted from a DC-DC converter

Piotr Musznicki 1 Jean-Luc SCHANEN 2

senior member IEEE,

Bruno Allard 3 member IEEE

Piotr J Chrzan 1

1 Politechnika Gdanska, Wydzial Elelktrotechnki i Automatyki KMiE ul. SOBIESKIEGO 7; 80-216 GDASK pchrzan@ely.pg.gda.pl 2 Laboratoire d'Electrotechnique de Grenoble, CNRS UMR 5529 INPG/UJF, ENSIEG BP.46, F-38402 SMH cedex GRENOBLE jean-luc.schanen@leg.ensieg.inpg.fr 3 Centre de Genie Electrique de Lyon CNRS UMR 5005 CEGELY INSA, 2 av Jean Capelle, 69621 Villeurbanne cedex Bruno.allard@cegely.insa-lyon.fr

Abstract This paper illustrates how to account for all parasitic due to the layout of a power converter (inductive and capacitive), in order to forecast Electromagnetic Interferences (EMI). The method is generic, and is validated here in the simple example of a DC-DC converter, realized on different technologies: Insulated Metal Substrate (IMS), Printed Circuit Board (PCB). In addition, several layouts aspects will be investigated. Conclusions are given on the influence of layout and all other components on EMI.

I. INTRODUCTION The significant advances in power electronics have lead

to improved performance DC-DC converters. One of the major challenges is to predict, with satisfying accuracy the levels of the electromagnetic interference (EMI). To achieve this task, precise models of all components of the converter must be used. Indeed, the magnitude of emitted noise runs on the electrical performance which depends on geometrical structures of the system. So models for realistic electromagnetic compatibility (EMC) including non ideal electrical interconnections (parasitic capacitance and inductance) passive components and semiconductors devices are needed in design process. This paper will show temporal simulations using the most possible precise models for each component, including connections. These models are described in section II, and simulation is validated in comparison with measurement (section III)

However, a global EMC simulation is not sufficient to understand all phenomena which originate EMI. Therefore, it is not easy to modify the converter if the EMC behaviour is not satisfying. Hence, a sensitivity study is carried out in section IV, in order to determine the influence of each element of the converter in the EMI spectrum

A 100W 14 V- 42 V boost converter, realized on IMS (Insulated Metal Substrate) has been chosen to illustrate the method. Switching frequency of MOS transistor is f = 100 kHz with duty cycle 0.75. It has been designed for typical automotive applications where this two level of voltage can be found. The technology (IMS) is also compliant with automotive constraints (easy cooling substrate). This well known power electronics application has been chosen because it is relatively simple, what allows to understand and simulate EMC. Fig. 1 shows the converter and its electrical circuit.

Fig. 1. Fundamental schematic diagram of considered boost converter (with LISN). Top: picture of the converter (without the LISN)

II. EMC MODELS To get accurate EMC behavior of a power converter

necessitates precise determination of switching waveforms. Semiconductors devices such as diode and MOS can be identified as disturbances sources and must be perfectly modeled. However, even if a model can be provided, it is also necessary to provide it with parameters. This is not the simplest task. A generic parameter extraction method from CEGELY has been used for this purpose [1].

For passive components (inductor and capacitors), impedance measurement bridge has been used to get an electrical equivalent circuit.

All interconnection parasitic (inductance, capacitance) cannot be easily measured. Indeed, low values are encountered, and it is furthermore hard to split an inductive behavior into different contributions. Therefore, PEEC modeling method [2] has been used to account for non ideal interconnects.

Inductor Diode

Output capacitors

Input capacitors (one electrolytic and

two paralleled ceramic)

MOSFET

Complete electrical equivalent circuit can thus be obtained, including the converter itself with all complex models of components and parasitic, but also the measurement equipment (Line Impedance Stabilization Network LISN) and the cabling impedance between LISN and converter.

A. Semiconductor parameters determination: The semiconductors used in this converter (MOSFET and

shottky diode) are unipolar components and therefore are not the most difficult power components to be modeled: bipolar effects are the most tricky (PIN diode [3], IGBT [4]).

Saber conventional models can easily be used for EMI prediction. However, the model parameters must be determined with sufficient accuracy.

This is not a so simple task, since manufacturer data are not always sufficient. A specific parameter extraction method has been used for this purpose [1]. It is not the aim of this paper to describe the method in details, thus it is just briefly reminded here.

MOSFET Saber model needs ten parameters, and diode four. Among these parameters, some have an influence on static behavior, and can be deduced from a curve tracer. Care must be taken in order to avoid self heating of the component during this static measurement: the current pulse must be short. Conventional tracers are often not adapted, and specific experimental setup has been built.

The method is based on an optimization algorithm, which minimize the error between measured static characteristic and simulated one, by modifying MOSFET or Diode parameters.

Once "static" parameters have been determined, a dynamic test (in a chopper cell) is used with the same optimization approach, in order to find the other parameters, which acts on dynamic behavior only (as capacitance).

This methods is especially interesting since it is general and can apply to any power component. For the power MOSFET and the Shottky diode used in the studied boost, parameters are listed hereafter.

Fig. 2. Principle of optimization method to determine semiconductor

parameters. 1st step: static measurement, 2nd step, dynamic measurement.

1) For MOSFET VTO - Zero-bias threshold voltage 3.76v[V], IS - Bulk p-n saturation current 1[pA], KP Transconductance 83.2, RD - Drain ohmic resistance 1[m], RS - Source ohmic resistance 1[m], RG - Gate ohmic resistance 10 [], THETA - Mobility modulation 2.025[1/V], Crss = 1.14[nF] , Coss = [0.1n] Ciss = 14.8[nF] Cgs = 14.8[nF] and Cds = 2.09[nF] - the inter-electrode capacitances.

2) For diode Is - saturation current injection 1[nA], rs - series resistance 7.4[m], Vj - voltage injection 0.3[V], Cj - junction capacitance 1.47[nF]

B. Passive components models As far as conducted EMI are considered, a typical ESL-

ESR representation is sufficient for capacitors [5]. The ceramic decoupling capacitors are of high quality and present a only 1 nH Equivalent Series Inductance (ESL). Electrolytic capacitors have been characterized with an impedance bridge, and values of 30 nH and 50 m have been obtained.

The measurement of the SESI 18K 1WR [6] shows that a simple paralleled R-L-C representation is suitable until several tenth of MHz. Parameters are given for a 8 A DC bias:

L = 14.7 H R = 3.96 k C = 13 pF Measurements have been carried out using an HP 4194A

impedance bridge, using the suitable test fixture, which allows a DC current polarization.

C. Cabling model This DC-DC converter has been realized on IMS. This

technology allows to obtain low inductive interconnections between components, but also high value of parasitic capacitance.

Values of self and mutual inductance of copper tracks have been calculated using InCa software based on PEEC method [7-8]. This method provides a very powerful approach to account for non ideal behavior of conductors. It is based on analytical formula to compute inductance and coupling, provided a uniform current density. However, due to bondings on copper tracks (see Fig. 1), current direction is unknown, and a complete inductive meshing must be used, contrary to other modeling used in the past [9].

According to PEEC method, a capacitive meshing can also bee used (Fig.3).

Using strictly results from PEEC method in circuit simulator is really too heavy for computer simulation. Therefore, a model reduction for inductive aspects has been proposed. The idea is similar to scattering matrix from the microwave theory: the complete geometry is seen from input-output only. For each conductor linking several

MOSFET Diode simulation

(static or dynamic)

MOSFET Diode parameters

Optimization Measurement

(static or dynamic)

access, one is taken as reference. The differential equations linking voltages (with respect to this reference point) to currents are deduced from global PEEC model.

=

+=

n

j

ijijijiji dt

dILIRV

1

This reduction is illustrated Fig. 3. There are as many reference points as conductors in the converter. In the considered boost, 5 tracks are considered (including bondings and screws see Fig. 4-), therefore, 5 reference points are considered. To be noted that driver track (n6 in Fig.4) has not been taken into account.

The parasitic capacitance behavior of layout is as important as inductive one and has impact on EMI spectrum. In PEEC method conductor is partitioned into segments where capacitive meshing is added to inductive one. Since we want to simplify the representation, less complex model of connections will be used. The solution is to add capacitor to each reference point of the structure, which represents capacitance of one whole conductor. It allows reducing number of capacitors to acceptable magnitude in Saber simulation.

Values of parasitic capacitance are dependent on the geometry of physical structure. In general dielectric and insulate material properties (dielectric constants) thickness and width of conductors are known. However, experimental measurements on few representative copper tracks have been carried out in order to confirm dielectric thickness and effective dielectric constant. Then, all stray capacitances have been calculated using Wheeler/Schneider [10] formula. This formula works pretty well and also appreciates edge effects. Six capacitors have thus been computed, and connected to the inductive model of the structure.

Fig. 4. InCa modelling of the Boost converter interconnections. 5 power tracks (including bondings and screws) + 1 drive track (not taken into

account in the electrical circuit).

III. MODEL VALIDATION.

A. Simulation. Powerful simulator is needed, which allows to make

simulations with a sufficient small time step, according to EMC frequency range required. Because of very high time constants of basic boost converter and of LISN, long simulations have been carried out to obtain steady state of working application.

Simulations of complex model of the considered converter have been carried out using the circuit simulator. On Fig. 5. is presented equivalent Saber scheme of boost converter, LISN and load where all interconnections of converter (IMS, screws and bondings) are summarized into a single macro components, using all differential equations.

In this case different calculation time step can generated different results what could occur especially in the high frequencies range, so adequately small step should be chosen (for frequencies of emitted EMI above 30 MHz it must be smaller than 0.016 s). Therefore, this operation takes a lot of computer power, memory and of course time.

Exploitation of simulation in term of EMI, is simply achieved with FFT of LISN voltage. Results are presented on Fig. 6.

Another validation can be made by comparison on MOSFET voltage (Fig. 7). The correlation is very good, taking into account all measurement inaccuracies (voltage cannot really be measured directly across the chip).

Fig. 5. Saber circuit including boost converter and LISN

Fig. 3 Reduction of PEEC model of one interconnection

1

2

3 4

5

6

Fig. 6 LISN voltage spectrum obtained with Saber

Fig. 7 MOSFET power voltage comparison

B. Measurement An experimental investigation on conducted EMI of DC-

DC boost has been carried out. The converter has been connected to power supply through a LISN, on which spectrum from 50 kHz to 30 MHz has been measured in typical conditions of work (Fig. 8). Many difficulties were encountered to achieve accurately this measurement. Indeed, half of the noise emitted from the converter was due to the driver. Voltage quick variations of the driver (15 V) are synchronous with the one of power circuit (42 V). Reducing of noise emitted by driver has been realized using common mode filter and by disposing it far from the ground plane (Fig. 9), but total decreasing of it was not possible.

Another difficulty was the modeling of the load: the 100 W resistors exhibit a large capacitor with the reference plane, which has not been taken into account in the Saber model. Therefore, even if the global shape of the EMI spectrum is similar, a 10dB difference is still noticeable between simulation and measurement.

The comparison of power MOSFET voltage however is nearly perfect (Fig. 7), what validates the semiconductor and inductive modeling.

The EMI level on the two legs of the LISN (plus and minus leg) are exactly the same. This remark can be made in both simulation and measurement. This phenomenon can be attributed to the use of nearly perfect input capacitors (ceramic), which leads to a symmetrical path between plus and minus tracks.

Fig. 8. LISN voltage spectrum obtained with spectrum analyszer.

Fig. 9. Disposition of driver and converter.

IV. SENSITIVITY STUDY To better understand the contribution of each parameter

of the model to EMI spectrum, a sensibility study has been carried out. Each time, one element of the complete EMC model has been replaced with an ideal behavior. The comparison of this new result with the complete one allows to determine the influence of the modified component. In this study, the main goal is to determine the importance and the frequency range of each component (semiconductors, interconnections).

A. Semiconductors influence Model of the MOSFET transistor - including parameters

presented in section II-A- used in the complete simulation have been constructed with Saber tool Model Architect. Next, it has been changed by ideal MOSFET (...no charge storage & ideal, simple characteristics...) model and new simulations have been done. Voltages of LISN from these two kinds of circuits are presented on Fig. 10.

It can be seen that there are different in range above 8 MHz. When the dynamic parameters of MOSFET responsible for turn on and turn off process are altered,

Driver

Common mode filter

LISN

envelope of spectra in mentioned range also changes. Other series of simulations have been made by replacing

power diode model with ideal diode model. It can be noticed looking at Fig. 11, that the differences in EMI spectrum starts at 3 MHz, but is especially noticeable around 10 MHz.

Consequently, semiconductor models are of great importance in the high frequency part of the spectrum.

Fig. 10. Influence of MOSFET parameters on LISN spectrum.

Fig. 11. Influence of power diode parameters on LISN spectrum.

B. Layout influence Inductive modeling of the tracks is not negligible in

accurate EMI forecasting. Stray inductance plays a great role in commutation and has an effect on semiconductor behavior. In addition, voltage ringing occur which contribute to EMI generation. The influence of stray inductance is shown on Fig. 12, where comparison of results of simulation with and without Saber macro component generated in InCa is presented.

On the other side parasitic capacitances of copper track can be removed. For IMS circuit, stray capacitance are of great values and their role is consequently significant. Removing parasitic capacitance gives great different on spectra for all frequency range of emitted EMI. It can be said that parasitic capacitance are important parameter for EMC behavior description.

In the same idea, it is easy to replace IMS substrate with conventional PCB. This results in an increase of stray inductance, and a decrease of stray capacitance (between 1 pF to 50 pF for PCB and from 100 pF to 1000 pF for IMS). This difference is due to the modified distance to the ground plane (~1.3 mm for PCB and ~70 m for IMS).

The comparison of LISN spectrum for these two technologies is presented Fig. 14. Disturbances generated in PCB layout circuit are much lower than those from IMS, due to lower stray capacitance.

Fig. 12. Influence of inductive connections on LISN spectrum.

Fig.13. Influence of parasitic capacitances on LISN spectrum.

Fig.14. Spectrum of LISN voltages for PCB and IMS (real)

V. SIMPLIFICATION OF BOOST MODEL FOR FORECASTING EMI

A better knowledge of EMC generation mechanisms has allowed the construction of a simple circuit to forecast EMC behavior of the boost converter. Since inductive characteristics of the interconnects and the semiconductor switching characteristics acts in the high frequency range only, the MOSFET and the diode have been replaced with a simple voltage source, and no inductive parameters have been taken into account for interconnects. On the contrary, all capacitive aspects have been kept in the modeling.

In this simple scheme (Fig. 15) can be found the LISN (Cn, Rn and Ln), the input capacitors (C1, C2 and C3), the input inductor (L, CL, RL). Two main parasitic capacitance due to the layout are also taken into account. They are computed from the complete capacitive model.

This leads to a very light linear model, and therefore to fast and easy calculations. It can be implemented in a simple Mathcad document. Spectrum obtained with Mathcad sheet is presented on Fig 16. It can be noticed that for frequencies below 10 MHz results are similar to those from measurement and Saber simulation.

Fig. 15 Simplified equivalent circuit for EMI estimation.

Fig 16. EMI spectrum obtained with simple model.

This scheme was used in [11] to improve the capacitive

layout of this converter, with respect to EMC constraints. It is thus shown in this paper that its accuracy is sufficient under 10 MHz. The modeling and prediction of conducted EMI using this model could be useful as a fast design and optimization tool for Power Electronic applications.

VI. CONCLUSION Simulations and experimental investigations on

conducted EMI of hard switching boost converter were carried out. A generic method to account for stray elements of power layout (inductive and capacitive) in simulation has been presented. An automatic coupling with SABER

allows forecasting EMI spectrum. After validation, with experimental results, a sensibility study allows a better understanding of the contribution of each stray element on the global EMC spectrum. Information - about each component influence - has been used to improve simple model for disturbances forecasting.

VII. REFERENCES [1] Allard, B.; Garrab, H.; Mi, W.; Ammous, K.; Morel, H. "Switching parameter maps-a new approach to the validity domain of power device models"; IEEE PESC'03, 15-19 June 2003 Pages:1220 - 1225 vol.3 [2] A.E.Ruehli "Inductance Calculation in a Complex Integrated Circuit Environment", IBM Journal of Research and Development, Sept. 1972

[3] Morel, H.; Gamal, S.H.; Chante, J.P "State variable modeling of the power pin diode using an explicit approximation of semiconductor device equations: a novel approach " Power Electronics, IEEE Transactions on , Volume: 9 Issue: 1 , Jan. 1994 Page(s): 112 120

[4] Berning, D.W.; Hefner, A.R., Jr.; "IGBT model validation" IEEE Industry Applications Magazine, ,Vol 4 , Issue: 6 , Nov.-Dec. 1998.

[5] JL.Schanen, J.Roudet, W.Teulings "Prediction of current distribution between paralleled capacitors in Hard Switching operation: application to decoupling capacitor" IAS'97 New Orlean, Louisiana, 5-9 oct 97

[6] Microspire, 16 parc d'activit du Beau Vallon, 57970 ILLANGE France - +33 3 82 59 13 33

[7] C.Martin, JL.Schanen, R.Pasterczyk "Power Integration: electrical analysis of new emerging package", EPE'03, sept 03, Toulouse, France

[8] M.Besacier, JL.Schanen, J.Roudet, P.Suau, JC.Crebier "PSPICE compatible electrical equivalent circuit for busbar" IEEE APEC 2000, Feb 6-10, New Orleans, Louisiana

[9] W.Teulings, J.L.Schanen, J.Roudet "Switching performances and EMI noise generation of a chopper on Insulated Metal Substrate" EPE'97 Trodheim, 8-9 sept. 97, pp 134-138

[10] E.Bogatin "Design rules for microstrip capacitance" IEEE trans on components, hybrids and manufacturing technology, Vol 11 n3, sept 1988.

[11] JLSchanen, L. Jourdan, J.Roudet "Layout Optimization to reduce EMI of a Switched Mode Power Supply", IEEE PESC'02, june 2002.