Interactions Between An Input Emi Filter And A Power Supply

503

INTERACTIONS BETWEEN AN INPUT EMI FILTER AND A POWER SUPPLY Stéphane Brehaut1, Jean-Charles Le Bunetel1, Didier Magnon1, Antoine Puzo2, Alfonso Santolaria3, David González3, Javier Gago3, Josep Balcells3 1

Laboratoire de Micro-électronique de puissance Tours, France

2

3

SAFT POWER SYSTEMS GROUP Chambray-Lès-Tours, France

Universitat Politècnica de Catalunya, Electronics Engeenering Department, Terrassa, Spain

Email: [email protected] Abstract — In power electronic, the presence of filter is dictated by the EMI gabarit. So, the passive filters are largely used in order to reduce the EMI pollution produce by the power supply. However, there are interactions between the passive elements and the power supply. This paper present the different phenomena which create the many loops of EMI resonance. The resonances produce the peaks of pollution. On the resonance’s frequency, the EMI pollution is largely above the gabarit.

I. INTRODUCTION Generally, the power supplies used in power electronic produce EMI pollution. They can be in radiated or in conducted mode. Many techniques are developed to reduce these disturbances. The first one is commonly the use of filters to reduce the conducted mode. The choice of filter is function of differential or common couplings. These couplings are different in function of the used of power supplies. Others techniques are to act on the interference sources. Between 150 kHz and 30 MHz, the conducted interferences of power supplies are represented by the differential and common mode. The disturbance current of differential mode depends of the variation of current dI/dt in the switch. It is the high frequency harmonics of the commutation of current. The disturbance current of common mode is function of the variation of voltage dV/dt of the switch. This variation applied on the parasitic capacitances creates a current going from the circuit to the ground. In the case of the power supplies in communication domain, the common mode is the principal pollution. This power supply must respect the hard susceptibility standards. A ground plane is added for the immunity. The result is the increase of the parasitic capacitive between the track and the ground. So, we have an increase of the common mode current. In order to reduce the disturbance currents, the classical approach is the use of passive EMI filters. The filter is necessary at the input and at the output of the device. If there is only an input filter, then the pollution go to the load, and can stop the operation of the load. The filter size is function of the attenuation value in order to respect the standard. The choice of filter takes the interactions between the input and

output filter into account. It takes the interactions between the filters and the power supply into account, too. To modify the propagation paths, we can act on the layout of circuit [1]. If the ground plane is changed, we can decrease the parasitic capacitances. So we reduce the common mode disturbances. This method permit to decrease the size of the filter. An other method is to act directly on the perturbation sources that are the switches. The voltage variation dV/dt can be reduce. If we increase the value of the grid resistance, the common mode current is lower. In this case, a compromise must be found between pollution level and the switching losses. The use of frequency modulation techniques permit to obtain a spectrum with a lower EMI level than with the conventional constant frequency square signal [2]. In fact, the amplitude of each harmonic is distributed on the modulation frequency band around the harmonic. We want to understand the problem of the interferences between a standard EMI filter and a power supply, figure 1. We do not take other techniques of reduction into account. We used a simulation tool in order to understand the EMI phenomena. A common mathematical tool as MATLAB£ is developed in frequency domain. We can predict with accuracy the conducted EMI before building a power supply with filter. The frequency range studied is 150 kHz-30 MHz. We consider in the simulation tool’s, the Line Impedance Stabilization Network (LISN), the converter and the load. We take all active and passive parasitic elements into account.

C56

L1

C3

C5

C2

C1 C54

Fig. 1.

C7

C4

C6

Full input EMI filter with the power supply

Power Supply + Load

504

Firstly, we present the simulation tool used in our research. Secondly, we apply this tool to simulate the pollution of industrial power supply. The power supply studied is constituted of both cascaded converters in series with the input filter. It is used like a battery charger in telecommunication domain. We show the input filter influence on the EMI spectrum. II. DESCRIPTION OF THE POWER SUPPLY The power supply used in this work is a power factor correction (PFC) of Boost type associated to an inverter. The PFC converter permits to obtain a input current always sinusoidal in phase with the network voltage. The inverter with a transformer and a rectifier reduces the voltage output and insures the isolation between the mains and the load, figure 2. The specifications of each module is given table I. We determine the conducted EMI in dBµV.

and the full bridge are responsible of the pollution. The impedance of the network is endless. The impedance’s variations of the components due to the thermal phenomena are negligee. B. Sources of pollution, propagation paths and receptor The propagation paths correspond to the electrical connections between interference components and the commutation cell. The study of propagation paths requires a previous knowledge of the high frequency behaviour of every components and interference elements. These elements are the parasitic capacitances and inductances. The passive components, the circuit layout and the bridge rectifier form the full propagation path. All these components and layouts were carefully measured. The switch and the diode are two sources of pollution. These sources produce the pollution in the commutation cell. We suppose that the rectifier does not generate disturbances but the impedances of the diodes are included in the simulation. C. Modelisation of the PFC in series with the inverter For the modelisation of the converter, we use a matrix method designed with Kirchoff law [3]. The equation 1 is considered. U=Z.I

Network 230ac 50Hz

PFC

Inverter

Fig. 2. Equivalent schematic with the PFC in series with the inverter and with the load

Input Voltage Output Voltage Vbus Input Current Output Current Switching frequency of PFC Switching frequency of Full Bridge

(1)

Load

230 VRMS 48 VDC 382 VDC 3.2 ARMS 12 ADC 50 kHz 112 kHz

U : sources of pollution produced by the commutation cell. Z : Impedance matrix representing the converter. I : Parasitic Currents produced by the sources of pollution across the impedance matrix. Each element of the circuit respects the impedance variation along the frequency range 150 kHz-30 MHz. The layouts are assimilated to inductances and the ground board is represented by capacitances.

TABLE I SPECIFICATIONS OF THE FULL POWER SUPPLY

IV. WHOLE SIMULATION SYSTEM.

III. EMI TOOL DESCRIPTION

To study the behaviour of EMI filter, we must be able to simulate the whole system without the filters. In this part, the simulation results concern the PFC in series with the inverter, figure 1.

This section is a description of the proposed tool. We work directly in the frequency domain. The software enables to determine the conducted EMI in dB/µV.

Simulation

A. Assumptions. For the conducted EMI prediction, we propose a complete electrical equivalent circuit including the full converter itself with all the parasites, the measurement equipment which is the LISN. We do not forget the cabling impedance between LISN and converter. We do many hypothesis in order to limit the complexity of the modelisation. We do not take the saturation effect of the boost inductance of the PFC into account. We consider that just the switches of the Boost converter

Template EN 55022

Measurement

Fig. 3. Template, simulation and measurement of the PFC in series with the inverter and C2, C5, C6 and C7 on the range 150 kHz30 MHz.

505

Figure 3 shows the result of the simulation of the full power supply that is compared to the experimental result. The simulation permits to calculated the envelop of electromagnetic spectrum. In high frequency, the simulation has computed three peaks that exist too in practical. These peaks are the result of resonance phenomenon at frequency 8, 12 and 20 MHz. Loop A

represent the measured impedance, figure 7. The calculated impedance will placed in ground wire.

Loop B

C6 C2

C5 C7

Fig. 4.

Loop C

Inverter + Load

Fig. 6. Measurement and simulation of the common mode choke on the range 150 kHz-30 MHz.

Resonant circuits A, B and C. LMC/2

These peaks are produced by resonance loops stimulated by the PFC pollution voltage source. The MOSFET in series with the inductance and with capacitances C5, C6 and C7 compose the three loops A, B, C, figure 4. The loops impedance calculation has allowed to identify the loops creating the resonance phenomenon [4].

Neutral

IMC/2 LMC

LMC/2 IMC/2

RMC

IMC

RMC/2

B

D

2CMC

Ground

CMC

The impedance of the leaking inductance of common mode filter is measured and presented, figure 8. We measure the impedance between A and C. From the measurement, we calculate the model of this impedance with the same method, figure 9. measure model

4

Impedance

10

2

10

10

6

10

6

Frequency (Hz)

10

7

10

7

100 50 0 -50

-100

Frequency (Hz)

Fig. 8.

IMD

B

C

RMC/2 2CMC

Phase

In this part, we describe the model used of the common filter inserted at input of the converter. We choose an common mode inductance because it is the common mode noise which is predominant. Each winding is in series with the network and the two winding are coupled, figure 5. The common mode currents are leakage currents that go through the ground. Also the common mode current IMC creates a flux in two winding what adds itself in magnetic material, and the differential mode current IMD creates a flux in the two winding what subtracts itself. We obtain a large inductance value for the common mode current, and a low inductance value for the differential mode. This inductance associated to the capacitances C6, C7 creates a filter for the common mode current.

A

A

Phase IMC/2

Fig. 7. Model of the common mode choke of L101 with LMC=5.6 mH, CMC=6.5 pF et RMC=2 kȍ.

V. INSERTION OF THE COMMON MODE FILTER

A

IMC/2

Measurement and simulation of the leaking inductance

C IMC/2

C

IMD

D

IMD B

IMC/2

A

LMD/2 RMD/2

C

2CMD

D

LMD/2

Fig. 5.

IMD

Schematic of the common mode choke B

The impedance of the common mode filter is measured and presented figure 6. The leg A and B are directly short-circuited like the legs C and D to make the measurement. The electric model is computed to

RMD/2 2CMD

D

Fig. 9. Model of the differential mode of L101 with LMD=35 µH, CMD=7 pF and RMD=30 kȍ

The full model is the mixing of common mode model and the leaking inductance model [5].

506

VI. SIMULATION RESULTS

VII. FULL EMI FILTER

Figure 10, the equivalent schematic of the common mode choke is inserted in the power supply model. Each parameter of common filter is presented in each impedance Zmd/2, Zmc et Zmd/2.

We add in the input filter two capacitances of common mode C54 and C56 and a capacitance of differential mode C1, figure 1. The input filter is now full. We can see that the level of the peaks at 1.2, 8 and 12 MHz have largely reduced. There is a drop of the EMI pollution on the lower frequency. On the other hand, the spectrum in HF stay the same, figure 13.

Zmd/2

C3

Zmc

C5

C2

C6

L.I.S.N

Power supply + Load

C4

Simulation

C7

Zmd/2

Fig. 10.

HF schematic of the filter of the common mode L1

In order to show the influence of filter, we simulate the full system and we compare at the measurement. We show a great reduction of the EMI pollution at low and medium frequency, figure 11.

Measurement

Fig. 13. Measurement and simulation of the full input filter on the range 150 kHz-30 MHz . Simulation

VIII. CONCLUSION

M easurement

Fig. 11. Simulation and measurement of the EMI pollution with the common mode filter C2, C3, C4, C5, C6 C7 and L1 on the range 150 kHz-30 MHz.

We can observe a new peak of resonance at 1.2 MHz in measurement and in simulation. We have identified this loop by the calculation of the impedance of different loop. The resonant loop E, figure 12, is created by the two capacitances in parallel C2 and C5 in series with the Boost inductance. This peak didn’t appear before, the pollution was higher than it. This peak just appears now because the EMI pollution generates the power supply are above this peak.

In summary, we have presented the power supply studied and the problem of EMI pollution and the different solutions of EMI reduction. We have shown that the EMI simulations of two industrial converters between 150 kHz and 30 MHz are similar to the measurement. Thanks to the validation of the simulation tool, we can studied the influence of various parameter in order to see their influence on the EMI spectrum like the influence of resonance loops. Now, it is easier to change the value of the component of the filter in order to optimize the size of the filter. REFERENCES [1]

[2]

[3]

C2 C4

[4]

C7

C3

C5 C6

Loop E

Fig. 12.

Equivalent schematic showing the loop E

Inverter + Load

[5]

L. Rosseto, S. Buso, G. Spiazzi, “Conducted EMI Issues in a 600-W Single-Phase Boost PFC design”. IEEE Transactions on industry application, Vol. 36, NO. 2,pp.578-585, March/April 2000. A. Santolaria, J. Barcells, D. Ganzáles, J. Gago “EMI Reduction in Switched Power Converter by Means of Spread Spectrum Modulation Techniques ”, PESC 2004 IEEE 35th, Aachen, Germany, 2004 S. Brehaut, J-C. Le Bunetel, D. Magnon, A. Puzo. “A conducted EMI model for an industrial power supply full bridge”, PESC 2004 IEEE 35th, 2004 S. Brehaut, J-C. Le Bunetel, D. Magnon, A. Puzo. “Analysis EMI of a PFC on the band pass 150kHz-30MHz for a reduction of the electromagnetic pollution”, APEC 2004 IEEE, Anaheim, USA, feb 2004. B.Revol, J. Roudet. “EMC modeling of a three phase inverter”, EPE 2003, Toulouse, France, 2003.