Vous êtes sur la page 1sur 6

Analysis EMI of a PFC on the band pass 150kHz-

30MHz for a reduction of the electromagnetic


pollution

S. Brehaut, M. Ould El Bechir, J.-C. Le Bunetel, A. Puzo.


D. Magnon SAFT Power Systems Group
Laboratoire de Microélectronique de Puissance 37173 Chambray-lès-Tours, FRANCE
LMP
7 avenue Marcel Dassault, 37200 Tours, FRANCE
stephane.brehaut@etu.univ-tours.fr

Abstract— The boost PFC circuit, widely used to provide a good which gives an electrical schematic of the circuit. We keep a
power factor correction is developed at low and medium power. part of the EMI filter because we need the capacitor above the
However, PFC generates electromagnetic interferences in the bridge rectifier for the tripping of the voltage measured by the
power converters. In order to have a better comprehension of the controller. The designed specifications include : output power
causes of the EMI pollution in the circuit, we use a simulation (Po), input voltage (Vin), line frequency, output voltage
tool. A prototype with a new layout is created to have an (Vout). The controller components are not included in Fig. 1.
experimental validation of the reduction of the EMI. The constant-frequency average-current-mode control for
continuous-current-mode operation is the control strategy for
Keywords- Power factor correction, EMI, Simulation,
the switch. The characteristics of the converter used for this
Resonance, Tracks
application are given in table I.
I. INTRODUCTION
TABLE I. SPECIFICATIONS
Lots of works have been committed with the aim to
understand the high frequency comportment of the power Input Voltage (VRMS) 230
factor corrector (PFC). Many researches have been performed Output Voltage (VDC) 382
about the simulation of the PFC in high frequency [1][2][3].
Switching frequency (kHz) 40.5
These researches have ended on a good modelisation from
150kHz to 1MHz. The influence of the spectrum have been Input Current (ARMS) 3.2
done with the addition of a modulation on the commutation Inrush
frequency of the switch or with the addition of screen on the limiter circuitry

source of MOSFET [4][5][6].


The purpose of this paper is to understand the whole C 10 6
spectrum between 150kHz and 30MHz. We have conceived a C 102
Load

simulation tool in order to recreate the high frequency C105 Control Capacitor
filter
comportment of the PFC on the band-pass 150kHz-30MHz. It
is easier to understand the causes of pollution’s phenomenon C 10 7

because we can interfere on each electrical parameter (passive, Ac Capacitor


230Vac 50Hz filter
active) and also on parasitic elements. This tool permits to source
quantify the effects of the layout on the electromagnetic
spectrum. A new layout has been studied and has ended in a Figure 1. Boost PFC Stage Schematic
new prototype.
III. BOOST EMI MODELLING
The document is presented as follows. Firstly, we describe
the studied converter, the simulation tool and its applications Our simulation tool is a resolution of an electrical scheme
for the PFC. Secondly, we show the propagation of the in the frequential domain. This schema is constituted of a
disturbances in function of the pollution spectrum. Then, the LISN, a PFC and a load. The schematic of PFC is completed
influence of the layout on the electromagnetic spectrum is by the parasitic inductances and capacitances and the parasitic
studied and is validated by experimental results. capacitances between the radiator and the active components,
Fig. 2. The switches are modelised by sources of voltage with
II. DESCRIPTION OF THE PFC. an impedance to their bounds. The load is a simple winded
resistance. In order to simplify the tool, we do not consider the
The system to be designed consists in a boost PFC
converter with a part of an EMI filter as showed in Fig. 2,

0-7803-8270-6/04/$17.00 (C) 2004 IEEE


effect of saturation of the core or the thermal effects which modify the characteristic of the passive and
active components. We admit that the impedance of the efficiency is greatly improved in radiated mode, from 5dBµV
network is endless. Each parasitic capacitance and inductance to 20dBµV. In fact, the ringing frequencies are above 30 MHz.
has been measured by a impedance meter on the conducted We can conclude that the commutation cell pollutes in
range frequency under study. conducted EMI with the slope of voltage and pollutes in
Z59
radiated mode with the ringing of voltage.
Commutation cell.
Z66 Z67 Load.
LISN.
Z63 Si diode
Z37 90
Z57 Z58 Z68 Z69
Z4 Z8 Z12 Z51 Z17 Z21 Z24 Z27 Z33 Z41
Z30 SiC diode
Z43 I36
Z56 Z47 Z62 Z Z Z73
Z38

Magnetude (dBµV)
Z2 Z5 Z9 Z52 Z13 Z18 Z22 Z28 Z31 Z34 48 72
80

Z50
Z45
Z49 Z25 Z44 Z64 Z65
Z16 70
Z61
Z46
Z55

Z3 Z6 Z10 Z26 Z39 Z74 Z75


Z53 Z14 Z19 Z32 Z35
60

Z36 Z40 Z42 10


6
Frequency (Hz) 10
7
Z7 Z11 Z15 Z54 Z20 Z71 Z70
Propagation path.
Figure 4. Pollution spectrum in conducted mode with two various diodes.
Figure 2. Equivalent scheme for EMI modelling, including parasitic
components, LISN and converter.
80 Si diode
We have ended in a frequency modelisation. The commutation SiC diode
cell represents the source of pollution. The layout, the passive
60
elements and the parasitic ones are the propagation paths. The
LISN is the receptor of the pollution. We can notice that the
simulation is nearer of the measurement, Fig. 3. The pollution 40
generated by the PFC is above the standard 55022 [9], because
we take off a part of the EMI filter. Firstly, the spectrum is
constituted by slope between 150 kHz and 7 MHz. Secondly,
3.107 2.108
we can see three peaks of resonance at 8 MHz, 12 MHz and 18 Frequency (Hz)
MHz. The robustness of this modelisation has already been
Figure 5. Pollution spectrum in radiated mode(30MHz-200MHz)with two
demonstrated previously [9]. various diodes.

The parasitic capacitances to the bounds of switches


Simulation
represent the parasitic capacitance between the active
component and the radiator, and the capacitance between the
strip and the component. The Fig. 7 shows the effect of
parasitic capacitances reduction. This reduction is achieved by
the suppression of the capacitance between active component
and the radiator. The interference level is reduced to 5 dBµV in
Measurement low and medium frequency because the pollution is due to
Standard CEI
CISPR 22 slope of current and voltage switching. The use of a effective
screen between switch and radiator, connected to the source of
the switch should allow to reduce the pollution at low and
Figure 3. Simulation and measurement for the complete HF scheme medium frequency.
(150kHz-30MHz)
w ith rin g in g .
IV. STUDY OF THE POLLUTION SPECTRUM.
A Influence of voltage ripples influence and parasitic
capacitances of active components
We study the effect of the ripples in EMI . In simulation, w it h o u t r i n g i n g .
Fig. 6, we can see that without the voltage ripples, the spectrum
level is lower above 15 MHz. The influence of the voltage
ripples is limited for the conducted EMI. Experimentally, this
influence can be showed with the change of the diode of
commutation cell (Si by SiC diode). The spectrum in
conducted mode is the same, Fig. 4, whereas the spectrum in
Figure 6. Simulation of EMI between 150KHz and 30MHz with and without
radiated mode is more attenuated, Fig. 5, above 50MHz. The ringing on the switch.

0-7803-8270-6/04/$17.00 (C) 2004 IEEE


w ith p a ra s itic Module of the

Magnétude
c a p a c ita n c e . boost inductance.

Phase of the boost


w ith o u t p a ra s itic inductance.

Phase
c a p a c ita n c e .

Figure 7. Simulation between 150KHz and 30MHz with and without Figure 10. Module and phase of three loops A,B and C between 1MHz and
capacitance MOSFET- radiator 10MHz

B Study of three peaks : 8, 12 and 18MHz


Loop A
The PFC is composed of a whole RLC components which

Magnetude
form several resonant circuits. These circuits favour the
Loop C
propagation of the disturbances at particular frequencies. The
three peaks identified, on the Fig. 8, correspond to three
circuits in parallel, Fig. 9. The impedance, module and phase, 10 6 Frequency (M Hz) 10 7
of these resonant circuits shows several changes of phase 0

corresponding to the peaks at 8, 12 and 18 MHz. -20


Loop C

Phase
-40
-60
-80
Loop A
7

100 2 10

10 6 Frequency (M Hz)
8 MHz
Magnitude dBµV

12 M Hz
Figure 11. Module and phase of three loops A,B and C between 1MHz and
18 M Hz
80 10MHz

The three resonant circuits are been recognized by the study


60
of the impedance of all propagation paths. The phase of their
impedance pass by zero at the resonant frequency. The
MOSFET, the boost inductance and the filter capacitances of
106 107 the common mode (C106, C107) and of the differential mode
F re q u e n c y H z
(C102, C105) form the whole of the paths with the layout. The
Figure 8. The 3 peaks: 8, 12 and 18 MHz Fig. 11 shows an example of the module and the phase of the
loops A and C between 1MHz and 10 MHz.
LoopB The Fig. 11 permits to determinate the loops which
intervene at each resonance (Table II). The loop C is
responsible for the peak at 8 MHz, the loops A and C are
C 106 L BOOST
responsible for the peak at 12 MHz. The thirdly peak is created
C 102
C 105 by the loop B. It means that if we remove the capacitance C105,
LoopA we cut the loop B and we eliminate the peak at 18 MHz.
C 107

TABLE II. RESONANCE LOOPS RESPONSIBLE OF THE PEAKS AT 8, 12 AND


18MHZ
LoopC
Peak at Peak at Peak at
8MHz 12MHz 18MHz
Figure 9. Resonant circuits A, B, C
Loop A resonance
If we study separately the three circuits A, B and C, we
have several resonant circuits. At frequencies studies, the
impedance of components varies a lot. It is due principally to Loop A in parallel
resonance
Loop B
the impedance of the boost inductance which presents many
changes of phase, Fig. 10. Indeed, the inductance becomes
capacitive at 170 kHz, and presents three consecutive changes Loop C resonance resonance
of phase which are increased by the other elements of the
circuit.

0-7803-8270-6/04/$17.00 (C) 2004 IEEE


elements. Lm symbolizes the inductive contribution of ground
Simulation with
changes of phase. plane. This scheme puts in evidence a low-pass filter of second
order with a transfer function following :

H(p)= 1 (1)
1+ pRbCb + p 2(Lb − Lm)CT

The equivalent impedance of the loop is :


Simulation with
perfect boost.

Ze= Rb + j((Lb + Lm)ω − 1 ) (2)


Cbω
Figure 12. Simulation with a model boost with and without changes of phase
When the phase passes by zero, there is a resonance for the
If we compare the different changes of the boost frequency :
inductance’s phase, Fig. 10, with the peaks of pollution at 8
and 12MHz, we can see the changes of phase at the same 1
frequency. At 18 MHz, the peak is due to the interaction f= (3)
2π (Lb + Lm)Cb
between the loops A and B. The main element acting on the
resonant peak’s, is the boost inductance.
The damping factor is :
We have replaced the boost model with the consideration of
phase’s change by a perfect boost in the simulated PFC,
Fig. 12. The resonances at 8MHz and 12MHz are eliminated. Rb Cb
The comportment of the real boost under 12MHz is inductive z= (4)
2 (L b + L m )
like for the perfect boost. So, the peaks at 18MHz remains
unchanged.
Then, the frequency and the amplitude of peak are function
So, the comportment of the boost inductance has an great of Rb, Lb, Cb and Lm values. So, the tracks and passive
influence on the pollution in conducted mode. On the one hand, elements intervene too. To damp the ripple of the resonance,
the solution should be to make an inductance without any the damping factor must be superior to 0.5. This factor depends
changing of phase. On the other hand, we can note that the first to the characteristic impedance of the tracks :
change of phase at 170KHz is without consequence on the
EMI, because the pollution generated by the commutation cell
is above the peak created by the change of phase at 170KHz of Zc = Lb + Lm (5)
the Boost inductance. So, the second solution is to bring the Cb
change of phase at low frequency or higher than 30MHz.
B Optimisation of tracks
V. INFLUENCE OF TRACKS ON ELECTROMAGNETIC With the (4), to augment z, we can increase Rb, or Cb or
INTERFERENCES the term (Lb+Lm).To minimize the common mode
A Effect of tracks on resonant loops interferences, the technical is to increase Cb only on area of
circuit and to decrease (Lb+Lm). The layout’s modification of
We use a method of recognition of resonant loop which circuit allows to act on the parasitic capacitance Cb and on the
consists to have a simplified electric view with a RLC circuit parasitic inductances (Lb+Lm).
[9][10]. This circuit takes the main parasitic elements of
passive components and tracks into account. The tracks The simulation permits to put in evidence the various
constitute the propagation paths of electromagnetic resonant loops. We have seen before that the peaks are due to
interferences which are modelled with the basis elements R, L the phase’s change of the boost inductance. The effect of track
and C. An equivalent electric scheme of resonant loop is is visible each time that the boost inductance changes the
represented by the Fig. 13. phase. Then, the tracks can modified the time of changes of
phase. Indeed, the track’s impedance isn’t negligible, compared
Rb Lb to the full impedance in the loop. We know that all the loops
determinated by simulation pass by the MOSFET, the boost
impedance, and the input capacitances. The tracks which
compose this propagation path, are the elements which must be
Vinteference Cb modified in order to control the amplitude and the frequency of
Lm
pollution peaks.
VI. EXPERIMENTAL STUDY
Figure 13. Equivalent electric model of a loop
The following results are done with a PFC which has the
Rb, Lb and Cb, represent respectively the resistive, same characteristics that industrial PFC presented previously.
inductive and capacitive contribution of tracks and passive Only the layout of circuit is changed. We keep the same

0-7803-8270-6/04/$17.00 (C) 2004 IEEE


components to can study the influence of layout on the the impedance of the inductance if we add a capacitance in
electromagnetic interferences of PFC. The uncoupling parallel to shift this frequency to the radiating spectrum.
capacitances C105, C106 and C107 are removed. This allows to
remove the loop A and B, and the peak at 18 MHz. The PFC reference.
reference PFC has the same layout that the industrial PFC.
A Tracks identification
In order to make the tracks less inductive, its must be wide
and short. To limit the capacitive comportment of tracks, we
removed the ground plane on some areas. This allows to
dissociate the capacitive effect and the inductive effect on the PFC with a new layout
studied track. These effects are linked by the characteristic without ground plane.
impedance (5). The sensible tracks which intervene anyway in
the common and in the differential mode are the tracks of the
resonant loop.
Figure 15. Reduction of EMI by the reduction of inductive and capacitive
B Experimental results effects of tracks with out ground plane
By the reduction of the inductive effect on tracks and by the
increase of the capacitive effect, we can see a spectrum C Limits of influence of the tracks and of the part of passive
attenuation of 3 dBµV between 150kHz and 1MHz, Fig. 14. components
The decrease of the first peak overtakes 7 dBµV, but the We have showed that the effects of tracks aren’t always
amplitude of second peak doesn’t change much (1 dBµV). sufficient to control the frequential comportment of the
There is only a shift of the frequency. resonant loops. Indeed, we use the wide of tracks to modify the
capacitive and inductive effects. For technological, thermal and
PFC reference. electrical reasons, the wide of tracks is limited.
To modify the impedance of resonant loops, the
comportment of passive components must be adapted to
minimize the peaks of resonance. For example, the
comportment of the boost inductance and the input filter
capacitor can be improved. The choice and the place of the two
PFC with a new layout elements allows to control the equivalent impedance of the
with a ground plane. resonant loops, and the peaks of pollution.
VII. CONCLUSION
We have analysed the electromagnetic interferences of a
Figure 14. Reduction of EMI by the reduction of the inductive effect of tracks PFC on the band pass 150 kHz-30 MHz to reduce the
with a ground plane electromagnetic pollution. The EMI modelling takes the real
impedance of components integrated in the industrial PFC and
The shift of the second peak is principally due to the the parasitic elements produced by the layout into account. The
decrease of inductance (LP and Lm).This is confirmed by the frequential study of this EMI model allows to determinate the
(3). Indeed, parasitic capacitances between tracks and the electromagnetic interferences in simulation. So, the
ground plane don’t intervene in resonant loops. A second consideration of each active, passive and parasitic element in
circuit confirms this affirmation. In the second circuit, we have HF, allows to reproduce the full conducted EMI.
reduced the parasitic capacitances between tracks and the
ground plane with a new layout without ground plane. The The simulation tool permits to see and to understand the
Fig. 15 shows that the parasitic capacitances don’t modify the influence of various parameters like the voltage ripples or the
frequency of peaks. parasitic capacitances of active components. The simulation
defines the propagation paths and permits to know the
The suppression of the ground plane on the bottom face of influence of boost inductance in series with filter capacitances.
circuit allows to reduce the interference level until 10 MHz. Now, we can identified the cause of the resonant peaks.
The decrease overtakes 8 dBµV between 150kHz and 4 MHz
and 10 dBµV for the first peak. But the second peak increases After to have identified the propagation paths by
of 1 dBµV. simulation, we can eliminate one resonant loop by removing
one component. The experimental circuit checks this
The impedance of boost inductance explains this simulation result. We could validate the influence of layout on
phenomenon. Indeed, at 12MHz, the boost inductance becomes the resonant peaks. The experimentation and the simulation
progressively inductive. The parasitic inductances of the tracks show the essential incidence of the boost inductance and the
are negligible compared to the impedance of the boost limits of the layout modifications on the resonant peaks.
inductance. Then, the layout of the tracks doesn’t modify the Finally, the next studies will concern the passive components,
interferences level at this frequency. However, we can modify in order to reduce the electromagnetic level without addition of
filter components.

0-7803-8270-6/04/$17.00 (C) 2004 IEEE


REFERENCES [6] L. Rosseto, S. Buso, G. Spiazzi, «Conducted EMI Issues in a 600-W
Single-Phase Boost PFC design». IEEE Transactions on industry
[1] J-C Crebier, « Contribution à l’étude des perturbations conduites dans application, Vol. 36, NO. 2,pp.578-585, March/April 2000.
les redresseurs commandés», Thèse de L’Institut National Polytechnique
de Grenoble, 1999. [7] L. Rosseto, S. Buso, G. Spiazzi, «Conducted EMI Issues in a Boost PFC
Design», Proc. of Int. Telecommunications Energy Conf. (INTELEC),
[2] S. Busquets-Monge, J. C. Crebier, S. Ragon, E. M. Hertz, J. Wei, J. San Francisco, October 1998, pp.188-195.
Zhang, D. Boroyevich, Z. Gurda, P. K. Lindner, A. Arpilliere,
«Optimization Techniques Applied to the Design of a Boost Power [8] CEI CISPR 22, «Radio disturbance characteristics – Limits and methods
of measurement (edition 3)», 1997.
Factor Correction Converter», PESC 2001 IEEE 32nd Annual, Volume:
2, 2001 pp. 920-925. [9] S. Brehaut, J-C. Le Bunetel, A. Schellmanns, D. Magnon, A. Puzo.
[3] J-C Crebier, M. Brunello, J. P. Ferrieux, «PFC full bridge rectifier EMI «Development of a Conducted EMI Model for an Industrial Power
forecast analysis», EPE 99. Factor Correction» EPE 2003.
[10] A. Pons, «Optimisation de la fonction de filtrage dans les convertisseurs
[4] S. Wang, F.C. Lee, W.G. Odendall, «Improving the performance of
de traction», Alcatel, 1998.
Boost PFC EMI filters», IEEE 18nd Annual APEC 2003.
[5] E. M. Hertz, «Thermal and EMI modelling and analysis of a boost PFC [11] A. Puzo, «CEM Chargeur, Méthode de dépollution à la source», Alcatel
Astom recherche, 1997.
circuit designed using a genetic-based optimisation algorithm», Thèse de
l’Institut polytechnique de Virginie, 2001. [12] M. Ould El Bechir, «Influence du routage d’un PFC sur la CEM»,
rapport de stage d’ingénieur de l’Ecole Polytechnique de l’université de
Tours, 2003.

0-7803-8270-6/04/$17.00 (C) 2004 IEEE

Vous aimerez peut-être aussi