Conducted EMI in PWM Inverter for Household Electric Appliance

Yo-Chan Son and Seung-Ki Sul

School of Electrical Engineering & Computer Science #024

Seoul National University

Kwanak P.O.Box 34, Seoul, Korea (ZIP 151-742)

http://eepel.snu.ac.kr

e-mail:

[e-mail]

Abstract- This paper presents the characteristics of the con-

ducted EMI in the PWM inverter for household electric appli-

ances and proposes a new active common-mode EMI filter. The

conducted EMI spectrums are measured with changing the ac-

tive/passive filter stages and the common-mode EMI is extracted

from the total conducted EMI spectrums in order to identify the

dominant source of the conducted EMI. For better understand-

ing of the conducted EMI, the noise sources are identified and

the characteristics of the noise source and its propagation paths

are investigated from the conducted EMI spectrums measured.

Based on the measurement and analysis of the conducted EMI

in the PWM inverter, the effective method and procedure of

designing filter stages are developed in order to attenuate the

conducted EMI effectively.

I.

I

NTRODUCTION

As the progress of the power electronics technology, its

applications are being adopted not only in the industry, but
also in the home appliances. The existing constant speed op-
eration with on/off control is being replaced by the variable
speed operation for better efficiency and quietness. For the
operation of the ac motor PWM inverters are used to change
the input frequency of the motor. Despite of these conven-
iences, the radio frequency (RF) disturbance signal generated
by the PWM inverter makes its application be difficult. Large
amount of high frequency noise is generated by the switching
operations of the PWM inverter. They are in form of con-
ducted or radiated emissions. In case of the conducted emis-
sion, high frequency signals generated by the one equipments
might interfere with other equipments that is connected at the
point of common coupling (PCC), and invoke the malfunc-
tion of the victim. Also the radiated emission might disturb
the operation of wireless installations or add high frequency
conducted noise to the victim coupled with the conduction
loops in the system. There has been a strong demand on the
regulation of the electromagnetic interference (EMI) by
equipments in the customer electronics since 1990, and each
of products should be compatible with national or interna-
tional standards such as IEC (International Electrotechnical
Commission), CENELEC (the European Committee for

Electrotechnical Standardization), etc [1]. For an example,
‘EN55 014’ of CENELEC limits RF disturbances by electri-
cal motor-operated appliances for household purposes [2].

Power line filters are usually attached in front of the

equipment in order to suppress these side effects of the PWM
inverter, but they result in additional cost and volume of the
system. These filters are intended for the attenuation of the
unwanted high frequency signals, but their design is quite
difficult due to the identification of noise source and nonline-
arity of filtering elements at the high frequency region.
Without better understanding of the noise source and the
characteristics of the system at the high frequency, the EMI
mitigation efforts could require excessive costs with degraded
performance. In this paper, the conducted EMI spectrums of
the PWM inverter for household appliances are investigated.
The implementation and modification of filter stages are de-
veloped and the common-mode EMI spectrum is extracted
from the total conducted EMI in order to separate the domi-
nant noise source. Based on the measurement and analysis of
the conducted EMI in the PWM inverter, the effective meth-
od of designing filter stages are developed in order to attenu-
ate the conducted EMI effectively. A new active common-
mode EMI filter is introduced in order to increase the attenu-
ation performance of overall filter stages.

II. G

ENERATION AND

P

ROPAGATION OF

C

ONDUCTED

EMI

A. Fast dv/dt of Output Voltages

Fig. 1 shows the general structure of the PWM inverters for

single-phase ac input system. In order to meet the harmonic
regulations, the power factor correctors (PFC) are usually
used with a front-end single-phase diode-bridge rectifier [3].
Ac motors and compressors are used as load machine of
PWM inverter. PWM inverters for motor drive applications
are operated at 1 ~ 20[kHz] of switching frequency and usu-
ally IGBTs are used as their power semiconductors switches.
Fast switching operations of IGBTs generate unwanted high
frequency voltages and currents coupled with system para-

lg

C

g

a

b

c

n

s

sg

C

PCC

3-

φ

Load

l

L

−

+

AC

v

ll

C

Fig. 1. PWM inverter connected to a single-phase ac input system.

Dashed line: path of normal-mode current, dotted line: path of common-mode current.

0-7803-7114-3/01/$10.00 (C) 2001

sitic components.

High frequency normal-mode currents are induced by the

abrupt voltage transition of output line-to-line voltage cou-
pled with line-to-line capacitance C

ll

and stray inductances

such as L

l

and some of them can be drawn from ac power

source as indicated by the dashed line in Fig. 1 if the imped-
ance of dc-bus capacitor is finite. This normal-mode current
can be a source of conducted EMI, and also can be a source
of radiated EMI coupled with loops in output cables or power
stage layouts. Also switching operation of PFC can be a sour-
ce of high-frequency normal-mode currents. Because the
semiconductor switch of the PFC is operated synchronously
with the line frequency only to maintain the continuous con-
duction of diode rectifier for harmonic reduction, the low-
frequency operation of PFC is expected not to be a major
source of high-frequency normal-mode current compared
with others.

When it comes to common-mode EMI, the PWM inverter

is the major source of the common-mode voltage, and its
common-mode voltage can be defined by its switching con-
dition. Except the operation of the PFC, the common-mode
voltage in Fig. 1 can be represented as (1) and (2).

(

)

2

3

dc

c

b

a

dc

sn

V

S

S

S

V

v

−

+

+

=

(1)

and

ng

sn

sg

v

v

v

+

=

,

(2)

where S

a

, S

b

and S

c

represent switching states of inverter

bridges. v

ng

changes slowly compared to the variation of v

sn

if

the operating frequency of PFC is quietly low, and most of

dv/dt comes from PWM switching operation in (1) as shown
in Fig. 2(a), where the system parameters and operating con-
ditions are given in Table I. Fast dv/dt of the common-mode
voltage excites parasitic components such as junction-to-
heatsink capacitances in power semiconductor switches, line-

to-ground capacitances C

lg

and stator-to-ground capacitances

C

sg

. The high frequency current as shown in Fig. 2(a) is gen-

erated on the path indicated by dotted line in Fig. 1. As well
as the high frequency normal-mode current, the common-
mode current is drawn from the ac power source and can be
emitted by loops made by safety earth wires or ground return
paths. Especially loops for the common-mode current are
rather larger than those of the normal-mode current, which
act as good antennas for radiated EMI [4].

B. Inductive Load Current Switching

Fig. 2(b) shows the waveform of inverter input current

when the induction motor or any other inductive load machi-
ne is connected in Fig. 1. The inverter input current i

inv

is de-

termined by the switching state of inverter as (3).

cs

c

bs

b

as

a

inv

i

S

i

S

i

S

i

+

+

=

,

(3)

where i

as

, i

bs

and i

cs

are phase currents of the inverter. The

common-mode voltage v

sn

is depicted in Fig. 2 to represent

the switching state of the inverter. During the active voltage
applied in a PWM period, the inverter current follows the
phase current with respect to the switching state, and is
clamped to zero when the zero voltage vector ([S

a

, S

b

, S

c

] =

[0,0,0] or [1,1,1]) is applied in the PWM period. This inverter
input current waveform is analogous to that of the dc-dc Buck
converter, which has large harmonic contents of the high fre-
quency current. Large dc-bus capacitors are supposed to han-
dle this high frequency current, and electrolytic capacitors are
used for that purpose. But they are relatively large effective
series resistance (ESR) and inductance (ESL), and have the
limited high-frequency capability. Thus some portion of the
high frequency inverter input current should be drawn from
the ac power source, and it can be a source of conducted EMI
[5]. Unlike the dv/dt problems, this is strongly dependent
upon the operating condition and the voltage modulation
method. It is well known that the rms value of the inverter
input current remains same regardless of the modulation
method if the switching frequency of PWM inverter is much
higher than that of output voltage and thus the magnitude of
the output current ripples can be negligible with respect to
that of the fundamental output current [6]. But the distribu-
tion of the frequency components of the inverter input current
in the frequency domain are clearly expected to be varied
with different modulation methods as in the case of [7].

III. M

ITIGATION OF

C

ONDUCTED

EMI

Fig. 3 shows the PWM inverter system considered in this

paper. It is a compressor drive unit of an air conditioner, and
its filter stages are modified in this paper for EMI analysis.
PFC in this system is operated at the ac line frequency. Its
switch is turned on at the moment of zero crossing point of
the input voltage, and remained as the on-state during some
fixed time only for reduction of the harmonic pollution. Thus
the inductor in PFC should be very large in order to allow the
continuous conduction of input current, and 20[mH] of dc
inductor with laminated silicon-steel E-I core is used in this
system. From the ac input terminal to dc-bus capacitor, all
components including control circuits and SMPS circuits are
installed within a single PCB board and the IPM board for the
IGBT inverter is connected to dc-bus capacitor in the PCB
board by wires for the convenience of the system layout. A
snubber capacitor C

s

is installed on the IPM board in order to

suppress the voltage spike at the power devices. A PFC

(a)

1

g

i

sn

v

g

i

(b)

inv

i

as

i

sn

v

Fig. 2. High-frequency current generated by PWM inverter

operations. (a) Leakage current by common-mode voltage

(Top: motor leakage current i

g1

, 200[mA/div]. Middle:

common-mode voltage v

sn

, 100[V/div]. Bottom: total leak-

age current i

g

, 50[mA/div]. Time: 5[

µ

s/div].), (b) Inductive

load current switching (Solid line: inverter input current

i

inv

, 2.5[A/div]. Dashed line: a-phase current i

as

, 2.5[A/div].

Common-mode voltage v

sn

, 100[V/div]. Time:

200[

µ

s/div].)

0-7803-7114-3/01/$10.00 (C) 2001

LISN 50

Ω

-50

µ

H

s

50

Ω

50

µ

H

1

µ

F

0.1

µ

F

2

CM

L

3

x

C

y

C

Inverter

Motor

ACEF

1

x

C

2

x

C

4

x

C

s

C

Fig. 3. Configuration of experimental system.

inductor and a 3.7[kW] induction motor are connected to PCB
board by wires, and a ferrite bead (not included in Fig. 3) is
installed on inverter output cables to the induction motor for
the reduction of common-mode radiated EMI.

EMI filters are usually located between the ac input termi-

nal and the diode rectifier, and provide the attenuation of high
frequency signals in order to suppress the conducted EMI
emission. In most cases, noise suppression is implemented
using LC low-pass filters, and the adjustment of the insertion
loss is done by changing their cutoff frequencies and in-
creasing the number of filtering stages. But there are some
limitations in implementing filter stages. In case of normal-
mode LC filter, the filter inductor should be able to handle
the rated current of the system without the saturation of its
magnetic core. With this fact the core size should be in-
creased if the larger inductance is required, which are directly
related to the system cost and volume. The value of the nor-
mal-mode filter capacitor (X-capacitor) is also important in
the low voltage system such as switched-mode power sup-
plies (SMPS), because the power factor of the system can be
degraded by X-capacitor. But they are usually negligible in
high voltage system, and the adjustment of the normal-mode
insertion loss should be done by the increase of X-capacitor
without changing the filter inductor to some extent. There are
similar limitations in implementing the common-mode filter.
The value of the common-mode filter capacitor (Y-capacitor)
is limited to several nF because of the safety purpose [8]. It is
possible to make the common-mode filter inductor provide
large inductance because the flux induced by the common-
mode current is usually small. Thus the adjustment of the
insertion loss is usually done by the increase of the common-
mode inductor size. Usually a common-mode inductor con-
tains not only the common-mode inductance but also the
normal-mode inductance resulting from its leakage induc-
tance. Normal-mode filtering is also done by the common-
mode inductance only with the aid of X-capacitor in small
power systems, but the separate normal-mode filter inductor
is required for greater attenuation in larger power systems.
Unlike the case of the filter design of signal processing appli-
cations, there are not only LC limitations mentioned above,
but also the difficulties in predicting the performance evalua-
tion of the designed filters because of unpredictable parasitic

effects from components and system layout, which makes the
high frequency modeling of the system be almost impossible.
There are several research results in which various active
filtering techniques are tried [9 ~ 12]. The purpose of the ac-
tive EMI filter is to provide greater attenuation of high fre-
quency signals under these limitations. Especially in case of
the active common-mode EMI filter, the signal to be sup-
pressed is the relatively high frequency signal compared to
the normal-mode current ripple in SMPS applications [9, 10],
and the compensation circuit should be fast enough to elimi-
nate it without any phase delay.

In this section, the variation of conducted EMI spectrum is

investigated with respect to the variation of the filter stages.
In order to suppress the common-mode noise effectively, a
new active common-mode EMI filter is introduced. Detailed
operating conditions and system parameters of the given sys-
tem are listed in Table I. A LISN (Line Impedance Stabiliz-
ing Network) is connected between the ac power source and
the inverter system in order to provide a stable 50[

Ω] imped-

ance to the inverter system in frequency range of 150[kHz] ~
30[MHz] [2, 8]. Measuring the total conducted EMI level of
the system, one of 50[

Ω] resistors in Fig. 3 is used as a dum-

my resistive load, and the other as the input impedance of the
spectrum analyzer. The total conducted EMI is the sum of the
normal-mode EMI and common-mode EMI, and one cannot
tell the origin among different noise sources apart only with
the result of the total conducted EMI spectrum. A differential
mode rejection network (DMRN) is used in order to measure
the conducted common-mode noise separately as shown in
Fig. 4 [13].

A. Active Common-mode EMI Filter (ACEF)

Fig. 5 shows the proposed active common-mode EMI filter.

The proposed circuit is based on the ripple current elimina-
tion technique measuring the source current ripple [9 ~ 11].

BNC

BNC

Spectrum

Analyzer

LISN

LISN

50

Ω

50

Ω

16.7

Ω

16.7

Ω

16.7

Ω

Fig. 4. Differential Mode Rejection Network (DMRN) for

single-phase ac power source[13].

12V

300V

b

r

c

C

CM

L

0

C

0

C

Fig. 5. Proposed active common-mode EMI filter(ACEF).

0-7803-7114-3/01/$10.00 (C) 2001

In order to minimize the phase delay caused by the filter cir-
cuit the single stage of the push-pull amplifier as in [11] and
[12] has been designed. The filter circuits in [11] and [12]
utilizes the dc bus voltage as the power source of the push-
pull amplifier. In this case, both transistors should handle the
full dc voltage, but the high voltage pnp transistor is hard to
obtain in the market and its manufacturing is not favorable as
the dc bus voltage increases. The proposed circuit uses the
additional 12[V] voltage source that is used as the control
power supply in the inverter system. Due to the use of the
low voltage power supply, it is possible to use faster devices
in the filter circuit and it can be applicable in any inverter
application with any dc bus voltage compared to [11]. Also
the proposed circuit does not care about the common-mode
voltage generated but tries to eliminate the common-mode
current compared to [12].

The source common-mode current makes the high frequen-

cy ripple flux in the common-mode choke depicted in Fig. 5,
which makes the high frequency voltage at an auxiliary
winding. The high frequency voltage is then converted to the
high frequency current signal by the base resistor r

b

and the

input impedance of the push-full amplifier. It excites the base
current of the amplifier and then the amplifier provides the
low impedance path for the high frequency signal between
the inverter and the motor with respect to the sensed signal.
In order to provide the high frequency current path between
the dc bus and the control power supply, two coupling ca-
pacitors C

0

are used. These coupling capacitors should be

small enough to isolate the 12[V] power supply from the dc
link at low frequency. Thus the low-frequency common-
mode current is not suppressed with the proposed circuit,
where its path is blocked by the coupling capacitor. Another
coupling capacitor C

c

should be implemented in order to

provide the low impedance path between the earth ground
and the output of the filter circuit while blocking the path of
the low-frequency signal. Overall the series impedance of C

0

and C

c

should be limited in order to keep the supply voltage

and to withstand the high voltage test.

B. Mitigation of Common-mode EMI

Fig. 6 shows the conducted EMI spectrum of the system

without any filtering elements between ac input terminal and
diode rectifier in Fig. 3. The peak detector is used in the EMI
measurement [2, 8], and each waveform is video-averaged for
10 scans. The EMI characteristic above 10[MHz] strongly
relies on the experimental layout of the LISN and the EUT,
and its filtering is not an easy job without the fundamental
redesign of the system. Thus it will not be discussed in this
paper that focuses on the filtering technique. Both conducted
EMI waveforms in Fig. 6(a) and (b) continuously droop be-
tween the frequency bandwidth from 150[kHz] to 30[MHz].
The waveform of the total EMI in Fig. 6(a) is slightly greater
than that of the common-mode EMI and one cannot tell that
which noise of the normal- and common-mode EMI is domi-
nant. The common-mode leakage current waveform is shown
in Fig. 6(c) that is directly returned to the source.

A common-mode inductor L

CM

of Fig. 5 is installed in the

system of Fig. 6 without the push-pull amplifier circuit. As
can be seen in Fig. 7(a) and (b), the total EMI is also reduced,
but the reduction ratio is not same as that in the common-
mode EMI because the leakage inductance of the common-
mode inductor is just 7.5[

µ

H]. With the increased impedance

of the common-mode current path, the peak value of the
leakage current is reduced as shown in Fig. 7(c) and also its
frequency is more sluggish than that in Fig. 6(c). The differ-

(a)

(b)

(c)

Fig. 7. Conducted EMI when L

CM

is installed only. (a) To-

tal EMI, (b) Common-mode EMI, (c) Leakage cur-

rent(200[mA/div], 5[

µs/div])

(a)

(b)

(c)

Fig. 6. Conducted EMI when no EMI filter installed. (a)

Total EMI, (b) Common-mode EMI, (c) Leakage cur-

rent(200[mA/div], 5[ms/div]).

0-7803-7114-3/01/$10.00 (C) 2001

ence between the total EMI and the common-mode EMI is
less than 10[dB

µ

V] and is not enough to isolate the mitigation

effort of the normal-mode EMI.

Fig. 8 shows the effect of the proposed ACEF. High volt-

age capacitors of 10[nF] are used for output and coupling
capacitors. The push-pull amplifier circuit is added in the
system of Fig. 7 and it suppresses the entire common-mode
EMI waveform by 10[dB

µ

V] at least. Fig. 8(b) shows that the

proposed circuit still provides fine attenuation until 10[MHz].
In Fig. 8(c), most of high-frequency signals in the returning
leakage current i

g

are disappeared by the proposed circuit,

and there is only some low frequency signal (less than
100[kHz]) that cannot be suppressed by the circuit. Because
the ACEF forces the motor leakage current i

g1

kept in the

system and provides the low impedance path for that, the
circulating leakage current between the inverter and the mo-
tor is slightly increased compared with that in Fig. 6 (c). The
total EMI is also reduced and now the common-mode EMI is
clearly less than the total EMI and the dominant EMI source
is the normal-mode EMI. At this moment, the level of the
common-mode EMI is the ultimate low limit of the total EMI
if the appropriate filtering is provided for the normal-mode
EMI. But the level of the common-mode EMI is still close to
the limit line and more attenuation is required.

An additional passive filter stage is added to the system of

Fig. 8. A common-mode inductor L

CM2

and small Y-capacitors

C

y

are used. After adding the additional common-mode filter,

the level of the common-mode EMI is quite less than the
limit line in Fig. 9(b) but the total EMI at low frequency is
slightly increased in Fig. 9(a) compared with that in Fig. 8(a).
It follows that the small leakage inductance of L

CM2

with Y-

capacitors forms a normal-mode filter which resonant fre-

quency is around 800[kHz]. Because the ferrite core has the
high Q-factor, it makes the resonant peak around that fre-
quency and the normal-mode EMI is boosted in that region.

With the aid of the ACEF and the additional passive com-

mon-mode filter stage the common-mode EMI is much re-
duced but the total conducted EMI is not reduced to meet the
limit line. The change of the total EMI spectrums shown in
Fig. 6 ~ 9 shows that the dominant source of the conducted
EMI is in the form of the normal-mode EMI after the appro-
priate common-mode mitigation, and the strategy for the re-
duction of the normal-mode noise is required in order to pro-
ceed the filter design effectively.

C. Mitigation of Normal-mode EMI

When common-mode chokes are implemented in the pre-

vious section, two small normal-mode inductors are intro-
duced as the leakage inductance of them. Thus it is possible
to build the multi-stage normal-mode filter. Total conducted
EMI spectrums according to the increase of the normal-mode
filter stages are shown in Fig. 10 ~ 12.

At first C

x1

and C

x2

are added to the system of Fig. 9 in Fig.

10. These capacitors are required in order to minimize the
coupling of the common- and normal-mode noises caused by
the circuit asymmetry of the PFC and also to keep the nor-
mal-mode ripple current circulating within those capacitors.
Each capacitor is the polypropylene capacitor of 330[nF].
The entire total EMI is much reduced due to the 2

nd

order

low-pass filter formed by those X-capacitors and the leakage
inductance of common-mode chokes, and it is lower than the
limit line almost frequency range except 1[MHz]. However
there is little margin from the limit line below 1[MHz] and
the EMI level even exceeds at 1[MHz]. Thus more insertion
loss is required in order to secure enough margins.

In order to construct the 4

th

order multistage filter for the

normal-mode EMI an additional X-capacitor C

x3

is inserted

between L

CM

and L

CM2

, and the total EMI waveform is meas-

ured as in Fig. 11. Unlike the expectation, the total EMI
waveform is not decreased at all compared with the result of
Fig. 10. This is because the source impedance, the LISN im-
pedance, is not so small compared to the input impedance of
the normal-mode filter stage, and thus the filter stage cannot

(a)

(b)

(c)

g

i

1

g

i

Fig. 8. Conducted EMI when ACEF is installed. (a) Total

EMI, (b) Common-mode EMI, (c) Upper: motor leakage

current i

g1

, Lower: source leakage current i

g

.

(a)

(b)

Fig. 9. Conducted EMI when the additional passive filter

stage is added to the system of Fig. 8. (a) Total EMI, (b)

Common-mode EMI.

0-7803-7114-3/01/$10.00 (C) 2001

provide the enough attenuation. That is, the front-end imped-
ance of this filter stage is 33[

Ω] at 150[kHz] (the leakage in-

ductance of L

CM2

), which is greater than the normal-mode

impedance of the LISN, 100[

Ω]. In this case the additional X-

capacitor should be provided near the source impedance in
order to provide enough insertion loss with the mismatched
impedance condition. After adding the front-end X-capacitor

C

x4

the total EMI is much lower than the limit line and ap-

proaches closely to the ultimate limit of the attenuation, the
level of the common-mode EMI as shown in Fig. 12. Alt-
hough the additional C

x4

slightly increases the total EMI be-

tween 2 ~ 10[MHz], it does not matter because the total con-
ducted EMI in that region is pretty low enough.

Until now, the total EMI has been effectively suppressed

with the reduction of the normal-mode EMI after reducing
the common-mode EMI first. But experimental results reveal
that additional filtering elements may not help the perfor-
mance of existing filters as shown in Fig. 11. Moreover as the
resonant peak shown in Fig. 10 ~ 12, there can be some un-
expected results from the presence of parasitic components in
filtering elements and PCB layout, which makes the analysis
and application of filtering techniques difficult.

IV. D

ISCUSSION

Fig 13(a) shows the connection of earth wires in the given

system. A safety earth wire is connected at the earth terminal
of the drive system including EMI filters, and the other wire
is routed to the earth terminal of the LISN. As mentioned
before, the PWM inverter is the obvious common-mode volt-
age source of v

com

, and the impulsive leakage current is gen-

erated as shown in Fig. 2(a), 6(c), 7(c) and 8(c). The sum of
currents in paths ‘ 2’ and ‘ 3’ in Fig 13(b) affects the common-

mode EMI voltage v

CM

across the 25[

Ω] resistor that is the

parallel resistance of the LISN. The proposed ACEF provides
the low impedance path ‘ 0’ as indicated in Fig. 13(b) in order
to minimize i

CM

. The remaining high frequency current is

captured by the Y-capacitor. In this system the ACEF lowers
the source leakage current, which keeps the additional com-
mon-mode choke L

CM2

from being saturated in case of the

excessively large leakage current. Because the proposed cir-
cuit is not working at the low frequency, the low frequency
leakage current is only dependent upon the Y-capacitor and
the system parasitic capacitance. That is, the proposed circuit
does not increase the low frequency leakage current that is
different from the behavior of the Y-capacitor. The additional
common-mode choke L

CM2

blocks the high-frequency leakage

current that helps the internal circulation by the Y-capacitor
(path ‘ 1’ ) and ACEF (path ‘ 0’ ). In the 2

nd

order low-pass fil-

ter formed by L

CM2

and C

y

, there can be some oscillatory cur-

rent of the low frequency, which results from the high Q fac-
tor of the ferrite core. It can be reduced by the use of the
lossy material of the magnetic core but the inductance can be
lowered because the relative permeability is much less than
that of ferrite. Or some proper damping method can be ap-
plied such as in [14] and [15] without worsening the total
attenuation capability.

As far as the normal-mode EMI is concerned, the major

source of the normal-mode EMI is the input current of the
PWM inverter and its propagation of the normal-mode EMI
can be represented as shown in Fig. 13(c). The input current
of the PWM inverter is highly pulsating and can contain more
high frequency components than the output current of the
PWM inverter. Some of high frequency normal-mode cur-
rents are absorbed by the snubber capacitor C

s

that is attached

to the IPM board. But its attenuation performance is limited
by its capacitance, and this cannot be increased as the in-
crease of the capacitance of the dc-bus electrolytic capacitor.
The electrolytic capacitor has relatively large ESR and ESL,
and is expected not to handle the high frequency normal-

Fig. 10. Total conducted EMI with C

x1

and C

x2

.

Fig. 11. Total conducted EMI of multistage filter.

Fig. 12. Final Result of total conducted EMI.

(a)

s

Inverter

1

g

i

g

i

2

g

i

LISN

Motor

EMI

Filter

−

+

AC

v

Drive System

(b)

s

−

+

CM

v

y

C

2

+

−

com

v

1

3

2

0

2

CM

i

CM

i

ACEF

(c)

−

+

DM

v

s

C

dc

C

inv

i

4

x

C

Fig. 13. Propagation of conducted noise.

(a) Layout of measurement, (b) Propagation of common-

mode EMI, (c) Propagation of normal-mode EMI.

0-7803-7114-3/01/$10.00 (C) 2001

mode EMI effectively. Thus most of high frequency normal-
mode currents should be blocked by external normal-mode
filtering elements. Stray inductances and capacitances can
make some unpredictable modes that worsen the attenuation
performance of the normal-mode filter as shown in Fig. 10
and Fig. 11. The 1[MHz] resonant peak has been made by the
resonance between the snubber capacitor C

s

and the dc-bus

wire. The length of the total dc-bus wire is about 20[cm],
which makes the stray inductance of 250[nH]. It is not pre-
sent in the common-mode EMI waveform as shown in Fig.
9(b) and can be suppressed only by the normal-mode filtering
elements as shown in Fig. 12. If the coupling of the common-
mode choke is very high, which is the usual case, its leakage
inductance can be rather smaller than that of the input imped-
ance of the system and the source impedance as in the case of
this system. Thus the desired attenuation cannot be obtained
because of the load effect of the filter. In order to minimize
this effect, appropriate X-capacitors should be installed in
front of the source and the system [8, 15]. In this system, X-
capacitors, C

x2

and C

x4

are functioning in this manner. After

minimizing the load effect the desired insertion loss can be
obtained as in the case of Fig. 12.

Another aspects to be considered is that the asymmetry of

the PFC may increase the interference between the common-
mode EMI and the normal-mode EMI, which makes the filter
design more difficult [7].

V. C

ONCLUSION

In this paper, the characteristics of the conducted EMI of

the PWM inverter for household appliances have been pre-
sented and its mitigation efforts are introduced. A new active
common-mode EMI filter is proposed in order to produce the
effective insertion loss in the common-mode circuit without
the limitation of the passive filter. The conducted EMI spec-
trums have been measured with changing and verifying the
effect of each filtering elements. The proposed circuit
provides good attenuation results without increasing the low-
frequency leakage current. The spectrums of the common-
mode EMI have been separately measured in order to distin-
guish the type of the dominant noise source, which enables
the contribution of the normal-mode EMI over the total EMI
to be examined. From these results the origin and character-
istics of the noise source have been analyzed, and the cou-
pling paths of the conducted EMI have been identified which
leads to the way of the effective mitigation techniques of EMI.

R

EFERENCES

[1] T.Williams, EMC for Product Designers, 2

nd

ed., Newnes, 1996.

[2] EN 55 014 : 1993, “Limits and Methods of Measurement of

Radio Disturbance Characteristics of Electrical Motor-operated
and Thermal Appliances for Household and Similar Purposes,
Electric Tools and Electric Apparatus”

[3] L. Rossetto, P. Tenti and A. Zuccato, “Electromagnetic Com-

patibility Issues in Industrial Equipment,” IEEE Industry Appli-

cations Magazine, Nov./Dec. 1999, pp. 34 ~ 46.

[4] G. Skibinski, R. Kerkman, and D. Schlegel, "EMI Emissions of

Modern PWM ac Drives," IEEE Industry Applications Magazine,
Nov/Dec 1999, pp. 47 ~ 81.

[5] S. Chen, “Generation and Suppression of Conducted EMI from

Inverter-Fed Motor Drives,” in Conf. Rec. of IEEE IAS, 1999, pp.
1583 ~ 1589.

[6] J. Kolar, T. Wolbank and M. Schroedl, “Analytical Calculation

of the RMS Current Stress on the DC Link Capacitor of Voltage

DC Link PWM Converter Systems,” in IEE Conf. Rec. of Elec-

trical Machines and Drives, 1999, pp. 81 ~ 89.

[7] L. Rossetto, S. Buso, and G. Spiazzi, "Conducted EMI Issues in

a 600W Single-Phase Boost PFC Design," IEEE Trans. Ind. Ap-

plicat., vol. 36, no. 2, Mar/Apr 2000, pp. 578 ~ 585.

[8] L. Tihanyi, Electromagnetic Compatibility in Power Electronics,

IEEE Press, 1995.

[9] L. Lawhite and M. F. Schlecht, “Design of Active Ripple Filters

for Power Circuits Operating in the 1-10MHz Range,” IEEE

Trans. Power Electron., vol. 3, no. 3, Jul 1988, pp. 310-317.

[10] N. K. Poon, J. C. P. Liu, C. K. Tse and M. H. Pong, “Tech-

niques for Input Ripple Current Cancellation: Classification and
Implementation,” IEEE Trans. Power Electron., vol. 15, no. 6,
Nov 2000, pp. 1144-1152.

[11] I. Takahashi, A. Ogata, H. Kanazawa, “Active EMI Filter for

Switching Noise of High Frequency Inverters,” in Conf. Rec. of

IEEE PCC-Nagaoka `97, 1997, pp. 331-334.

[12] S. Ogasawara and H. Akagi, “Circuit Configurations and Per-

formance of the Active Common-Noise Canceler for Reduction
of Common-Mode Voltage Generated by Voltage-Source PWM
Inverter,” in Conf. Rec. of IEEE IAS, 2000, pp. 1482-1488.

[13] M. J. Nave, “A Novel Differential Mode Rejection Network for

Conducted Emissions Diagnostics,” IEEE National Symposium
on Electromagnetic Compatibility, 1989.

[14] S. Ogasawara and H. Akagi, "Modeling and Damping of High-

Frequency Leakage Currents in PWM Inverter-Fed AC Motor
Drive Systems," IEEE Trans. Ind. Applicat., vol. 32, no. 5,
Sep./Oct. 1996, pp. 1105-1114.

[15] M. J. Nave, Power Line Filter Design for Switched-Mode Pow-

er Supplies, Van Nostrand Reinhold, 1991.

T

ABLE

I.

S

YSTEM PARAMETERS OF GIVEN SYSTEM

.

Power rating

Single-phase 220[V], 1.5[kW].

PWM inverter

60[Hz] fixed output frequency,

discontinuous PWM with 2.5[kHz] switching

frequency.

Switching device

Mitsubishi DIP-IPM PS21205

Load machine

3-phase 3.7[kW] induction motor

PFC

Line frequency switching

for harmonic reduction.

Common-mode choke

L

CM

: 2[mH], N = 13, ferrite torroid,

7.5[

µH] of leakage inductance

L

CM2

: 6[mH], N = 27, ferrite torroid,

35[

µH] of leakage inductance

Y-capacitor and

coupling capacitor

High voltage capacitor (2[kV])

C

y

: 2.2[nF],

C

c

and C

0

: 10[nF]

X-capacitor

Polypropylene capacitor

C

x1

, C

x2

and C

x4

: 470[nF],

C

x3

: 680[nF]

C

s

: 100[nF]

Push-pull amplifier

NEC 2SC3840 (pnp), 2SA1486 (npn)

0-7803-7114-3/01/$10.00 (C) 2001