Generation of High-Frequency Common Mode Currents in Machines of Inverter-Based Drive Systems
Generation of High-Frequency Common Mode Currents in Machines of Inverter-Based Drive Systems
A. Muetzel' A. Binder
DARMSTADT UNIVERSITY OF TECHNOLOGY
Landgraf-Georg-Strasse 4,
64283 Darmstadt, Germany
Tel.: +49 / (6151) 16.21.67, Fax: +49 / (6151) 16.60.33
- -
E-Mail: m abinderWewAu-darsttde
URL: -darmstadt.de/ewhtml
1now with the University of Wisconsin-Madison, WI, USA
Acknowledgements
This work was supported in part by the Arbeitsgemeinschaft Industrielle Forschung (AiF), Germany,
Grant 12584N.
Keywords
<<Adjustable speed drive>>, <<EMC/EMI>>, <<Modeling>>, <<Simulation>>, <<Variable speed drive>>
Abstract
The additional HF (High Frequency) common mode current of inverter-based drive systems can cause pa-
rasitic effects such as HF bearing currents and wide band EMI (Electro-Magnetic Interference). This pa-
per presents a straightforward approach to prediction of the HF common mode current based on the
dVLgldt at the motor terminals. Only design parameters of the machine are used. No additional
measurements are required. The HF common mode current generation is described by a distributed-
parameter network. This model allows anticipation of the HF common mode current in order to better
predict possible parasitic effects. For a given frame size, the number of parallel winding branches has a
strong influence on the amplitude of the HF common mode current.
Introduction
If no additional measures are taken, a fast switching IGBT-inverter (Insulated Gate Bipolar Transistor) is a
common mode voltage source vcorm that exposes the motor terminals to high dvldt. This causes an addi-
tional high frequency common mode current icorm, mainly because of the interaction of the high dvldt at the
motor terminals and the capacitance between motor winding and frame Cwf. The frequencies f of these
currents icorn range from 100 kHz up to several MHz. The parasitic effects that can be caused by the flow
of these common mode currents include wide band EMI, interference with the ground fault protection sys-
tems in industrial facilities, induction of high frequency circulating bearing currents, and bearing currents
due to rotor ground current flow (Fig. 1) [1-9].
Due to the occurrence of these parasitic phenomena, prediction of the HF common mode current icom is
important to anticipate the possible endangerment of a drive-system, for example due to bearing currents.
Furthermore, anticipation of interference with ground protection systems in industrial facilities and with
other signals can be important. Hence, a general view of the understanding of HF common mode current
generation is given here.
The sinusoidal, damped shape of the measured HF common mode current motivates use of a simplified li-
near RLC-network to model the motor-cable configuration similar to what is done in [10] and [11]. In this
LeCro_ .1
-
r-I-
Common
___ ___ __ ___
....
.. .... __ _ _ .... UIT1tllLt Rcs~~~~~~~~~~~~
.\ ,. mode
11
current
20 A/Div
H-H H-H~~~~~~ curen
1,1
llt^ 9,44
73 t
I - i4;-+ -<
__ 4
! J
Bearing
E, "L;C1116 5-I-
currents
currents NDE/DE
NDE .HLv,*g H-- XH- f~H~ HH
._._._.
-
-. 5 A/Div
DE Rotor
5 A/Div ground
current
10 A/Div
(a) 1 p/sDiv (b) 1 ps/Div
Fig. 1: a) Circulating bearing currents at motor speed n = 3000 /min, b) Bearing currents due to rotor
ground currents at motor speed n = 15 /min at rotor grounded on the drive-end side; squirrel-cage induc-
tion motor, frame size 400 mm, 500 kW rated power, bearing temperature Tb 70°C (DE: Drive End,
NDE: Non-Drive End)
model, L is related to the cable only, and R and C are related to the motor only. Using this approach as a
starting point, a relationship between the dvLgldt of the line-to-ground voltage at the motor terminals VLg
and the stator winding-to-frame capacitance Cf is derived to estimate the common mode current icom (1)
[12], where dvLgldt can be correlated to the dvcom/dt of the common mode voltage vcorn.
dVL C
dt
(1)
This approach implies that icorn is not a conduction current, but a mere displacement current. It was shown
that (1) is only reliable for a rough estimate of the common mode current [12]. Fig. 2 shows two mea-
sured waveforms of the common mode current icom, along with the line-to-ground voltage at the motor
terminals VLg of two four-pole induction motors with 110 kW rated power. If icom was a merely capacitive
current, it would be almost constant along the rise of VLg. However, icom reaches its maximum value before
VLg reaches its maximum value. The described observation of the correlation between icorn and VLg is also
true for the many other investigated configurations of inverter-based drive systems with 11 kW through
500 kW rated power [13]. Therefore, inductive elements in the path of the common mode current inside
the machine cannot be neglected.
LeCro4 LeCro
Common Common
mode mode
current current
10 A/Div 10 A/Div
H-H-S *r I-H-H- H-I-H -H H-I-
Line-to- Line-to-
lll ll ll .ll .ll ..,,1.,,1..,,.1 ground ground
voltage voltage
0.5 kV/Div .. ns,,</niN 0.5 kV/Div
(a) (b)
nV...S .tN/L-nV
-/ni 1)
V..) ItNLfv
Fig. 2: Measured common mode currents and line-to-ground voltages, two four-pole 400 V induction-mo-
tors a) and b) with 110 kW rated power. Both motors are supplied with the same voltage source inverter
(Vdc = 560 V) via a shielded motor cable with 80 m length.
The superposition of ground currents resulting from two switching instants of the IGBT-inverters at differ-
rent phases is not considered, as the time-lag T, between two switching instants is larger than the charac-
teristic time constant of the HF common mode current. For example, if the time lag T, between the swit-
ching instants of the three phases is assumed to be the same, at 3 kHz switching frequency the time lag is
Tc = 1/3.(3000)-'s = 0.1 1ms, whereas the characteristic time constant of the HF common mode current is
typically in the range of some 10 js.
The common mode voltage as given by the individual line-to-ground voltages vugj vvgS and v,g is used as a
starting point to describe the phenomenon (2).
com
vug + vvg + vWg (2)
3
The three line-to-ground voltages vug, vvg, and vwg are replaced by the product of the phase currents i, iv,
and iw, and the common mode impedance Zo ("zero sequence impedance") which is to be determined, and
is the same for all three phases (3). Then, the HF common mode current icom can be computed from the
line-to-ground voltages at the motor terminals (Fig. 3).
Equ. Distributed>
VLg (2) Parameter x 3 io
~~VCOM Network 10
quency of MHz.
The model contains a distributed parameter network of qserie, segments (a), "a" parts in parallel, to con-
sider the overall phase winding. The number of segments (a) in series equals the number of coils in series
qseries where each segment (a) represents one coil. The value of qseries is determined from the number of
poles 2p and the number of parallel winding branches "a"c of the modeled machine (4). One different
segment, segment (b) is added to consider the parameters of the path of the common mode current that
are not related to specific coil (Fig. 4). The segments are explained in the following.
a
qseries q a
(4)
EPE
P.3
22005 Dresden
5
ISBN
E PE 0 0 5 - licensed
Authorized D re s de I S B N
usen limited to: Seoul National University. Downloaded on9 January
O - 7 5 825,2023
1 5 - O 8at- 07:03:01
5 UTC from IEEE Xplore. Restrictions apply. P 3
Generation of High Frequency Common Mode Currents in Machines of Inverter-Based Drive Systems MUETZE An:nette
icO/13 segment
segment -a
(a) a parts in
lle
usegm t dpara
I(a)
\g segmetht
.(a) seg ment
q e i s segments-
Fig. 4: Sketch of distributed-parameter network for modeling of HF common mode current generation
Coil Model
Two different circuits are used as segments for the distributed-parameter network (Fig. 5). The first cir-
cuit, resulting in segment (a)ais used to represent an individual (b)~~~~~~~~~~~~~~~~~a
coil of the machine. The second circuite
giving segment (b) is associated to components of the path of the ground current that are not related to
any specfoc coil.
Tcoml icm2 icoml icom2
F = R s LKd_s e>
iLg Rf
VLg l
Cs Cg _ L2 VLg I VLg2
(a) (b}
Fig. 5: Segments of the distributed-parameter network representing (a) one coil of the machine, (b) part of
the common mode current path not related to an individual coil
+ The parameter L ', represents the inductance of the stator winding per coil that acts in parallel to the
series capacitance C. This inductance corresponds approximately to the stator leakage inductance per coil
at 1 MHz frequency. Only the stator leakage inductance is considered, as the high frequency flux is con-
fined to the stator [15, 17]. From a simplified point of view, this can be understood as follows: For an in-
duction machine, the slip Sk of the fundamental air gap field excited by the current harmonic k, where
k >> 1. The rotor seems to be at stand-still, as it is Sk =
(1 s)Ik 1. At sI 1, the air gap fields
caused by the stator current iLk and by the induced rotor harmonic current are in opposition so that the
respective fluxes cancel each other out. Thus, only HF leakage flux needs to be considered.
Next, a simplified equivalent circuit of the capacitances of a machine at high-frequencies is considered.
The capacitance parallel to the stator winding-to-frame capacitance Cwf is given by the series connection
of the stator winding-to-rotor capacitance C,, and the sum of rotor-to-frame capacitance Crf and bearing
capacitance Cb (Fig. 6). Considering the ratios of the capacitances [18], the ratio of the stator and rotor
common mode currents is,comfir,com is obtained from (7), where Cb,NDE = Cb,DE = Cb, Crf 15 Cwr, Cb Cwr,
and Cwf 10 Cwr. As a result, the contribution of the rotor to the generation of the high frequency ground
current is negligible and only the stator leakage inductance L', is considered in the model.
stator winding
1s'com ircom
Cwr
VCOM V
_
co m 3 Cwf
~~~
rotor 4
_ frame
Fig. 6: Parallel current paths in stator and rotor for HF common mode current generation
=,.cor + 1 /I2
=
2/8 <2.8w10 = 28
ir.con Cwr Crf+ 2Cb 3Cwf Cwr 15C\<r + 2Cwr 3Cwf Cwr 3Cwf Cwr
The value of the parameter L'GS is derived as follows: The stator leakage inductance LS, as given in the
equivalent circuit diagram for fundamental frequency, comprises of three components:
0 L(7S,Sl: stator slot leakage inductance,
O L,ys,ew: stator end-winding leakage inductance,
0 L,,S,k: stator harmonic leakage inductance.
0 L6S,Sl: The slot leakage inductance at 1 MHz frequency is calculated using formulas given in [19]. Con-
sidered are Nj turns per coil one upon the other that all carry the same current. Here, several turns
arranged side by side in a slot can be considered as one, where bs,c, is the width of this resulting conductor.
The parameter as (9) represents the inverse of the skin depth 5s (8) of current displacement in the wires
due to eddy-currents. The parameter ,6 (10) is defined as the product of as and the conductor height hs,CU.
The parameter k, describes the decrease of the slot leakage inductance due to eddy-currents. It is a func-
tion of ,8 and the number of conductors upon each other Ns, (11), where the functions ?'(p) and Vf'() are
given by (12) and (13) respectively. For,B> 2, (12) and (13) simplify to (14) and (15).
J
Jf/K
bcu
bS
(8) P ass.cu (10)
<, k ~~~~~~~~~~~~~~~~~~~~~9(f) + VI(f)(Nsl 1) (
_
(9)Nj
'fIo
The values of the parameters discussed in (9) to (15) at 1 MHz frequency are calculated for six squirrel-
cage induction motors with 11 kW through 500 kW rated power. The value of kx is below 0.001 for all
motors. The slot leakage inductance at MHz frequency is therefore reduced to negligible small value a
by HF eddy currents.
O Lys,ew: The value of the end-winding leakage of a machine depends on the shape of the winding over-
hang. Mostly, the value of the end-winding leakage, used for calculations, is derived from experimentally
obtained data at (50... 60) Hz. The exact calculation of the HF flux distribution in the end-winding is
complicated, because the effect of the adjacent coils and phases on each other and the neighboring iron
parts has to be considered. For simplification, it is assumed that the end-winding leakage flux does not
change significantly with the frequency, because the winding is not embedded in iron.
O Ls,,k: The harmonic leakage accounts for the self-induced voltage in the winding due to air-gap field
space harmonics. The space harmonics result from the non-sinusoidal flux distribution along the circum-
ference of the air gap that is due to the finite number of slots, the phase belt, and the slot openings. At
1 MHz frequency, the air gap flux consists only as zig-zag leakage flux. Calculation of this HF zig-zag
leakage flux was not done. This part of the leakage flux is neglected, because the leakage flux of the win-
ding overhang, which decreases at HF, is fully included on the other hand.
Because of the different described factors that render exact calculation very difficult, the stator leakage in-
ductance at MHz L,S(1MHz) is approximated by one third of the stator leakage inductance per phase at
fundamental frequency L, (16). Therefore, the parameter L, which is the inductance L,,(IMHz) per
coil, is given by (17).
The parameter R ', represents the resistance of the stator winding per coil at 1 MHz frequency. The in-
crease of the resistance with frequency is calculated using again formulas given in [19]. Here, for the part
of the winding that is embedded in the slot, a parameter kr is introduced, which gives the increase of the
resistance of the stator winding due to current displacement, using the parameters a, and ,8 as defined by
(9) and (10). The value of kr is given in (18), where q' (/) and Vz(/) are given by (19) and (20)
respectively. For,B> 2, (19) and (20) simplify to (21) and (22), hence (18) simplifies to (23).
kr =(+/ 13N (18) _17>2:y(fl)P) (21)
3
Due to complicated distribution of the HF flux in the end-winding, the increase of the resistance in the
winding overhang is difficult to predict. Using (23) to estimate the increase of the stator winding resis-
tance embedded in the stator iron, and approximating the part of the winding overhang to have the same
order of magnitude, the stator resistance per coil R at 1 MHz of the mentioned test motors is about 500
'
times larger than the resistance at fundamental frequency per coil Rs*a/qsenes (24).
EPE
E PE 22005 Dresden
0 0 5 - licensed
D re s de I S B N
75815 08 5 P.6
P 6
Authorized usen limited to: Seoul National University. Downloaded on9 January
O - 7 5 825,2023
1 5 - O at
8 -07:03:01
5 UTC from IEEE Xplore. Restrictions apply.
Generation of High Frequency Common Mode Currents in Machines of Inverter-Based Drive Systems MUETZE An:nette
+ The parameter Re considers HF dielectric losses in the insulation. When neglecting these small losses,
infinite current following a voltage pulse with lim(dvldt) -X co would result from the modeling. The resis-
tance Re is not calculated from the dielectric losses, which is difficult to do, but it is set constant to a very
small value of 1 mQ. Thus, the resistance Re is negligible when compared with the other elements of the
equivalent circuit and does not affect the simulation results significantly.
Elements of Segment (b)
Next, the parameters of segment (b) are explained. These parameters are associated with components of
the path of the common mode current and are not assigned to a specific coil.
+ The parameter Rf signifies the damping of the HF common mode current due to the motor frame resis-
tance, including skin effect. The resistance of the path of the common mode current through the stator
lamination is derived in [20] analytically and is in the order of 0.1 Q to a few Q. A according to [1 1], the
HF motor frame resistance in series is in the order of some tens of Ohms, which was obtained empirically.
Therefore, the value of Rf is set constant to 20 Q.
+ The parameter Lg represents the internal inductance of the path of the ground current through the stator
lamination. The method of calculation is discussed in [20].
Table I summarizes the values of the different parameters for the six different squirrel-cage induction mo-
tor test motors with 11 kW through 500 kW rated power.
Table I: Calculated parameters for the squirrel-cage induction motor test motors
Rated Power 11 kW 11 kW 110kW 110kW 500kW 500kW
Motor A Motor B Motor C Motor D Motor E Motor F
Cwf [nF] 1.9 1.8 12.1 9.9 28.0 23.0
2plls a 4111 4112 4111 4112 6113 211 1
Cg [nF] 0.238 0.317 1.008 0.990 2.333 2.875
Cs [nF] 1.107 2.157 3.051 3.935 7.523 4.516
L6S [mH] 6.7 9.0 0.77 0.82 0.25 0.82
RS(1000C) [Q] 1.993 1.551 0.046 0.061 0.021 0.008
L Gsqseriesla [mH] 2.233 3.0 0.257 0.273 0.083 0.273
R'sgserJesla [.] 996.6 775.5 22.8 30.7 10.3 4.1
Lg(1MHz) [mH] 0.043 0.042 0.111 0.116 0.122 0.231
a higher number ofparallel winding branches respectively. As the steep voltage rise at the motor termi-
nals is attenuated as it travels through the coils, the more parallel winding branches the machine has, the
larger part of the winding is exposed to the initial steep voltage rise.
Differences between the simulated and measured waveforms are supposed to be due to aspects that are not
considered in the model, including cross-coupling between different phases and different coils in the win-
ding overhang. Furthermore, the waveforms of the dvLgldt at the motor terminals are subject to a certain
variation. Of course, the values of the parameters have been calculated also with the above mentioned un-
certainty. As voltage and current were not measured at the same time, the voltage used for simulation and
the "real" voltage for the measured current may have differed. However, the simulation of the first current
peak has achieved a satisfactory result.
0 .... .... .... .... ....~ .... .... .... .... ....
-2 _'K._.
-4 '92- 0 if1U
2 4 6 8 4A 10A
(a) (b) us [LS
Fig. 7: (a) Simulated, (b) measured common mode current, four-pole 400 V induction motor A with
11 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 50 m length, a = 1
2
HHHHi HH HH H! HH HH HH Hi
0 J Al
-2
2 4 6 28 4 106 8 10
(a) (b) [us [Is
Fig. 8: (a) Simulated, (b) measured common mode current, four-pole 400 V induction motor B with
11 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 50 m length, = 2 a
10
--;
-10
------------
-~~~~~~~~~~~~~~~~~1
(a)
2 4 6 28 4 6
10 8 10
(b) pS
Fig. 9: (a) Simulated, (b) measured common mode current, four-pole 400 V induction motor C with
110 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 10 m length, = 1 a
!E:=1+
Simulated current / A Measured current / A
40
10
20
5
0
0
-20 .. ... .. ... .. ..
2 4 6 8 10 2- 4 U O 1U
(a) us (b) [LS
Fig. 10: (a) Simulated, (b) measured common mode current, four-pole 400 V induction motor D with
110 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 10 m length, a = 2
Simulated current / A Measured current / A
onX
oiu F
40
0
.. !. ..
-40
(a)
2 4 6
10 8 10
[us (b) [Is
Fig. 11: (a) Simulated, (b) measured common mode current, six-pole 400 V induction motor E with
500 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 50 m length, = 3 a
20
-20
-10
2 4 6 28 4 6
10 8 10
(a) ps (b)
Fig. 12: (a) Simulated, (b) measured common mode current, two-pole 400 V induction motor F with
500 kW rated power, supplied with a 560 Vdc inverter, shielded motor cable with 50 m length, = 1 a
Conclusion
The HF common mode current can be described by a distributed-parameter network based on the design
parameters of a machine. This model allows determination of the HF common mode current in advance
of building the whole drive set-up. Thus, possible parasitic phenomena can be better predicted. For a gi-
ven frame size, a higher number of parallel winding branches results in a larger amplitude of the HF com-
mon mode current. As the steep voltage rise occurring at the motor terminals is attenuated as it travels
through the winding, a larger part of the winding is exposed to the high frequency components of the vol-
tage rise if the machine has more parallel winding branches.
References
[1]. S. Chen, T.A. Lipo, D. Novotny, "Circulating type motor bearing current in inverter drives," Proc. IEEE-
Industry Application Society Annual Meeting, vol. 1, pp. 162-166, 1996.
[2]. J. Ollila, T. Hammar, J. Lisakkala, H. Tuusa, "A new reason for bearing current damages in variable speed
drives," Proc. 7th European Conf. on Power Electronics and Applications (EPE), pp. 2539-2542, Trondheim 1997.
[3]. D. Busse, J. Erdman, R. Kerkman, D. Schlegel, G. Skibinski, "Characteristics of shaft voltage and bearing
currents," IEEE Industry Applications Magazine, vol. 3, pp.21-32, November/December 1997.
[4]. A. von Jouanne, H. Zhang and A. K. Wallace, "An evaluation of mitigation techniques for bearing currents, EMI
and over-voltages in ASD applications," IEEE Transactions on Industry Applications, vol. 34, pp. 1113-1122,
September/October 1998.
[5]. S. Bell, T. J. Cookson, S. A. Cope, R. A. Epperly, A. Fischer, D. W. Schlegel, G. L. Skibinski, "Experience with
variable-frequency drives and motor bearing reliability," IEEE Transactions on Industry Applications, vol. 37, pp.
1438-1446, September/October 1998.
[6]. P. Link, "Minimizing electric bearing currents in ASD systems," IEEE Industry Applications Magazine, vol. 5,
pp. 55-66, July/August 1999.
[7]. A. Binder, R. Aust, A. Schrepfer "Bearing currents - A danger to inverter-fed AC-motors ?," Iron and Steel
Engineer, pp. 47-52, July 1999.
[8]. H. E. Boyanton, G. Hodges, "Bearing fluting," IEEE Industry Applications Magazine, vol. 8, pp. 53-57,
September/October 2002.
[9]. R.F. Schiferl, M. J. Melfi, "Bearing current remediation options," IEEE Industry Applications Magazine, vol. 10,
pp. 40-50, July/August 2004.
[10]. S. Ogasawara, H. Akagi, "Modeling and damping of high-frequency leakage currents in PWM inverter-fed AC
motor drive systems," IEEE Transactions on Industry Applications, vol. 32, pp. 1105-1114, 1996.
[11]. I. Boldea, S.A. Nasar, Electric Drives, CRC Press LLC, Florida, 1999.
[12]. A. Muetze, A. Binder, "High Frequency Ground Currents of Inverter-Fed Squirrel-Cage Induction Motors up to
500 kW," Proc. 10 European Conference on Power Electronics and Applications (EPE), Toulouse, 2003, paper
number 027 (CD-Rom).
[13]. A. Muetze, Bearing currents in inverter-fed AC machines, Shaker Verlag, Aachen, 2004.
[14]. E. Zhong, T.A. Lipo, S. Rossiter, "Transient modeling and analysis of motor terminal voltage on PWM
inverter-fed AC motor drives," Proc. IEEE-Industry Application Society Annual Meeting, vol. 1, pp. 773-780, 1998.
[15]. G. Skibinski, G. Kerkman, D. Leggate, J. Pankau, D. Schlegel, "Reflected wave modeling techniques for PWM
AC motor drives," Proc. Applied Power Electronics Conference and Exposition (APEC), vol. 2, pp. 1021-1029,
1998.
[16]. M. Berth, Elektrische Belastung der Wicklungsisolierung pulsumrichtergespeister Niederspannungsmotoren,
PhD thesis, Technische Universitaet Dresden, VDI-Verlag, series 21, no. 247, 1998.
[17]. G. Suresh, H.A. Toliyat, D.A. Rendusara, P.N. Enjeti, "Predicting the transient effects of PWM voltage
waveform on the stator windings of random wound machines," IEEE Transactions on Power Electronics, vol. 14, pp.
23-29, 1999.
[18]. A. Muetze, A. Binder, "Calculation of Motor Capacitances for Prediction of Discharge Bearing Currents in
Machines of Inverter-Based Drive Systems," Proceedings of the IEEE 5h International Machines and Drives
Conference (IEMDC), to be published, San Antonio, TX, 2005.
[19]. K. Vogt, Berechnung Elektrischer Maschinen, VCH Verlagsgesellschaft mgH, Weinheim, 1996.
[20]. A. Muetze, A. Binder, "Calculation of circulating bearing currents in machines of inverter-based drive
systems," Proceedings of the IEEE 39' IAS Annual Conference, paper no. 041AS19p4, Seattle, WA, 2004.