Analytic Calculation of the DC-Link Capacitor Current

for Pulsed Three-Phase Inverters

Folker Renken
Automotive Systems Powertrain
Siemens VDO Automotive AG
Siemensstrae 12
D-93055 Regensburg
Phone: ++49 941 790 5385 Fax: ++49 941 790 90541 E-Mail: [email protected] http://www.siemensvdo.com



Abstract - For pulsed three-phase inverters with symmetrical load
the capacitor current in the dc-link circuit is analytically
calculated. These calculations can be applied for a constant dc-
voltage as well as for sinusoidal modulated voltages and sinusoidal
currents at the output. The additional load of the dc-link capacitor
caused by harmonic currents of the filter circuit or by switching
transitions of semiconductors is examined, too. At last, the
calculations are examined by practical measurements.

I. INTRODUCTION
Pulse inverters in uninterruptible power suppliers became
generally accepted with the further development of
disconnectible power semiconductors in the last years [1].
The dc-link capacitors contribute substantially to the volume,
to the weight and to the costs of these inverters. For this
reason the necessary expenditure of capacitors must be
determined exactly to prohibit over design if possible. In most
applications the dc-link capacitor effort is dependent on the
load current. For pulsed three-phase inverters with
symmetrical load the capacitor current in the dc-link circuit is
analytically calculated. The power stage of the pulse inverter
is presented in the following figure.

u
E
C
d
L
d i
d
i
E
S
H1
S
H2
S
H3
S
L2
S
L1
S
L3
i
P1
i
P2
i
P3
L
P1
L
P2
L
P3
C
P1
C
P2
C
P3
Y
u
P1-
u
P2-
u
P3-
L
1
L
2
L
3
i
Cd

Fig. 1: Power stage of a pulsed three-phase inverter
The power stage consists out of three inverter-legs, an input
filter circuit with dc-link capacitors and a three-phase filter
circuit on the alternating voltage side.

II. PULSE CONTROL SCHEME OF THE INVERTER
For the calculation of the dc-link capacitor current, three sine-
modulated phase voltages u
P1-
, u
P2-
and u
P3-
are assumed,
whose fundamental part has the same amplitude and a phase
shift angle of 120 to each other. Beyond that a symmetrical
load with any phase shift angle
P1
as well as three-phase
currents i
P1
, i
P2
and i
P3
at the output of the inverter are
presupposed. The dc-link voltage u
d
at the dc-input of the
inverter bridge is assumed as constant.

In figure 2 the pulsed control scheme of the three-phase
inverter is presented. Above the three 120-shifted sinusoidal
control voltages u
s1
, u
s2
, and u
s3
are compared with a higher
frequency triangle shaped modulation voltage u
m
. If one phase
of the sinusoidal control voltages is higher than the
modulation voltage the high-side switch of this phase is
turned-on. In the other case the low-side switch of this bridge-
leg is connected through. In this way the frequency of the

u
s3
u
s1
u
s2
u
m u

1
t
u
P1- G
u
P1-
u
d

1
t
u
P12 G
u
P12
u
d
-u
d

1
t
u
P3- G
u
P3-
u
d

1
t
u
P23
u
d
-u
d

1
t
u
P2- G
u
P2-
u
d

1
t
u
P31 G
u
d
u
P31
-u
d

1
t
90 360 180 270

Fig. 2: Pulse control scheme of the three-phase inverter




triangle shaped modulation voltage u
m
determines the pulse
frequency of the inverter. The fundamental frequency is given
by the frequency of the control voltages.

In the middle of the figure three half-bridge voltages u
P1-
, u
P2-
,
and u
P3-
with the fundamental oscillation frequency are
shown. The lower part of the figure shows the connected
voltages u
P12
, u
P23
and u
P31
between the phases. It can clearly
be seen - that with subtraction of the phase voltages during
one period - two pulses are formed. These voltage waveforms
at the load have approximately sinusoidal output currents in
the individual phases of the inverter as consequence.

For ideal sinusoidal phase currents the generation of the input
current i
d
in a three-phase inverter is presented in figure 3.
The modulation factor amounts thereby to m = 0.8 and the
fundamental phase shift angle is
P1
= 45. Above the output
voltages u
P1-
, u
P2-
and u
P3-
are shown with the sinusoidal
output currents i
P1
, i
P2
and i
P3
of the inverter bridge-legs.
These sinusoidal currents divide themselves in each case in
the high side and low side transistor of the bridge-legs. In the
figure the current pulses are shown, while the high side
transistor of the respective bridge-leg leads. The sum of the
current pulses of all three-phases results in the input current i
d

of the inverter bridge.

u
P1- G
u
P1-
u
d
i
P1 i
P
-i
P

1
t
u
d
i
P
-i
P
u
P2- G
u
P2-
i
P2
90 360 180 270
u
d
u
P3- G
u
P3-
i
P3
i
d
i
d-
i
P
-i
P
i
P

P1
= 45

1
t

1
t

1
t

1
t

1
t

1
t

Fig. 3: Generation at the inverter bridge input current i
d

In figure 4 the pulse generation of the input current i
d
in the
inverter for different fundamental phase shift angles is
presented. From the drawn envelopes it is clearly
recognizable, that the waveforms of the curves after 1/6 of the
fundamental period repeat always. The average value of the
input current i
d-
is positive with small phase shift angles
(motor operation), equal to zero with
P1
= 90 and takes with
larger angles to negative values (generator operation). The
difference between the input current i
d
and their average value
i
d-
flows in the dc-link capacitors [2].

30 180 90 120

1
t
i
d
i
d-
i
d
i
d-
i
P
i
P
i
P
-i
P
-i
P
-i
P
60 150
i
d
i
d-
i
d
i
d-
i
d
i
d-

1
t

1
t

1
t

1
t

P1
= 0

P1
= 90

P1
= 45

P1
= 135

P1
= 180

Fig. 4: Input current i
d
with different phase shift angles
P1


III. CAPACITOR LOAD FOR SINE WAVE CURRENTS
Now the rms-current in the dc-link capacitors is analytically
calculated for sine wave output currents with arbitrary phase
shift angles f
P1
and modulation factors m. For this purpose
the three sinusoidal desired waveforms u
S1
, u
S2
and u
S3
are
presented in the following illustration above. With pulse
width modulation these three sinusoidal desired curves are
compared with the triangle modulation voltage u
m
. The
frequency of the sinusoidal desired voltages determines the
fundamental frequency and those of the modulation voltage
the pulse frequency of the inverter.

It is presupposed that the pulse frequency of the inverter is
very large, the three duty cycle curves follows in principle the
same waveforms as the desired voltages, because of the
comparison with the triangle modulation voltage (figure 5).



With the indicated duty cycle waveforms the output of the
bridge-leg is connected in each case to the positive pole of the
dc-input, so output current flows in the high side transistor of
this bridge-legs.

1
t

1
t
u
s3
u
s1
u
s2
u
m u

1
t
90 150 210
i
P1
i
P2
i
P3
i

P1
t
e
T
P
t
e2
T
P
t
e3
T
P
t
e1
T
P
0.5
1.0
180 120 240

Fig. 5: Duty cycle waveforms and output currents
of the three-phase inverter

Duty cycle waveforms of the phase voltages:

[ ] ) sin( 1
2
1 ) (
1
P
e1
t m
T
t t
+
(1)

[ ] ) 120 sin( 1
2
1 ) (
1
P
2 e
+ t m
T
t t
(2)

[ ] ) 120 sin( 1
2
1 ) (
1
P
3 e
+ + t m
T
t t
(3)
In the figure below the three currents at the output of the
bridge-legs are presented. These currents all have the same
amplitude and are shifted to the voltage by the same phase
angle. (Symmetrical load).
Fundamental currents waveforms of the phases:
( )
1 P 1 P P1
sin

) ( t i t i (4)
( )
1 P 1 P P2
120 sin

) ( t i t i (5)
( )
1 P 1 P P3
120 sin

) ( + t i t i (6)
As shown in chapter II the waveform of the bridge input
currents i
d
repeats after 1/6 of fundamental period. For this
reason the dc-link current in the range 90 = t = 150 can be
calculated. Within this range the following condition applies:

( ) ( ) ( )
P
3 e
P
2 e
P
1 e
1
T
t t
T
t t
T
t t

(7)
Now for the elementary pulse period T
PX
the input current i
d

of the inverter within the range 90 = t = 150 is derived. In
figure 6 the time durations of the different output currents in
the high side switches of the bridge-legs in the elementary
period T
PX
is presented. Thereby altogether four time
durations have to be differentiated.

T
PX
t T
e2 PX
( )
t T
e3 PX
( )
t T
e1 PX
( )

Fig. 6: The different time durations of the output currents
in the high side switches at the pulse period T
PX

In table I the input currents in the different time durations of
the elementary period T
PX
of figure 6 are summarized. If all
high side switches are in the blocking state, the current in the
dc-input is zero. If only high side switch S
H1
is turned on, the
phase current i
P1
flows in the dc-input. If the elements S
H1
and
S
H2
switched on, the input current i
d
can be calculated from
the sum of phase current i
P1
and i
P2
. If all high side switches
are switched on, then the current in the dc input consists out
of the sum of all three-phase currents. With not connected
neutral point the input current becomes zero.

Table I
Current i
d
in the different time durations of T
PX


Now all periods in the range 90 = t = 150 can be divided
according like elementary period T
PX
into time durations. If it
is considered that the sum of the three currents at the output
of the bridge-legs is zero, the time durations in all periods can
be divided into three time durations with the following
different current composition:
1. ( ) ( ) ( ) t t t t T t t
3 e 1 e P 1 3 e
+ (8)
( ) 0
31 d
t i (9)
2. ( ) ( ) ( ) t t t t t t
2 e 1 e 12 e
(10)
( ) ( ) t i t i
1 P 12 d
(11)
3. ( ) ( ) ( ) t t t t t t
3 e 2 e 23 e
(12)
( ) ( ) t i t i
3 P 23 d
(13)
with ( ) ( ) ( ) 0
3 P 2 P 1 P
+ + t i t i t i
Time durations Input current i
d
:
) (
PX 1 e PX
T t T 0
) ( ) (
PX 2 e PX 1 e
T t T t ( )
PX 1 P
T i
( ) ( )
PX 3 e PX 2 e
T t T t ( ) ( )
PX P2 PX 1 P
T i T i +
( )
PX 3 e
T t ( ) ( ) ( )
PX 3 P PX 2 P PX 1 P
T i T i T i + +




The duty cycles can be calculated by the switching on times
regarded to the elementary period T
P
. Using the duty cycle
and the current waveform during the fundamental period the
three ranges results in the following equations:
1.
( )
( ) + + 60 cos
2
3
1
1
P
1 3 e
t m
T
t t

(14)
( ) 0
31 d
t i (15)
2.
( )
( ) 60 cos
2
3
1
P
12 e
t m
T
t t

(16)
( ) ( )
1 P 1 P 12 d
sin

t i t i (17)
3.
( )
( ) t m
T
t t

1
P
23 e
cos
2
3

(18)
( ) ( )
1 P 1 P 23 d
120 sin

+ t i t i (19)
Now with these equations the rms input current I
d
of the
inverter bridge can be determined. First the rms-value at the
elementary pulse periods is determined (parentheses). The
total rms-current can by calculated by integration of the rms-
values of all pulse periods.
( ) [ ]
( )
( ) [ ]
( )

'

+
150
90 P
23 e
2
23 d
P
12 e
2
12 d d
3
t
t
dt
T
t t
t i
T
t t
t i I

(20)
( )
1
]
1

2
1 P P d
cos 4 1
4
3

m
i I
(21)
The dc-component of the current flows over the dc input of
the inverter bridge. This component can be calculated if at
first the average value at the elementary pulse periods is
determined (parentheses). Subsequently the average value
within the range 90 = t = 150 will be determinate.
( ) [ ]
( )
( ) [ ]
( )

'

+
150
90 P
23 e
23 d
P
12 e
12 d d
3
t
t
dt
T
t t
t i
T
t t
t i I

(22)
( )
1 P P d
cos

4
3

m i I
(23)
For ideally smoothed current in the dc-input of the inverter
bridge i
E
= i
d-
, which occurs arises by the geometrical
subtraction of total rms-current I
d
and dc-component I
d-

results in the current of the dc-link circuit capacitors.
2
d
2
d Cd
I I I
(24)
( )
2
1 P
2
P Cd
cos
16
9 3
4
3

,
_

m m m
i I
(25)
The result shows: The rms-value of the current in the dc-link
capacitors is depended on the modulation factor, the
fundamental phase shift angle and the height of output
current. The maximum rms-current in the dc-link capacitors
mounts to:
P P P
2
Cd
65 . 0

46 . 0

12
25
I i i I

(26)
with 0
1 P
and
613 . 0
9
3 10

m

In figure 6 the current in the dc-link capacitors normalized to
the peak value of output current as a function of the phase
shift angle is presented for different modulation factors.

0.0
0.2
0.3
0.4
0.5
0.1
0 45 90 135 180 -45 -90 -135 -180
m = 0.2
m = 0.4
m = 0.6
m = 0.8
m = 1.0
I
Cd
i
P

P1

Fig. 7: Current in the dc-link capacitors as a function
of the fundamental phase shift angle
It can clearly be recognized that with small modulation
factors the load of the dc-link capacitor current is much larger
with ohmic load as with inductive or capacitive load. With
increasing modulation factor m the rms-value of the dc-link
capacitor current for all phase angles
P1
becomes larger.
Only if the capacitor current with more rising modulation
factor reaches its maximum, the current reduces again with
ohmic load but with inductive or capacitive load continues to
rise.

IV. CAPACITOR LOAD WITH HARMONICS
The calculated dc-link capacitor current in chapter III is made
for a sinusoidal current i
P
at the output of the inverter bridge.
But by the pulse width modulated voltage however also
harmonic current in the filter circuit occurs. This currents and
the fundamental currents together forms the output current of
the inverter bridge. The additional load in the dc-link
capacitors of these higher-frequency currents must to be
determined with the help of the Fourier analysis.

i
P1
i
P2
i
P3
L
P1
L
P2
L
P3
C
P1
C
P2
C
P3
Y
u
P1-
u
P2-
u
P3-
L
1
L
2
L
3
u
P1Y
u
P2Y
u
P3Y
u
PY-
L
P1
L
P2
L
P3
i
Sou1
i
Sou2
i
Sou3
i
P1
i
P2
i
P3
L
P1
L
P2
L
P3
Symmetrical

Load
C
P1
C
P2
C
P3
L
1
L
2
L
3
u
P1-
u
P2-
u
P3-
u
P1Y
u
P2Y
u
P3Y
u
PY-
Y
L
P1
L
P2
L
P3

Fig. 8: Filter circuit with symmetrical load (above) and filter circuit
with zero load and substitute load at the input (below)



In figure 8 (above) the filter circuit of the three-phase inverter
is presented. During the design of the filter circuit elements a
compromise between large losses with large inductance and
small capacity as well as with better rule dynamics of the
inverter with small inductance and large filter circuit capacity
must be reached. For the inductance design it is usual to use
the permissible maximum current in the inductance during the
pulse period referred to the peak value of the nominal output
current.
( )
AN max LP

3 . 0 1 . 0 i i (27)
The harmonic current which is limited by the inductance,
flow over the capacitor and produces a voltage ripple which is
superimposing the output voltage. This static voltage ripple is
specified to a maximum permissible value by the design of
the filter capacity. A maximum voltage ripple of smaller than
1% of the peak value of the output voltage is usual [3].

AN max AP
01 . 0 u u (28)
For this small voltage ripple during the pulse periods
sinusoidal voltage and current waveforms at the output of the
filter can be accept in very good approximation. The
sinusoidal current at the output forms together with the filter
capacitor currents the output current of the bridge-legs.

Now for the calculation of the input current i
d
the currents in
the bridge-legs are to be derived. For this the alternate filter
circuit is presented in the figure 8 below. In this mathematical
model the filter circuit of the inverter is without load, so that
only the idle current flows in the filter circuit. The load of the
inverter is reproduced thereby by the current sources arranged
parallel to the filter circuit. The following equation shows the
current waveform in the source of phase 1.
( ) ( )
Sou 1 Sou 1 Sou
sin

t i t i (29)
The current waveforms of the other sources have a phase
angle of 120 to the current in equation 29, so that all sources
together forms a symmetrical load. The addition of the
currents of filter circuit and current sources results in the
output current of the bridge-legs. For the calculation of the
dc-link capacitor current first the pulse signal presented in the
following figure is to be mathematically described with the
help of the Fourier analysis.
u
d
2
u
0 2
t
3 4

Fig. 9: Pulse waveform with constant pulse width

In the figure a pulse waveform with the constant pulse width
2 and the height u
d
is presented. This waveform can be
mathematically described by the following functions [4].
( ) ( )
1
]
1

+

+

... ) 2 cos(
2
2 sin
) cos(
1
sin
2
2
) (
d
t t
u
t u
(30)

( )
1
]
1

) cos( sin
1
2
2
) (
1
d
t
u
t u
(31)
With the pulse width modulation the pulse width 2 changes
over the fundamental period. This is shown by the duty cycle
waveform in equation 1. With this function the waveform of
the pulse width 2 (
1
t) can be determined.

( ) t
m
T
t
t
1
P
e1
1
cos
2 2
) (

+

(32)
The pulse waveform in an elementary period is indicated by
the size t. This size can be calculated as a function of the
fundamental period t(
1
t). The factor indicates the
number of pulses per fundamental period =
P
/
1
.

t t
T
t
t t
1 P
P
1
2 ) (
(33)
These equations now result in the voltage waveform of the
pulse width modulated signal at the output of the bridge-legs.
The variable in the cosine function indicates the phase shift
angle of the pulses in the pulse periods. The height of the
voltage pulses amounts to u
d
.

[ ]

'

+ +

) cos( 1
2
1
) (
1 d 1 1 P
t m u t u
(34)
[ ] ( ) [ ]

+
1
]
1

t p t m
1 1
1
cos ) cos( 1
2
sin
2

In figure 10 the half bridge voltage u
P1-
and their fundamental
portion is presented. The modulation factor of this voltage
amounts to m = 0.8 and the pulse number per fundamental
period is = 45. Below in the figure the spectra of this half
bridge voltage can be seen. The standardized dc-component
of the voltage amounts to 0.5 and the fundamental component
are according to the modulation factor equal 0.8. Beyond that
by the modulation of the voltage in each case frequency bands
occurs with the pulse frequency and with multiples of the
pulse frequency. All harmonics of the waveform are odd.




Fig. 10: Half-bridge voltages u
P1-
with the fundamental
oscillation frequency u
P1- G

-0.2
0.8
1.2
1.0
0.6
0.4
0.2
0.0
u
u
d

45 0 135

1
t
180 90
m = 0.8
u
P1- G
u
P1-

0.0
0.4
0.6
0.5
0.3
0.2
0.1
u
u
d

90 45 0 225

P
/
1

180 135
m = 0.8
u
P1-




The voltage at the neutral point u
PY-
can be determined as the
sum of all three half bridge voltages divided by three. To
calculate this voltage first the voltage u
P2-
and u
P3-
must be
determined. These voltages have a phase angle of 120 to the
voltage u
P1-
.

However it must be noted with the calculation
that the pulse position is equal. For this reason the expression

1
t must stay unchanged. In such a way the same pulse
position is reached as shown in figure 2 and 3.
[ ] ) ( ) ( ) (
3
1
) (
1 P3 1 P2 1 P1 1 PY
t u t u t u t u

+ +
(35)
( ) ( ) [ ] + + +

120 120 ) (
3
1
) (
1 P1 1 P1 1 P1 1 PY
t u t u t u t u
(36)
The following figure shows the calculated waveform of the
neutral point voltage u
PY-
with the average value for the
modulation factor m = 0.8 and for the pulse number = 45.
Below in the figure again the individual spectra of the voltage
are presented. The standardized dc-component of the voltage
amounts to 0.5. Furthermore, with this voltage all by three
dividable spectra from figure 10 arises.




Fig. 11: Neutral point voltage u
PY-
with the average value

The phase voltage u
P1Y
is the difference between half bridge -
and neutral point voltage. Exactly as with the calculation of
the neutral point voltage there also must paid attention to the
correct phase shift angle and pulse position.
) ( ) ( ) (
1 PY 1 P1 1 P1Y
t u t u t u


(37)
( ) ( ) [ ] +

120 120 ) ( 2
3
1
) (
1 P1 1 P1 1 P1 1 P1Y
t u t u t u t u
(38)
To calculate the phase current i
P1
with the voltage u
P1Y
the
cosine function in the argument of the sine - and cosine
functions by means of the Bessel functions must be dissolved.
The following relations apply:

[ ] ( )

,
_

cos ) (
2
sin ) cos( sin
0
J k
(39)

[ ] ( )

,
_

cos ) (
2
cos ) cos( cos
0
J k
(40)
with
... 3, 2, 1, if 2
0 if 1

k

With the help of these Bessel functions now first the phase
voltage u
P1Y
can be calculated. With the result that in the
following equation is presented the individual spectra are
recognizably. This makes the following calculation of the
phase current possible.
+

) cos(
2
) (
1
d
1 P1Y
t
m U
t u
(41)
( )

,
_

,
_

1
]
1

,
_

'

2 2
sin
3
2
cos 1
3
2
1 0
d
m
J k
U

( ) [ ] ( ) [ ]

1
]
1

+ + + +
1 1
cos cos

With the determination of the phase current i
P1
the individual
voltage components are in each case divided by the absolute
value of the resistance and provided with the appropriate
phase shift angle. For the fundamental component from the
equation 41 only the "no-load" operation of the filter circuit is
considered. As additional fundamental load the current source
i
Sou1
is used as shown in figure 8.
( ) +

+ ) cos(
2
sin

) (
PF 1
PF
d
Sou 1 Sou 1 P1
t
Z
m U
t i t i
(42)
( )

,
_

,
_

1
]
1

,
_

'

2 2
sin
3
2
cos 1
3
2
1 0
d
m
J k
U

( )
( ) [ ]

+ +

) , ( cos
,
1
PO 1
PO
Z

( )
( ) [ ]

1
1
]
1
+ +

+
+
) , ( cos
,
1
PO 1
PO
Z

In equation 42 the following absolute values of the resistances
Z
P1
and Z
POt
(, ) as well as the phase shift angles
P1
and

POt
(, ) are to be used:

2
P
P
2
LP PF
2
1
2

,
_

+
C
L R Z

,
_

LP
P
P
PF
2
1
2
arctan
R
C
L


( )
( )
2
P
P
2
LP PO
2
1
2 ) , (
1
]
1

t
t +
t
C
L R Z


( )
( )
1
1
1
1
]
1

t
t

t
LP
P
P
PO
2
1
2
arctan ) , (
R
C
L

In figure 12 the phase voltage u
P1Y
and the phase current i
P1
is
presented with half nominal load and with a phase shift angle
of
P1
= 45. The fundamental current with superimposed
harmonics, is clearly recognizable thereby. In the figure below
-0.2
0.8
1.2
1.0
0.6
0.4
0.2
0.0
u
u
d

45 0 135

1
t
180 90
m = 0.8
u
PY-

0.6
0.0
0.4
0.5
0.3
0.2
0.1
u
u
d

90 45 0 225

P
/
1

180 135
m = 0.8
u
PY-




the individual spectra of the voltage u
P1Y
and the current i
P1

can to be seen.

From the spectra presented in figure 10 only the not by three
divisible voltage spectra remain. Each of these harmonics
produced a current in the filter circuit. The more largely the
harmonic is the more the current is absorbed by the filter
circuit.




Fig. 12: Voltage u
P1Y
at the load and the output current i
P1

without and with harmonics (
P1
= 45)

With the phase current and the output voltages of the bridge-
legs now the current i
d
in the input of the inverter bridge can
be calculated. As graphically presented in figure 3, the
respective phase current is thereby multiplied in each case by
the appropriate pulse pattern of the bridge-leg voltages and
the results afterwards added. For this calculation first the
currents i
P2
and i
P3
and the voltages u
P2-
and u
P3-
must be
derived. These currents have a phase shift angle of 120 to the
current i
P1
and the voltages to voltage u
P1-
. However here in
the calculation it must be noted that the pulse position is
equal. For this reason the expression
1
t must remain
unchanged. In such a way the same pulse position can be
reached as presented in figure 3.

d
1 - P3
1 P3
d
1 P2-
1 P2
d
1 - P1
1 P1 1 d
) (
) (
) (
) (
) (
) ( ) (
u
t u
t i
u
t u
t i
u
t u
t i t i

+

(43)

+

+

d
1 - P1
1 P1
d
1 - P1
1 P1 1 d
) 120 (
) 120 (
) (
) ( ) (
u
t u
t i
u
t u
t i t i
(44)

d
1 - P1
1 P1
) 120 (
) 120 (
u
t u
t i
+
+

In figure 13 the current i
d
is presented for a sinusoidal phase
current i
P
and a phase shift angle
P1
= 45. This curve of the
current corresponds in principle to the drawn waveform of i
d

in figure 3. However number of the pulses ( = 45) is larger in
this example. The course of the curve repeats in each case
after 1/6 of the fundamental period. This is also shown by the
spectra in the figure below, which are divisible by six in each
case. The dc-component corresponds to equation 23 (half
nominal current). Beyond that, frequency bands occur with
the pulse number and with multiples of the pulse numbers.




Fig. 13: Current i
d
in the dc-link circuit without
harmonic output current (
P1
= 45)




Fig. 14: Current i
d
in the dc-link circuit with
harmonic output current (
P1
= 45)

In figure 14 the current i
d
is presented with superimposed
harmonic phase currents i
P
. The phase shift angle of the
fundamental waveform is
P1
= 45. In opposite to the curve
in which is presented figure 13 still another additional higher
frequency current is superimposed to the already with
harmonics influenced input current. The course of the curve
0.0
0.4
0.8
0.6
0.2
-0.2
-0.4
-0.6
-0.8
u
u
d


i
i
P1G

45 0 135

1
t
180 90
m = 0.8
i
P1

u
P1Y

90 45 0 225

P
/
1

180 135
0.0
0.4
0.6
0.5
0.3
0.2
0.1
i
i
P1 G

m = 0.8
i
d

1
t
45 0 135 180 90
i
d

-0.2
0.4
0.6
0.2
0.0
i
i
P1 G

m = 0.8
90 45 0 225

P
/
1

180 135
0.0
0.4
0.6
0.5
0.3
0.2
0.1
i
i
P1 G

m = 0.8
i
d

45 0 135

1
t
180 90
i
d

-0.2
0.4
0.6
0.2
0.0
i
i
P1 G

m = 0.8
0.0
0.4
0.6
0.5
0.3
0.2
0.1
u
u
d

i
i
P1G

90 45 0

P
/
1

180 135
m = 0.8
u
P1Y

i
P1

225



repeats also here in each case after 1/6 of the fundamental
period. This is shown at the spectra, which are presented in
the figure below.

Now the influence of the harmonic filter currents (figure 12)
to the load of dc-link capacitor is calculated. Thereby it is
presupposed, that an ideally dc-current flow i
E
= i
d-
in the
input of the inverter (worst case). So the individual input
current harmonic components of the inverter bridge can be
calculated in each case with sinusoidal phase currents and
with harmonic filter current that are superimposing the
sinusoidal components. At first the rms-values of the
amplitudes of all harmonics in equation 45 are to be derived
and then the calculated individual portions are added
geometrically.

,
_

1 n
2
n Cd
d
2

i
I
(45)
The calculations with additional harmonic filter currents are
related to them with sinusoidal current. In figure 15 the results
from the calculations as a function of the load are presented
with different modulation factors. The maximum current
ripple specified by the dimensioning of the filter circuit
amounts in this case to 0.3
AN
. In the following figure the
influence of the harmonic filter circuit current is presented
with
1
= 0 (above) and with
1
= 45 (below).




Fig. 15: Influence of the harmonic output filter current
on the dc-link capacitor current

In the figure is clearly recognizable, that the harmonic ripple
current of the filter circuit only stresses the dc-link capacitors
considerable with small output currents i
P
< 0.3i
AN

additionally. With larger output current the dc-link capacitor
load corresponds to that with sinusoidal output current of the
inverter bridge in very good approximation. For the
dimensioning of the capacitors in the dc-link circuit for
instance, these large loads are important. That means that
sinusoidal output currents for the dimensions of the dc-link
capacitors are sufficient. The load of the capacitors for
sinusoidal output currents was presented in figure 7.

Beside the harmonious currents at the filter circuit still current
harmonics arise by switching processes in the bridge-legs.
This current was examined in [5, 6, 7] for single-phase
inverters. The tendentious results have also their validity for a
three-phase inverter. With small output load of the inverter
the additional current of the switching processes in the dc-link
circuit can substantially contribute to heating up of the dc-link
capacitors. During higher load usually again the calculated by
sine wave output current caused capacitor current dominated.
But no statement for all semiconductor types can be made.
Beyond that this additional current contributed considerably
by the pulse frequency of the inverter.

V. COMPARSION OF THE CALCULATIONS WITH
PRACTICAL MESUREMENTS
Now the calculated capacitor current in the dc-link circuit is
to be compared with practical measurements of a MOSFET-
inverter. The dc-link circuit of the inverter consists out of
7 electrolytic capacitors. For the 6 switches in each case
8 MOSFETs are connected in parallel. The inverter works
with a pulse frequency of f
P
= 8 kHz. The nominal output
power of the inverter in continuous operation amounts to
P
N
= 6kW and the input voltage is u
d
= 48V [8, 9, 10]. At the
inverter output an electrical machine is attached, which can
work both in the motor and in the generator operation.

The currents in the individual capacitors of the dc-link circuit
are measured with current probes, whereby the individual
measured values are transfer over current transformers at first.
The waveforms of the individual capacitor currents are
presented and the rms-values are calculated with the
oscilloscope.

Figure 16 above shows the current waveforms in three
capacitors of the inverter with an output current of I
P
= 158A
and with a phase shift angle of
P1
= 130. At this phase angle
the electrical machine works in the generator operation.

The rms-value of the current in each individual capacitor
amounts to I
Cd1
12A. Beyond that the respective portions of
the peak output current
P 1
32A and of the dc-current
I
d- 1
-12A can be seen in the diagram waveforms apart from
the period duration T
1
10.2ms. The sum of the current
components in all 7 capacitors results in each case in the total
current value. Below the current diagrams the most important
electrical parameter of the inverter are presented.

Below in the figure the current waveforms in the individual
capacitors with an output current of I
P
= 238A and a phase
shift angle of
P1
= 125 are presented. During this mode the
rms-current in each individual capacitors amounts to
I
Cd1
18.5A. The duration of the fundamental period is
T
1
11.8ms, each peak value of output current is
P 1
48A
I
P
I
P N

2.0
4.0
1.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0
I
Cd LC
I
Cd Sin

m = 1.0
m = 0.8
m = 0.6
AN max LP

3 . 0 i i

1
= 45
2.0
4.0
1.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0
I
Cd LC
I
Cd Sin

I
P
I
P N

m = 1.0
m = 0.8
m = 0.6
AN max LP

3 . 0 i i

1
= 0



and each input dc-current is I
d- 1
-19A. Below the current
waveforms also the most important electrical parameters of
the inverter are presented.




Fig. 16: Current in three dc-link capacitors of the inverter
with different loads (practical measurements)

The current of the practical measurements corresponds in
principle to the waveforms of the capacitor current in figure 3.
The rms-value of this current depends to the size of output
current, on the phase shift angle and of the modulation factor.
The rms-value of the capacitor current is the same as the
calculated current which is presented in figure 7.

VI. CONCLUSION
Today pulse inverters are used world-wide within many
ranges for ac power supply, for drive train engineering and for
frequency conversion. The dc-link capacitor in the power
parts contributes thereby substantially to the volume, to the
weight and to the costs of such inverters. For this reason the
necessary size of capacitors must be determined exactly to
avoid over design. In most applications the dc-link capacitor
size is dependent by the current load.

In this publication the dc-link capacitor current for a three-
phase inverter was analytically calculated. At first an ideal
smoothed input dc-voltage as well as sinusoidal currents at
the output of the inverter bridge is presupposed. The results
show that the input current of the inverter bridge consists out
of a dc-component with higher-frequency portions.

While the dc-component flows in the input of the inverter, the
higher-frequency current portion in reality flows completely
over the dc-link circuit capacitor. This portion can amount
maximally to a value I
Cd
= 0.46
P
.

In chapter IV the influence by current harmonics, which result
from the output filter circuit and from switching processes in
the inverter bridge is examined. It shows that with small
output power a clear capacitor load results by these
harmonics. But with higher output power again the calculated
load caused by sine wave current dominates. The load of the
dc-link capacitors by this harmonics can be neglected.

At the end the calculated dc-link currents are compared with
practical measurements of an MOSFET inverter. The current
of the practical measurements corresponds in principle to the
theoretical waveforms. Beyond that a very good agreement
between the calculated and measured values exists.

REFERENCES
[1] J. Schmidt, "Einsatzbereiche, Anforderungen und Konzepte
netzunabhngiger Stromversorgungen," Seminar Haus der Technik e.V.
Essen 1999
[2] K. Heumann, "Grundlagen der Leistungselektronik," Lehrbuch Teubner
Verlag Stuttgart 1975, 1. Auflage S. 178-182
[3] F. Renken, "Einphasige Wechselrichter mit hart und weich schaltenden
Transistoren fr statische unterbrechungsfreie Stromversorgungen",
Dissertation Univ. der Bw. Hamburg, VDI Verlag 1999, Fortschritt-
Berichte ISBN 3183271214, S. 35-82
[4] H.-J. Bartsch, "Taschenbuch mathematischer Formeln," Harry Deutsch
Verlag 1982, 6. Auflage S. 413
[5] F. Renken, "Analytic Calculation of the DC-Link Capacitor Current for
Pulsed Single-Phase H-Bridge Inverters", 10th EPE Toulouse 2003
[6] F. Renken, "Analytic Calculation of the DC-Link Capacitor Current for
Pulsed Single Phase H-Bridge Inverters", EPE Journal Volume 13
No 4 - September - October - November 2003
[7] Renken F. "Inverter with Soft-Switched Transistors for Uninterruptible
Power Supplies", 10. EPE Meeting/ Proceedings: Papers on CD ISBN
90-75815-07-7, Toulouse, France 2003
[8] V. Karrer, F. Renken, "Power Electronics for the Integrated Starter
Generator", Conference: Optimization of the power train in vehicles by
using the Integrated Starter Generator (ISG), Haus der Technik e. V.
Munich 2002, Proceedings Expert-Verlag ISBN 3-8169-2977-2, pp.
222-247
[9] F. Renken, V. Karrer, "Leistungselektronik fr den Integrierten Starter
Generator," 22. Tagung Elek-tronik im Kraftfahrzeug", Haus der
Technik Essen, Tagungs-Nr. E-H030-06-056-2, Stuttgart 2002
[10] Skotzek P., Reindl T., Poisel M., Eichenseher V., Karrer V., Allwang
R.: Steuergert fr den Integrierten Starter Generator im 42V-Bordnetz.
21. Tagung Elektronik im Kraftfahrzeug, Haus der Technik - Essen,
Mnchen Mai 2001
U
PY
= 13.9V U
d
= 48V f
1
= 100Hz
P1
= 130
I
P
= 158A I
d-
= -86A I
Cd
= 84A m = 0.8

P 1
i
Cd 1
I
d- 1

i
Cd 2

i
ch 3

40A
40A
40A
800s
U
PY
= 15.7V U
d
= 48V f
1
= 85Hz
P1
= 125
I
P
= 238A I
d-
= -130A I
Cd
= 124A m = 0.9

P 1
i
Cd 1
I
d- 1

i
Cd 2

i
Cd 3

40A
40A
40A
800s