81

Analytical Calculation of the RMS Current Stress on the DC Link

Capacitor of Voltage DC Link PWM Converter Systems
Johann W. Kola, Thomas M. Wolbank and Manfred Schrodl

Technical University Vienna

Dept. of Electrical Drives and Machines
Gusshausstr. 27/372, A-1040 Vienna, AUSTRIA
Tel.: +43-1-58801-37225, Fax.: +43-1-58801-37299
email: [email protected]

Abstract. In this paper a simple analytical expression for the end seal is reduced and thus the drying-out of the ca-
the current stress on the DC link capacitor of a voltage DC pacitor is delayed.Therefore, in order to meet a demanded
link converter system, as caused hy the load-side inverter MTBF (Mean Time Between Failure) of the inverter besides
is derived. There, a sinusoidal inverter output current and selecting an appropriate rated voltage of the capacitor also
a constant value of the DC link voltage are assumed. The the correct thermal design is of major importance.
DC link capacitor current rms value is determined by the
modulation depth and by the amplitude and the phase an- The temperature being important for the working life is the
gle of the inverter output current. Despite the neglection can-temperature T, (cf. Fig.8 in [3]). T,is determined by
of the output current ripple the results of the analytical the ambient temperature T, and by the power dissipation in
calculation match very well with a digital simulation even the capacitor as caused by the capacitor rms current I C . ~ , ~ ,
if the output current ripple is relatively high as given, e.g.,
for low-frequency IGBT inverter systems. Thus, the a p
proximation can advantageously be used for designing the (RESRdenotes the equivalent series resistance of the capac-
DC link capacitor of P W M converter gystems. itor which represents the sum of the frequency sensitive re-
sistance of the oxid dielectric, the temperature sensitive re-
sistance of the electrolyte and the relatively constant small
1 Introduction contributions of the foil, the tabs and the terminals [3], [4];
denotes the heat transmission resistance between
Voltage DC link inverters are frequently fed via uncon- the capacitor can and the ambience).
trolled rectifier bridges from the single-phase or the three
phase mains. There, in the DC link usually aluminium elec- Remark In connection with Eq.(l) it is important to point
out that an ahnolute maximum rating IC,,.,, of the current
trolytic capacitors (connected in series and/or in parallel)
s t r m on an electrolytic capacitor (as specified in the data sheet)
are used
is determined by the maximum allowable capacitor hot spot tem-
to compensate the difference between the power re- perature T h > T, (cf. Fig.8 in p]).(With rising temperature the
quirement of the inverter (whose mean value is con- electric strength of the dielectric material is lowered and in ad-
stant in steady state operation) and the output power dition chemical reactions between the electrolytic material and
of the input rectifier bridge varying with two or six the aluminium oxide of the anode foil take place). Hence it is
t i e s the mains frequency, neewary to emure in any ease I c , ~ - c I C , ~ , ~ , ~even ~ { ifT ~ )
to supply the input current of the inverter with pulse the working life is not of special importance.
frequency, Therefore, for a correct thermal design of the capacitor
to reduce the spreading of current harmonics with the rms value IC,.,. of the DC link capacitor current is of
pulse frequency into the mains, paramount importance. The methods for calculating Ic,rms
as proposed in the literature so far are based on a spectral
to supply transient power peaks and
analysis of the phase quantities (cf. 151, Fig.4(h) in [6], 171,
to protect the inverter from transient peaks of the [SI, 191, [lo], 1111) and/or rely on digital simulations. Be-
mains voltage. sides the considerable amount of time required for estab-
lishing a simulation model one drawback of these methods
As explained, e.g., in [l]the operating voltage and especial- is the missing possibility of demonstrating the influence of
ly the working temperature (in most cases the temperature the operating point of the inverter (which is characterized
of the capacitor-can is taken as reference value) take signif- by the modulation depth and by the amplitude and the
icant influenceon the working life of electrolytic capacitors. phase displacement of the inverter output current and volt-
If, e.g., an aluminium electrolytic capacitor is operated at age fundamentals) on the capacitor current stress as the
0.9.~ -
rated voltaw the failure rate is lowered to 60% as com- simulation is valid always only for a single set of operating
pared to an operation at rated voltage (cf. Fig.7 in 111). If parameters.
the working temperature is lowered, the typical working life
doubles for every 10°C below the rated temperature (Arrhe- However, as shown in this paper, it is also possible to calcu-
nius Law, cf. p. 6 in [2]), as due to the lower temperature late the rms current of the DC link capacitor in analytical-
the d i h i o n of the g-us parts of the electrolyte through Iv
, closed form in the time domain with su5cient accuracy.

Ninth International Conference on Electrical Machines and Drives, Conference Publication No. 468, 0IEE, 1999
 :R

N s

Fig.1: Basic structure of the power circuit of a voltage DC link converter. The AC machine fed by the inverter is
considered by a simple equivalent circuit formed by leakage inductances L and machine countereleetromotive forces
gN,,. The inductors LN at the input (and/or an inductor at the output of the input rectifier bridge) have to be
provided for limiting the peak values of the input currents and/or the &e& on the maim and the current s t r e on the
DC link capacitor C [12],

The effort necessary for designing the capacitor regarding The relative on-times of the switching states (010) and
the current stress thus is reduced to the evaluation of a sim- (W,
ple mathematical expression. In addition the knowledge of
the functional influence of the operating parameters on the
capacitor current rms value can easily be used for deter-
mining the worst case current stress.

2 Basic Considerations
For preparing the analytical calculation of the capacitor can be determined by simple geometrical corniderations (cf,
current stress a brief review of the fundamentals of the in- Fig.2) and are defined by the location 'pu of the respective
verter control and of the formation of the inverter input pulse half period within the fundamental period of the in-
current i will be given in the following. Due to the sym- verter output voltage and by the relative amplitude of the
metries of an ideal three-phsse voltage system and due to inverter output voltage fundamental and/or the modulation
the phasesymmetric structure of the converter power cir- index
cuit there the analysis can he limited to a :-wide interval
of the inverter output voltage fundamental period.
(5)

2.1 Space Vector Modulation (With reference to the conditions given for a typical prac-
tical applications, the DC link voltage is assumed to be
For stationary operation the reference value of the inverter constant in the following.) In contrast the distribution of
output voltage can be represented by a space vector the total on-time of the non-voltage-forming (free-wheeling)
switching states (000) and (111)
= Ou e x p ( j w ) cpu = wNt (2)

of constant magnitude UU and constant angular speed

UN = 2 r f ~ (f,v denotes the inverter output frequen-
cy). go,(l)is approximated within each pulse half period
t , E ( 0 , T p ) by switching between the immediately neigh- between the beginning and the end of e h pulse halfperiod
bouring inverter output voltage space vectors. For the angle can be freely chosen without influeneing the fundamental
interval cpu E ( f , $) (cf. Fig.2) which is considered in the output voltage phasor xu,(]).This degree of freedom
following we, therefore, have as switching state sequence

I,=0(000) - (010) - (110) ~ (lll)lt,=)Tp (3)

(111) - (110) - (010) - (O0O)l,,=rp ...
where each inverter switching state is characterized by the
triple (SR,SS,ST)of the associated phase switching func-
tions s,, i = R , S , T . (We define s; = 1 if the output e.g., can be used for minimizing therms value of the inverter
voltage of phase i referred to the fictitious center point output current ripple with switching frequency [13]. Alter-
of the DC link is positive, uu,; = +;cl., and si = 0 if natively, one also could use a suboptimal even distribution
the output voltage of phase i is negative, uu,; = -$'o). 6(0m) = 6(lI1) and/or 6(l~l),r= 0.5 under consideration of
 83

h 3 Calculation of Average and RMS Value

UU.IO10) of the Inverter Input Current
For the sake of concentrating to the essentials a purely si-
nusoidal shape of the inverter output currents

Z N , R . ( ~ )= f N CO+U - V)
2n
-
Re iN,S,(l) = IN C d 9 U --
3
- 9)
~U.10111 YU.IlWI
iN,T,(l) = fNCoS(9tJ + -2n3- V ) (10)

corresponding to a space vector

iN,(1) = fh'eqb(9u - V)) (11)
is assumed in the following (9 denotes the phase displace-
ment of the fundamentals of the inverter output voltage
and the inverter output current f , ( l ) ) . This is quiv-
alent to only considering the fundamentals of the phase cur-
Fig.2 Space vectors of the inverter output voltage ahsigned rents which in a 6rst approximation show a constant d u e
to the active inverter switching state8 and to the non-voltage within a pulse half period. The error caused by the n e
forming switehig states (Ow) and (111); furthermore shown: glection of the output current ripple will be calculated in
inverter output voltage reference value %,(1) which has to be section 5 and remains small even if tbe ripple amplitude
formed in the average over a pulse half period of position y u ; reaches relatively high values (as compared to the ampti-
xu,(l) is associated with the fundamentals of the pubwidth tude of the phase current fundamental).
modulated inverter output phase voltages u",i, i = R , S , T .

a low realization effort of the inverter control [14] 3.1 Mean Value of the Inverter Input Current

Based on the considerations in section 2.2 we have for the

2.2 Inverter Input Current Formation local mean value of the inverter input current (related to a
pulse half period)
As becomes immediately clear by considering the inverter
bridge legs as two-pole switches between positive and neg-
ative DC link bus (si = 1 and/or 0) we have for the input
current i of the inverter
,i = & i*Tp
. i dt, (12)

i = 3 R i N , R + 3SiN.S +3TiN.T (8)

(for the considerations in this paper the dead-time interval
which bas to be considered for the gating of the transistors
of a bridge leg in practice and the reverse recovery current
of the freewheeling diodes which take minor infiuence.on
the shape of i 1151, [16]are neglected). Dependent on the
inverter switching state i is formed by segments of the in-
verter output phare currents i ~ , ; .E.g., for the switching
state (OlO), the output terminal of phase S is connected to
the positive DC bus and phases R and T are switched to
the negative DC link rail, therefore, i and ~ N , Sshow iden-
tical time behavior (cf. Fig.S). For switching state (110)
n 91 m n
5s -
there results i = - ~ N , T , where 'T-
IO& 1I1011'
11101 IIVl
~N.R + is,s + ~ N , TE 0, (9)
(as being valid due to the isolated neutral point N of the
AC machine, cf. Fig.1) has to be considered. Only for the
free-wheeling states (000) and (111) no current appears at Fig& Illustration of the formation of the inverter input current
the converter input. The distribution 6(111),~of the free i in dependency on the inverter switchingstate ( S R , S S . S T ) ; local
wheeling states between the beginning and the end of a time behavior of the inverter input current i, of the inverter out-
pulse half period therefore in a fvst approximation only Dut phase current i N , s and of the phase switching functions 8 ~ ,
takea iduence on the position of the current conduction $5 and S T ; furthermore shown: simplified shape of the inverter
intervals i = i N , s and i = - ~ N , T within a pulse half period, input current resulting for the negieetion of the phase current
but does not influencethe shape of i in principle. (A further ripple and/or exclusive consideration of the ph- current h-
minor influence of d ( l ~ l ) , ron the shape of i is due to the dmentals (shown dashed and pointed out by the dotted area).
ripple of the phase currents ~ N , Sand ~ N , Twhich shows a The simplied current shape gives the basis ofthe calculation of
different shape for different d u e s of ~ ( I I I ) , ~ ) . the DC Lik capacitor current strepg in section 3.
 84

Eq.(12) can be expressed as the width of the current pulses (cf. Eq.(4)) appearing a t
the inverter input for increasing modulation index M or by
iavg ..T
= 6(OlO)iN.S - ~ ( ~ I o ) ~ N (13) the increase of the inverter output power due to the pro-
With this, and Eqs.(4) and (10) we receive portional increase of the fundamental output voltage as a
consequenceof the higher modulation index M (cf. Eq.(5)).
3"
,i = I,, =-1~Mcosp.
4 (14) It is important to note that besides the (global) mean value,
Corresponding to the constant power supplied by a sym- 2ideally does not contain any low frequency harmonics but
metrical and sinusoidal threephase voltage/current system only harmonics with Switching frequency.
the inverter input current shows a constant local average
d u e,,i = Iw, i.e., local and global (related to the funds- 3.2 R M S Value of the Inverter Input Current
mental of the output voltage) mean value are identical. For
m p = 0 the generates Only fundamental reactive For the local r m value
~ of the inverter input current
power and/or no power is supplied from the DC link (har-
monic power components are neglected). Accordingly, we
have I,, = 0, the inverter input current i is d e h e d by seg-
ments out of the vicinity of the zero crogsings of the output
2rne - 2
TP
l'Tp i2dt, (15)

phase currents Z N , , with alternating signs (cf. Fig.l(a)). We have in analOgY to Eq.(I4)

On the contrary, for cos9 = 1 we have purely active pow- c.= 6(0lO)i%,S + 6(110)i%,T (16)
er operation and/or maximum current is supplied by the Therefore, with
DC link. There, i is formed by segments of identical signs
out of the vicinity of the maxima of the phase currents i ~ , ,
(cf. Fig.l(b)). For a given value of the fundamental dis-
placement factor cos v of the output currentlvoltage system
6.= :J, $ 2
i r d w, (17)

In connection with Eq.(18) one has to point out that the

r m value
~ of the inverter input current i in a first approxi-
mation (as calculated by exclusivelyconsidering the output
phase current fundamentals) is independent of the inverter
control scheme. Various control schemes differ only con-
cerning the distribution 6(111),~of the freewheeling states
1131 (being characterized by i = 0) between beginning and
end of each pulse half period but not concerning the relative
on-times 6(010) and 6(110)of the votageforming states (cf.
Eq.(4)). If the fundamentals of the output phase currents
ZN,$,(I) and - ~ N , T , ( I ) are assumed to be approximately con-
stant within t, € [0, fTp), therefore, there is no influence
of the actual position of the switching state intervals 6(010)
and 6(110)on the value of irms.Therefore, the only (minor)
influence of 6 ( 1 l l ) ,on
~ irma and Irmsis by the ripple compo-
nent of ~ N , Sand - ~ N , T which, however, has been neglected
for the derivation of &.(18).

ly defining S ( l ~ l )[17].
, ~ This kind of optimization is of special
importance if foil capacitors (which are characterized by a hiph-
er admissible current stress and higher service life as compared
to electrolytic capacitors, but show a eonsiderahly lower eapaci-
t a m ) are employed for realizing the DC link capactior.

(b) 4
DC Link Capacitor Current
RMS Value
Fig.4 Time behavior of the inverter input current a and the
output phase current i N , R within a fundamental period of the The DC link current is defined by the difference
converter output voltage for cosv = 0 (cf. (a)) and cosq = 1
(cf. (b)). ic =' i - i (19)
 a5

of the inverter input current i and the output current ir. of 0.750
the converter input stage which in most cmes is realized as
uncontrolled three-phase rectifier bridge.

By splitting up i and ir. into DC and AC components (des-

t 0.625

ignated by subscript 'ac'), I",,!C.M

0.s00
i = I,,, + ,i
i~ = IL+,., +i ~, , ~ (20) 0.375
one now can replace Eq.(19) by
. .
zc = ZL.& - 2.c .
0.250
(21)
As is immediately obvious the load on the DC link capacitor 0.125
is defined solely by the AC components of i and i ~ the ,
DC component Iavsof a can be considered do be supplied
directly by the input rectifier bridge. (a) 0.m
0.0 0.2 0.4 0.6 0.8 1.0 1.2
With Eq.(20) and
H-

one receives for the global rms value of the DC link capac- 0.625
itor current
brW.1
0.500

0.375
The problem with a mathematical evaluation of this equa-
tion usually m i s t s in the calculation of the integral being
dependent on ,i and i ~ , If,~however,. as in the cme at 0.2s0
hand the currents i~ and i do not contain harmonica in the
same frequency range - harmonics of higher amplitudes are
usually only present for ordinal numbers n 4 50 in the out-
put current of uncontrolled rectifier bridges (cf., e.g., Fig.13
in 1181); the frequency range being occupied by harmonics
of i is defined by the inverter switching frequency and shows
typically a lower bound of n w 200.. .5M) for a realization
of the turn-offpower semiconductors by IGBTs - we have

and we receive for the current streps on the DC link capac-

itor
2
= I&,# + . (25)
Therefore, the value of the DC link capacitor current
can clearly be divided into a contribution being determined
by the mains-commutated input rectifier and a contribu-
tion caused by the P W M inverter (cf. Fig.7 in [4]). For
the sake of brevity in the following only the calculation of
Ic,r,,,.,l = Im,cm*will be discussed in detail. Concerning
the calculation of = IL,,,,,, we would like to refer
to, e.g., [19].

Rsrnark: If the DC link is fed via a selfammutated rectifier,

e.g., a three-phase P W M rectifier system with high pulse fre-
quency (cf,,e.g., p. 201 in [ZO]) the DC link capacitor current
results from a superposition of harmonies of the DC link input Fig.& Dependency of the inverter-side contribution ZC,~,,,~,~(cf.
and output currents. The current str- an the DC link capac- Eq.(27)) to the global DC link capacitor current rms value I C . , ~
itor therefore is dependent on the amplitudes and on the phase on the inverter modulation index M and on the phase displace-
relations of the harmonics of equal ordinal number resulting in a ment 9 of the fundamentalsof the inverter output phase voltages
higher or lower amplitude of the respective harmonics in the DC uU,, and the -dated output phase currents i~,,.Parameters
link capacitor current. Therefore, IC,,,,,~ can be minimbed by of the family of CUNW cacp (cf. (a)) and M (cf. (b)); ( c ) :
an appropriate coupling and/or synchronization of the rectifier threedimensional representation.
 86

and inverter controls (cf. section 3 in [Zl] and [ZZ]).The maxi- the resulting current stress I C , ~is~independent
,I of the
mum current stress on the DC link capacitor current occurs for value of cas 9.
eompletly 'uncorrelated' operation of the input- and output-side
systems (related to the power flowinto and out of the DC link). According to Fig.6 for a coarse worst case estimation of
As shown in [23] (cf. Eg.(50) in [23])in this case the current s t r e s on the DC link capacitor as a basis for
= L,m+ I L , ~ , ~ , ~
Ic3rms3ma (26) the capacitor dimensioning one has to start from
CM be given a8 a worst case estimate. For a more accurate
determination of I C , ~ one
, ~ would have to refer to c digital sim-
ulation.
independent of the value of the fundamental displacement
For the r m value Ic,.,.,~ of the AC component ,i of i we factor cos 9.
have in general
2 2
= 1:-
Ic,rms,l - I,, . (27) 5 Accuracy Estimation
and with Eq.(14) (251 In order to be able to derive a simple expression for the
on C and for being in-
inverter-side current stress Ic.rms.l
Ic,r-,i
d3 A
= I N J ~2 ~M [ - +cos2 q(- -9
-M)] , (28) dependent of details of the inverter control scheme, the r i p
4n ?r 16 ple of the inverter output currents iN,i has been neglected
Eq.(Z8) represents the wanted functional dependency of the and/or only the fundamentals ~ N , s , ( I )of the phase currents
current stress on C on the inverter operating parameters (cf. i ~ ,have i been considered so far. Thus, a very clear approx-
Fig.5). The proportional relationship of I C , ~ , , , ~and
,I imation being independent of the absolute values of the DC
and/or a can be explained clearly by considering the link voltage (lo,the pulse frequency f p and the stray in-
fact that i is formed by segments out of the phase currents ductance L of the AC machine (cf. Fig.1) has been derived.
with a width being linearily dependent on the modulation Now in the following the error which is caused by the ne-
index M (cf. Eq.(8)). (For M = 0 the inverter remains in glection of the phase current ripple as compared to an exact
the &wheeling state within the entire pulse half period; calculation by digital simulation is determined and/or the
therefore, we then have i = 0 or ic,rms,l= 0, respectively). limit of the applicability of Eq.(28) for the design of the DC
link capacitor is determined.
However, if now MS 9 = 1 is assumed, besides Irms also the
DC component IDvgincreases with increasing M, according There, for characterizing the ripple of the phase currents a
to Eqs.(l4) and (18). Therefore, according to Eq.(27) a
maximum of the DC link capacitor current rms value I c , ~ ~ ~
(33)
occurs about in the middle of the modulation range
is de&&. Thus all operating parameters being of influence
on the formation of the output current ripple are combined

This can be explained clearly by considering the rms value mental. (w

and related to the amplitude of the output current funda-
represents the ripple amplitude resulting
of the AC component of a unipolar square wave signal with for the application of a symmetrical square wave voltage of
variable duty ratio a which reaches a maximum for a = frequency f p , duty cycle OL = 0.5 and amplitude $U, to
0.5. For operating the inverter at maximum modulation L.) Furthermore, a symmetrical distribution 6<111),r= 0.5
depth Mmaxthe shape of i still shows intervals with zero of the freewheeling switching state (cf. section 2.1) is as-
+
current (of relative width 6 p q 6(111), cf. Eqs.(4) and
sumed which is frequently applied in practice due to the
low calculation effort required for the determination of the
(6)). Accordingly, the inverter input current in any case
shows a AC component i , and/or Ic3rms,1 does not reach inverter switching instants 1241.
0 for values of M clase to M,-. For feeding the DC link via an uncontrolled rectifier bridge
In the case of cos9 FE: 0 again a sequence of phase cur- the value of the DC link voltage is determined directly
rent segments showing increasing widths for increasing M by the mains voltage and usually shows a very low r i p
is switched into the DC link. However, there is no reduc- ple. Therefore, for the following considerations a con-
tion of Ic.,,,,.~ as compared to b,. due to the missing DC stant DC Link voltage of U0 = 540V is assumed come
component I,, of i (no power is supplied by the DC link sponding to the voltage value resulting for the operation
and/or, as discwed in section 3.1, the phase current seg- of the converter system in the European low voltage three
ments occuring within a pulse half period show opposite phase mains (line-to-line voltage: 4M)V,,.) and ideal c b
signs). Therefore, I C , coincides
~ ~ with I,,, (cf. Eq.(18)) pacitive smoothing. For the switching frequency w e define
and the maximum of I c , ~ - occurs for maximum modu- fp = & = lO!dIz. Thus, hy considering output induc-
lation depth Mmax = 3
and/or turns into a boundary tance values in the range of L = 0.3375.. .1.35mH in con-
nection with an amplitude of the output phase currents be-
maximum. This is true ~ngeneral for
ing set to I N = 20 A a range of n = 0.25. . .1.0 is inspected,
1 where the output current shape changes form being a p
cos29 5 -
4($ - 1) proximately sinusoidal (K = 0.25) to having a relatively
high ripple component ( K = 1, cf. Fig.6). Furthermore,
In this connection we would like to point out that for the conditions for cos9 = 1 and cos9 = 0 are investigated
as in these cases qualitatively different shapes of i and/or
i , (cf. Fig.4) occur.
 87

in practice for operating an IGBT inverter at rated pow-

XiO.25 er. Therefore, the dimensioning can be based on Eq.(28)
and/or a time-amuming calculation of I C . , ~ by
~ ,digital
~
simulation can be omitted.
Remark Harmani-optimal modulation (i.e., a modulation
(a) scheme which is optimized with respect to minimum harmonic
current rms value by extending simple sinusoidal modulation by
addition ofa thud harmonic 1131) in afirst approximation results
in an almost identical time behavior of the output current ripple
and/or an equal harmonic rms value of the phase current r i p
ple as given for the (suboptimal) control scheme which has been
taken as a basis for the considerations in this paper (cf. Fig.6
(b) in [14]. Therefore, Fig.7 can be considered BP being valid also
I ' I ' I ' I ' I ' I for optimized control. For purely sinusoidal modulation there
results a higher ripple of the phase currents 1131, [24] and in
Xi0.75 eonsequence the analytical calculation shows a higher deviation
from the result8 of the digital simulation for a given value of (c.
However, bs this control scheme due to its limited modulation

I
- f l%l

Fig.6 Time behavior of an inverter output phase current for

different values of the parameter n which characterizesthe rela.
tive amplitude (related to the amplitude f N ofthe output current
fundamental)of the output current ripple; n = 0.25,0.5,0.75 and
1.0 (cf. (a) - (d)); simulation parameters: cosrp = 1, M = 1.0

The dependency of the relative error

0.0 0.4 0.8
- M

of the analytical calculation of I C , ~ , , . ~based

, ~ on Eq.(28),
, ~ ~ ~ the inverter-side current stress on C as
I C , ~ - , I denotes
determined by digital simulation) on K and M is shown
in Fig.?'. For values of M close to 0 the phase currents
show a very low ripple (cf. Fig.16 in [25]). Therefore, the
error caused by assuming a purely sinusoidal output current
shape remains limited to very small values. For increasing
modulation depth M a bigher relative error of the analytical
approximation occurs due to the increasing output current
ripple amplitude.

In the case of cos9 = 0 we in general have a lower accu-

racy than for cos9 = 1. This can be explained clearly by
the fact that for cos9 = 0 the inverter input current (and
correspondingly also ,i Eq.(21)) is formed by segments
out of the vicinity of the zero crossings of the output phase
currents; thus a current ripple of given amplitude K takes a 0.0 0.4 0.8 4.2
relatively higher intluence on the shape of i as for forming i - U
by segments out of the vicinity of the maxima of the phase
currents and/or for mrp = 1 (cf. Fig.4). Fig.7: blative error f of the analytical calculation of the ca.
pacitor currant streas Zc.,ms,i as caused by the P W M inverter
In case, e.g., a minimum degree of accuracy according to (d.Eq.(28)) *om the results of a digital simulation in d--
If I < 5 . . ,lo% is set for the dimensioning the application deney on the modulation index M and the relative amplitude U
range of Eq.(28) is limited by K % 0.75. This, howev- of the inverter output current ripple for -9 = 1 (cf. (a)) and
er, certainly includes the conditions being typically p r s e n t e~sp = 0 (cf. (b)).
range M 5 1 is of very limited practical importance this will not 7 Conclusions
he diseupned in more detail here.
In this paper a simple analytical expression for the current
stress on the DC link capacitor of a three-phase PWM con-
6 Dimensioning of the DC Link verter system as caused by the inverter stage is derived.
Capacitor As checked by a comparison to the results of a digital sim-
ulation the calculated expression meets the accuracy re-
For the dimensioning of C regarding the current stress the quirements for designing the DC link capacitor for practical
data sheet specifies a maximum permissible current stress IGBT inverter systems (switching frequency f p > IOkHz,
Ic,rms,dim for a required working life of the aluminium elec- relative ripple of the output current n < 0.75).
trolytic capacitor in dependency on the ambient tempera- The topic of further research will be the analytical calcula-
ture T.. Ic,rma,dim represents a fictitious current rms value tion of the DC link capacitor current stress in the case of
(cf. p. 9 in [26]) which considers the decrease of the RESR over-modulation and of the DC link capacitor current rms
of an aluminium electrolytic capacitor with rising frequen- value as occurring for three-level (cf. 1281 and/or 1291) and
cy. In most cases the value of this resistance is given for a single-phase 1211 PWM inverter systems.
frequency o f f = lOOHz (REsn,looH.). Therefore, we have
for the losses occurring in the capacitor
ACKNOWLEDGMENT
Pc = RESR,IWH.
I&-,dim (35)
The authors are very much indebted to the Jubilamfonds
der OESTERREICHISCHEN NATIONALBANK which
generody supports the work of the Department of Elec-
and trical Drives and Machines of the Technical University of
Vienna.
(37)
where Ic,(fi),rms
denotes the rms value of the spectral com-
ponent of ic with frequency f j ; REsR,f; defines the d u e References
of the equivalent series resistance being given for this he-
quency. [l] MeManus, R.P.: The Effects of Operating Pamme-
tws on Aluminium Electmlytie Capacitor Reliability in
According to the data sheets of aluminium electrolytic c b Switched Mode Converters. Proceedings of the 6th Nation-
pacitors, about a constant value of kfd>loreai~0.45 is given al Solid-state Power Conversion Conference, Miami Beach,
for frequencies f > 1OkHz (cf., e.g., p. 35 in [27])). For a May 2 4 , pp. 13-1 - 13-5 (1979).
realization of the turn-off power semiconductors of the in- [2] VALVO/Philips: AL-Elektrotytkondetrpatoren. Technis-
verter by IGBTs (as assumed in this paper) the switching che Informationen 890427 (in German).
frequency typically is set to values f p > 1OkHz. There- [3] Schnabsl, W.: Aluminium Electrolytic Capacitors for
fore, the effective current stress on the DC link capacitor Power Electronics. P r m d i of the 5th International
as caused by the inverter simply can be calculated via Power Electronics Conference (electronica90), Munich, pp.
117-129 (1990).
2
IC,rme.l.dim z%0.45IZ.,rm.,1 . (38) [4] Gasperi, M.L.: A Method fov Predicting the Ezpected
Life of Bus Capacitors. Record of the 32nd IEEE Industry
There, Ic.rms.l denotes the capacitor current ~lllsvalue re- Applications Society Annual Meeting, New Orleans, Oct.
sulting for the maximum amplitude of the inverter output 5-9, Vol. 2, pp. 1042-1047 (1997).
current. For the low frequency components of the capacitor 151 Rosa, J.: The Harmonic Speetnrm of DC Link Currents
current IC,rms,l,dim which are caused by the uncontrolkd in Inverters. Proceedings of the 4th International Confer-
rectifier bridge at the system input a different weighting of ence on Power Conversion, San Rancisco (CA), March 2%
the harmonics has to be performed according to the capac- 31, pp. 3&52 (1982).
itor data sheet (cf., e.g., Fig.3 in [4] or p.37 in [27]) and/or 161 Ziogas, P.D., Wiechmann, E.P., and Stefanovic,
I ~ , ~ ~ , 2 , d ihas
, , , to be calculated by a digital simulation. V.R.: A Computer-Aided Analysis and Design Approach
for Static Voltage Soume Inverters. IEEE Transactions on
By comparing the admissible current stress of a capacitor Industry Applications, Vol. IA-21, No. 5 , pp, 1234-1241
with the actually occurring total current stress the num- (1985).
ber of capacitors to be connected in parallel can be de- 171 TauRq, J.A., Goodman, C.J., and Mellitt, B.: Rail-
termined and/or a capacitor having a sufficiently high ca- way Signalling Compatibility ojlnnverter Fed Induction Mo-
pacitance and, therefore, a relatively high maximum cur- tor L%ws fov Rapid Damit. IEE Proceedings, Vol. 133,
Pt. B, No. 2, pp. 71-84 (1986).
rent rating can be chosen. Due to the larger snrface of ca-
pacitors of higher capacitance (and/or higher volume) the 181 'IhuAq, J.A., Xiaoping, J., Allan, J., and Burdett,
S.: A Simple Method of Calculating Inverter DC Side Cur-
heat transmission resistance Rth,=-* shows a lower value. rent Harmonies. P r o d i n g s of the 3rd European Confer-
Therefore, for a given can temperature T, a higher power ence on Power Electronies and Applications, Aachen, Ger-
loss I 2C , r m s , d i m of
R ~the
~ ~capacitor can be tolerated (cf. many, Oct. S12, Vol. I, pp. 93-97 (1989).
Eq(1)). In connection with this it is important to note [9] Evans, P.D., and Hill-Cottingham, R.J.: DC Link
that the capacitance of the DC link capacitor of practi- Current in P WM Inverters. IEE Proceedings, Vol. 133, Pt.
cal systems usually is determined by the effectivecapacitor B, No. 4, pp. 217-224 (1986).
current streas and not, e.g., by a given maximum admis- [lo] Habetler, T.G., and Divan, D.M.: Rectifier/Inwertcr
sible value of the amplitude of the DC link voltage ripple. Reactive Component Minimization. Proceedings of the
This confirms the assumption of a constant DC link voltage 22nd IEEE Industry Applications Society Annual Meeting,
Atlanta, Oct. 1&23, pp. 648-657 (1987).
made in chapter 2.1.
 89

(111 Zhang, H., Wheeler, N.,and Grant, D.: Switching [25] Kolar, J.W., Ertl, J., and Zach, F.C.: Calculation of
Harmonics i n the DC Link C u m n t in a PWM AC-DC-AC the Passive and Active Component Stress of Three-Phase
Conweder. Proceedings of the 30th Industry Applications PWM Converter Systems with High Pulse Rate. Proceed-
Society Annual Meeting, Orlando (FL), Oct. G12, Vol. 3, ings of the 3rd European Conference on Power Electroniw
pp. 2644-2655 (1995). and Applications, Aachen, Germany, Oct. 4-12, Vol. 111,
1121 Ray, W.F., and Davis, R.M.: The Definition and Im- PP 1303-1311 (1989)'
podance oJPower Factor for Power Electronic Converters. [26] RIFA: Eleetmlytic Capacitors 94/95, Geneml Technical
Proceedings of the 3rd European Conference on Power Elee- Information,
tronirs and Applications, Aachen, Germany, Oct. +12, pp.
799-805 (1989). 1271 SICSAFCO: Geneml Technical Data.
[ZS] Kolar, J.W., Drofenik, U,, and Zach, F.C.: Cur-
1131 Kolar, J.W., Ertl, H., and Zach, F.C.: Analytically rent Handling Copability of the Neutml Point of a Three-
Closed Optimization of the Modulation Method of a PWM
Phase/Switch/Level B o o s t - m e PWM (VIENNA) Rectifi-
Retifier Syatem with High Pulse Rate. Proceedings of the
er. Proceedings of the 27th lEEE Power Electronics Spe-
17th International Conference on Intelligent Motion, Mu-
cialists Conference, Baveno, Italy, Vol. 11, pp. 1324-1336
nich, Germany, June 25-29, pp. 370-381 (1990). (1996).
1141 K o h , J.W., Ertl, H., and Zach, F.C.: Influenee of I291 Kolar, J.W., Drofenik, U., and Zach, F.C.: On the
the Modulation Method on the Conduction and Switching
Interdependence of AC-Side and DC-Side Optimum Con-
Losses of a PWM Converter System. IEEE lkansactions
trol of Three-Phose Neutd-PointClamped (Three-Level)
on Industry Applications, Vol. IA-37, No.6, pp. 1063-1075
(1991). PWM Rectifier Systems. Proceedings of the 7th Interna-
tional Power Electronics and Motion Control Conference,
[15] Green, T.C., and Williams, B.W.: Control oflnduc- Budapest, Hungary, Sept. 2 4 , Vol. 1, pp. 40-49 (1996).
tion Motors Using Phase Current Feedback Derived from
the DC Link. Proceedings of the 3rd European Conference [30] Ham, A.M., Kerkman, R.J., and Lipo, T.A.: Simple
on Power Electronics and Applications, Aachen, Germany, Analytical and Graphical Tools for Camier Bared PWM
Oct. 4-12, Vol. 111, pp. 1391-1396 (1989). Methods. Proceedings of the 28th lEEE Power Electronics
[16] Chan, C.C., Chau, K.T., Li, Y.,andChan, D.T.W.: Specialists Conference, Vol. 11, pp. 1462-1471, (1997).
A Unified Analuis of DC Link Current in Space-Vetor
-
PWM Drives. Proceedines of the International Conference
on Power Electronics and Drive Systems, Singapore, May
26-29, Vol. 2, pp. 762-768 (1997).
117) Dahoho, P.A., Sato, Y.,and Kataolte, T.: An Anal-
y s u of h e Ripple Componenls of the Input Cunent and
Voltage oJPWM Inverters. Procmd~ngaof the Internation-
al Cotikrencr on Power ElectronicJ and Drive Systems, Sin-
gapore, Feb. 21-24, Vol. I , pp. 323-328 (1895).
1181 Kolar, J.W., Ertl, H., Zach, F.C., Blasko, V., Kau-
ra, V., and Lukaszawskl, R.: A Novel Concept JOT
Regenemlive Braking of PWM-VSI Dnves Employing (I
h s s - f k e Bmkng Iieaislor. Proceding8 of the 12th IEEE
Applied Power Electroncis Conkrenu, Atlanta, Feb.23-27,
Vol. I , pp. 297-305 (1997).
[ I Y ) Mohan, N.. Undeland, T.M., and Calm, J.M.: lnpur
Current tlomonics in 7hnswturs lnueriers Jor Induction
Motor Dnues. International Power Electronim Conference,
Tokyo,pp. 64W359 (1983).
1201 Mohan, N., Undeland, T.M., and Robblns, W.P.:
Pouer Elecironica: Convhters, Applrcaltom and Deign.
2nd Edition, New York: John Wiley & Sons (1995).
1211 Kolar, J.W., Ertl, H., and Zach, F.C.: Apprunnlote
Detrrmrnotion of the Cuwent Rms Value. of the DC Link
Copocitor of Stngle-Phase and Three-phase P W M Con-
wrter Systems. Proceedings of the 23rd Internntiwal In-
telligei,t Motion Conference, Nuremberg, Germany, Junc
22 24, pp 532-551 (1993)
1221 Sack, L.: A C C u m n t of the 1)C Link Copootor $1, R d -
r~rfionolPower Converters. Proceedings of the 34th Inrer-
nalional Power Conversion Coiiference, Nurernbcrg. Qer-
many, June 10-12, pp. 293-300 (1997).
[2RI Kolar, J.W., Ertl, H., and Zach, F.C.:A Corriprehsn-
awe Design Approaeh JOT (1 Threc.Phose fltgh-&quency
Single-Switch Dlacontmuous-Mode Boost Power Fuclor
Corrector Based on Analyttcally Denued Normahzed Con-
verter Component Ratings. IEEE Transactions on lndusiry
Applications. Vol. 31, No. 3, pp. 56It582 (1995).
1241 Holtz, J.: Pulse Width Modulation tn Electronic Pow
Convermon. Power Electronics and Variabla Requrn-
cy Drives - techno lo^ and Applications (edited by B.K.
B w ) , pp. 138- 203 (Chapter 4), IEEE Press: New York
(1997).