See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242535533

Approximate Loss Formulae for Estimation of IGBT Switching Losses through

EMTP-type Simulations

Article

CITATIONS READS

56 3,840

3 authors, including:

A.D. Rajapakse A.M. Gole

University of Manitoba University of Manitoba
113 PUBLICATIONS   3,419 CITATIONS    137 PUBLICATIONS   4,158 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Stability Prediction and Control using PMUs View project

MEMS Field mill sensor View project

All content following this page was uploaded by A.D. Rajapakse on 19 May 2014.

The user has requested enhancement of the downloaded file.

 1

Approximate Loss Formulae for Estimation of

IGBT Switching Losses through EMTP-type
Simulations
A.D. Rajapakse, Member, IEEE, A.M. Gole, Member, IEEE, and P. L. Wilson, Member, IEEE

Abstract-- This paper presents a method for simulating the switching process, which lasts only about several hundreds
switching and conduction losses of an Insulated Gate Bipolar of nanoseconds. Simulating this process considering the
Transistor (IGBT) device in an electromagnetic transient detailed physics as in the number of models described in [3]
program (emtp) without recourse to an unreasonably small time-
requires a very small time-step and iterations within a time-
step. A set of equations are derived for calculating switching
losses of an IGBT using the device switching characteristics step. This could result in an unacceptably large CPU time
approximated with piece-wise linear functions. These loss when simulating large multi-device power electronic systems
equations are integrated to a power electronic switch model of an (such as FACTS devices) using fixed time-step emtp-type
emtp-type program and used for the simulation of losses in the simulation programs.
device. This approach can be used to determine the heat One approach proposed for accurate estimation of power
generation of IGBT devices in a large class of Voltage Sourced
losses is the use pre-defined scalable switching functions
Converter (VSC) systems.
obtained through measurements to guide the simulation during
Keywords: Insulated gate bipolar transistors, Semiconductor switching transients [4]. However, this method still requires
device switching losses, Estimation of switching losses, the simulation to be carried out with very small time steps.
Simulation of power electronic systems, Pulse width modulated Another approach that has been proposed for estimation of
power converters. switching losses is the use of simple functions derived for
losses based on the typical switching waveforms [5],[6]. This
I. INTRODUCTION method was extended in [1] by deriving a set of formulae for

T HE insulated gate bipolar transistor (IGBT) is popularly

used in high power, high frequency power-electronic
applications such as pulse width modulated (PWM) inverters.
switching losses based on the predicted device’s current and
voltage waveforms. The method differs from the simple use of
lookup tables or fitted curves [7],[8] because the predicted
These applications require well designed thermal management waveforms conform to the physics of the switching process
systems to ensure the protection of IGBT s, which operate and take into account the dependency of the switching losses
with smaller safety margins due to economic considerations. on various factors such as the switching voltage, switching
Hence, tools for accurate prediction of device power current, stray inductance and the reverse recovery process of
dissipation and junction temperature become important in the freewheeling diode.
achieving optimized designs. At high switching frequencies, This paper further refines the approach of [1] and develops
switching losses constitute a significant portion of the device a simpler set of loss formulae that uses less number of
power dissipation. Therefore, accurate calculation of parameters that need to be extracted from the published data
switching losses is an important step in the thermal sheets (or test waveforms). Validation of the model was
management system design [1]. conducted using a simple laboratory set-up and by comparison
Most electromagnetic transient programs (emtp-type) with published loss curves by the manufacturer. The paper
model the power electronic devices in a circuit as on-off also presents an application example of loss determination in a
switches or two state resistances [2]. This simple PWM inverter.
representation is sufficiently accurate to simulate the system-
level electrical behaviour. However, determination of II. DERIVATION OF SWITCHING LOSS FORMULAE
switching losses requires the consideration of the physics of
A. Approach Used for Modelling of Device Losses
The losses in a power-switching device constitute of
This work was supported by the Manitoba HVDC Research Centre and the
Natural Sciences and Engineering Research Council (NSERC), Canada. conduction losses, off-state blocking losses, turn-on switching
A.D. Rajapakse and A. M. Gole are with the Department of Electrical and losses, and turn-off switching losses. The conduction loss is
Computer Engineering, University of Manitoba, Winnipeg, MB, Canada, R3T calculated in a straightforward manner as the product of the
5V6 (email: athula @ee.umanitoba.ca, gole @ee.umanitoba.ca)
P. L. Wilson is with the Manitoba HVDC Research Center, Winnipeg, MB,
device current and the forward saturation voltage; and the
Canada, R3J 3W1 (email: plwilson@hvdc.ca) blocking loss is the product of the blocking voltage and the
Presented at the International Conference on Power Systems leakage current [1].
Transients (IPST’05) in Montreal, Canada on June 19-23, 2005 The device switching losses depend on the application
Paper No. IPST05 - 184
circuit. Thus, a circuit configuration of a voltage source
 2

converter, which is the most common application of IGBT, is larger emtp time-step of several microseconds. Thus the emtp-
considered. In this configuration, each switch consists of an type simulation can be conducted with a larger time-step, with
IGBT in anti-parallel with a freewheeling diode. It is assumed a formula providing an estimate of the loss after each
that the switching devices are subject to ‘hard switching’ and switching event.
the load is inductive. The turn-on of the IGBTs is significantly The loss equations proposed in [1] are complex and
affected by the reverse recovery behaviour of the freewheeling involve several parameters that need to be found using test
diode and the parasitic inductances. A test circuit of Fig 1 waveforms. In the absence of test waveforms, these
closely represents one leg of a voltage source converter. parameters must be simultaneously tuned using the data sheet
specified switching losses at the rated conditions. This may be
Same as L a difficult task for an inexperienced user. Therefore, more
(D.U.T.)
R simplified set of loss formulae is presented in this paper.
Vcc
Moreover, the number of tuning parameters involved with
C Device Under Ic each formula is restricted to one. This is achieved by
Test (D.U.T.) following the general modeling approach of [1], but using
Vce piece-wise linear waveforms to approximate the variations of
the current and voltage waveforms during the switching
Fig. 1. Inductive switching loss test circuit events.
Figure 2 shows measured current and voltage waveforms
of a hard switched IGBT device during the (a) turn-on and (b) B. Diode Turnoff Loss Formula
turn-off transients. Switching process completes within a few In modern fast recovery diodes used with IGBTs, the turn-
hundred nanoseconds, and therefore, simulation of switching on loss is negligible (less than 1%) compared to the turn-off
losses is difficult without resorting to extremely small loss. However, the reverse recovery during the turn-off causes
simulation time steps. The type of simulation programs appreciable amount of losses in the diode and also affects the
considered in this paper, that is emtp-type programs, uses a turn-on losses of the incoming IGBT.
much larger time-step (tens of microseconds) in simulating Idealized approximation of diode turn-off waveforms are
large systems. shown in Fig. 3. In a free-wheeling diode, the initial rate of
fall of the diode current dId/dt, is determined by the turn-on
150 (a) speed of the IGBT in the opposite leg. During the initial phase
Vce (V), 5*Ic (A)

of reverse recovery, negative current through the diode

100 increases and reaches its peak value, Irrm. At this point, the
diode starts to rapidly gain the reverse voltage.
50 Vce
Ic Id0

0 Id(t) trr
-200 0 200 400 600
trra trrb
time (ns)
Vds krr⋅ trrb
300 0.1Irrm
(b) Vce t
250
Ic
Vce (V), 5*Ic (A)

200 Irrm
0.1Vd0
150 0.1Vd0
Vd(t)
100 -Vd0
t0(rr) t1(rr) t2(rr) t3(rr) t4(rr)
50
Fig. 3. Approximated waveforms of diode turn-off transient
0
-400 -200 0 200 400 600
time (ns) The waveforms can be characterized using the parameters
Fig 2. Measured IGBT (a) turn-on and (b) turn-off waveforms.
Irrm (peak reverse recovery current), and trr (the reverse
A method to meet these apparently contradictory recovery time). In the interval [t0(rr), t2(rr)], Id is linearly
requirements was proposed in [1] by developing algebraic decreasing, with Vd=Vds (on state forward voltage drop).
equations that represent the voltage and current waveforms Knowing the initial dId/dt , (which depends on the IGBT turn-
during the switching event. These algebraic equations are on rise time), the parameters trra, and trrb are estimated as:
based on the fact that current and voltage waveforms during
t rra = I rrm  dIdtd 
 (1)
the switching are principally a function of the pre- and post-  t1( rr ) 
switching voltages and currents [3]. A formula for switching
t rrb = 1.11(t rr − t rra ) (2)
energy can be obtained by integrating the product of the above
voltage and current equations that ‘fill-in’ the intermediate It is also assumed that diode voltage drops to 90% of its
sub-microsecond values of voltage and current within the reverse blocking voltage during [t2(rr), t3(rr)] = krr.trrb.
 3

An energy loss occurs in the diode during the reverse When the load current is fully transferred to the IGBT at
recovery, particularly during building up of reverse voltage. t2(on), the outgoing freewheeling diode starts to turn off and
This can now be computed as: forces its reverse recovery current through the IGBT. This
t4 ( rr ) current appears as an overshoot in Ic during [t2(on),t4(on)]. When
Wrec = ∫V
t0 ( rr )
d (t ) I d (t )dt the diode reverse recovery current reaches its peak value at
t3(on), the collector-emitter voltage Vce begins to fall (with the
I diode gaining its reverse voltage). The duration, during which
Wrec = 0.5Vds I d 0
( ) + 0.033V
d0
dI d
I
d 0 rrm rra t the reverse recovery current increases, can be computed by
dt t1( rr ) . (3) substituting (4) into (1):
+ Vd 0 I rrm (0.467 − 0.433k rr + 0.15k rr2 )t rrb I rrm
t rra = tr (6)
The parameter krr [0.0-1.0] can be determined by 0 .8 I c
substituting data sheet value of Wrec and the corresponding The time taken for current to fallback to on state value after
measuring conditions to (3). the peak is equal to trrb defined in (2). It is assumed that
C. IGBT Turn-on Loss Formula collector-emitter voltage rapidly falls to 10% of its off-state
value within a time of krr.trrb. The remaining voltage gradually
Fig. 4 shows approximate waveforms for the hard turn-on
drops and reaches the on-state saturation voltage, Vces at time
transient of an IGBT. In data sheets, the turn-on behaviour is
t5(on). The turn on energy loss can be computed as
characterized by the turn on delay time, td(on), the rise time, tr, t5 ( on )
and the turn-on energy, Won.
Vth
Won = ∫V
t0 ( on )
ce (t ) I c (t )dt
Vge(t)
Won = 0.05Vce 0 I c 0 t d ( on ) + 0.225Vce 0 I c 0 t r
+ Vce′ 0 I c 0 [0.394t r + t rra + 0.55k rr t rrb ]
trra trrb
Vce0
+ 0.5Vce′ 0 I rrm t rra + 0.5Vces I c 0 t vtail
Lp ( )
dI c
dt
krrtrrb Irrm
Ic(t) + Vce′ 0 I rrm (0.033 + 0.483k rr − 0.167k rr2 )t rrb
(7)
Ic0
0.9Ic0 + Vce′ 0 I c 0 k vtail [0.5(1 − k rr )t rrb + 0.05t vtail ]
tvtail
td(on) tr + 0.167Vce′ 0 I rrm k vtail (1 − k rr )t rrb
where
Vces
Vce(t) tvtail
0.1Ic0 0.1Vce0 k vtail = . (8)
t0(on) t1(on) t2(on) t3(on) t4(on) t5(on)
t tvtail + (1 − k rr )t rrb
Fig. 4. Hard switching turn-on transient waveforms for loss calculation
D. IGBT Turn-off Loss Formula
The turn-on gate pulse applied at t0(on) raises the gate
voltage Vge gradually, with a rate of rise determined by the
input capacitance of the IGBT and gate drive resistance. After Vge(t)
a time td(on), when Vge reaches a threshold voltage Vth, the
collector current, Ic, starts to rapidly rise and the load current
in the freewheeling diode (of the opposite leg) gradually
transfers to the IGBT. The rate of rise of collector current can Vce0
Cp ( )
dVce
dt
Vos
Vce(t)

be determined using the definition of rise time: Ic0

dI c 0.8 I c 0 0.9Ic0
= (4) koff⋅td(off)
dt tr
td(off) tf titail
During this current rise, the device (collector-emitter)
voltage Vce(t) experiences a drop due to parasitic inductance Vces
0.1Vce0 0.1Ic0
(Lp). The voltage drop due to parasitic inductance is not Ic(t)
t
instantaneous due to parasitic capacitance effects and the t0(off) t1(off) t2(off) t3(off) t4(off)
nonlinearities in the IGBT transconductance. It is assumed Fig. 5. Approximated hard switching turn-off transient waveforms
that the voltage Vce drops linearly over the period [t1(on),t2(on)]
The IGBT’s turn-off behaviour shown in Fig. 5 is
and reaches a plateau in the voltage waveform. This plateau
characterized in data sheets by the turn-off delay time, td(off),
voltage, V’ce, can be found as
fall time, tf, and turn-off energy, Woff. The turn-off process
dI c 0.8 I c 0 starts on the application of negative gate voltage at time t0(off).
Vce′ 0 = Vce 0 − L p = Vce 0 − Lp . (5)
dt tr The input capacitance of the IGBT discharges gradually
 4

reducing gate-emitter voltage, Vge, but the collector-emitter off losses at different collector currents. Two commercially
voltage, Vce, remains essentially unchanged until Vge drops available IGBTs (with anti-parallel diodes) from different
sufficiently to drive the IGBT out of saturation. This initial manufacturers were considered, the SNR13H2500 from ABB,
period [t0(off),t1(off)] is assumed to be equal to koff⋅td(off) as rated at 2.5 kV, 1300 A and the IXGK50N60BD1 from IXYS
indicated in Fig. 5. Thereafter, the collector-emitter voltage rated at 600 V, 50 A. The variation of calculated reverse
rises rapidly at a rate of recovery, turn-on and turn-off energy losses (as in(3), (7) and
dVce 0.9Vce 0 (12)) for these two devices are shown in Figs. 6 and 7
= . (9) respectively superimposed on the loss curves from the
dt (1 − k off )td ( off ) manufacturers’ data sheets. As can be seen, the equations
There could be a drop in the collector current due to the developed in this paper capture with reasonable accuracy, the
rate of rise of collector voltage, if the parasitic capacitance is variation of losses with current.
significant. Considering the idealized waveforms, the resulting
6
plateau in the collector current, I’c0, can be approximately Won (data) ABB SNR 13H2500 @125oC
determined as 5 Won (cal)

dVce 0.9Vce 0 Woff (data)

I c′0 = I c 0 − C p = 1− Cp (10) 4
Woff (cal)

(1 − k off )td ( off )

W , W , W (J)
dt Wrec (data)

rec
Wrec (cal)
3
When Vce reaches the forward blocking voltage Vce0, at

off
t2(off), the freewheeling diode become forward biased and starts

on
2
to take over the load current. The IGBTs internal construction
includes a MOSFET driving a bipolar transistor, and due to 1

the mechanisms involved in these devices, the collector

0
current Ic initially has a rapid fall; followed by a more gentle 0 250 500 750 1000
I (A)
1250 1500 1750 2000

drop towards extinction at time t4(off). The rapid drop in current Fig. 6. Variations of Won and Woff with Ic cfor ABB IGBT (Vce0=1250 V)
through the parasitic inductance produces an overshoot in the 10
Won (data) IXYS IXGK 50N60ED1 @125oC
voltage Vce . The peak overshoot voltage is determined from Won (cal)

dI c 0.8 I c′0
8
Woff (data)
Vos = L p = Lp
W , W , W (mJ)

(11) Woff (cal)

dt tf
rec

6 Wrec (data)
Wrec (cal)
off

It is assumed that the peak of the overshoot occurs at the

4
midpoint of rapid current fall. The tail current, which is
on

assumed as 10% of the on state collector current, dies down

2
after a period of titail at t4(off).
Based on the idealized waveform, an approximate
0
expression for the turn-off energy can be developed as 0 20 40
Ic (A)
60 80 100

t4 ( off )
Fig 7 Variations of Eon and Eoff with Ic for IXYS IGBT (Vce0=480 V)
Woff = ∫ Vce (t ) I c (t )dt
t0 ( off ) 2) Experimental Confirmation
An IGBT-diode package rated at 600 V, 25 A
Woff = 0.5Vces I c 0 k off t d ( off ) + 0.05Vce 0 I c 0 k off td ( off ) (International Rectifier IRG4PC40KD) was used for the
+ 0.275Vos I c′0 t f + 0.55Vce 0 I c′0 (1 − k off )t d ( off ) (12) experiment. The test circuit as in Fig. 1 was constructed with
the parameters as used in [1] (L=20 mH, C= 4500 µF). The
+ Vce 0 I c′0 [0.55t f + 0.05ttail ] on-state current in the device was adjusted by selecting the
resistance R appropriately.
E. Validation of Switching Energy Models
The current and voltage waveforms for IGBT turn-on and
The approach developed above was validated by turn-off transients were recorded using a high speed
comparison with published results from manufactures’ data oscilloscope. The tests were repeated at various combinations
sheets as well as with a laboratory setup of the circuit in Fig. of several different voltages (150 V and 120 V) and currents
1. (15 A, 20A and 25 A). The experimental energy losses were
1) Comparison with Manufactures Data Sheets obtained by integrating the product of the measured voltage
The possibility of using the derived expression for and current during switching. Table 1 compares these
switching energies with data sheet parameters was studied by experimental energy losses with the corresponding calculated
computing switching energy versus collector current curves energy losses from (7) and (12). It also gives the energy losses
for several IGBTs. The parameters that are not available in the calculated using the formulae proposed in [1]. It can be seen
data sheets were estimated by adjusting their values to match that the new formulae proposed in this paper are more
with the switching energy at the rated current. Those accurate with the maximum error less than ±10% for all cases.
parameter values are then used to predict the turn-on and turn-
 5

The number of non-standard parameters (which are not can be appropriately changed so that the average power loss
provided in manufacturers datasheets) necessary for the loss due to switching is ‘felt’ in the network.
equations (krr in (3), tvtail in (7) and koff and titail in (12) ) are The accumulated switching energy loss in the device at
less than number of such parameters used in the equations time t , W(t), is computed at each time step as
given in [1]. These parameters can be readily determined if W (t ) = [W (t − ∆t )e − ∆t / TW + Won (t ) + Woff (t )] (13)
test waveforms are available or otherwise determined
assuming that the accumulated energy decays exponentially
approximately as explained above, using the switching energy
values given at the rated conditions. with a time constant of TW. ∆t is the simulation time-step. The
decaying energy is reflected in the electric network as power
TABLE I: COMPARISON OF MEASURED AND CALCULATED POWER LOSSES FOR dissipation in the device. In order to achieve this, if the IGBT
IRG4PC40KD IGBT is in conduction, its forward voltage drop is adjusted
Vce0 Ic0 Won (mJ)
(V) (A) Meas *Cal[1] Error (%) Cal Error (%)
according to
150 15 0.193 0.216 11.9 0.194 6.4 W ( t ) + W (t )(1 − e − ∆t / TW ) , (14)
150 20 0.274 0.312 13.9 0.279 1.7 Vces = cond
150 25 0.353 0.417 18.1 0.375 0.6 ∆t ⋅ I c
120 15 0.155 0.173 11.6 0.155 0.0 and if the IGBT is off, its leakage current is adjusted so that
120 20 0.221 0.242 9.5 0.217 -1.7
120 25 0.310 0.325 4.8 0.295 -4.8 W ( t ) + W ( t )(1 − e − ∆t / TW ) . (15)
I cs = blok
Vce0 Ic0 Woff (mJ) ∆ t ⋅ V ce
(V) (A) Meas *Cal[1] Error (%) Cal Error (%)
150 15 0.183 0.203 10.9 0.177 -3.3
A similar approach is adapted for the diode too.
150 20 0.289 0.318 10.0 0.287 -0.5
150 25 0.394 0.478 21.3 0.425 7.9 Device Loss
Model
120 15 0.112 0.125 11.6 0.120 7.0 Electrical Device
120 20 0.183 0.205 12.0 0.184 0.4 parameters Parameter temperature
120 25 0.289 0.321 11.1 0.285 -1.4 Calculation
*Cal[1]: Calculated using the loss formulae given in [1] Ambient
Rest of Modified Thermal conditions
model
the Switch Path
parameters
III. REPRESENTATION OF SWITCHING LOSSES IN EMTP-TYPE Electrical Model Model
SIMULATIONS Network Loss
Operating Estimation Power loss
The general switch model of the emtp-type program conditions
considered (PSCAD/EMTDCTM) consists of an ideal switch in Fig. 8. Interface between the device model and the network model
series with a resistance whose value dependent on the switch
state and a constant dc source representing the forward IV. APPLICATIONS
voltage drop. A loss estimation module is integrated to the
existing switch model as an additional calculation layer. Thus A. Example: VSC with Hysteresis Current Control
its traditional structure is retained in the network simulation. The voltage source converter (VSC) example given in [1]
The losses in the device (including conduction, blocking and was simulated with the loss estimation methods developed
switching losses) are estimated at each time step after the above. The inverter shown in Fig 9 uses six IGBT/Diode
network solution by observation of the pre- and post- modules (parameters used in the model corresponds to IXER
switching currents and voltages. 35N120D1 IGBT/diode module). The converter is controlled
Several manufacturers provide information on the using hysteresis current control, which maintains the current
temperature dependence of certain switching parameters (i.e., in any phase within a specified tolerance band around the
Vces and switching times) and this information is also reference current setting. Unlike conventional sinusoidal
represented in the developed model. Once the losses in the PWM, the switching events and their frequency are highly
device are known, its internal temperature can be found using load dependent and so an a-priori estimation of switching
a dynamic model of the thermal path, such as the model losses is not possible.
proposed in [1]. The computed device temperature is then
used to change the parameters of the switch loss model for the T1 T3 T5
E D1 D3 D5 R L
next time-step.
Because the losses are analytically computed separately in
the loss calculation module they do not appear as losses in the
electrical circuit simulation. Since the network simulation time E T4 T6 T2
step is much larger than the switching times, it is impossible to D4 D6 D2
represent instantaneous power losses due to switching in the
Fig. 9. Hysteresis controlled voltage source inverter
network. However, the average power losses due to switching
can be represented in the network simulation by slightly Figure 10 shows the load (L=5mH, R=5Ω) currents in three
modifying the simple switch model as shown in Fig. 8. The phases when the hysteresis band is set to 5A. The currents and
values of the voltage source representing the forward voltage voltages across the IGBT T1 and diode D1 are shown in Fig.
drop and the current source representing the leakage current 11. Figure 12 shows the total power losses in IGBT T1 and
 6

diode D1 as functions of time. The curves PT1cal and PD1cal have significant effect on the switching losses and thus they
show the instantaneous power losses calculated using the are included in the proposed loss equations. Determination of
methods described in Section II. In IGBT, the conduction loss the values of these components is discussed in [9] and [10].
is proportional to the magnitude of the current. The spikes at
the beginning and end of conduction periods correspond to V. CONCLUSIONS
turn-on and turn-off losses respectively. In the diode, reverse This paper presented an improvement to a method
recovery energy losses appear as spikes at the end of developed for estimating switching losses of an IGBT
conduction periods. The curves denoted as PT1meas and PD1meas subjected to hard switching conditions without resorting to
show the respective device power dissipations as measured in very small time step simulations. This is achieved by deriving
the network. Due to the averaging, the peaks in the power a set of formulae for approximately estimating the turn-on and
dissipation (due to switching) are smoothen and distributed turn-off switching losses using predicted trajectories of the
over a period of time. The time constant TW should be selected device current and voltage variations between their respective
so that it facilitates energy dissipation without buildup in the pre- and post-switching values.
accumulated energy W(t). However, a too small time constant According to test results, the formulae presented in this are
can cause overly large values of Vces and Ics. more accurate than those of [1] and at the same time, reduced
0.04 Ia Ib Ic the number of non-standard parameters to be extracted from
manufacturer’s data sheets. The method suggested for
0.02
representing the average effect of switching losses in the
I , I , I (A)
a b c

0
network works fairly well as demonstrated in the example of a
-0.02 PWM inverter simulation.
-0.04

0.3 0.305 0.31 0.315 0.32 0.325 0.33 0.335 0.34 VI. REFERENCES
time (s)
[1] A.D. Rajapakse, A.M. Gole and P.L. Wilson, “Electromagnetic transient
Fig. 10. Load currents simulation models for accurate representation of switching losses and
thermal performance in power electronic systems,” IEEE Trans. Power
600
Delivery, vol. 20, No.1, pp. 319-327, Jan. 2005.
400 [2] H. Selhi, and C. Christopoulos, “Generalized TLM switch model for
V (V), 10* I(A)

200 power electronics applications,” IEE Proc. Scientific Measuring

0 Technology, vol. 145, no.3, May 1998.
I
-200 T1
V [3] K. Sheng, B.W. Williams and S.J. Finney,”A review of IGBT
T1
-400
I
models,”IEEE Trans. Power Electronics, vol.15, no. 6, Nov. 2000, pp
-600 D1 1250-1266.
V
D1 [4] C. Wong, “EMTP modeling of IGBT dynamic performance for power
0.178 0.18 0.182 0.184 0.186 0.188 0.19 dissipation estimation,” IEEE Trans. Industry Applications, vol. 33, pp
time (s) 64 –71, Jan.-Feb. 1997.
Fig. 11. Device voltages and currents [5] J. Qian, A. Khan, A. I. Batarseh, “ Turn-off switching loss model and
250
P analysis of IGBT under different switching operation modes,” in Proc.
200
T1meas
21st International Conference on Industrial Electronics, Control, and
P
D1meas Instrumentation, 6-10 Nov. 1995, vol.1, pp. 240 –245.
150 P [6] M. Bland, P. Wheeler, J. Clare and L. Empringham, “Comparison of
(W)

D1cal
P calculated and measured losses in direct AC-AC converters,” in Proc.
LOSS

100 T1cal
IEEE 32nd Annual Power Electronics Specialists Conference, 17-21
P

50 June 2001, vol.2, pp. 1096 –1101.

[7] S. Azuma, M. Kimata, M. Seto, X. Jiang, H. Lu, D. Xu, and L. Huang,
0 “Research on the power loss and junction temperature of power
0.178 0.18 0.182 0.184 0.186 0.188 0.19 semiconductor devices for inverter,” in Proc. of the IEEE International
time (s) Vehicle Electronics Conference, 6-9 Sept. 1999, vol.1, pp. 183 -187
Fig. 12. Power losses in T1 and D1 [8] F. Blaabjerg, U. Jaeger, S. Munk-Nielsen, and J.K. Pedersen,
“Comparison of NPT and PT IGBT-devices for hard switching
applications,” in Conference Record of the 1994 IEEE Industry
The estimated total losses (using loss formulae) in the Applications Society Annual Meeting, vol.2, pp. 1174 –1181.
IGBT T1 is 24.39W (conduction=17.62W, turn-on=3.12W, [9] M.C. Caponet, F. Profumo, R.W. De Doncker, and A. Tenconi, “Low
turn-off=2.43W, and blocking=1.23W) and that of diode D1 is stray inductance bus bar design and construction for good EMC
performance in power electronic circuits”, IEEE Trans. Power
7.10W (conduction= 5.63W, turn-off=1.20W, and Electronics, vol.17 , no. 2, Mar. 2002, pp. 225 – 231.
blocking=0.26W). The average power dissipations measured [10] K. Xing, F.C. Lee, and D. Boroyevich, “Extraction of parasitics within
in the network are: IGBT T1=23.26 W and diode D1=5.67W. wire-bond IGBT modules,” Proc. of Applied Power Electronics
Conference and Exposition, 15-19 Feb. 1998, pp 497 – 503.
The above results show that the device dissipations are fairly
closely reflected in the network simulation. This example also
shows that the presented method can be used for analyzing the
contribution of different type losses to device heating and
determining the device losses under different loads, different
control settings and with different device types.
As indicated in Sections IIC and IID, parasitic inductances

View publication stats