No 99 Electrical Engineering 2019 DOI 10.21008/j.1897-0737.2019.99.0003
___________________________________________________
* Cracow University of Technology
Wojciech MYSIŃSKI*, Bartłomiej SYSŁO*
COMPARISON BETWEEN THERMAL SIMULATION RESULTS GENERATED BY PLECS SOFTWARE
AND LABORATORY MEASUREMENTS
This article deals with the subject of simulation of power losses and thermal process- es occurring in semiconductors, as illustrated by an example of a DC/DC buck converter.
The simulations were performed in PLECS software. The results obtained from the pro- gram were compared with measurement results of a laboratory converter model.
The physical model is based on the same components as assumed in the simulation.
Similarly, the parameters of the transistor control signal were the same. During operation of the converter, the temperature changes were analyzed using a K-type thermocouple.
Based on the obtained results of the temperature measurement in the steady state of the converter operation, the correctness of the simulation carried out in the PLECS program was verified and confirmed.
KEYWORDS: Thermal simulation, PLECS, buck converter, thermal time constant, IGBT, Diode, power losses.
1.INTRODUCTION
Power losses in semiconductor elements and related thermal phenomena are an important element of the design process of electronic and power electronic devices. Semiconductor manufacturers provide increasingly accurate datasheets of manufactured elements, comprising accurate equivalent thermal circuit dia- grams. On the other hand, there is software available on the market that allows for efficient use of information provided by manufacturers and for determining the temperature values of individual components. One of such programs is PLECS, dedicated for power electronics devices, being a product of the Swiss company Plexim. Its distinguishing feature is the ability to combine electric, mechanical, thermal and magnetic elements in the simulation process, which allows, for example, for observing entire drive systems, including temperature changes, in which power semiconductor junctions work under given load and ambient conditions.
In general, power losses in semiconductor switching components can be divided into conduction losses and switching losses [3]. In the case of an IGBT transis- tor, conductive losses depend on voltage drop on the conductive collector- emitter junction and on collector current:
1
( ) 0
1 ( ) ( )
t
cond T C CE
P i t v t dt
T
(1)where:
T – transistor switching period, t1 – transistor conducting time,
iC(t) – instantaneous value of collector current,
vCE(t) – instantaneous value of collector-emitter voltage.
On the other hand, switching losses are provided by manufacturers as energy losses characteristics during turn-on and turn-off, dependent on the switched-off transistor collector-emitter VCE, collector current of turned-on transistor IC, junc- tion temperature TJ and switching times, that are influenced indirectly by the value of gate resistance RG:
( ) ( ( , , , ) ( , , , ))
sw T sw on CE C j G off CE C j G
P f E V I T R E V I T R (2) where: fsw – transistor switching frequency, EON – energy loss during turn-on, EOFF – energy loss during turn-off.
Switching loss characteristics, provided by manufacturers for transistors with an integrated anti-parallel diode, also take into account losses from transistor tail current and reverse current of the anti-parallel diode. Therefore, no switching losses are calculated separately for the anti-parallel diode. The conduction losses for the diode are determined by the following formula:
1
( )
1T ( ) ( )
cond D F F
t
P i t v t dt
T
(3)where: iF(t) – instantaneous value of diode forward current, vF(t) – instantaneous value of diode forward voltage.
Total power losses can be defined as:
( ) ( ) ( )
tot cond T sw T cond D
P P P P (4)
Depending on power dissipated on semiconductor junctions, it is possible to determine the temperature at which the junctions operate, at known ambient temperature and a certain thermal resistance between the semiconductor junction and the ambient – Rthj-a. An example of a thermal model and equivalent chain of thermal resistances and capacities is shown in Figure 1.
Manufactu thermal m thermal n accordanc
where:
ΔT – temp Ptot – diss Rth – therm For the ca
where: Rth mal resist temperatu Due to th possible p known, an DC/DC b current-ty ing chara transistors switching sembly th the lack o
Fig
urers of sem model from t
etwork [4, 5]
ce with formu
perature diffe ipated power mal resistanc ase from Figu
Tj a
hj-c – junctio tance (influe ure (generally he comparativ power electro nd so that it buck convert ype output ci acteristics fo s used in th g transistor a
hat is an ON of a stabilized
. 1. Thermal mo
miconductor the junction t ]. For a stead ula no. 5 [2]
ference, r, ce.
ure 1, it will
a PtotRthj a
n to case the nce of therm y defined in ve aim of th onics system t is simple i ter [1] was
rcuit. It is al or power tra his experime and the freew N Semicondu
d high-voltag
odel of a semic
devices pro to the case, dy state, basi
.
tot th
T P R
be:
a Ptot(Rthj
ermal resista mal grease or heat sink dat his article, it m so that the p
in practical i chosen, with so the basic ansistors are ent [6]. In wheeling dio uctor NGTB2 ge DC source
onductor devic
ovide accura in the form ic calculation
h
c Rthc s R
ance, Rthc-s – r pad), Rths-a
tasheet).
was decided phenomena o implementat h a voltage- configuratio determined the configu ode are base
25N120FL3W e with suffic
e [3]
ate informati of the Cauer ns can be car
ths a) R
case to heat – heat sink t d to select th occurring in tion. For this type input c on in which t d, including uration used,
ed on the sa WG. Howev ient current e
ion on the r or Foster rried out in (5)
(6) t sink ther-
to ambient he simplest it are well s reason, a circuit and the switch-
the IGBT , both the ame subas-
ver, due to efficiency,
it was de former an circuit is p sistor Q1 gate is sho
The fi PLECS pr and Q2, w sufficient Q1 and Q
Fig
To enter NGTB25N lated. To
ecided to sup nd a single- presented in is constantly orted with th
F
2.TH rst stage inv rogram, to a which allowe temperature Q2 from overh
g. 3. Turn-on en
r the trans N120FL3WG
determine t
pply the me -phase bridg Figure 2. In y in the block he emitter, wh
Fig. 2. Schemati
HERMALS volved initi assess the lev ed for the sel e, and at the heating durin
nergy losses ch
sistor mode G, the charac the losses du
asurement s ge rectifier.
this applicat king state, wh
hich prevent
ic of a laborator
SIMULATI al simulation vel of power
ection of hea same time p ng measurem
aracteristics of
el with the cteristics list uring switch
system using The measur tion, the anti hile in the ca ts its switchin
ry buck convert
IONSINPL n of the con r losses occu at sink Rths-a prevents the j ments.
NGTB25N120
e integrated ted in the dat hing the tran
g a variable rement and
i-parallel dio ase of transis
ng.
ter
LECS nsidered syst urring at tran at a level tha junctions of
0FL3WG transis
d anti-paral tasheet shoul nsistor on an
autotrans- simulation de of tran- stor Q2 the
tem in the nsistors Q1 at achieves transistors
stor
llel diode ld be tabu- nd off, the
characteri tion temp losses, ch should be ceed in th junction t characteri ode datash
The equiv also be d scribed in is present
istics EON,OFF
eratures need haracteristics e entered. In he same way
temperature.
istics are intr heet. Examp
Fig. 4. Ou
valent therma determined.
n detail in the ed in Figure
F = f(UCE) an d to be enter s UCE = f(IC
order to dete y, entering th As mention roduced in th les of charac
utput characteris
al chain mod The individu e datasheet. T
5.
Fig. 5. Simu
nd EON,OFF = red in a look
C) for differe ermine cond he characteri ned earlier, he case of a cteristics are
stics of NGTB2
dels of both ual elements The simulati
ulation schemat
= f(IC) for di kup table. To ent transistor uction losses istics Uf = f
no separate transistor w shown in Fig
25N120FL3WG
the transisto s of the the ion scheme i
tic in PLECS
fferent trans o determine c r junction te s of the diod f(If) for diffe e diode switc with an anti-p
gures 3 and 4
G transistor
r and the dio ermal networ in the PLEC
istor junc- conduction emperature de, we pro- erent diode
ching loss parallel di-
4.
ode should rk are de- S program
Based on laboratory were dyna of alumin manufactu der the tr insulation
ac
thc s
V
R
Figures 6-
The simul data mem termine th with the a heating pr
the known c y system we
amic enough num alloy A urer, equal to ransistors wa n purposes). T
300 , 2 0.4 ,
D
s
V C K R W
-8 show the s
Fig. 6
lation of ther mory, so it w
he temperatu appropriate h rocess of the
capabilities o ere experimen
h. The labor A6060, with o 3.7 K/W. T as also taken
The paramet 3.1 ,
ths a 3.7
mF L R K
W
simulation re
6. Voltage and c
rmal phenom was decided t
ure of the c hardware ca e system up to
of the bench ntally select atory model
a known th The thermal n into accou ers of the sim
1.6 ,
, 25
O
o a
L mH
K T C
W
esults.
current wavefor
mena require to use the St components apabilities, it o a steady st
instruments ted so that th
uses an A5 hermal resist resistance o nt at 0.4 K/
mulated circu
( )
44.28
, 0
O
T
R
C DC
rms of buck con
es a substanti teady-State A
in the stead t is possible
ate.
, the parame he thermal p 723/3 heat s tance specifi of the therma /W (used for uit are as foll
, 10
.5
fS k
nverter
ial amount o Analysis opt dy state. Nev
to simulate
eters of the phenomena sink, made ied by the al pads un-
r electrical lows:
, kHz
of time and tion to de- vertheless,
the entire
3.V In orde formulas occurring
Fig. 8. Jun
VERIFICAT er to verify t
introduced i on semicond
( 2 cond T
out
P
I
( )
sw T sw
P f
Fig. 7. Power
ctions and heat
TIONOFP the correctne in the literat ductors were
)
(25 )
( (
( o
T OUT
CE C
I V
r
on off ou ref
E I
I
losses – Steady
t sink temperatu
POWERLO ess of the co ture [3] for e calculated i
0(25 )
( 2
CE oC
r j
V T
TC T
Ki K
ut in
f ref
V V
y state analysis
ures – Steady st
OSSESCA onducted sim the buck co independentl
(
( 25
5 )))
o
V j
o
T
TC T C DC
(1
Ki
TCEsw
tate analysis
ALCULATI mulations, ba
onverter, pow ly.
) o ))
T
C
( ))
w TjTref
ION ased on the
wer losses
(7)
) (8)
( ) 0(25 )
2 (25 ) ( )
( ( ( 25 ))
( ( 25 )))
o
o
o
cond D OUT F C V j
out F C r j o D
P I V TC T C
I r TC T C DC
(9)
( )D 1 ( )T
DC DC (10)
where: IOUT – average load current, DC(T), DC(D) – transistor/diode duty cycle, TCV, TCr – temperature coefficients of the on-state characteristic, Iref, Vref, Tref – reference values (datasheet), Ki – exponents for the current-dependency of switching losses (T:1, D:0.6), Kv – exponents of voltage-dependency of switch- ing losses (T:1.35, D:0.6), TCEsw – Temperature coefficients of the switching losses (0.003), TCErr – temperature coefficients of the diode switching losses (0.006).
The table below shows a comparison of calculated power losses with simulation.
Temperatures are calculated referring to formula 6. Junction to case thermal resistances of IGBT and anti-parallel diode are: Rthj-c(T) = 0.43 K/W, Rthj-c(D) = 0.78 K/W.
Table 1. Comparison of results obtained from simulation and formulas.
Power losses Temperature
Transistor Diode Total IGBT Diode Heatsink Conduction Switching Conduction Tj(T) Tj(D) Ts Simulation 1.77 W 0.92 W 1.52 W 4.21 W 42.9 oC 42.5 oC 40.6 oC
Formulas 1.69 W 0.89 W 2.31 W 4.89 W 44.2 oC 44.9 oC 43.1 oC
As shown in Table 1, there are slight discrepancies between the results of simu- lations and calculations. These differences may be a result of different methods of using data from datasheets, inaccuracies in the selection of factors for the calculation method, as well as from the fact that not all parameters in the datasheet are well described for small currents. Generally, the convergence of results can be considered acceptable.
4.LABORATORYMODELMEASUREMENTS
The final stage is to examine the laboratory model. The power circuit is based on the mentioned NGTB25N120FL3WG transistors, connected in a configura- tion such as shown in the schematic in Figure 1. All circuit parameters are the same as in the case of PLECS simulations (chapter 2).
The transistor Q1 is driven with an isolated gate driver based on the STGAP2S [7] IC. Control pulses are generated using a digital waveform generator. A dif- ferential probe was used to measure collector-emitter voltage, while the current probe was used to measure the load current. The temperature of the heat sink was also measured by a K-type thermocouple contacting the heat sink, addition-
ally secur timeter to ture acqui was affixe is shown i
The meas Below are was exper perature fs = 10 kH Figures 1
red with a sp ogether with isition, the da ed between t in Figure 10
Fig. 9. Drawi
surements we e examples o rimentally d after ca. 3 Hz, DC(T) = 0
1 and 12 pre
pecial thermo an optical co ata was proc the transistor
.
ing of the labor
Fig. 10. P
ere carried o of recorded w determined th 30 minutes.
0.5, VD = 30 sent the resu
o-conducting ommunicatio cessed in ded
rs, as shown
atory model of
Photo of measu
out at an amb waveforms. B hat the heat
For the c 00 V, the he ults of one of
g adhesive: A on interface dicated softw in Figure 9.
buck converter
uring stand
bient temper Based on sev
sink reache case under eat sink temp
f the measure
AG Termoglu was used fo ware. The ther
. The measur
r power circuit
rature of app veral measur es its steady- consideratio perature reac ements.
ue. A mul- r tempera- rmocouple ring bench
prox. 25oC.
rements, it -state tem- on, i.e. at
ched 58oC.
Fig. 11. Y
There tained on on the ba phenomen by the ma ter, it dep surface fin To determ is necessa
Yellow – Q1 con
5.I is a large d the basis of asis of measu
non is the fa anufacturers pends, among
nish, as well mine the ther
ary to specify
ntrol signal, blu at fs = 10kH
Fig. 12
INTERPRE discrepancy b f PLECS sim
urements (58 act that the th is approxim g other thing as the dissip rmal resistanc y a thermal ti
ue – Q1 collecto Hz, DC(T) = 0.5,
. Heat sink tem
ETATION between the mulation (40.
8.3oC). Und hermal resist mate. In pract gs, on the ori pated power ce of the hea ime constant
or-emitter volta VacA=300 V
mperature
OFRESUL temperature 6oC) and the oubtedly, wh tance of the tice, this is n ientation of t
[4, 5].
at sink under t [8].
age, pink – outp
LTS e of the hea e temperatur hat contribut
heat sink (R not a constan
the heat sink r specific con
put current,
at sink ob- re obtained
ted to this Rths-a) given nt parame- k, air flow,
nditions, it
th R m cth s R Cth th
(11)
where:
τth – thermal time constant, m – mass of the substance, cs – specific heat of the substance.
Assuming a simplification that the thermal capacity of the heat sink significantly exceeds the thermal capacity of the other elements (as well as thermal resistanc- es), based on the heat sink's time constant with the known thermal capacity, its thermal resistance was calculated.
The thermal time constant of the heat sink under is determined graphically at the level τth = 457 s. Specific heat of A6060 aluminum alloy is: cs = 898 J/(kg.K).
The mass of the heat sink with assembly screws is: m = 78 g.
3
457 6.52
78 10 898
th th
s
R K
m c W
(12)
The calculated thermal resistance significantly exceeds the value assumed in simulations (3.7 K/W). Table 2 shows the comparison of simulation results in PLECS with real measurements, taking into account indicatively determined sink-ambient thermal resistances.
Table 2. Temperature comparison between PLECS simulation and laboratory model after independent Rths-a correction.
Ro =44.28 Ω, Lo = 1.6mH, CD = 3.1mF Io Avg
[A]
Corrected Rths-a
[K/W]
Heat sink temperature [oC]
Conditions PLECS simulation Measurements
fs = 10kHz, DC(T) = 0.5,
Vac Amp=300 V 3.12 6.52 52.5 58.3
fs = 5kHz, DC(T) = 0.5,
Vac Amp=300 V 3.25 6.44 50.4 52.9
fs = 20kHz, DC(T) = 0.5,
Vac Amp=300 V 3.26 6.47 58.1 69.4
fs = 10kHz, DC(T) = 0.1,
Vac Amp=300 V 0.63 7.90 28.4 35.3
fs = 10kHz, DC(T) = 0.9,
Vac Amp=300 V 5.65 5.75 77.5 77.4
fs = 10kHz, DC(T) = 0.5,
Vac Amp=250 V 2.57 6.60 44.4 49.9
fs = 10kHz, DC(T) = 0.5,
Vac Amp=350 V 3.92 6.62 61.6 67.7
6.CONCLUSIONS
As shown in Table 2, there are some discrepancies between the measure- ments of the real system and the simulations in the PLECS program, neverthe- less the correlation is quite good.
The simulation results are in most cases slightly understated, which is most like- ly the effect of inaccurately determining the characteristics of losses during switching and conduction for small currents in the datasheets, they were approx- imated in this interval (Figure 3, 4). An excellent confirmation is the measure- ment at fs = 10 kHz, DC(T) = 0.9, Vac Amp = 300 V, in which the discrepancy be- tween the PLECS simulation and the measurement, taking into account the cor- rect thermal resistance of Rths-a, is only 0.1oC. Then, the average output current was at the level of 5.65 A, which was the highest achieved value. To confirm this verdict – the greatest differences occurred at fs = 10 kHz, DC(T) = 0.1, Vac Amp = 300 V, when the average output current was at the level of only 0.63 A.
Better accuracy of measurements can be obtained by conducting tests in a ther- mal chamber to eliminate the instability of external conditions such as tempera- ture and air flow. Nevertheless, the simulation results in PLECS presented in this article and measurements of the actual model confirm good convergence be- tween results and provide valuable proof of the usability of PLECS simulations.
REFERENCES
[1] Rashid, Muhammad H., Power Electronics Handbook. San Diego, Academic Press, 2001.
[2] Nowak M., Barlik R., Rąbkowski J., Poradnik inżyniera energoelektronika 2. War- szawa, WNT, 2014.
[3] Wintrich A., Ulrich N., Werner T., Reimann T., Application Manual, Power Semi- conductors. Nuremberg: Semikron, 2015.
[4] Górecki K., Zarębski J., The influence of the selected factors on transient thermal impedance of semiconductor devices, 2014 Proceedings of the 21st International Conference Mixed Design of Integrated Circuits and Systems (MIXDES), Lublin, 2014, pp. 309–314.
[5] Gorecki K., Zarebski J., Nonlinear Compact Thermal Model of Power Semiconduc- tor Devices, in IEEE Transactions on Components and Packaging Technologies, vol. 33, no. 3, pp. 643–647, Sept. 2010.
[6] ON Semiconductor, IGBT, Ultra Field Stop, NGTB25N120FL3WG datasheet, Rev.
5, 2017.
[7] STMicroelectronics, Galvanically isolated 4 A single gate driver, STGAP2S datasheet, June 2018.
[8] Szekely V., Rencz M., Thermal dynamics and the time constant domain, in IEEE Transactions on Components and Packaging Technologies, vol. 23, no. 3, pp. 587–
594, Sept. 2000.
(Received: 04.02.2019, revised: 05.03.2019)