Further Improvements in the Reliability of IGBT Modules

Further Improvements in the Reliability of IGBT Modules
Further Improvements in the Reliability of IGBT Modules
Thomas Schütze, Hermann Berg, Martin Hierholzer
eupec GmbH & Co. KG
Max-Planck-Straße 5
59581 Warstein, Germany
Abstract- This paper gives a survey of the measures and
the resulting improvements of IGBT module reliability
reached by eupec during the introduction of IGBT high
power modules.
10.000.000
IHV (traction)
Since their market introduction in the beginning of
1995, eupec IGBT high power modules (IHM) got a quick
access to several applications in the lower and medium
voltage range because of their obvious advantages with
regard to controllability, isolation and mounting.
When introducing high voltage 3.3 kV IGBTs (IHV)
intended for the upper power range and traction
applications additional demands are madeto the electrical
and mechanical characteristics of the modules. These
growing requirements will be met by continuous
progresses in the module design in the areas of substrate,
bond wire and base plate technology, partial discharge
immunity as well as improvements of the electrical chip
characteristics.
POINTS OF IMPROVEMENT
Bonding
A high power IGBT module comprises approx. 450
wires together with 900 wedge bonds. For many years the
reliability of this contact technology has been a concern
especially for traction applications. Considerable work,
e.g. in the LESIT-program, has been concentrated on
accelerated power cycling tests, analysis of failure
mechanisms and improvements in bonding technology.
Disconnection of bond wires due to heel cracks, bond liftoffs, reconstruction of Al-metallization on the chips and
corrosion of wires were step-by-step identified as reliability
limiting weak points. Development activities on
•
•
•
•
•
composition of wire
shape of bonding tool
bonding parameters
metallization of chips and leads
protective coatings
have led to considerable improvements in the reliability
of the bond contact. Test results of short time power
cycling on IGBT modules with up to 24 paralleled IGBT
chips are shown in Fig. 1, comparing the number of cycles
versus the junction temperature swing. The „IHM
1.000.000
no. of cycles
INTRODUCTION
--- estimated curve
l test points
100.000
IHM (standard)
10.000
30
40
50
∆ Tj (K) 60
70
80
Fig. 1. Short time power cycling
(standard)“ modules are designed for the needs of standard
industrial applications while the curve labeled „IHV
(traction)“ represents the results for modules applying all
the above mentioned improvements, therefore fulfilling
even the severest requirements of traction applications.
These modules are available in the traction relevant IGBT
voltage classes 1700 / 2500 / 3300 V.
The criteria for failure was an increase of forward
voltage by more than 5%. The tests with temperature
swings of 40°C, 50°C and 60°C have been performed to
get reliable data for practical operating conditions. The
runs at delta Tj = 70°C and 80°C have been made to gain
information about accelerating factors.
The test at ∆Tj = 40°C with 20 modules under test took
one year while the run at ∆Tj = 60 °C could be finished
within 4 weeks. It is worthwhile to mention that these two
tests were carried out with the same production lot of
IGBT modules on the same test equipment controlled by
the same team. Diode parts and IGBT parts of high power
modules were tested separately.
Concluding from the technological analysis of failed
modules out of the test programs it can be stated: there are
no more bond wire lift-offs and no corrosion of the wire to
be found. The failure mechanism has been changed
throughout. Even after 9 million cycles all bond wire
connections to the IGBT chips are still good An additional
overload test of the IGBTs with 8 kA subsequent to power
cycling was passed without failure.
Base plate
The use of copper as base plate material is common for
its well known advantages with regard to high thermal
conductivity, easy mechanical handling, galvanic plating
and adequate pricing. Disadvantages are non reversible
changes of mechanical properties above 300°C and the
mismatch of the coefficient of thermal expansion (CTE) to
the ceramic substrate.
The soldering between substrate and base plate is
therefore a failure source. Because of different CTEs of the
materials thermal stress occurs and generates mechanical
strain on the solder. Repetitive, heavy load cycling will
create solder cracks and therefore an increase of the
thermal impedance between chip and base plate.
Efforts have been made to mitigate the bimetallic effect
of the soldered system metal / ceramic by an adequate
shaping. A machined convex bow as shown on the right
side of Fig. 2. clearly improves the heat transmission
between base plate and heat sink.
The divergence of approximately 50% to the thermal
conductivity of copper is obvious but not as significant as
it might look at first glance. The thermal resistance of the
base plate compared to the total thermal resistance of the
module is in the range of only 20%. When further
considering the renunciation of additional intermediate
layers when using CTE-matched materials, the increase is
even less. Furthermore the diminished bimetallic effect
results in a well-balanced contact surface to the heat sink.
The most outstanding advantage can be seen in the gain of
reliability. At highly accelerated cycling tests with ∆Tc =
80 K the solder layer between copper base plate and
ceramic showed a delamination at the edges of the
substrate after 4000 cycles. With the new Al/SiC base
plate and under the same test conditions we have reached
20.000 cycles so far without any signs of delamination.
Tests will be continued to define the exact factor of the
reliability improvement.
10.000.000
IHV KF2 (AlSiC)
--- estimated curve
l test points
plane part
bow stamped
bow machined
Fig. 2. Bending of the system metal / ceramic
before (top) and after (bottom) soldering
Al (Alloy)
Cu
Cu
CuMo
CuMo
Al/SiC
Al/SiC
CuW
CuW
Cu/Mo/Cu
Cu/Mo/Cu
Kovar
Kovar
0
5
10
15
20
100
200
300
400
Thermal Conductivity (W / mK)
Coefficient of Thermal Expansion (ppm / K)
Al (Alloy)
Al (Alloy)
Cu
Cu
CuMo
CuMo
Al/SiC
Al/SiC
CuW
CuW
Cu/Mo/Cu
Cu/Mo/Cu
Kovar
Kovar
0
5
10
Density (g / cm³)
15
20
IHM (standard)
1.000
30
40
50
60
∆ Tc (K)
70
80
90
Fig. 4. Load cycling capability
Fig. 4 shows this gain of reliability when comparing the
high voltage IGBT traction module „KF1“ (Cu base plate)
with the new generation „KF2“ (AlSiC base plate).
Partial Discharge
0
25
IHV KF1 (Cu)
100.000
10.000
A relatively stiff material with low deviation of its CTE
to the ceramic would solve both described problems. As
seen in Fig. 3 the metal matrix compound (MMC)
material Al/SiC offers an extreme stiffness and a CTE
close to that one of the AlN ceramic (7.3 ppm/K).
Al (Alloy)
no. of cycles
1.000.000
0
20
40
60
Specific Stiffness (GPa cm³ / g)
Fig. 3. Selection criteria for base plate materials
80
To estimate the lifetime of the insulation without the
need of high voltages as in the dielectric test, the „partial
discharge test“ has been introduced [1].
Partial discharge (PD) is a partial breakdown of the
insulation material. An example for a PD source is a small
void in ceramics. If the voltage exceeds the breakdown
voltage of the gas included, a sudden flash-over discharges
the void. The recharge can be measured. PD occurs when
increasing the voltage beyond the inception voltage and it
disappears when decreasing the voltage below the
extinction voltage.
Following the development steps between 1995 and
1997 the partial discharge level of the 3.3 kV IGBT
modules has been considerably reduced as shown in Fig. 5.
homogeneous temperature distribution. This conditions
can be reached by:
• rugged NPT chip technology
• narrow distribution of chip parameters
• a positive temperature coefficient of Vcesat.
1000
1995
Pârtial Discharge (pC)
100
FEB 1996
10
SEPT 1996
1
MAY
1997
0,1
0,5
1
1,5
2
2,5
3
3,5
4
4,5
5
5,5
6
U (kVrms )
Fig. 5. PD-Improvements on 3.3kV IGBT Modules
In 1995 we started using DCB on 0.63 mm thick Al2O3
ceramics. We had the low inception and extinction
voltages typical for this material and partial discharge
values in the range of 200 to 300 pC. We limited the
voltage during these first test runs on complete modules to
5 kVrms , 1 min. By changing the ceramics from Al2O3 to
AlN in February 1996 the inception voltage was increased
and the devices could meet the specification target 10 pC
as per IEC 1287. Further improvements of substrates and
silicone gel resulted in an even lower partial discharge.
The 10 pC at 6 kV target was reached in September 1996.
20
Tests have been performed to clarify within which limits
matching of chips is necessary.
In principle, matching is possible on single wafer level
(all chips within a module from same wafer), on lot level
(all chips within a module from same lot) and across all
production lots (mixing chip lots in modules), of course
with chips fulfilling proper data limits.
Statistic process control and practical experience have
shown us that staying within the same chip lot during the
assembly of modules results in reliable devices, even with
24 chips per device [2].
40,00%
35,00%
30,00%
mean= 3,45V
sigma=0,03V
25,00%
20,00%
15,00%
10,00%
5,00%
0,00%
V(CE)sat
Fig. 7. Distribution of VCEsat-values of IGBT chips
of one lot (50 A, 3300 V);
test conditions: Tvj = 25 °C, VGE = 15 V, Ic = 50 A
Partial Discharge (pC)
15
10
As an example Fig. 7 shows the VCEsat -distribution of
the individual chips out of a certain lot, there are no chips
outside the range of 3.45 + 0.09 V (3 sigma). The practical
consequence is that all these chips can be mounted without
further selection in a series production.
5
0
0
10
20
30
40
50
60
t (min)
40,00%
Fig. 6. Recording of PD-test of a 3.3kV IGBT module
with enhanced insulation capability;
test conditions: Up = 6 kVrms, t = 60 min.
35,00%
30,00%
mean= 3,51V
sigma=0,11V
25,00%
20,00%
In addition to the requirements of the IEC 1287standard we have considered the behavior of the modules
under long term high voltage stress. We recorded PD
during a one hour test at a voltage Up = 6 kV, see Fig. 6.
PD decreased and was significantly lower than the
required 10 pC.
Chip characteristics
In today’s high power modules up to 24 chips are
mounted in parallel. A prerequisite for a proper operation
of these modules are an equal current sharing and a
15,00%
10,00%
5,00%
0,00%
V(CE)sat
Fig. 8. Distribution of VCEsat-values of high voltage
IGBT modules of 8 lots (1200 A, 3300 V);
test conditions: Tvj = 25 °C, VGE = 15 V, IC = 1200 A
Measuring the VCEsat- values of ready mounted modules
out of 8 different lots, as plotted in Fig. 8, we find, as
expected, a wider distribution. This is due to differences
•
•
•
•
•
load current: 50% Inom
blocking voltage: 50% Vces
ambient temperature: 40°C
operating: > 1000 h
application class: N
resulted in a failure rate of below 500 FIT.
So far eupec has delivered more than 150.000 IHM and
IHV IGBT modules to customers. High attention has been
paid to the rejects from the customer and especially from
the field. In tight cooperation with key customers detailed
investigations of all failures have been performed.
With these new informations, based on more than 700
Mio. estimated hours of module operation, we can now
expect a future failure rate for our high power modules of
50 FIT.
modules * h
160
total FIT
600
140
500
120
400
100
80
300
60
200
40
100
20
Jan 98
Jul 97
Okt 97
Apr 97
Okt 96
Jan 97
Jul 96
Jan 96
Apr 96
Jul 95
0
Okt 95
0
FIT rate (electr. device failure)
180
700
Jan 95
As explained above, a lot of measures were taken to
insure the reliability of the modules and to meet the
customer’s requirements. Due to the low number of
devices under test, it is very difficult to predict a lifetime
for the modules operated under field conditions only from
the results of these accelerated reliability tests. First
estimates, based on an operation under the following
conditions:
200
Apr 95
ONCLUSION
800
modules operation hours / Mio.
between the chip lots but not the result of parameter
scattering within one individual chip lot.
Fig. 9. Operation hours and resulting FIT rates
REFERENCES
[1] J. Göttert, W. Köhler, K. Sommer, G. Lefranc
„Insulation voltage test and partial discharge test of
3.3kV IGBT modules,“ Proceedings PCIM 1997
Nuremberg, pp. 119-122
[2] K. Sommer, J. Göttert, G. Lefranc, R. Spanke,
„Multichip high power IGBT modules for traction
and industrial applications,“ Proceedings EPE 1997
Trondheim, pp. 1112-1116
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising