1. Polymeric Packaging of High P
N. S. Nobeen1*
1. School of Materials Scienc
2. Orient Gro
3. School of Mechanical and Manufa
4. Department of Mat
*
nadee
Abstract
For power thyristor devices used in h
current (HVDC) schemes, hermetic packag
used despite plastic packaging having
progress towards replacing them in variou
applications, e.g. aerospace and military. A
technologies have demonstrated an exc
reliability and performance, they offer sev
used for the thyristor devices intended
transmission schemes. This is because futur
will be required to have higher current/volt
meet the increasing energy demands from
developing countries, such as China, Braz
such systems will require thyristors
semiconductor wafer diameters. Consequen
the packages to be bigger, more fragile an
expected that by switching from the present
to a polymer material, the device will be mo
and lighter than the current hermetic config
such a shift also provides many challen
materials selection, design and manufacturin
include polymer permeability to moistu
voiding, crack formation which could all
failure.
To assess whether the performance an
polymer housing can be comparable to
packages, a polymeric package demonstrat
and tested. Along with electrical and t
studies performed to study the package be
activities, such as identifying appropriate m
and design concepts for the housing, and
manufactured polymer housing, were als
develop the housing. It is these activities th
this paper. For instance, from the materia
polymers, such as polyimide (PI) and epox
as being ideal candidates for the high ope
and good electrical performance required
From the reliability study which compris
cycling accelerated life tests, the manu
housing was concluded to be reliable.
1. Introduction
Compared to high voltage alternating
high voltage direct current (HVDC) tran
efficient over long distances for over-gr
transmission. HVDC is therefore expected to
in the future to meet the expanding energ
large and highly populous countries, such
Brazil, etc., and the increasing need to m
Power Semiconductor Devices: Material Se
Assessment
*
, K. Ahmad2
, D. C. Whalley3+
, D. A Hutt3
and B Haw
ce & Engineering, Nanyang Technological University
oup of Companies, 26KM, Multan Road, Lahore, Paki
facturing Engineering, Loughborough University, Loug
terials, Loughborough University, Loughborough, LE
esh.nobeen@gmail.com; +
d.c.whalley@lboro.ac.uk
high voltage direct
ges are still being
made successful
us high reliability
Although hermetic
cellent history of
veral drawbacks, if
for future HVDC
re HVDC schemes
tage ratings e.g. to
large and rapidly
zil, India, etc., and
based on larger
ntly, this will drive
nd expensive. It is
t ceramic housings
ore robust, cheaper
guration. However,
nges in terms of
ng. Potential issues
ure, delamination,
l lead to thyristor
nd reliability of a
o current ceramic
tor was developed
thermal modelling
ehaviour, different
material candidates
d reliability of the
so carried out to
hat are discussed in
al selection study,
xy, were identified
rating temperature
from the housing.
ed of temperature
ufactured polymer
current (HVAC),
nsmission is more
round and subsea
o play a major role
gy requirement of
h as China, India,
make national and
regional grids more robust to th
renewables, such as wind and
proposed that transmission ra
3000 A at ±500 kV to 4000 A
future requirements will be able
In order to transmit at these
trade-offs will have to be ma
economic design for the thyri
HVDC transmission. For insta
offs between thyristor paramet
the design of the solid state va
fitted, and the design of the c
valves will be used will also n
the most critical change concern
used to enable the AC/DC
thyristors, e.g. 125 mm or even
and improved performance
implementation for future highe
A cross-sectional view of a
thyristor (also known as a hoc
Fig. 2. Such a device is c
sandwiched between molybde
pieces that act both as catho
electrical connection, and as t
cool the device. A gate lead
connection between the gate e
external circuit to trigger the se
housing is used to enclose th
maintain a clean, dry and non
the wafer, preventing environ
mechanical robustness. It als
convolutes whose shape is as
tracking index of the housing m
tracking along the housing exte
Fig. 1: Photo of a
hockey-puck thyristor
Fi
Ceramic based thyristor ho
excellent history of reliability
their hermetic and degradatio
current ceramic housing design
the large diameter thyristors in
and weight as well as poor re
election & Reliability
worth4
y, Singapore 639798
stan
ghborough, LE11 3TU, UK
11 3TU, UK
he intermittent generation from
d PV. Studies, e.g. [1], have
tings must be increased from
A at ±800 kV to ensure these
e to be fulfilled.
e higher power ratings, various
ade to achieve a practical and
stor-based converters used for
ance, optimisation of the trade-
ters will be needed, along with
alve modules in which they are
converter system in which the
need to be addressed. However,
ns the thyristor devices (Fig. 1)
conversion. Larger diameter
150 mm, with low failure rates
e will be required for
er power transmission levels.
typical press pack high voltage
ckey-puck device) is shown in
omprised of a silicon wafer
enum discs and copper pole
ode and anode terminals for
the main heat transfer path to
is also provided to establish
electrode of the wafer and an
miconductor device. A ceramic
he entire assembly in order to
n-corrosive atmosphere around
nmental attack, and providing
so consists of a number of
sociated with the comparative
material to avoid failure due to
rior.
ig. 2: Cross-sectional view of a
hockey-puck thyristor
ousings have demonstrated an
y and performance because of
on resistant nature. However,
ns offer many disadvantages for
terms of higher processing cost
ecyclability and concerns over
118978-1-4799-2834-7/13/$31.00 c 2013 IEEE

2. robustness. By switching from a ceramic housing to a
polymer-based configuration, it is expected the thyristor
device will be cheaper, lighter more robust and will offer a
high level of design freedom afforded by the injection
moulding process. However, such a change also presents
many challenges in terms of their manufacturing and long
term performance due to the following polymer properties: (1)
they are non-hermetic and can lead to partial discharge failure
in high voltage systems [2], (2) they have complex
manufacturing routes and, (3) varying properties over the
required wide operating temperature range.
To address these issues, a small (50 mm wafer diameter)
demonstrator polymeric package was developed and tested as
part of a larger research programme exploring routes to higher
power thyristors. By adopting this approach, the polymeric
demonstrator performance and reliability could be
benchmarked against a similar size commercial ceramic
housing, before the development of larger size housings was
initiated. Tasks undertaken within the project to develop the
housing included the selection of an appropriate design
concept and material candidate for the polymer housing,
modelling studies of the polymer housing behaviour due to
electrical stresses and temperature changes, identification of
appropriate manufacturing parameters for the polymer
housing, and manufacture and evaluation of the prototype
housings.
Previous work discussing the electrical behaviour of a
polymer based housing for the thyristor device and the
influence of different design parameters on the electric field
strength, as well as application of the Taguchi method of
experimental design to the optimisation of polymer package
design has been previously reported by the authors [3, 4]. In
the present paper, work focussed on selecting the appropriate
design concept and material candidates for the polymer
housing is described. Together with these, results from
different thermo-mechanical simulation studies and
temperature cycling tests conducted to assess the integrity of
the manufactured polymer housing are also reported.
Section 2 discusses the different housing configurations
studied in order to select the most appropriate housing design
for the 50 mm demonstrator, while work performed to identify
appropriate polymer candidates is reported in section 3.
Temperature cycling tests on the manufactured prototypes and
the thermo-mechanical simulations performed are overviewed
in section 4, before the key research findings are discussed
and conclusions drawn in section 5.
2. Housing design selection
For the development of the 50 mm demonstrator, it was
first necessary to identify the most appropriate housing design,
because the design would influence the selection of suitable
polymer candidates, the manufacturing method and the
subsequent modelling studies performed within the
programme. Following a detailed review of previous
inventions in the area, the design configurations that were
identified for study included the ‘replacement’ design, the
‘overmoulded non-integrated’ design and the ‘overmoulded
integrated’ design concepts, as shown in Fig. 3. For the
‘replacement’ design concept the thyristor assembly was
similar to those using existing commercially available
ceramic-based hockey-puck packages where the different
components are stacked on each other and the metal flanges
are cold welded to seal the device and provide environmental
protection. However, the ceramic housing was here directly
replaced by a polymer material. On the other hand, no housing
cavity existed between the polymer material and the assembly
of copper pole pieces, molybdenum discs and silicon wafer in
both the ‘overmoulded integrated’ and ‘overmoulded non-
integrated’ configurations. The copper pole pieces,
molybdenum discs and silicon wafer were stacked and
overmoulded with the polymer material of the housing.
Compared to the ‘overmoulded integrated’ design, a rubber
edge passivation was present around the silicon wafer
circumference in the ‘overmoulded non-integrated’ housing
concept.
(a): Replacement design (b): Overmoulded integrated
design
(c): Overmoulded non-integrated design
Fig. 3: Overview of different housing concepts
Various criteria were considered to identify which of these
was the most appropriate design concept. These included yield
and ease of testing, failure modes and mechanisms,
manufacturing feasibility and marketing edge of the device. In
this work, the merits and drawbacks of the candidate design
concepts were assessed and compared against the different
criteria, and the ‘replacement’ design was found to be the
most appropriate for the 50 mm demonstrator.
3. Polymer candidates for the thyristor housing
Selection of appropriate polymer candidates for the
polymer housing was facilitated by use of the Cambridge
Engineering Selector (CES) software package, as it contains
relevant properties information about all material classes and
allows rapid selection/comparison of materials against a
number of performance criteria [5]. Using this tool, sub-
classes of relevant materials and specific materials among
these sub-classes could be selected. As part of the polymer
selection process, because some performance criteria for the
polymer housing would be influenced by combinations of
material properties rather than a single property, a number of
performance indices (PIs), which are a combination of
properties, were derived to select the housing material
candidates. These included the electrical performance, thermal
shock resistance, thermo-electrical performance, external
excitation resistance, explosion resistance and fatigue
resistance, hermeticity and maximum service temperature
(MST). They were calculated using a combination of
Silicon waferMolybdenum discs
Edgepassivation Polymer
Anode
Cathode
Gate
Silicon waferMolybdenum discs
Over-moulded
Polymer
Anode
Cathode
Gate
Silicon waferMolybdenum discs
Edgepassivation Over-moulded
Polymer
Anode
Cathode
Gate
2013 IEEE 15th Electronics Packaging Technology Conference (EPTC 2013) 119

3. properties. For example, the thermo-electrical PI was
determined as a product of electrical resistivity and thermal
conductivity, whilst the thermal shock resistance was
calculated by dividing the thermal conductivity by the
coefficient of thermal expansion.
These PIs were used along with different property limits
defined in the CES software to select a shortlist of appropriate
polymers. The property limits represent the minimum and
maximum values of the important properties, and in this work
only materials having their property values in between these
limits were selected from the generated graphs. Because
different materials may have some limitations, as well as
advantages, a trade-off among the properties was used to
identify the most suitable polymer for the housing.
Consequently, different scores were attributed to the different
polymers based on their merits and drawbacks before they
were ranked to select the housing material candidates. An
overview of the different polymers identified on the basis of
these different PIs is next reported.
3.1 Summary of shortlisted polymers from CES
An example of a typical graphical plot showing the
different polymers identified from the CES software after
applying the property limits for the electrical performance and
MST PIs is presented in Fig. 4. In this case, the polymers that
were identified as providing good electrical performance and
higher temperature operating capabilities included
polyetheretherketone (PEEK), polyimide (PI), Novolak (glass
filled epoxy), and some grades of liquid crystal polymer
(LCP). Of these, the LCP (40% GF + 10% mineral filled)
polymer showed the best electrical performance.
Fig. 4: CES plot showing the electrical performance PI
plotted against maximum service temperature
Similarly, other polymer candidates based on other
criteria, namely thermal shock resistance, mechanical
performance, hermeticity and cost, were identified before they
were compared and ranked to select the right candidate for the
polymer housing. For example, 40-45% glass and mineral
filled polybutylene terephthalate (PBT) was seen to have the
best thermal shock resistance, LCP the best hermeticity, and
Novolak was most suitable when cost and overall performance
are considered. With regards to identifying polymers with
high mechanical performance at high MST, polyphenylene
sulfide (PPS) and PEEK were found to have the best
properties followed by PI and PBT.
As highlighted earlier, the various polymers identified
from the CES software had their merits and drawbacks, e.g.
LCP had the best overall properties, but its glass transition
temperature was below the MST of the thyristor; PI had good
electrical properties and a glass transition temperature above
the MST limit, however its hermeticity is poor compared to
the other candidate materials, particularly LCP. Therefore,
different scores were attributed to the different PIs to weight
their importance and allow a ranking process to identify the
best polymer material overall from a short list including LCP,
PPS, Novolak, PI, PBT, PEI and PEEK. In this paper, the
different scores attributed to the different PIs are not discussed
for confidentiality reasons. Instead, the overall scores for the
selected materials are shown in Fig. 5. From these results,
polyimide and glass-filled epoxy were concluded to be the
best material for the thyristor housing. Although LCP had
excellent electrical and hermeticity properties, as is the case
with PPS and PBT, it could not be used due to its low glass
transition temperature.
Fig. 5: Graph illustrating the shortlisted materials ranking
Following on from this initial work to select appropriate
polymers and design concepts for the housing, and after
technical and economic reviews, a high performance glass-
reinforced polyimide polymer grade was selected for the
prototype manufacture. Different development activities were
then undertaken to develop and manufacture the 50 mm
prototypes, before its reliability was assessed as is next
discussed.
4. Performance assessment of developed prototype
To assess the expected performance of the manufactured
prototypes, thermo-mechanical based studies were performed.
These were important because thermo-mechanical stresses,
which occur due to the thermal expansion mismatches that
occur when the temperature changes, are recognised as a
major cause of failure for electronic packages. As observed
from studies, e.g. [6], [7], the occurrence of high thermo-
mechanical stresses can lead to defects, such as delamination
and crack formation. In the case of the high voltage thyristor,
these thermo-mechanical stresses may occur as a result of
temperature change events, such as at the end of the moulding
process when the housing is cooled back to ambient
temperature, service conditions i.e. when the thyristor device
is periodically turned on and off, and during the
environmental changes the device will experience during
transportation and storage.
The thermo-mechanical performance of the prototype was
assessed through both simulations and experimentally cycling
sample housings between two temperature levels. The
simulation studies were used to predict the internal stress
distribution in the housing over a range of temperatures and
the likely failure locations. Comsol Multiphysics was the
63
58
63
68
48
59
55
0
10
20
30
40
50
60
70
80
LCP PPS-GMNovolak PI PBT-GM PEI PEEKMaterialscores(%)
120 2013 IEEE 15th Electronics Packaging Technology Conference (EPTC 2013)

4. finite element solver used to compute the thermal stresses. On
the other hand, the thermo-mechanical experiments consisted
of exposing the polymer housings to a number of temperature
cycles according to the IEC 60068-2-14 standard temperature
profile. In it, the high and low temperature limits were taken
to be 125 °C and -40 °C respectively, while the transition time
between these two temperatures and the dwell time for this
cycle were specified as 2 minutes and 2 hours respectively.
The temperature cycling (TC) experiments were also
complemented with partial discharge (PD) and dye penetrant
tests before and after exposure to the temperature cycles to
inspect the housing integrity. The PD testing prior to the TC
test aimed to identify any defects (e.g. voids, delamination)
resulting from the moulding process that may cause long term
housing failure, while the PD test repeated after the TC
experiment investigated possible degradation in the electrical
performance of the housing following the thermal exertions,
e.g. due to cracks or delamination. The dye penetrant tests
aimed to identify specific defects, such as delamination and
surface cracks, both before and after the TC tests.
To measure the PD activity in the polymer housing, the
IEC 60270 test method was used, and the PD activity was
measured using a ‘Biddle Series 27000’ partial discharge
detection system from Biddle instruments. The test regime
consisted of increasing the applied voltage up to 12 kV (above
the required specification of the thyristor). As the voltage was
gradually increased, the partial discharge inception voltage
(i.e. voltage at which repetitive PDs are first observed when
the applied voltage is increased from a lower value) was
observed at the minimum discharge level. In this case, the PD
limit was 5 pC. The voltage was then further increased to the
maximum specified voltage level and thereafter decreased
back to zero. As the voltage was reduced, the partial discharge
extinction voltage, i.e. the voltage at which the PD activity
becomes less than 5 pC, was also noted.
For the dye penetrant test, a coloured dye, which consists
of an enhanced epoxy resin and hardener system, was used to
mount the samples using vacuum impregnation. After the
mounting and curing process, the samples were sectioned for
polishing and inspection under an optical microscope.
The results of these TC simulation studies and experiments
are discussed in the following two sections.
4.1 Thermo-mechanical simulation studies
For these studies, the thyristor housing was simulated
using a 2D axisymmetric FE model to take advantage of its
inherently axi-symmetrical geometry. Two models, A and B,
were studied as part of the investigation. FE model A
consisted of only the polyimide housing and metal flanges,
while model B (illustrated in Fig. 6) represented the fully
assembled thyristor device with the copper pole pieces joined
to the metal flanges. From model A, the aim was to study the
stresses that occur due to the metal flanges and polymer
material differential thermal expansion. Through the study of
model B, the aim was to investigate the potential stress
transfer across the housing due to the other components of the
thyristor device when the temperature was varied. For the
simulation, the Young’s modulus (E) and coefficient of
thermal expansion (CTE) of the different subdomains of the
models were required to be defined. The Young’s modulus
and CTE used for the different subdomains were as follows:
(1) copper flange/pole piece: E = 117 GPa, CTE = 16.9x10-6
K-1
; (2) molybdenum discs: E = 329 GPa, CTE = 4.8x10-6
K-1
;
(3) silicon wafer: E = 185 GPa, CTE = 2.6x10-6
K-1
; (4)
silicone rubber: CTE = 8.1x10-6
K-1
. For the polyimide
polymer, because the visco-elastic properties and exact CTE
were not available, constant CTE values and temperature-
dependent linear elastic material property values supplied by
the selected polyimide polymer supplier were instead used,
and are as follows: CTE = 52×10-6
K-1
; E = 10.2 GPa (-40 °C),
10.2 GPa (23 °C), 6.5 GPa (100 °C), 6.0 GPa (150 °C). A
purely elastic simulation of a temperature change between -40
°C and 125 °C was performed. From different studies, e.g. [7],
because high tensile or shear stress regions are identified as
locations where delamination or crack growth is initiated,
identification of such areas inside the housing was the aim of
this simulation. Following the moulding process, because
residual stresses would also develop inside the housing during
its cooling from the moulding temperature, the ‘stress-free’
state, where no residual stresses are present in the polymer
package, was also required to be identified for the thermo-
mechanical simulation. For this study, this state was
considered to be the moulding temperature of the material
(210 °C in this case). Together with this, because the
simulation results were found to be mesh independent, the
default second order Lagrange elements in Comsol
Multiphysics were used. The normal and shear stress
parameters were studied to identify the likely regions of
delamination, whilst the von Mises stress was analysed to
predict onset of plasticity.
Fig. 6: FE model of assembled device
Example contour plots of the tensile and compressive
stresses (represented by the normal stress in the horizontal r-
axis), the shear and von Mises stress, when the temperature is
reduced to 25 °C in housing model A, are shown in Fig. 7. As
observed from the plots, because localized regions of high
normal and shear stress were observed to occur at both the
cathode and anode end flange/housing interface regions, they
were regarded to be likely failure sites in the package. A
close-up view of the von Mises stress distribution around the
cathode end flange/housing interface region was also studied
and is shown in Fig. 7d. From the plot, the highest stress
magnitude was observed to occur at location L1.
Copperpolepiece
(Cathode end)
Copperpolepiece
(Anodeend)
Molybdenum disc
Molybdenum disc
Siliconwafer
Polyimide
package
Air
Silicone
rubber
Curved
copper
flange
17mm
13.2mm
23.5mm
1.45mm
1.59mm
Curved
copper
flange
2013 IEEE 15th Electronics Packaging Technology Conference (EPTC 2013) 121

5. (a): Normal stress in housing (b): Shear stress
(c): Von Mises stress (d): Von Mises stress
Fig. 7 Contour plots of stresses across the polymer housing
when the thermal load is reduced from 210 °C to 25 °C
On the other hand, the variation of von Mises stress when
the thermal load is reduced from the ‘stress-free’ state to 125
°C and then to -40 °C is illustrated in Fig. 8. As depicted in
the graph, the highest von Mises stress magnitude was
observed to occur at -40 °C, and the difference in the variation
of the stress magnitude in FE models A and B were seen to be
insignificant. This suggested the expansion of the different
components in the assembled device did not contribute any
additional stresses to those that occur in the housing as a result
of the temperature change.
For the onset of plasticity prediction in the polymer, the
von Mises stress variation at locations L1, L2, L3 and L4 in
the housing were also investigated and then compared with the
tensile strength of the selected polyimide polymer for the
housing. For this purpose, a linear variation of the tensile
strength with temperature was assumed using tensile strength
values provided by the supplier, namely 165 MPa at 23 °C and
106 MPa at 150 °C. Because sufficient data relating to the
yield criterion of the selected polymer was unavailable, the
von Mises yield criterion was chosen to predict the onset of
plasticity.
The graph illustrating the von Mises stress variation due to
the temperature change against the tensile strength of the
polyimide material at housing locations L1 – L4 (Fig. 7d) is
shown in Figure 9. As shown in the graph, when a linear
elastic behaviour is assumed for the polyimide material, the
limiting temperature, where plastic flow starts in the housing,
was 65 °C at location L1 only. Onset of plasticity at regions
L2, L3 and L4 was considered to be unlikely because the von
Mises stress was observed to always be lower than the
polyimide tensile strength. Because the linear elastic model
used for the glass-filled polymer in this study is likely to only
be approximate as a result of shrinkage and the orientation of
its glass fibre filler during the moulding process, the stress
levels predicted in this study could be either lower or higher
than actually occur in practice. This would then change the
limiting temperature for plasticity at various locations in the
housing. Additional materials properties data might allow
further work in the future to more accurately predict the
limiting temperature for plasticity onset, thereby minimising
the polymer degradation and early package failure.
Fig. 8: Von Mises stress variation at housing location L1
when the temperature is changed in models A and B
Fig. 9: Von-Mises stress at housing locations L1, L2, L3
and L4 due to temperature change
4.2 TC experimental investigation
Photos illustrating the distribution of the cured epoxy resin
around the convolutes and at the flange/housing interface of a
manufactured prototype are depicted in Fig. 10a and 10b
respectively. The optical microscope inspection revealed no
cracks or defects over the housing surface area. However,
some delamination was noticed in the flange/polymer contact
regions at both the cathode and anode ends of the housing; this
delamination did not extend to the end of the curved flange
used in the prototype. An illustration of the delamination at
the polymer and metal flange interface is shown in Fig. 10c,
while the distribution of the coloured resin around the end of
the curved flange is depicted in Fig. 10d. In this project, the
root causes for the occurrence of the observed delamination
were not identified. Further investigation would be needed to
identify the causes, which may either relate to the surface
finish of the flange or the moulding parameters for the
housing.
In the partial discharge tests performed before the TC
experiments, no PD activity of more than 5 pC was measured
for any of the 148 manufactured housings when a 12 kV
voltage was applied. This suggested the manufactured
housings did not contain any voids or other defects that could
result in the housing failure due to internal discharge when
exposed to high electrical stresses. Six of these tested
housings were then used for the TC tests.
L1
L2
L3
L4
0
5
10
15
20
25
30
35
40
-50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
VonMisesstress/x107Pa
Temperature/oC
FE model A FE model B
122 2013 IEEE 15th Electronics Packaging Technology Conference (EPTC 2013)

6. (a): Potted resin around the
housing convolutes
(b): Potted resin around the
flange/housing contact region
(c): Delamination around the
curved flange contour
(d): No delamination around
the flange end
Fig. 10: Photos depicting the distribution of the coloured
resin on the mounted housing samples before the TC tests
Following the TC tests the PD measurements were
repeated on the tested housings, and they were not observed to
have degraded as a result of the cycling i.e. when the voltage
was increased up to 12 kV, a PD activity of less than 5 pC was
again measured. For the dye penetrant tests, delamination
patterns identical to those observed before cycling were again
observed. This suggested the tested polymer housings did not
degrade due to the TC experiments.
5. Conclusions
Thyristor devices intended for future HVDC transmission
systems will need to have large diameters and low failure rates
to match the required higher voltage and current transmission
ratings. Because existing ceramic housings for these devices
will offer drawbacks, such as higher processing cost, weight
and poor recyclability, the potential to develop a large
polymeric housing for these thyristor devices was explored in
a collaborative research programme. An overview of some of
the activities in that work are reported in this paper - namely
the material and design concept selection, and a performance
study of the manufactured demonstrator housings.
In the study of the housing design concept, different
housing designs were compared against different criteria, such
as ease of testing, manufacturing, marketing edge, and a
‘replacement’ design configuration was found to be
appropriate. This design configuration is similar to current
ceramic-based thyristor devices available commercially, with
the exception of the ceramic housing which is directly
replaced with a polymer material. For the selection of the
polymer, a number of performance indices (PIs), such as
electrical, maximum service temperature, etc., were used to
compare the merits of different polymers. A high performance
glass-reinforced polyimide polymer was selected for the
polymer housing.
In the reliability test programme performed on the
manufactured polymer housings, the thermo-mechanical
behaviour of the package was studied by using both
temperature cycling experiments and simulation studies. The
simulation studies studied the internal stress distribution and
the likely failure areas due to temperature changes. On the
other hand, the temperature cycling conducted between -40 °C
and 125 °C temperature levels aimed to identify defects due to
the moulding process and electrical stress exposure. From the
simulation study, likely areas for temperature induced failure
to originate, were found to be at the flange/housing interface;
whilst from the temperature cycling experiments, occurrence
of defects due to the moulding process and electrical stress
exposure were not observed, thus concluding the developed
prototype to be reliable.
Acknowledgments
The authors wish to acknowledge the funding support
from the UK Technology Strategy Board (TSB) and
Engineering and Physical Sciences Research Council
(EPSRC), together with Dynex Semiconductors Ltd, DuPont
and Alstom T&D for their technical cooperation in this work.
References
1. V. F. Lescale; U. Astrom; J. Nunes; L. Weimers and D.
Wu. (2006). Power transmission with HVDC at
800kV.The ABB Group. [Online]. Available:
http://www05.abb.com/global/scot/scot221.nsf/veritydispl
ay/8e4000a3468ecad9c12571d90031659e/$File/B4_106.
pdf
2. L. G. Vettraino; Risbud; H. Subhash, "Current trends in
military microelectronic component packaging," IEEE
Transactions on Components and Packaging
Technologies, , vol.22, no.2, pp.270,281, Jun 1999.
3. N. S. Nobeen, D. C. Whalley, D. A. Hutt and B. Haworth,
“Computational modelling of electrical field intensity for
high voltage semiconductor package design,” IMAPS
International Symposium on Advanced Packaging
Materials: Microtech. APM '10., pp.54,59, Feb. 28-March
2 2010.
4. N. S. Nobeen; D. C. Whalley; D. A. Hutt; B. Haworth,
"Finite Element Analysis of a high voltage semiconductor
polymeric package design using a Taguchi based
experimental design," 4th
IEEE Electronic System-
Integration Technology Conference (ESTC) , , pp.1,7, 17-
20 Sept. 2012
5. K. K. Ahmad, “Novel polymer-based packaging
technologies for high power semiconductors,” M. Phil
edn., Department of Materials, Loughborough University.
Loughborough, UK, 2009
5. K. Xue; J. Wu; H. Chen; Y. Sun; K. Kwan.; J. Yuen.; A.
Lam., "Numerical analysis of interfacial delamination in
thin array plastic package," Electronic Packaging
Technology & High Density Packaging, 2008. ICEPT-
HDP 2008. International Conference on , vol., no., pp.1,5,
28-31 July 2008
6. Lin, T.Y.; Xiong, Z.P.; Yao, Y. F.; Tok, L.; Yue, Z. Y.;
Njurnan, B.; Chua, K.H., "Failure analysis of full
delamination on the stacked die leaded packages,"
Proceedings of the 53rd IEEE Electronic Components and
Technology Conference, pp.1170 - 1175, May 27-30, 2003
7. T. D. Moore, J. L. Jarvis, Improved reliability in small
multichip ball grid arrays, Microelectronics Reliability,
Volume 41, Issue 3, March 2001, Pages 461-469
Coloured dye
Glass-filled
polyimide housing
Coloured dye
Curved flange
Glass-filled
polyimide housing
Glass-filled Vespel
Flange
Delaminatedregion around
curvededge of flange
Polyimide housing
Nodelamination around
thecurvedend of flange
2013 IEEE 15th Electronics Packaging Technology Conference (EPTC 2013) 123