applied
sciences
Article
Root Cause Analysis of a Printed Circuit Board (PCB) Failure in
a Public Transport Communication System
Cher-Ming Tan 1,2,3,4,5 , Hsiao-Hi Chen 1 , Jing-Ping Wu 1 , Vivek Sangwan 1, * , Kun-Yen Tsai 1
and Wen-Chun Huang 1






Citation: Tan, C.-M.; Chen, H.-H.;
Wu, J.-P.; Sangwan, V.; Tsai, K.-Y.;
Huang, W.-C. Root Cause Analysis of
a Printed Circuit Board (PCB) Failure
in a Public Transport Communication
System. Appl. Sci. 2022, 12, 640.
https://doi.org/10.3390/
app12020640
Academic Editor: Alfio Dario
Grasso
Received: 13 December 2021
Accepted: 7 January 2022
Published: 10 January 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Appl. Sci. 2022, 12, 640. https://doi.org/10.3390/app12020640
1 Center for Reliability Sciences and Technologies, Chang Gung University, Taoyuan 33302, Taiwan;
cmtan@cgu.edu.tw (C.-M.T.); emilychen1129@gmail.com (H.-H.C.); cpwu@alumni.nctu.edu.tw (J.-P.W.);
tsai.kunyen.hw@gmail.com (K.-Y.T.); weichunhuang2020@gmail.com (W.-C.H.)
2 Department of Electronic Engineering, Chang Gung University, Taoyuan 33302, Taiwan
3 Institute of Radiation Research, Chang Gung University, Taoyuan 33302, Taiwan
4 Center for Reliability Engineering, Ming Chi University of Technology, New Taipei City 24301, Taiwan
5 Department of Radiation Oncology, Chang Gung Memorial Hospital, Taoyuan 33302, Taiwan
* Correspondence: sangwanvivek81@gmail.com
Abstract: A printed circuit board (PCB) is an essential element for practical circuit applications and
its failure can inflict large financial costs and even safety concerns, especially if the PCB failure occurs
prematurely and unexpectedly. Understanding the failure modes and even the failure mechanisms
of a PCB failure are not sufficient to ensure the same failure will not occur again in subsequent
operations with different batches of PCBs. The identification of the root cause is crucial to prevent
the reoccurrence of the same failure. In this work, a step-by-step approach from customer returned
and inventory reproduced boards to the root cause identification is described for an actual industry
case where the failure is a PCB burn-out. The failure mechanism is found to be a conductive anodic
filament (CAF) even though the PCB is CAF-resistant. The root cause is due to PCB de-penalization.
A reliability verification to assure the effectiveness of the corrective action according to the identified
root cause is shown to complete the case study. This work shows that a CAF-resistant PCB does not
necessarily guarantee no CAF and PCB processes can render its CAF resistance ineffective.
Keywords: 3D X-ray; bias temperature-humidity reliability test; conductive anodic filament (CAF);
de-penalization; finite element analysis
1. Introduction
Equipment and devices can fail during their operation in the field and this may
occur prematurely. Such a premature failure can render inconveniences and even safety
concerns because the failures are unexpected. To prevent the reoccurrence of the failures,
a root cause analysis is important. There are differences between failure modes, failure
mechanisms, and root causes. An example is an open circuit (failure mode) caused by
conductor corrosion (failure mechanism) due to an incomplete protective coating (root
cause). Whilst the identification of failure modes can be easy because they can be seen or
measured, the finding of failure mechanisms and subsequent root causes are not trivial.
Printed circuit boards (PCBs) are essential for all electronics. They provide mechanical
support and electrical connections to the electronic components of an electronic system.
As electronic controls and the internet become essential, many public transport systems
contain several PCBs. However, the operating environment in public transport can be harsh
due to cyclic temperature and humidity, especially in regions with four different weather
seasons.
PCB failures range from circuit malfunctions to propagating PCB faults. PCB failures
that may appear similar can originate from many different root causes. A propagating
circuit board fault is usually considered universally to be a high severity thermal event.
https://www.mdpi.com/journal/applsci
 Its initiating mechanism starts when a resistive path forms between two traces or planes
in a PCB that are at a different electric potential. This resistive path can form due to an
external heat source [1], insulation breakdown [2], arcing [3], or contamination [4]. The
power dissipated at the fault region should be sufficient to generate enough heat to further
Appl. Sci. 2022, 12, 640 2 of 13
damage the PCB and sustain the thermal event [1]. Slee et al. [1] described various causes
of propagating faults in PCBs; namely, resistive heating, interconnect overheating, con-
tamination, electrochemical migration, tin whiskers, insulation failure, and component
Its initiating mechanism starts when a resistive path forms between two traces or planes
failure.
in a PCB that
In this are aatpropagating
work, a different electric potential.
fault reported Thistoresistive
was due path canmigration.
electrochemical form dueWe to
an external
report heat of
our steps source
tracing[1], insulation
from the failurebreakdown
modes to [2], arcing
the root [3],ofora contamination
cause PCB used in public [4].
The powerfor
transport dissipated
an ethernet at the fault region
connection. should
After their be sufficient
operation forto1.5
generate
months,enough heat to
the operations
further damage the PCB and sustain the thermal event [1]. Slee et al.
of the ethernet device stopped and the burn-out area at the bottom edge of the PCB was [1] described various
causes
found,of aspropagating
shown in Figure faults1.in PCBs; namely, resistive heating, interconnect overheating,
contamination,
These PCBs electrochemical
have multilayers migration, tin whiskers,
that consist insulation
of 8 layers with 2failure,
mm ofand totalcomponent
thickness.
failure.
Layer 4 is the electrical ground plane and layer 5 is the power plane (VIN) where 12 V is
In this to
connected work, a propagating
it. Its base material fault reported
is FR4 (flamewas due to electrochemical
retardant). This PCB hadmigration.
superior CAF We
report our steps of tracing from the failure modes to the root cause of a
(conductive anodic filament)-resistance (anti-migration) properties and a high glass tran-PCB used in public
transport for an ethernet
sition temperature connection.
(Tg) (minimum 170After
°C), their
whichoperation
makes itfor 1.5 months,
a better the in
candidate operations
the com-
of the ethernet device stopped and the burn-out area at the
parison of a standard FR4 for applications where high mechanical and chemicalbottom edge of the PCB was re-
found,
sistancesas shown
to heat in
and Figure 1. properties are required.
moisture

Figure1.1.Visual
Figure Visualinspection
inspectionof
offailed
failedPCBs:
PCBs:(a)
(a)top
topview;
view;(b)
(b)side
sideview.
view.

These PCBs have

2. Identification of themultilayers
Failure Modesthat consist of 8 layers with 2 mm of total thickness.
LayerA4 PCB
is thewas
electrical ground plane
returned from customers and layer
due to5 isburn-out,
the power asplane
shown (VinIN )Figure
where1.12Upon
V is
connected to it. Its base material is FR4 (flame retardant). This PCB had
detailed measurements, it was found that the impedance between VIN and GND signifi- superior CAF (con-
ductive anodic filament)-resistance
cantly dropped. A similar burn-out (anti-migration)
as in Figure 1 was properties
observedandonaahigh
few glass transition
originally good
temperature (Tg ) (minimum ◦
170themC), which makes it a better
inventory boards after testing with the system in thecandidate
factory. Thein the comparison
supply current
of a standard FR4 for applications where high mechanical and chemical resistances to heat
was monitored to avoid severe damage to the testing PCBs. It was found that, after the
and moisture properties are required.
initial short-circuited event, the burn-out expanded due to the over current protection
(OCP)
2. mechanisms
Identification in the
of the power
Failure module. These OCPs are for the protection of the power
Modes
supply. Two mechanisms are employed for the OCP; namely, the cycle-by-cycle current
A PCB was returned from customers due to burn-out, as shown in Figure 1. Upon
limit and the cycle skip.
detailed measurements, it was found that the impedance between VIN and GND signifi-
The cycle-by-cycle current limit turns off the power if the maximum current within
cantly dropped. A similar burn-out as in Figure 1 was observed on a few originally good
each pulse width modulation (PWM) cycle is exceeded. The company set the maximum
inventory boards after testing them with the system in the factory. The supply current was
current threshold to 8 A as the continued working current. However, the actual maximum
monitored to avoid severe damage to the testing PCBs. It was found that, after the initial
current could be larger than this value due to the response delay; the magnitude depends
short-circuited event, the burn-out expanded due to the over current protection (OCP)
on the load condition.
mechanisms in the power module. These OCPs are for the protection of the power supply.
The cycle skip mechanism is activated by the continuous occurrence of a cycle-by-
Two mechanisms are employed for the OCP; namely, the cycle-by-cycle current limit and
cycle current
the cycle skip. limit condition. The PWM enters into a silent period then restarts the pow-
eringThe
cycle again. The company
cycle-by-cycle set the
current limit continuous
turns occurrence
off the power time and the
if the maximum silentwithin
current period
to 30pulse
each ms and 250 ms,
width respectively.
modulation (PWM) cycle is exceeded. The company set the maximum
current threshold to 8 A as the continued working current. However, the actual maximum
current could be larger than this value due to the response delay; the magnitude depends
on the load condition.
The cycle skip mechanism is activated by the continuous occurrence of a cycle-by-cycle
current limit condition. The PWM enters into a silent period then restarts the powering
cycle again. The company set the continuous occurrence time and the silent period to 30
ms and 250 ms, respectively.
Appl. Sci.
Appl. Sci. 2022,
2022, 11,
11, xx FOR
FOR PEER
PEER REVIEW
REVIEW 33 of
of 13
13

Appl. Sci. 2022, 12, 640 3 of 13

These power
These power supply
supply protection
protection mechanisms
mechanisms can can continuously
continuously deliver
deliver 8.5
8.5 W
W under
under aa
short-circuit
short-circuit condition.
condition.
These power supply protection mechanisms can continuously deliver 8.5 W under a
short-circuit condition.
3. Identification
3. Identification of
of the
the Failure
Failure Mechanism
Mechanism
3. Identification
To understandof the
understand the Failure
the failure Mechanism
failure mechanism,
mechanism, aa detailed
detailed failure
failure analysis
analysis was
was performed.
performed. To To
To
begin,
To the structure
understand of
the the PCB
failure near the
mechanism, burn-out
a area
detailed was studied,
failure analysis as shown
was
begin, the structure of the PCB near the burn-out area was studied, as shown in Figure 2. in Figure
performed. To2.
A V-groove
begin, the penalization
structure of the was
PCB used
near here;
the it can
burn-out protect
area the
was PCBs from
studied, as shocks
shown
A V-groove penalization was used here; it can protect the PCBs from shocks and vibration and
in vibration
Figure 2. A
experienced
V-groove
experienced during paste
penalization
during paste printing
wasprinting to it
used here;
to component assembly,
can protectassembly,
component soldering,
the PCBssoldering,
from shocks and
and even
and testing
vibration
even testing
[5].
experienced
[5]. during paste printing to component assembly, soldering, and even testing [5].

Figure 2.
Figure 2. Cross-section
Cross-section of
of the
the board
board near
near the
the V-cut
V-cut scoring
scoring area.
area.
Figure 2. Cross-section of the board near the V-cut scoring area.

After the
After
After thePCB
the PCBfabrication
PCB fabrication processes
fabrication processesare
processes arecompleted,
are completed,de-penalization
completed, de-penalization is
de-penalization isperformed
is performed to
performed to
to
split them
split them
split into
them into individual
into individual PCBs.
individual PCBs. Currently, there
PCBs. Currently, there are are various
are various types
various types
types ofof de-penalization;
of de-penalization;
de-penalization;
namely,hand
namely,
namely, handbreak,
hand break,pizza
break, pizzacutter/V-cut,
pizza cutter/V-cut, punch,
cutter/V-cut, punch, router,
router, saw,
saw,and
saw, andlaser
and laser[5].
laser [5]. In
[5]. Inthis
In thiswork,
this work,aaa
work,
pizza
pizza cutter/V-cut
pizzacutter/V-cut was employed for the de-penalization.
cutter/V-cut was employed for the de-penalization.
de-penalization.
Upon
Upon careful examination of the
Upon careful examination of the PCBs that
careful examination of the PCBs
PCBs that were
that were not
were not short-circuited
not short-circuited or
short-circuited or burned
or burned out,
burned out,
out,
the
the presence
the presence
presence of of many
of many crack
many crack lines
crack lines could
lines could
could be be seen
beseen (Figure
seen(Figure
(Figure3).3). These
3). These crack
These crack lines
crack lines were
lines were found
were found
found
after the
after
after the de-penalization.
the de-penalization. With
de-penalization. With these
these crack
these crack lines,
crack lines, moisture
lines, moisture from
moisture from the
from the environment
the environment could
environment could
could
penetrate
penetrate
penetrateand and diffuse
anddiffuse
diffuseintointo the
intothe PCB
thePCB easily.
PCBeasily.
easily. In
InInthethe presence
presence
the presence of
of of the
thethe applied
applied
applied voltage,
voltage,
voltage, the
thethe elec-
electric
elec-
tric
field field
could
tric field could
cause
could cause
cause the
the coppercopper
the copper ions
ionsions
from from the
the the
from PCB
PCBPCB copper
copper
copper planes
planes to
to drift
planes drift across
across
to drift either
either
across as
as as
either a
a dendrite
dendrite formation
formation [1,5–8]
[1,5–8] or asor
a as
CAF a CAF formation
formation [9,10].
[9,10]. This This
was was
believed
a dendrite formation [1,5–8] or as a CAF formation [9,10]. This was believed to be the believed
to be to
the be the
failure
failure mechanism.
mechanism.
failure mechanism.

Figure
Figure3.
Figure 3. Cross-section
3. Cross-sectionof
Cross-section ofthe
of the PCB:
the PCB: (a)
PCB: (a) potential
(a) potential crack
potential crack location;
crack location; (b)
location; (b) after
(b) after pizza
after pizza cutter
pizza cutter (V-cut)
cutter (V-cut)de-panel
(V-cut) de-panel
de-panel
process.
process.
process.
Dendritic growth
Dendritic growth usuallyoccurs
growth usually
usually occurswhenwhenmetal
metal ions
ions move
move intointo
the the electrolyte
electrolyte layerlayer
near
Dendritic occurs when metal ions move into the electrolyte layer near
near
the the
anodeanode
and and
thenthen deposit
deposit nearnear
the the cathode
cathode and and
growgrow
in a in a tree-like
tree-like or or needle-like
needle-like for-
the anode and then deposit near the cathode and grow in a tree-like or needle-like for-
formation.
mation. IPC-5704
IPC-5704 definesdendritic
defines dendriticgrowth
growthasasthe
thegrowth
growth ofof conductive
conductive metal
metal filaments
filaments
mation. IPC-5704 defines dendritic growth as the growth of conductive metal filaments
on a PCB through an electrolytic solution under the influence of a DC voltage bias and it is
Appl. Sci. 2022, 12, 640 4 of 13

usually formed on the surface of a PCB [11]. Heat can be generated from the electrical ionic
conduction current during the migration process or from the electrical current conducted
through the dendrites when they bridge the conductors. Eventually, the dendrites can
develop into substantially low resistance with significant heat dissipation and can result in
a propagating fault in a PCB [1].
The CAFs are formed as anode metal gradually becomes ions and begins to migrate
under the effects of the bias voltage. It is formed within the PCB insulation materials.
Unlike dendrite formation, the conductive filament growth forms from the anode to the
cathode, hence the name conductive anodic filament [1]. Failure symptoms of a CAF
range from intermittent resistive faults under current limit conditions to catastrophic
propagating circuit board faults if the fault current is not sufficiently limited [1]. Basically,
CAF formation follows a two-step process; namely, the formation of a microcrack that
bridges two conductors to establish the migration path and an electrochemical reaction with
the presence of a DC bias and moisture to drive the migration along the cracks between the
two conductors [12]. This was observed in our experiments, as will be shown later.
The formation of dendrites and CAFs can seriously influence the reliability of electronic
systems. From our failure analysis observations, the failure mechanism was likely to be a
CAF as the short-circuit paths were not on the surface although this PCB was CAF-resistant.

4. Mechanism of the Copper Particles in the FR4 during CAF

Due to the presence of moisture and an electric field, copper ions will migrate along
the glass fiber. This is a well-established phenomenon in PCBs known as CAF. As copper
ions move, and as it is known that conductive particles exposed to an applied electric field
will be charged [13,14], these copper ions will also be charged. As the charged particles
move, they also induce a current through the external circuit [15] and, in the case of partial
discharge (which will be described later), a very short pulse of current will be induced as
there will be a charge transfer between the particle and the facing electrode, which is the
cathode [16].
Generally, conductive particles in the presence of an electric field and fluid such as
water will be subjected to various forces including a drift electric force, fluid force, and
gravity force. The fluid force includes a mass force, drag force, and Basset force. For a
detailed study of the forces, one may refer to the work by Pan et al. [17]. Although the work
by Pan is for conductive particles in liquids, several of the underlying physical mechanisms
considered can be applied. In the situation studied in this work, due to its mass and density,
the copper particles were only subjected to the drift electric force and no rebound between
the two electrodes occurred.
As the conductive particles are close to the cathode as driven by the electric field from
the anode to the cathode, partial discharge can occur due to the intensification of the electric
field between them [17]. It has been found that the electric field is enhanced by 4.2 times
for a spherical particle [16]. In the presence of multiple conductive particles, as in our case,
the frequency of the partial discharge can be much higher, as shown by Li et al. [18].
However, partial discharge does not necessary lead to the breakdown of the dielectric
but it will lead to the reduction of its breakdown voltage. As the size and concentration of
the particles increase, the insulating performance of the dielectric is further inhibited [19–21].
With the reduction in the insulating performance, a leakage current can flow through
the FR4 and generate heat. It is worth noting that if the temperature around the FR4
exceeds more than 65 ◦ C, heat dissipation becomes a major issue for the FR4; beyond
this temperature, the failure mechanism changes to a thermal breakdown from an electric
breakdown [2].
When the particle concentration is relatively high, the formation of a particle bridge is
possible. It was found that a DC field is beneficial to such a formation, as shown by the
work by Li et al. [22].
After a partial discharge, it is suspected that a short carbonization path is formed that
connects the conductive particles to the nearby electrode [2,23]. In a sense, the electrode
 When the particle concentration is relatively high, the formation of a particle bridge
is possible. It was found that a DC field is beneficial to such a formation, as shown by the
Appl. Sci. 2022, 12, 640 work by Li et al. [22]. 5 of 13
After a partial discharge, it is suspected that a short carbonization path is formed that
connects the conductive particles to the nearby electrode [2,23]. In a sense, the electrode is
now
is nowextruded.
extruded. Hence, the conductive
Hence, particles
the conductive that are
particles thatfurther away can
are further away now canalso
now have a
also
partial discharge with the subsequent formation of a carbonization path.
have a partial discharge with the subsequent formation of a carbonization path. When the When the re-
sistance
resistancereduces
reducestotoaacertain
certain value that corresponds
value that correspondswithwithananincrease
increasein in
thethe leakage
leakage cur-
current,
rent, the Joule heating can drive out the surrounding moisture and an
the Joule heating can drive out the surrounding moisture and an increase in resistance can increase in re-
sistance can be expected, as was observed in our experiments (shown
be expected, as was observed in our experiments (shown later in Section 7) However, the later in Section 7)
However, the charge
charge relaxation relaxation
on the on the
conductive conductive
particles particles
may also may also
be reduced andbethus
reduced and thus
the movement
the movement of the charged conductive particles can move further.
of the charged conductive particles can move further. At the same time, this localized At the same time,
this localized heat enlarges the crack. Hence, external moisture can now
heat enlarges the crack. Hence, external moisture can now diffuse further and in a greater diffuse further
and in a into
amount greaterthe amount
FR4. Asinto the FR4. Asparticles
the conductive the conductive particles
move closer move
to the closer to
extruded the ex-
electrode,
truded electrode, partial discharge occurs again. This iteration continues until
partial discharge occurs again. This iteration continues until the effective distance between the effective
distance between
the extruded the extruded
electrode and the electrode and thesoanode
anode becomes close becomes so close
that the field that the
between them field
is
between
so high, itthem is so the
exceeds high, it exceedsfield
breakdown the breakdown
of the FR4, field
which ofisthe FR4, which
already reduced is already
due to there-
duced
presencedueofto the presence
conductive of conductive
particles particles and carbonization.
and carbonization.

5. Experimental
5. Experimental Results
Results and
and Discussion
Discussion
To confirm the proposed mechanism, a Thermo Scientific ELITE VX system was used
To used
to identify
to identify the
thehot
hotspot
spotofof
thethe
thermal emissions
thermal andand
emissions a 3DaX-ray was performed
3D X-ray using ZEISS
was performed using
XradiaXradia
ZEISS 520 Versa
520on the hot
Versa spot.
on the hotFigure 4 shows
spot. Figure the enlarged
4 shows X-ray micrograph
the enlarged at the hot
X-ray micrograph at
the hot spot in the burn-out area and we can clearly see that the short-circuit was copper
spot in the burn-out area and we can clearly see that the short-circuit was due to due to
particlesparticles
copper at the edge of the
at the edge PCB along
of the thealong
PCB de-penalized line. Fromline.
the de-penalized the From
cross-sectional view
the cross-sec-
of the X-ray
tional view ofmicrograph, one can see that
the X-ray micrograph, onethe trace
can see of copper
that is indeed
the trace from the
of copper VIN (anode)
is indeed from
toward
the the GND
VIN (anode) (ground).
toward the GND (ground).

Figure 4.
Figure A 3D
4. A 3D X-ray
X-ray of
of the
the PCB
PCB that
that shows
shows the
the cross-section
cross-section of of the
the PCB.
PCB. The
The propagation
propagation of of the
the
cracks and
cracks and the movement
movement of copper from the VIN layerto
IN layer tothe
the GND
GND layer
layer due
due to
to the
the presence
presence of of the
the
(a) shows
water layer and electric field can be seen: (a) shows thethe formation
formation of
of crack
crack lines
lines before
before the
the resistance
resistance
between them begins to decrease; (b,d) are the plane of the GND and VIN layer, respectively, as
extracted from the 3D X-ray; (c) is the plane in the middle of the GND layer and VIN layer as extracted
from the 3D X-ray where the white arrows represent copper particles; (e) side view of the PCB
showing the short point in between the VIN layer and GND layer.
 colors in Figure 6. The histogram of the volume of these cracks is shown in Figure 7a,
which shows that the majority of the cracks (25) had a volume less than 1 × 105 (μm3). No
crack had a volume greater than 4 × 105 (μm3). Figure 7b shows all the identified cracks
and the cracks with the same volume are shown with the same colors in the 3D view. It
could be observed that large cracks occurred at the edge of the PCB and they progressed
Appl. Sci. 2022, 12, 640 toward the inner parts of the PCB along with the GND and VIN layers. All these cracks 6 of 13
were due to excessive high strain at a high strain rate induced during the V-cut de-panel-
ization as measured experimentally with the results shown in Figure 8. The experimenta-
tion To
is discussed in theand
further analyze nextcharacterize
section. the cracks formed in the PCB, as shown in Figure 4,
the PCB Thearea
largewith
cracks
the observed
cracks was were concentrated
dissected out andinexamined
a certain area
under ofan
theX-ray
PCB Micro
insteadCT of
being distributed
(computed uniformly.
tomography) Thissystem
scanner non-uniform distribution
(model number of the crack sizesThe
CTLab-HX130-2-E). was believed
dissection
to beneeded
was due todue
the inhomogeneous strain
to the size limitation ofdistribution along the
the X-ray system. For edge of the
a better viewPCB during
of the the
cracks,
V-cut
the imagede-panelization as simulated
dataset was reoriented with using
a new aimage
finiteplane,
element analysis,
as shown which
by the pinkwill
colorbe dis-
plane
in the top
cussed in left
the side
next image
section.inThere
Figurewere
5. Inalso
Figure
large5 (top
cracksleft), ‘a’, ‘b’,
present andfrom
away ‘c’ arethe
marked
edge but to
show the front,
they were top,
closed to and left sides,
the largest respectively,
cracks, indicatingand their
the respective
weakening of 2D
theimages are shown
mechanical integ-
in theofsame
rity figure.
the board after the large cracks occurred.

Figure 5.5. Image

Figure Image dataset
dataset reoriented
reoriented with
with aa new
new image
image plane
plane as
as shown
shown byby the
the pink
pink color
color in
in the
the 3D
3D
image (top left side). The 2D images are taken at the: front side (a); top side (b); left side
image (top left side). The 2D images are taken at the: front side (a); top side (b); left side (c). (c).

A total of 78 cracks were identified, and these cracks are shown by their individual
colors in Figure 6. The histogram of the volume of these cracks is shown in Figure 7a, which
shows that the majority of the cracks (25) had a volume less than 1 × 105 (µm3 ). No crack
had a volume greater than 4 × 105 (µm3 ). Figure 7b shows all the identified cracks and the
cracks with the same volume are shown with the same colors in the 3D view. It could be
observed that large cracks occurred at the edge of the PCB and they progressed toward
the inner parts of the PCB along with the GND and VIN layers. All these cracks were
due to excessive high strain at a high strain rate induced during the V-cut de-panelization
as measured experimentally with the results shown in Figure 8. The experimentation is
discussed in the next section.
The large cracks observed were concentrated in a certain area of the PCB instead of
being distributed uniformly. This non-uniform distribution of the crack sizes was believed
to be due to the inhomogeneous strain distribution along the edge of the PCB during the
V-cut de-panelization as simulated using a finite element analysis, which will be discussed
in the next section. There were also large cracks present away from the edge but they were
closed to the largest cracks, indicating the weakening of the mechanical integrity of the
board after the large cracks occurred.
Appl. Sci. 2022, 11, x FOR PEER REVIEW 7 of 13
Appl.
Appl.Sci. 2022,12,
Sci.2022, 11,640
x FOR PEER REVIEW 77 of 13
13
Appl. Sci. 2022, 11, x FOR PEER REVIEW 7 of

Figure 6. A total of 78 cracks are identified and represented by individual colors.

Figure6.
Figure 6. A
A total
totalFigure
of 78
of 78 cracks
cracks areof
6. A total
are identified and
78 cracksand
identified represented
are identified
represented by
andby individualby
represented
individual colors.
individual colors.
colors.

(a) (a) (b) (b)

(a) (b)
Figure
Figure 7. (a) Volume
Volume 7. (a) Volume
histogram
histogram ofofthehistogram
the cracksshown
cracks of the
shown incracks
in shown
Figure
Figure in cracks
6; (b)
6; (b) Figure
cracks 6; the
(b) the
with
with cracks
samesamewith theare
volume
volume same volum
Figure
are 7.
shown (a) Volume
are histogram
shown with of
the the
same cracks
color. shown in Figure 6; (b) cracks with the same volume
shown withwith
the the
samesame color.
color.
are shown with the same color.

Figure 8. Strain gauge measurement data show that a part of the principal strain exceeds the IPC
Figure 8.
Figure StrainWP-011
8. Strain gaugemeasurement
gauge measurement
standard limits data
during
data show
show that
thethat
cuttingaprocess.
a partpart of principal
of the the principal strain
strain exceeds
exceeds the
the IPC-
Figure 8.
IPC-WP-011
WP-011 Strain
standard gauge
standard measurement
limits
limits during
during data
the showprocess.
cutting
the cutting that a part of the principal strain exceeds the IPC-
process.
WP-011 standard limitsTo during
furtherthe cuttingthe
confirm process.
proposed mechanism, 5 samples of the PCB that had crac
To further
To confirm
furtherlines due tothe
confirm proposed
proposed mechanism,
de-penalization
the were subjected
mechanism, 55 samples
samples of the
theRH.
to 65 °C/95%
of PCBTable
PCB that 1had
that had crack the te
summarizes
crack
lines due to de-penalization were subjected to 65 ◦ C/95% RH. Table 1 summarizes the test
To further
lines due confirm
results. Thethe
to de-penalization testproposed
condition
were mechanism,
was
subjected selected
to 65 °C/95% 5 samples
so thatRH. of the
a continuous
Table PCB
filmthat had
of water
1 summarizes crack
could
the test be forme
results.
lines dueThe
to test condition was
de-penalization selected
were so that
subjected to a65continuous
°C/95% film
RH. of water
Table 1 could be formed
summarizes the test
results. The test condition was selected so that a continuous film of water could be formed
in the PCB.
results. TheThe
test test resultswas
condition showed clearly
selected that
so that these cracksfilm
a continuous allowed moisture
of water could beto diffuse
formed
into the PCB and reduced the impedance between the GND and VIN .
Appl. Sci. 2022, 11, x FOR PEER REVIEW 8 of 13
Appl. Sci. 2022, 11, x FOR PEER REVIEW 8

Appl. Sci. 2022, 12, 640 8 of 13

in the PCB. The test results showed clearly that these cracks allowed moisture to diffuse
into the PCB andinreduced
the PCB.the
The test results
impedance showedthe
between clearly
GNDthat
andthese
VIN. cracks allowed moisture to di
into the PCB and reduced the impedance between the GND and VIN.
Table 1. Change in resistances after temperature/humidity test.
Table 1. Change in resistances after temperature/humidity test.
Table 1. Change
InitialinResistance
resistances afterMeasured
temperature/humidity
Resistance test.
Initial Resistance be-VINMeasured Resistance after Test Duration
Sample # between and after the Specified
Initial Resistance be- Measured Resistance Test after
Duration
(Hours)
Sample # tween VIN and GND GND (Ω) the Specified TestTest
Duration (Ω)
Duration Test Duration
Sample # (Hours)
tween VIN and GND the Specified Test Duration
1 (Ω) Over 108 (Ω) 58.6 160 (Hours)
(Ω) (Ω)
1 2 Over 108 Over 108 58.6 30.9 K 160
113
1 Over 108 58.6 160
2 Over 108 8 30.9 K 113
3 2 Over
Over 10 108 316.2 30.9 K 90 113
3 Over 108 316.2 90
4 3 Over
Over 108 10 8
1.63 K 316.2 232 90
4 Over 108 1.63 K 232
5
4 Over
Over 108
10 8 1.63 K 232
5 Over 108 46.5 46.5 141
141
5 Over 108 46.5 141

From
Fromthe
theabove observations,
aboveFrom
observations,
the abovethe
thefailure
failuremechanism
mechanism
observations, could
could
the failure be
besummarized,
could beas
summarized,
mechanism asshown
shown
summarized, as sh
ininFigure
Figure9.9.The flow
Theinflow chart
chart and
and its summary table
table to
to represent
represent the
the research
research methodology
methodology
Figure 9. The flow chart and its summary table to represent the research of methodo
ofthis
thiswork
workare
areshown
shown
of thisinin Figure
Figure
work are1010 and
and
shown inTable
Table 2, 2,
Figure respectively.
respectively.
10 and Table 2, respectively.

Figure9.9.Failure
Figure Figure
process9.
Failureprocess andFailure
and process and
theunderlying
the underlying the underlying mechanisms.
mechanisms.
mechanisms.

Figure 10. Flow chart of the failure analysis process.

Figure10.
Figure Flowchart
10.Flow chartofofthe
thefailure
failureanalysis
analysisprocess.
process.
Appl. Sci. 2022, 12, 640 9 of 13

Table 2. Summary of the failure analysis methodology.

Failure Analysis
Steps Details and Findings
Methodology
1 Reception of Failed Device Customer reported unavailability of ethernet

• Burn-out area at the bottom edge of the PCB found

through naked eyes.
• Presence of cracks observed at the burned-out area
Detailed Examination and using a low- to high-definition optical microscope,
2
Failure Mode Inspection thermal emissions, 3D X-ray, and X-ray Micro CT
Scanning.
• Cracks are found in between the L4~L5 layers at
the PCB edge.

• Cracks are present at the PCB edge from where the

moisture penetrates inside the PCB, which causes
the short-circuit; the burned-out area can be
observed.
Failure Mechanism • Conductive anodic filament (CAF) is a proposed
3 Proposal and failure mechanism.
Identification • Strain produced during the V-cut de-panelization
process is more than IPC-WP-011 standard limits
and verified using a strain gauge measurement
and an ANSYS multiphysics simulation, which are
the causes behind the cracks in the PCB.

PCBs are subjected to a 65 ◦ C/95% RH reliability test

Failure Mechanism and resistance degradation is observed, which signifies
4
Confirmation that the moisture penetrates into the PCBs through the
cracks that are present in between the L4~L5 layers.

• New de-panelization process is introduced.

• PCBs are prepared using the new de-panelization
Corrective Action and its
5 process subjected to 65 ◦ C/95% RH.
Effectiveness Verification
• No cracks and degradation in the PCB resistance
are observed

6. Possible Root Causes and Their Verification

Knowing that it was the crack lines that caused the observed failure mode, and that
these lines were generated because of the high strain developed during the de-penalization
process, a strain gauge measurement was performed on the PCB at three test locations, as
marked by the black color blocks in Figure 11, near the edge of the PCB during the V-cutting
process to ascertain the root cause. The strain gauge measurement results are depicted
in Figure 8. One can see the localized areas along the de-paneling line at the bottom side
with a high strain rate and high strain around 1400–2050, which was beyond the acceptable
criteria according to IPC-WP-011 [24].
ANSYS WorkBench 19.0 was also employed to examine the strain induced by de-
penalization. A 3D CAD model of the PCB was designed in ANSYS WorkBench and a V-cut
cutter was designed to mimic the real-world scenario in the simulation software where
mechanical strain was observed, as shown in Figure 11. Our simulation results showed
the maximum stress area (red area) distributed in strips along the edge of the PCB and
this coincided with the burn-out area. The maximum strains did not distribute uniformly,
which agreed with our X-ray micrograph and the strain gauge measurement results. Li [25]
also reported that the stress generated through the V-cut process is significantly high in
comparison with other cutting processes that cause mechanical cracks in PCBs.
Appl. Sci.
Appl. Sci. 2022, 12, x640
2022, 11, FOR PEER REVIEW 1010ofof13
13

Figure 11. PCB de-panel process and mechanical simulation result. Locations 1, 2, and 3 show strai
gauge measurement points.

In the pizza cut de-penalizing, the applied mechanical stress caused cracks on th
copper keep-out area of the PCB, as shown in Figure 12. However, without de-paneling
there was no crack on the copper keep-out area, as can be seen in Figure 13.
It was, therefore, clear that it was the pizza cutter de-penalization that caused th
cracks in the PCB, which then resulted in moisture diffusion into the PCB. As the cuttin
area was near the GND plane and VIN plane, as can be seen in Figure 2, the reported failur
Figure
Figure 11.
11.PCB
PCBde-panel
de-panelprocess
processand
andmechanical
mechanical simulation
simulation result.
result. Locations
Locations 1,
1, 2,
2, and
and 33show
showstrain
strain
modemeasurement
gauge was observed.
gauge measurement
We found that no damage was observed if a manual de-penalizatio
points.
points.
was performed. Thus, we believed that a proper V-groove design and machine settin
In the
could
In the pizza
solve cut
thiscut
pizza de-penalizing,
issue, but this was
de-penalizing, the beyond
the applied mechanical
applied mechanical
the scope of stress caused
thiscaused
stress cracks
work. cracks on the
With proper
on the setting
copper
copper
such as keep-out
keep-out area of the
area of the
the incident PCB, as shown
PCB,rotary
angle, as shown in Figure
in Figure
speed, 12. However,
12. However,
and cooling without
fluid,without
such damagede-paneling,
de-paneling,
can be mini
there
there was
was no
no crack
crack on
on the
the copper
copper keep-out
keep-out
mized, as in the case of wafer dicing [26]. area, as can be seen in Figure 13.
It was, therefore, clear that it was the pizza cutter de-penalization that caused the
cracks in the PCB, which then resulted in moisture diffusion into the PCB. As the cutting
area was near the GND plane and VIN plane, as can be seen in Figure 2, the reported failure
mode was observed. We found that no damage was observed if a manual de-penalization
was performed. Thus, we believed that a proper V-groove design and machine setting
could solve this issue, but this was beyond the scope of this work. With proper settings
such as the incident angle, rotary speed, and cooling fluid, such damage can be mini-
mized, as in the case of wafer dicing [26].

Figure12.12.Observation
Figure Observation of cracks in a PCB
of cracks in aafter
PCBde-paneling using anusing
after de-paneling Olympus BX51 microscope
an Olympus at 50
BX51 microscope a
×50magnification of: (b) location No. 1 and (c) location No. 2, as marked in (a), which is a horizontal
× magnification of: (b) location No. 1 and (c) location No. 2, as marked in (a), which is a horizonta
cross-section
cross-section of of
a PCB with
a PCB a pizza
with cutter
a pizza de-panel.
cutter de-panel.
It was, therefore, clear that it was the pizza cutter de-penalization that caused the
cracks in the PCB, which then resulted in moisture diffusion into the PCB. As the cutting
area was near the GND plane and VIN plane, as can be seen in Figure 2, the reported failure
Figure 12. Observation of cracks in a PCB after de-paneling using an Olympus BX51 microscope at
mode was observed. We found that no damage was observed if a manual de-penalization
50 × magnification of: (b) location No. 1 and (c) location No. 2, as marked in (a), which is a horizontal
was performed. Thus, we believed that a proper V-groove design and machine setting
cross-section of a PCB with a pizza cutter de-panel.
could solve this issue, but this was beyond the scope of this work. With proper settings
such as the incident angle, rotary speed, and cooling fluid, such damage can be minimized,
as in the case of wafer dicing [26].
Appl. Sci.
Appl. 2022,
Sci. 2022,11,
12,x640
FOR PEER REVIEW 11 of 13 11 of 1

Appl. Sci. 2022, 11, x FOR PEER REVIEW 11 of 13

Figure
Figure13. Absence ofofcracks in a PCB without de-paneling. Observations were were
performed using using a
Figure 13. Absence
13. Absence cracks
of cracks in a in a PCB
PCB without
without de-paneling.
de-paneling. Observations
Observations were performed performed
using an
an OlympusBX51
Olympus BX51 microscope
microscope atat5050× × magnification
magnification of: (b)
of: location
(b) No. No. 1 and (c)
location location No. 2, as No. 2, a
Olympus BX51 microscope at 50× magnification of: (b) location 1 and No. 1 and
(c) location (c) 2,location
No. as
marked
marked in (a),
marked in (a),which
(a),whichis
which a horizontal
is aishorizontal cross-section
a horizontal location
cross-section
cross-section of
locationlocationa PCB without
of a PCB
of a PCB without a de-panel.
without a de-panel.
a de-panel.
7. Corrective Actions and Reliability Verification
7.7.Corrective
Corrective Actions and Reliability Verification
Actions andofReliability
With the identification Verification
the root cause, another type of de-penalization was intro-
With the identification of the root cause, another type of de-penalization was intro-
ducedWithwhere the noidentification
crack was produced.
of theDue Due to
root confidential considerations, this new way
duced where no crack was produced. tocause, another
confidential type of de-penalization
considerations, this new way was intro
cannot
duced be disclosed.
where no crack was produced. Due to confidential considerations, this new wa
cannot be disclosed.
ToToverify the effectiveness of the corrective actions, reliability tests were performed
cannot be disclosed.
verify the effectiveness of the corrective actions, reliability tests were performed
along
alongwith the thenew de-penalization method.
method.The Thetesttest condition was
set set at ℃/95%
65 °C/95%RH RH
Towith
verify new
the de-penalization
effectiveness of the corrective condition
actions, was at 65
reliability tests were performed
sosothat
that a continuous water film could be formed if cracks existed. To speed up the test, atest, a
a continuous water film could be formed if cracks existed. To speed up the
along with
voltage ofof100 the newapplied
de-penalization method. VThe test condition was set at 65 ℃/95% RH
voltage 100VVwas across the
was applied across theGND
GND andVIN
and .IN .
so that
From a continuous
From Figure
Figure14, water
14, with
with the film
the could be formed
new de-penalization
new de-penalization if cracks
method,
method, thethe existed.
resistance
resistance Tobetween
speed
between the up
thethe test,
voltage
GND and
GND and V of
V 100 V was
remained applied
high: upacross
to 800 the
h GND
of testingand V .
consistently
IN for the
ININ remained high: up to 800 h of testing consistently for the 10 test samples. 10 test samples.
However,From
However, Figure
4 4PCBs
PCBswith 14,the
with with
the thede-penalizing
V-cut new de-penalization
de-penalizing methods
methods method,
showed
showed the resistance
a significant
a significant between th
reduction
reduction
in in
GND the
the resistance
resistance
and even before
even
VIN remained beforehigh:
400
400 hhup
of
oftesting,
800 except
testing,
to hexcept forfor
of testing oneoneboard. ThisThis
board.
consistently clearly showed
clearly
for the showed
10 test samples
thatour
ourcorrective
corrective action
that
However, 4 PCBsactionwithwas waseffective
the effective
V-cut in overcoming
in overcoming
de-penalizing the damage
the damage
methods introduced
showed adue
introduced to de-
due to
significant reduction
penalization.
de-penalization.
in the resistance even before 400 h of testing, except for one board. This clearly showed
that our corrective action was effective in overcoming the damage introduced due to de
penalization.

(a) (b)
Figure 14. Resistance changes over time under the reliability test mentioned above: (a) 10 samples
Figure 14. Resistance changes over time under the reliability test mentioned above: (a) 10 samples
with the new de-penalization; (b) 5 samples with pizza cut de-penalization and the resistance fluc-
with the new
tuation de-penalization;
as reported earlier can be(b) 5 samples
observed. with pizza
The resistance cutY-axis
in the de-penalization
is in GW. and the resistance
fluctuation as reported earlier can be observed. The resistance in the Y-axis is in GW.

(a) (b)
Figure 14. Resistance changes over time under the reliability test mentioned above: (a) 10 sample
with the new de-penalization; (b) 5 samples with pizza cut de-penalization and the resistance fluc
Appl. Sci. 2022, 12, 640 12 of 13

8. Conclusions
This work demonstrated a step-by-step failure analysis methodology for multilayer
printed circuit boards that led an observed failure mode to the root cause with verification
of the root cause and the effectiveness of the corresponding corrective action. Printed circuit
boards used in public transport systems were found to be burnt after a time of operation. A
detailed failure analysis showed that the failure mechanism was a propagating fault of a
CAF formation as moisture was diffused into the PCBs via microcracks introduced due
to a V-cut de-penalization. Although the PCB itself was CAF-resistant, the presence of a
microcrack due to de-penalization could render its CAF resistance ineffective.
With the identified root cause, a modified de-penalization method was developed and
no microcracks were observed with this new method, which verified the root cause. The
PCBs with the modified de-penalization method also underwent reliability tests under a
high temperature/humidity with a high voltage applied across and only small changes
in resistance between the previously affected planes were observed, which confirmed the
effectiveness of the corrective action. The detailed steps from the failure mode to the failure
mechanism and from the failure mechanism to the root cause were clearly described.

Author Contributions: Methodology, C.-M.T. and J.-P.W.; formal analysis, C.-M.T., H.-H.C., J.-P.W.,
and V.S.; investigation, C.-M.T., J.-P.W., and V.S.; resources, H.-H.C., K.-Y.T., and W.-C.H.; writing—
original draft preparation, V.S.; writing—review and editing, C.-M.T., J.-P.W., H.-H.C., and V.S.;
project administration, H.-H.C. All authors have read and agreed to the published version of the
manuscript.
Funding: This research received no external funding.
Acknowledgments: This research is an industrial collaboration with a company to solve a real
industrial issue, leveraging on the in-depth applied failure analysis theory developed by the Centre
for Reliability Science and Technology in Chang Gung University. This research did not receive any
specific grants from funding agencies in the public, commercial, or not-for-profit sectors. The authors
also acknowledge the support of C.Y. Hsueh from Jie Dong, Taiwan and Rigaku Corporation, Japan,
in assisting with the X-ray computed tomography analysis.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Slee, D.; Stepan, J.; Wei, W.; Swart, J. Introduction to printed circuit board failures. In Proceedings of the 2009 IEEE Symposium
on Product Compliance Engineering, Toronto, ON, Canada, 26–28 October 2009; pp. 1–8.
2. Zhou, Q.; Wen, M.; Xiong, T.; Jiang, T.; Zhou, M.; Ouyang, X.; Xing, L. Study on Insulation Breakdown Characteristics of Printed
Circuit Board under Continuous Square Impulse Voltage. Energies 2018, 11, 2908. [CrossRef]
3. Davis, J.H. Silicone protective encapsulants and coatings for electronic components and circuits. In Plastics for Electronics; Springer:
Dordrecht, The Netherlands, 1985; pp. 67–97.
4. Khandpur, R.S. Printed Circuit Boards: Design, Fabrication, Assembly and Testing; Tata McGraw-Hill Electronic Engineering: New
York, NY, USA, 2005; ISBN 0071589252.
5. Coombs, C.F.J.; Holden, H.T. Printed Circuits Handbook, 7th ed.; McGraw-Hill Education: New York, NY, USA, 2016.
6. Minzari, D.; Jellesen, M.S.; Møller, P.; Ambat, R. On the electrochemical migration mechanism of tin in electronics. Corros. Sci.
2011, 53, 3366–3379. [CrossRef]
7. Yang, S.; Wu, J.; Christou, A. Initial stage of silver electrochemical migration degradation. Microelectron. Reliab. 2006, 46, 1915–1921.
[CrossRef]
8. Lee, S.B.; Jung, M.S.; Lee, H.Y.; Kang, T.; Joo, Y.C. Effect of bias voltage on the electrochemical migration behaviors of Sn and Pb.
IEEE Trans. Device Mater. Reliab. 2009, 9, 483–488. [CrossRef]
9. Tan, C.M.; Narula, U.; Seow, G.L.; Sangwan, V.; Chen, C.H.; Lin, S.P.; Chen, J.Y. Moisture resistance evaluation on single electronic
package moulding compound. J. Mater. Chem. C 2020, 8, 1943–1952. [CrossRef]
10. Ready, W.J.; Stock, S.R.; Freeman, G.B.; Dollar, L.L.; Turbini, L.J. Microstructure of Conductive Anodic Filaments Formed during
Accelerated Testing of Printed Wiring Boards. Circuit World 1995, 21, 5–9. [CrossRef]
11. IPC. IPC-5704 Cleanliness Requirements for Unpopulated Printed Boards; IPC: Bannockburn, IL, USA, 2009.
12. Zou, S.; Li, X.; Dong, C.; Ding, K.; Xiao, K. Electrochemical migration, whisker formation, and corrosion behavior of printed
circuit board under wet H2S environment. Electrochim. Acta 2013, 114, 363–371. [CrossRef]
13. Pérez, A.T. Charge and force on a conducting sphere between two parallel electrodes. J. Electrostat. 2002, 56, 199–217. [CrossRef]
Appl. Sci. 2022, 12, 640 13 of 13

14. Drews, A.M.; Kowalik, M.; Bishop, K.J.M. Charge and force on a conductive sphere between two parallel electrodes: A Stokesian
dynamics approach. J. Appl. Phys. 2014, 116, 074903. [CrossRef]
15. Pan, C.; Wu, K.; Du, Y.; Tang, J. Comparison of Sato’s equation and Pedersen’s theory to obtain gas discharge current. IEEE Trans.
Dielectr. Electr. Insul. 2016, 23, 1690–1698. [CrossRef]
16. Tobazéon, R. Electrohydrodynamic behaviour of single spherical or cylindrical conducting particles in an insulating liquid
subjected to a uniform DC field. J. Phys. D Appl. Phys. 1996, 29, 2595–2608. [CrossRef]
17. Pan, C.; Tang, J.; Chen, G.; Zhang, Y.; Luo, X. Review about PD and breakdown induced by conductive particles in an insulating
liquid. High Volt. 2020, 5, 287–297. [CrossRef]
18. Li, J.; Hu, Q.; Zhao, X.; Yao, X.; Luo, Y.; Li, Y. Partial-discharge characteristics of free spherical conducting particles under AC
condition in transformer oils. IEEE Trans. Power Deliv. 2011, 26, 538–546. [CrossRef]
19. Zhang, J.; Wang, F.; Li, J.; Ran, H.; Huang, D. Influence of copper particles on breakdown voltage and frequency-dependent
dielectric property of vegetable insulating oil. Energies 2017, 10, 938. [CrossRef]
20. Wang, K.; Wang, F.; Li, J.; Zhao, Q.; Wen, G.; Zhang, T. Effect of metal particles on the electrical properties of mineral and natural
ester oils. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1621–1627. [CrossRef]
21. Wang, X.; Wang, Z.D. Study of dielectric behavior of ester transformer liquids under ac voltage. IEEE Trans. Dielectr. Electr. Insul.
2012, 19, 1916–1925. [CrossRef]
22. Li, Y. Motion Characteristics of Solid Particles in Insulating Oil and Their Influence on Breakdown Voltage of Insulating Oil.
Master’s Thesis, Chongqing University, Chongqing, China, 2017.
23. ECSS. European Cooperation for Space Standardization Space Engineering: High Voltage Engineering and Design Handbook; ECSS:
Noordwijk, The Netherlands, 2012.
24. IPC. IPC-WP-011 Guidance for Strain Gage Limits for Printed Circuit Assemblies; IPC: Bannockburn, IL, USA, 2012.
25. Li, W.; Sun, X. An analysis case on the failure of BGA solder joints. In Proceedings of the 18th International Conference on
Electronic Packaging Technology, ICEPT 2017, Harbin, China, 16–19 August 2017; pp. 731–734.
26. Shi, K.W.; Yow, K.Y. The characteristics and factors of a wafer dicing blade and its optimized interactions required for singulating
high metal stack lowk wafers. In Proceedings of the 2013 IEEE 15th Electronics Packaging Technology Conference, EPTC 2013,
Singapore, 11–13 December 2013; pp. 208–212.