International Journal of Fatigue 114 (2018) 206–216

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Evaluation of structural design methodologies for predicting mechanical T

reliability of solder joint of BGA and TSSOP under launch random vibration
excitation
⁎
Tae-Yong Parka, Jong-Chan Parkb, Hyun-Ung Oha,
a
Space Technology Synthesis Laboratory, Department of Aerospace Engineering, Chosun University, 309, Pilmun-daero, Dong-gu, Gwangju 61452, Republic of Korea
b
Launcher Systems Development Team, Korea Aerospace Research Institute (KARI), 169-84, Gwahak-ro, Yuseong-gu, Daejeon 34133, Republic of Korea

A R T I C LE I N FO A B S T R A C T

Keywords: The fatigue damage that accumulates on a solder joint under a launch vibration environment is a major cause of
Solder joint failure in spaceborne electronics. Thus, predicting the mechanical safety of the solder joint in the structural
Random vibration design phase is important to guarantee a successful space mission. This paper proposes a mechanical design
Electronics methodology to predict and ensure more reliable mechanical safety of the solder joints of a plastic ball grid array
Mechanical reliability
(PBGA) and a thin-shrink small outline package (TSSOP) under launch random vibration excitation. Random
vibration fatigue tests for a sample printed circuit board (PCB) assembly were performed to assess the fatigue life
of BGA and TSSOP solder joints. The effectiveness of the proposed structural design methodologies was eval-
uated by comparing the margin of safety calculated from various analysis approaches and the random vibration
fatigue test results of the PCB specimen.

1. Introduction of the pyrotechnic device for separation of stages, fairing, and satellite.
Among these mechanical loads, random vibration is one of the main
Advances in semiconductor and electronic packaging technologies causes of electronics failure. This is because the repetitive bending
have driven the trends in space engineering, as in other fields such as behavior of the PCB causes relative displacement between the package
automotive, home electronics, and medical engineering [1,2]. As a re- and PCB, which leads to fatigue fracture on the solder joint that con-
sult, the mission capabilities of satellites have been continuously in- nects the package and the PCB [5]. Although random vibration in an
creasing. In addition, the bulky packages developed in earlier genera- actual flight lasts for only several minutes, the electronics also experi-
tions have been replaced with highly integrated electronic packages ence vibration during on-ground validation tests at the component
because launch costs are proportional to the total weight of a satellite. level, payload module level, and satellite system level. Therefore, the
Surface mount-type packages such as a ball grid array (BGA) and small mechanical reliability of the fatigue life of a solder joint under random
outline package (SOP) are typical examples of these packages, which vibration should be predicted in the early design stage to avoid risks
are widely used in various space missions [3,4]. These packages have such as increased development period and cost. Reliability prediction is
higher component densities and many more electrical connections particularly important for surface-mount packages because their
within a smaller package size compared to conventional through-hole packaging structure is relatively weaker than that of through-hole
mounting-type packages. Therefore, they enable the implementation of packages in terms of fatigue damage [5].
higher functional performance, and efficiently use the accommodation Various methodologies have been proposed to predict the me-
area of the printed circuit board (PCB) installed in the spaceborne chanical reliability of the solder joints under random vibration [6–11].
electronics. Most of them are based on finite element analysis (FEA) to determine
The electronics in satellites are subjected to various forms of me- the stress and strain responses of critical solder joints under given
chanical loads during the launch phase, such as steady-state static load mechanical load conditions, and theoretical approaches to estimate
due to engine thrust, sine vibration due to engine cut-off, and random fatigue life. Wong et al. [6] developed a fatigue life prediction model for
vibration induced by the noise of exhaust and turbulent flows generated a BGA solder joint under random vibration using the empirical formula
along the launcher. In addition, shock load is produced by the actuation derived from universal slopes produced by high-cycle fatigue test data.

⁎
Corresponding author.
E-mail addresses: ppp3906@naver.com (T.-Y. Park), jcpark@kari.re.kr (J.-C. Park), ohu129@chosun.ac.kr (H.-U. Oh).

https://doi.org/10.1016/j.ijfatigue.2018.05.012
Received 14 November 2017; Received in revised form 30 March 2018; Accepted 9 May 2018
Available online 16 May 2018
0142-1123/ © 2018 Elsevier Ltd. All rights reserved.
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

To consider different levels of acceleration response to random vibra- integrated PCBs with various electronic packages.
tion, this formula was combined with a three-band technique derived This paper proposes a mechanical design methodology for more
from Gaussian distribution. Wu et al. [7] developed a methodology for reliable prediction of the mechanical safety of solder joints in the
estimating the fatigue life of BGA solder joints under random vibration structural design stage of spaceborne electronics. To validate the ef-
using Basquin’s power-law fatigue damage model and a linear super- fectiveness of the proposed methodology, fabricated PCB specimens
position method of Miner’s rule. Its effectiveness was validated by with plastic BGA (PBGA) packages and thin-shrink small outline
comparison of the fatigue life prediction results with those obtained packages (TSSOP) were exposed to the qualification level of random
using the commercial software CALCE PWA. By using a similar method, vibration loads of spaceborne electronics. The fatigue life of solder
Mathew et al. [8] performed fatigue life assessment on the electronic joints on the PCB specimen was also predicted using the commercial
unit of a solid rocket booster for a space shuttle under random vibration reliability and life prediction tool called Sherlock [16] after the fatigue
to determine the number of future missions in which the unit can be test. The time to failure predicted using Sherlock agreed with the ex-
used without failure. For the assessment, they used the vibration time perimental results, which indicates that Sherlock is an effective meth-
history collected during the actual flight as input data of FEA. Yu et al. odology for the structural design of electronics. This tool reduces the
[9] developed a methodology to evaluate the fatigue life of SAC305 and time and effort required to construct the FEM of a complex electronic
SAC405 solder joints of a BGA package under random vibration based PCB by using design files such as Gerber or ODB++ files. In addition,
on vibration tests and FEA. The results of fatigue life prediction of the the inherent failure mechanism of the electronics can be rapidly pre-
solder joint using rainflow cycle counting and Miner’s rule agreed with dicted based on the physics of the failure approach. However, even
the experimental results. However, these previously proposed meth- Sherlock requires considerable time to construct the FEM because the
odologies have limitations in terms of reliability prediction for the PCB detailed geometry and material information of the electronic packages
of spaceborne electronics. This is because the construction of a detailed is needed. In particular, these PCB design files can be obtained only
finite element model (FEM) requires increased time and effort as the after the design has progressed to some extent, such that the use of
number of electronic packages increase. In other words, it is nearly Sherlock less efficient in the initial design stage of the electronics.
impossible to construct the entire electronics with many integrated Thus, to find a more applicable methodology for predicting me-
PCBs. chanical safety of solder joints in the initial structural design stage of
In addition to the abovementioned previous studies, Steinberg’s spaceborne electronics, we propose methodologies based on margin of
fatigue failure theory [5] has been widely adopted for evaluating the safety (MoS ) calculations with respect to the allowable displacement of
mechanical safety of solder joints based on the dynamic displacement of the PCB and the critical strain acting on it. To validate the effectiveness
the PCB. For example, Jung et al. [12] evaluated the mechanical re- of the proposed methodologies, the results of MoS calculations were
liability of a remote drive unit under random launch vibration based on compared with those of experimental results. The comparison indicates
Steinberg’s theory. In addition, this theory was used for investigating that the MoS calculated using the critical PCB strain is the most effec-
the mechanical reliability of electronic PCBs for cube satellite appli- tive in evaluating the mechanical safety of solder joints when the
cations [13] and electronics for military applications [14]. However, average in-plane principal PCB strain with respect to eight nodes on the
Steinberg’s theory has theoretical limitations which cause errors in PCB was calculated by a simplified FEM using 0D lumped mass and
mechanical safety prediction results. This is because the empirical rigid beam element 2 (RBE2) under the equivalent static loads for
formula used for estimating the allowable PCB displacement is limited random vibration.
to explaining a rectangular PCB simply supported on the edges with the
ideal mode shape of a half-sine wave [15]. Therefore, complex mode 2. Fatigue tests of PCB specimen under random vibration
shapes of the PCB owing to asymmetric configurations, irregular con-
straints, or the presence of stiffeners cannot be clearly explained using In this study, various design methodologies based on MoS calcula-
this theory. The above relative position factor of package also has tions were proposed to evaluate the mechanical safety of solder joints at
theoretical limitations because it is calculated with respect to the lo- the initial structural design stage of spaceborne electronics. These
cation of the center of gravity of the package. This produces a larger methodologies will be described in detail later. The calculated MoS
error with respect to the estimated allowable displacement as the values were compared with experimental results for validating the ef-
package size increases because of the difference between displacements fectiveness of the proposed methodologies under a random vibration
occurring at the center and the package edge. In addition, the allowable environment.
displacement can be overestimated as long as the package position gets
closer to PCB edge. Nevertheless, Steinberg’s theory has been inevitably 2.1. Description of PCB specimen
used as there is no other theory to replace it so far.
For the structural analysis of electronics, a detailed FEM that in- Fig. 1 shows the configuration of the PCB specimen with PBGA
cludes the actual configuration of the solder interface is effective for packages and TSSOPs fabricated for fatigue testing under random vi-
predicting the dynamic responses of the PCB. However, constructing bration. The numbers and locations of each package are also shown in
such an FEM requires much time and effort, especially in the case of the figure. Five 324-pin PBGA packages (U1, U4, U5, U6, and U9) and
high-density packages with BGA, SOP, and ceramic column grid array four 48-pin TSSOPs (U2, U3, U7, and U8) were mounted on the PCB,
(CCGA). The computation time also increases with an increased number which was formed from FR-4. The total weight of the assembled PCB
of finite element meshes by modeling the detailed configuration of the was 65.6 g and the dimensions were 121 mm × 107.3 mm × 1.65 mm.
solder interface. Therefore, the use of a detailed FEM has limitations in The boundary conditions on the PCB include a total of 10 holes for M3
terms of the many tradeoff studies required to verify the effectiveness of screws. A solder material of eutectic Sn-Pb37 was used to mount these
the structural design of electronics in their initial design stage. If the packages onto the PCB by considering the space heritage. Table 1 lists
package is simplified into a rigid beam and 0D mass elements, the time the specifications of the electronic components mounted on the PCB
and effort required to develop the FEM can be reduced. However, this specimens. Table 2 lists the specifications of the input random vibration
incurs an unavoidable change in natural frequency and displacement for the specimen test, which correspond to the qualification level for
response of the PCB [15]. In particular, it increases with package size, spaceborne electronics.
which is one of the limitations in predicting the dynamic responses of a
PCB for mechanical safety evaluation based on Steinberg’s theory. 2.2. Random vibration fatigue test
Therefore, a more practical methodology is needed to evaluate the
mechanical safety of an entire electronic unit that includes many Prior to the random vibration test, nondestructive inspections of the

207
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Fig. 1. Configuration of PCB specimen with PBGA packages and TSSOPs.

Table 1
Specifications of electronic components.
Package No. Configuration Properties

U1, U4, U5,

U6, U9
• Package
PBGA
Type:

• Pin Count: 324

• Mount
Mount
Type: Surface

• Size(mm):
19 × 19 × 1.6
• Weight(g): 1.4
• Solder
Pb37
Material: Sn-

U2, U3, U7,

U8
• Package
TSSOP
Type:

• Pin Count: 48
• Mount
Mount
Type: Surface
Fig. 2. Representative results of nondestructive inspection of solder joints ((a)
• Size(mm):
12.5 × 6.1 × 1.1
X-ray micrograph of U5 package, (b) optical micrograph of U2 package).

• Weight(g): 0.283
• Lead Material:
Copper
• Solder
Pb37
Material: Sn-

Table 2
Specifications of random vibration.
Frequency (Hz) Power spectral density (PSD, G2/Hz)

20–60 +3 dB/oct
60–1000 0.273
1000–2000 −6 dB/oct
Overall 20 Grms

PCB specimens were conducted to check the manufacturing status of

the solder joint. Fig. 2(a) and (b) show the representative X-ray and
micro-optical inspection results for the U5 and U3 packages, respec-
tively. The results indicate that the quality of all solder joints was ac-
ceptable, with no unexpected voids and initial cracks.
In the fatigue test, two PCB specimens were used to guarantee the
reliability of the test results. The test results were also compared with Fig. 3. Setup for the random vibration fatigue test.
the predicted results using Sherlock, as discussed in Section 3. Fig. 3
shows the setup for the random vibration fatigue test of the PCB spe-
cimen. An electrodynamic shaker (IMV, J260/SA78M) was used to

208
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Fig. 5. Time profiles of daisy-chain resistance of each package ((a) first PCB
specimen of case 1, (b) second PCB specimen of case 2).

criterion for full cracking of the solder joint was defined as when the
daisy-chain resistance exceeds 10.5 kΩ, which is the maximum mea-
surement limit of the test equipment because the daisy-chain circuit of
the package can be considered as electrically opened at that value. In
addition, the partial crack was defined as a resistance threshold of 100
Ω, which is the same as that used in a previous study [17].
Fig. 5(a) and (b) show the test results of time histories of daisy-chain
resistances for the PCB specimens of cases 1 and 2. The first PCB spe-
cimen of case 1 was exposed to the random vibration environment
specified in Table 2 for 7.67 h. The resistance value of the U5 package
Fig. 4. Configurations of daisy-chain circuit ((a) PBGA package, (b) TSSOP). gradually increased after 5.89 h of random excitation and reached
10.5 k Ω after 7.42 h, which is defined as the failure of the solder joint.
apply the random vibration levels specified in Table 2. To measure the The resistance value of the U6 package increased to 1.1 k Ω during the
time to failure of each package during the tests, we used the in-situ test. To increase the reliability of test results, the second PCB specimen
resistance monitoring method based on the daisy-chain circuit, which of case 2 was tested for 16 h. The results of Fig. 5(b) indicate that the U5
connects the solder joints in series. Fig. 4(a) and (b) show the daisy- package reached failure after 6.89 h of excitation. This is close to the
chain circuit applied to the PBGA package and TSSOP, respectively. results obtained in case 1, with a difference of 7.14%, although re-
Data acquisition equipment (National Instruments, NI-9219) were used sistance variation exceeding 100 Ω was observed after 3.02 h. In ad-
to monitor the resistance of each package at a speed of 50 samples/s. By dition, the U6 package reached failure after 12.04 h. The other seven
considering the measurement error range of the equipment, the initial packages of both PCB specimens did not show any resistance variations
resistance of each package was set to approximately 50 Ω by adding during the test.
more resistors at the end of the electrical circuit. In the test, the failure Fig. 6(a) and (b) show the representative scanning electron

209
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

microscopy (SEM) cross-section micrographs of the corner-most solder vibration. The ∊c calculated using Eqs. (2) and (3) is used to predict the
joints of the tested PBGA package and TSSOPs on the second PCB number of cycles that reached failure Nf , on the solder joint [18]
specimen of case 2, respectively. None of the four TSSOPs showed any
b
cracking of the solder joints even after 16 h of excitation, as shown in ∊
Nf = Nc ⎛ c ⎞
⎜ ⎟

Fig. 6(a). In contrast, in the case of U5 and U6 packages, full cracks ⎝ ∊max ⎠ (4)
were observed along the boundary between the solder ball and solder
pad at the package side, as shown in Fig. 6(b). In addition, partial where Nc is the critical number of cycles required to reach failure,
cracks occurred at the U1, U4, and U9 solder joints, although no re- which corresponds to the 20 million cycles for random vibration, as
sistance variations of those packages were observed during the test. proposed by Steinberg; and b is the fatigue exponent of the Sn-Pb37
This is because the resistance measurement equipment with limited solder material obtained from the stress-life cycle curve. In the case of
accuracy could not detect slight variations in resistance due to micro- the maximum PCB strain ∊max , Sherlock used the two-sigma value of the
cracks. To determine the time to failure of the solder joints of the tested root mean square (RMS) strain selected from a simple approximation
packages, SEM inspections were also conducted for the case 1 specimen. method [5], which reduces the fatigue life calculation under random
Table 3 summarizes these results and the fatigue life of each package of vibration based on the three-band technique with a Gaussian prob-
both case 1 and 2 specimens. The results indicate that both specimens ability distribution.
showed the same crack propagation states on the solder joint of each Fig. 8 shows the natural frequency and mode shape of the PCB
package, except for the U6 package, which did not reach failure criteria specimen analyzed by Sherlock. The global bending mode of the PCB
on the case 1 specimen. These results show that the solder cracks of U1, was observed along the out-of-plane direction at the first eigen-
U4, and U9 packages initiated at some point during the 7.67 h test. In frequency of 632.26 Hz. This is the most dominant mode for the me-
the case of the U5 package, solder cracking initiated after 3.02 h in case chanical reliability of the solder joint owing to the largest relative
2. These test results were compared with the fatigue life prediction displacement between the PCB and the package.
results of the PCB test specimen described in the next section. Fig. 9 shows the maximum strain distribution of the PCB specimen
under 20 Grms of random vibration. The major strain responses oc-
3. Fatigue life prediction using the sherlock tool curred mainly at the PCB center, indicating that the BGA packages
adjacent to this area will be relatively vulnerable to fatigue failure
After fatigue testing of the PCB specimen, its fatigue life under the compared with the others located at the PCB edge. Fig. 10 shows the
random vibration excitation specified in Table 2 was predicted using predicted time to failure on each package under the random vibration
the commercial software Sherlock described in Section 2. Fig. 7(a) excitation. The results indicate that the U5 package at the PCB center
shows the Gerber design file of the PCB specimen provided by the PCB was the first to reach an almost 100% failure probability when the
manufacturer, which includes information on the PCB layer, package, exposure time to random vibration was 6.7 h. The rest of the packages
and solder joint. This was used as an import file in Sherlock for con- were expected to reach failure after at least 56.91 h.
structing the FEM. The FEM shown in Fig. 7(b) includes 324 BGA solder Based on the failure criteria for full cracking of solder joints defined
balls and 48 gull-wing lead frames of the TSSOP. The total number of in this study, the average difference in the time to failure between the
elements and nodes used for the FEM were 28,598 and 116,585, re- predicted and test results was 6.9%. Although the second PCB specimen
spectively. Table 4 lists the properties of the material used in the ana- showed a partial crack at 3.02 h of testing, the prediction results shown
lysis. in the Fig. 9 are still valid because they indicate a failure probability of
To predict the fatigue life of the PCB, Sherlock used the critical approximately 40% at that time. These results show that Sherlock is an
strain theory, which is based on an empirical formula modified from effective tool for predicting the mechanical reliability of the electronics.
Steinberg’s theory. In Steinberg’s theory, the allowable PCB displace- However, the solder cracks on the other four PBGA packages initiated
ment is defined as [5] within a much shorter time than the predicted time to failure, even
though there was still time until full cracking was reached. In this
0.00022B
Zallow =
Chr L (1)
where Zallow is the allowable displacement of the PCB; B is the length of
the PCB parallel to the component; C is an electronic packaging-type
constant derived from the numerous vibration tests and FEM-based
studies; and h , r , and L are the thickness of the PCB, relative position
factor of the package mounted on the PCB, and length of the package,
respectively. This theory states that the package can withstand at least
20 million fatigue cycles under the random vibration environment if the
maximum displacement is larger than the allowable value.
In the critical strain theory, the critical PCB strain ∊c , which corre-
sponds to the maximum principal in-plane strain from the area adjacent
to the critical solder joint, is defined as [19]
ζ
∊c =
C L (2)
where the variables r and B in Eq. (1) are replaced with an allowable
principal in-plane PCB strain ζ to compensate for the theoretical lim-
itations of Steinberg’s theory; and C and L are the same as those used in
Eq. (1). The value of ζ can be calculated using the equation presented in
the IPC-WP-011 guideline [19] as follows:
2.35
ζ= × {1900−300 × log(ε )̇ }
t (3)
Fig. 6. Representative SEM micrographs of the solder joints of second PCB
where t is the thickness and ε ̇ is the strain rate of the PCB under specimen of case 2 ((a) U2 package, (b) PBGA package).

210
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Fig. 6. (continued)

regard, we judge that the fatigue cycles of the solder joints of the four geometry and material information of the packages is required to obtain
PBGA packages were accelerated because the boundary conditions of proper analysis results. In addition, these PCB design files are typically
the adjacent U5 package were gradually varied according to solder available only when the design has progressed to some extent.
crack propagation during the test. Thus, the fatigue life predictions for Therefore, Sherlock might be less efficient for reliability evaluation
solder joints should be more conservative to guarantee the mechanical during the initial structural design stage of the electronics. This issue
safety of the electronics. will be addressed via the mechanical safety evaluation based on MoS
However, even Sherlock faced limitations in terms of reducing the calculations using the various methodologies proposed in this study.
time and effort required for FEM construction because detailed

211
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Fig. 7. FEM of PCB specimen constructed using Sherlock ((a) Gerber design file,
(b) FEM).

Table 3
Summary of crack propagation state and time to failure on each package.
Package No. Case 1 Case 2

Crack Time to Crack Time to Fig. 8. Modal analysis results of PCB specimen ((a) first mode: 632.26 Hz, (b)
propagation failure (h) propagation failure (h) second mode: 1210.6 Hz, (c) third mode: 1397.54 Hz).

U1 Partial crack < 7.67 Partial crack < 16

U2 No crack > 7.67 No crack > 16 4. Evaluation of structural design methodology
U3 No crack > 7.67 No crack > 16
U4 Partial crack < 7.67 Partial crack < 16
To obtain a more practical structural design methodology for elec-
U5 Full crack 7.42 Full crack 6.89
U6 Partial crack < 7.67 Full crack 12.04 tronics in the structural design phase, we propose evaluation ap-
U7 No crack > 7.67 No crack > 16 proaches to assess the mechanical safety of solder joints of various
U8 No crack > 7.67 No crack > 16 packages shown in the Fig. 11 according to the choice of FEM, struc-
U9 Partial crack < 7.67 Partial crack < 16 tural analysis method, and fatigue failure theory; these were derived
from MoS calculations based on Steinberg’s theory and the critical
strain theory described in Section 3. The MoS with respect to the al-
Table 4
lowable displacement calculated using Eq. (5) is as follows [12]:
Material properties used for the analysis.
Material Elastic Shear Poisson ratio Density
Zallow
MoS = −1
modulus modulus (kg/m3) SF × Zmax (5)
(MPa) (MPa)
where Zallow is the allowable displacement of the PCB which can be
PCB (FR-4) 31,893 13866.5 0.15 2477 calculated using Eq. (1). SF is a safety factor and Zmax is the maximum
PBGA Package Component 15,168 6320 0.2 1900
PCB displacement estimated from the three-sigma value of the RMS
TSSOP Component 11,700 4500 0.3 2940
Lead (Copper) 113,000 42164.2 0.34 8900 acceleration. For the methodology based on the critical strain theory,
Solder (Sn-Pb37) 29,379 10801.1 0.36 8490 the MoS with respect to the critical strain can be calculated as
∊c
MoS = −1
SF × ε pmax (6)

where ∊c is the critical strain which can be calculated using Eqs. (2) and
(3); and ε pmax is the maximum in-plane principal strain of the PCB. The

212
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

100

U1 U6
80 U2 U7
U3 U8
U4 U9

Faliure Probability (%)

60 U5

40

20

0
0 2 4 6 8 10 12 14 16
Time (h)
Fig. 10. Time profiles of predicted failure probabilities of each package.

Fig. 9. Maximum strain distribution of PCB specimen under random vibration

excitation.

mechanical safety of solder joints can be secured for 20 million cycles of

random vibration if the calculated MoS values using Eqs. (5) and (6) are
greater than zero. In this study, the SF of 1.11 was used for the MoS
calculation. This value is equ MoS ivalent to a safety factor of 2.0 in the
fatigue life of the Sn-Pb37 solder [5].
As a first step in evaluating the effectiveness of our proposed
methodologies, we constructed a detailed FEM for the PCB specimen, as
shown in Fig. 12(a). The of each package was calculated based on the
displacement and strain responses predicted from the random vibration
analysis with the random input profile specified in Table 2. In addition,
the calculated MoS values were compared with the fatigue test results
described in Table 3. Here, the methodologies based on Steinberg’s
theory and the critical strain theory are referred to as STT-RV-1 and
CST-RV-1, respectively.
In the analysis, the detailed FEM reflects the actual configuration of
the package, solder, solder pad, and lead frame. The model consists of
738,995 nodes, 496,906 CPENTA elements, 84,764 CHEXA elements,
and 20 RBE2 elements. As the boundary condition, three degrees-of-
freedom were constrained on the screw holes of the PCB. The modal
analysis results indicate the first eigenfrequency as 641.53 Hz. This Fig. 11. Evaluation scheme for validating the structural design methodologies.
value is almost the same as that calculated from the FEM using
Sherlock. To calculate the MoS based on the STT-RV-1 methodology, as shown in Fig. 6(b). On the other hand, the results obtained using the
the maximum displacement Zmax is calculated as [5]. CST-RV-1 methodology showed negative margin for all PBGA packages.
This accurately represents the fatigue test results, which showed cracks
250 × (3 × Z¨rms )
Zmax = on the solder joints of the PBGA packages. These results indicate that
fn2 (7) the CST-RV-1 methodology based on the critical strain theory is more
where Z̈rms is the RMS acceleration of the PCB derived from the ana- effective in evaluating the mechanical safety of solder joints compared
lysis, and fn is the first eigenfrequency of the PCB. with the STT-RV-1 methodology based on Steinberg’s theory.
To calculate the MoS based on the CST-RV-1 methodology, the We propose another design approach that calculates MoS values
maximum in-plane principal strain ε pmax is calculated as [20] based on quasi-static analysis. For this, we derived the random
equivalent quasi-static load of 83.45 Grms calculated by the Mile’s
⎛ ε xrms + ε yrms ε xrms−ε yrms 2 equation as [5]
ε pmax = 3 × ⎜
2
+ ⎛
⎝ 2 ⎠
(
⎞ + ε xyrms )2 ⎞⎟
⎝ ⎠ (8) π
Grms = ⎛ ⎞ (fn )(Q)(PSD fn )
⎝2⎠ (9)
where ε xrms and ε yrms are the RMS in-plane normal strains, and ε xyRMS is
the RMS in-plane shear strain. where Q is the amplification factor, and PSD fn is the input PSD accel-
Table 5 summarizes the MoS values of each package calculated eration at the first eigenfrequency of fn .
using the STT-RV-1 and CST-RV-1 methodologies. The calculated MoS By applying this methodology, the mechanical safety of the solder
values showed positive margin with respect to all TSSOPs. In case of joint can be more simply evaluated while reducing the computation
PBGA packages, the MoS values obtained using the STT-RV-1 metho- time compared with previous methodologies based on random vibra-
dology showed a positive margin for the U1 and U9 packages, although tion analysis. Here, the methodologies based on Steinberg’s theory and
these packages had partial cracks on the solder joint in the fatigue tests, the critical strain theory are referred to as STT-QS-1 and CST-QS-1,

213
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Fig. 12. Configurations of various FEMs ((a) detailed FEM, (b) simplified FEM).

respectively. The MoS values calculated using these methodologies are and solder joints shown in Fig. 12(b), respectively. The first eigen-
summarized in Table 6. The results indicate that only the U5 package frequency calculated from this model was 611.06 Hz, which shows a
showed a negative margin from the STT-QS-1 methodology. In contrast, difference of only 4.75% compared to that of the detailed FEM. The
the results based on CST-QS-1 methodology indicate a negative margin random equivalent static load of 80.46 Grms was used for the quasi-
with respect to all PBGA packages. This accurately represents the fa- static analysis. Here, the methodologies based on Steinberg’s theory and
tigue test results, which showed cracks on the solder joint of PBGA the critical strain theory are referred to as STT-QS-2 and CST-QS-2,
packages. In addition, these results are similar to those obtained using respectively.
the CST-RV-1 methodology although there are some differences in the Table 7 summarizes the results of MoS calculations based on the
calculated MoS values. This indicates that the CST-QS-1 methodology is STT-QS-2 and CST-QS-2 methodologies. The results indicate that only
also effective in evaluating the mechanical safety of solder joints, si- the U5 package showed negative margin when calculating the MoS
milar to the CST-RV-1 methodology, even though the analysis method is values based on the STT-QS-2 methodology. This is similar to those
much simpler than that of random vibration analysis. based on the STT-QS-1 methodology. In case of the CST-QS-2 metho-
However, the construction of a detailed FEM of the entire electro- dology, the MoS values showed negative margin with respect to all
nics package shown in Fig. 12(a) requires much time and effort. PBGA packages, which accurately represents the fatigue test results of
Moreover, the use of such a large-sized FEM for analysis at the elec- PBGA packages shown in Fig. 6(b). These results indicate that the CST-
tronic box level requires significantly longer computation time. QS-2 methodology is more effective for mechanical safety evaluation
Therefore, in this study, the detailed FEM was simplified using 0D than the STT-QS-2 methodology. Further, the simplified FEM is effec-
lumped masses and RBE2 elements to model the masses of the package tive for evaluating the mechanical safety of solder joints as the detailed

214
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

Table 5 Table 7
Comparison of MoS calculated using STT-RV-1 and CST-RV-1 methodologies. Comparison of MoS calculated using STT-QS-2 and CST-QS-2 methodologies.
No. Type STT-RV-1 CST-RV-1 No. Type STT-QS-2 CST-QS-2

Zallow Zmax MoS ∊c ∊max MoS Zallow Zmax MoS εc εmax MoS
(mm) (mm) (μ-strain) (μ-strain) (mm) (mm) (μ-strain) (μ-strain)

U1 PBGA 0.379 0.184 0.65 387 445 −0.31 U1 PBGA 0.379 0.127 1.39 387 531 −0.42
U2 TSSOP 0.737 0.19 2.11 662 208 1.55 U2 TSSOP 0.737 0.135 3.37 662 381 0.39
U3 TSSOP 0.739 0.193 2.06 662 211 1.51 U3 TSSOP 0.739 0.135 3.38 662 348 0.52
U4 PBGA 0.313 0.272 −0.08 387 503 −0.39 U4 PBGA 0.313 0.203 0.23 387 748 −0.59
U5 PBGA 0.22 0.379 −0.54 387 582 −0.47 U5 PBGA 0.22 0.254 −0.31 387 907 −0.66
U6 PBGA 0.314 0.278 −0.10 387 514 −0.40 U6 PBGA 0.314 0.203 0.24 387 769 −0.60
U7 TSSOP 0.689 0.19 1.90 662 208 1.55 U7 TSSOP 0.689 0.135 3.08 662 351 0.51
U8 TSSOP 0.688 0.193 1.85 662 211 1.51 U8 TSSOP 0.688 0.135 3.08 662 340 0.56
U9 PBGA 0.378 0.184 0.65 387 446 −0.31 U9 PBGA 0.378 0.127 1.38 387 531 −0.42

Table 6 Table 8
Comparison of MoS calculated using STT-QS-1 and CST-QS-1 methodologies. Comparison of computation time between various methodologies.
No. Type STT-QS-1 CST-QS-1 Methodology Modal Random Quasi-static Remarks
analysis vibration analysis
Zallow Zmax MoS εc ε pmax MoS (min) analysis (min) (min)
(mm) (mm) (μ-strain) (μ-strain)
CST-RV-1 6.28 38.47 – Detailed FEM
U1 PBGA 0.379 0.122 1.49 387 509 −0.39 CST-QS-1 6.28 – 9.52 Detailed FEM
U2 TSSOP 0.737 0.129 3.57 662 165 2.21 CST-QS-2 1.47 – 1.12 Simplified FEM
U3 TSSOP 0.739 0.129 3.58 662 166 2.19
U4 PBGA 0.313 0.17 0.47 387 615 −0.50
U5 PBGA 0.22 0.231 −0.24 387 650 −0.52
U6 PBGA 0.314 0.184 0.36 387 635 −0.51
U7 TSSOP 0.689 0.129 3.27 662 165 2.21
U8 TSSOP 0.688 0.129 3.27 662 165 2.21
U9 PBGA 0.378 0.122 1.48 387 509 −0.39

FEM shown in Fig. 12(a). Moreover, the time to failure of the solder
joint, estimated by dividing the 20 million critical fatigue cycles into
the first eigenfrequency of PCB, was approximately 9.09 h. Therefore,
the calculated MoS accurately represents the fatigue test results given in
Table 3 because all PBGA packages actually failed within 7.67 h of
excitation.
In previous studies [10,21,22] that predicted the mechanical relia-
bility of solder joints under vibration based on failure probability, at
least 10 specimens were used to validate the mechanical reliability
prediction results. This indicates that a much larger number of speci-
mens are needed for sufficient validation of the methodology proposed
in this study. However, the main objective of this study is to validate
the feasibility of the mechanical design methodology that is applicable Fig. 13. Random vibration fatigue test setup of PCB specimen with CCGA
in the initial design stage of electronics than the conventional Stein- package.
berg’s theory. Only the fatigue test results of two specimens clearly
show that methodologies using the critical strain theory accurately respectively. Fig. 13 shows the fatigue test setup. In the tests, the PCB
represent the experimental results compared with Steinberg’s theory. specimen was exposed to 28 Grms of random vibration for 20 min. In-
Table 8 summarizes the computation time of modal, random vi- situ monitoring of the daisy-chain resistance of the CCGA package was
bration, and quasi-static analyses for each methodology. By using the performed during the test. The failure criteria for the solder joint were
simplified FEM and quasi-static analysis approach, the CST-QS-2 same as those used in the test shown in Fig. 3. Fig. 14 shows the time
methodology needs much less computation time compared to the CST- history of daisy-chain resistances for the PCB specimen. The CCGA
RV-1 methodology. Therefore, it can be applied for the mechanical package rapidly reached the resistance value of 10.5 kΩ after ap-
safety evaluation of electronics that includes many integrated PCB with proximately 5.38 min, which is defined as the failure of the solder joint.
various packages. The optical microscope inspection results shown in Fig. 15 indicate full
To validate the effectiveness of the CST-QS-2 methodology in cracks on several solder columns located at the corner of the package.
evaluating the mechanical safety of a ceramic column grid array A simplified FEM was constructed as the form shown in Fig. 12(b).
(CCGA) package, we performed an additional fatigue test on the PCB The fn analyzed using this FEM was 350 Hz. The equivalent static load
specimen under random vibration excitation. The test results were also calculated from a PSD fn of 0.404 G2/Hz was 64.47 Grms. As the vari-
compared with the MoS values calculated from the CST-QS-2 metho- able C for the CCGA package has not been developed so far, we used the
dology. The PCB specimen used was made of FR-4 with dimensions value of 1.75 to calculate εc in Eq. (2); this value was originally used for
100 mm × 100 mm × 2 mm and total mass of 51.08 g. A daisy-chained the BGA package [5]. The calculated MoS value shown in Table 9 in-
624-pin CCGA package with dimensions 32.5 mm × 32.5 mm dicates a negative margin. The time to failure of this package predicted
× 4.88 mm, and mass of 13.28 g was mounted at the PCB center. The by Eq. (5) was 3.29 min. Therefore, these accurately represent the test
solder materials and solder column were Sn-Pb37 and Sn-Pb80, results of cracks on the solder joint, although there was a difference of

215
T.-Y. Park et al. International Journal of Fatigue 114 (2018) 206–216

the MoS values calculated based on critical strain were much more
effective in evaluating the mechanical safety of solder joints for all the
methodologies. In addition, the methodology based on the quasi-static
analysis of the simplified FEM using 0D lumped mass and RBE2 under
equivalent static load for random vibration was found to be applicable
for the structural design of electronics as a methodology based on the
random vibration analysis of a detailed FEM. The effectiveness of this
methodology was also validated for the CCGA package by comparing
the calculated MoS with the results of testing an additional specimen
under random vibration.

Acknowledgements

This research was supported by the Korean Space Launch Vehicle

Development Program of the National Research Foundation of Korea
(NRF) funded by the Ministry of Science, ICT (MSIT). (NRF-
2016M1A3A1A02021176).

References
Fig. 14. Time profiles of daisy-chain resistance of CCGA package.
[1] de Rooij NF, Gautsch S, Briand D, Marxer C, Mileti G, Noell W, Shea H, Staufer U,
and van der Schoot B. MEMS for Space. in: Proceedings of 2009 International Solid-
State Sensors, Actuators and Microsystems Conference. 2009:17–24.
[2] Ardebili H, Pecht MG. Encapsulation technologies for electronic applications. MD,
USA: Elsevier Inc. and William Andrew; 2009.
[3] Ghaffarian R, Kim NP. Ball Grid Array Reliability Assessment for Aerospace
Applications. Microelectron Reliab 1999;39(1):107–12.
[4] Helvajian H. Microengineering Aerospace Systems. 1st Edition CA, USA: The
Aerospace Press; 1999.
[5] Steinberg DS. Vibration analysis for electronic equipment. 3rd Edition NY, USA:
John Wiley & Sons Inc.; 2000.
[6] Wong TE, Reed BA, Cohen HM and Chu DW. Development of BGA Solder Joint
Vibration Fatigue Life Prediction Mode. in: Proceedings of 49th Electronic
Components and Technology Conference 1999:149–154.
[7] Wu M. Vibration-induced fatigue life estimation of ball grid array packaging. J
Micromech Microeng 2009;19:1–12.
[8] Mathew S, Das D, Osterman M, Pecht M. Virtual remaining life assessment of
electronic hardware subjected to shock and random vibration life cycle loads. J
Institute Environment Sci Technol (IEST) 2007;50(1):86–97.
[9] Yu D, Al-Yafawi A, Nguyen TT, Park S, Chung S. High-cycle fatigue life prediction
for Pb-free BGA under random vibration loading. Microelectron Reliab
Fig. 15. Representative optical micrograph of CCGA solder joints. 2011;51:649–56.
[10] Jang J, Jang G, Lee J, Cho Y, Cinar Y. Fatigue life estimations of solid-state drives
with dummy solder balls under vibration. Int J Fatigue 2016;88:42–8.
[11] Kim YK, Hwang DS. PBGA packaging reliability assessments under random vibra-
Table 9
tions for space applications. Microelectron Reliab 2015;55(1):172–9.
Results of MoS and time to failure of CCGA package calculated using CST-QS-2 [12] Jung IH, Park TW, Seo JH, Han SW, Kim SH. Structural vibration analysis of
methodology. electronic equipment for satellite under launch environment. Key Eng Mater
2004;270–273:1440–5.
Type εc (μ-strain) ε pmax MoS Predicted time to Remarks [13] Oh HU, Jeon SH, Kwon SC. Structural design and analysis of 1U Standardized STEP
(μ-strain) failure (min) cube lab for on-orbit verification of fundamental space technologies. Internat J
Mater Mech Manufact (IJMMM) 2014;2(3):239–44.
CCGA 268 871 −0.72 3.29 Actual time to [14] Chary GV, Habtour E, Drake GS. Improving the reliability in the next generation of
failure: 5.38 min US army platforms through physics of failure analysis. J Fail Anal Prev
2012;12(1):74–85.
[15] Beda MG. A Curvature-based Interpretation of the Steinberg Criterion for Fatigue
Life of Electronic Components in: Proceedings of 48th Annual International
2.09 min compared to the fatigue test results shown in Fig. 14. These Symposium on Microelectronics, 2015;1:1–6.
results indicate that the CST-QS-2 methodology proposed in this study [16] McLeish J. and Blattau N. CAE Apps for Physics of Failure Reliability & Durability
is also effective for evaluating the mechanical safety of solder joints of Simulations in: 60th IEEE Annual Reliability and Maintainability Symposium
(RAMS), 2014:1–6.
the CCGA package. [17] Perkins A, and Sitaraman SK. A Study into the Sequencing of Thermal Cycling and
Vibration Tests. in: Proceedings of 58th IEEE Electronic Components and
5. Conclusions Technology Conference 2008:584–592.
[18] Caswell G. 17 Equations that changed the world – There’s More Than That!!! Part 1.
DfR Solutions 2014:1–13.
This study aims to find a more practical structural design metho- [19] IPC-WP-011. Guidance for Strain Gage Limits for Printed Circuit Assemblies.
dology for evaluating the mechanical safety of solder joints of space- Association Connecting Electronics Industries, (2011) 1–4.
[20] De Clerck J, Epp DS. Rotating machinery, hybrid test methods, vibro-acoustics &
borne electronics during the initial structural design phase. Various laser vibrometry. Switzerland: The Society for Experimental Mechanics Inc.; 2016.
design methodologies were proposed based on MoS calculations with [21] Grieu M, Maire O, Massiot G, Munier C, Bienvenu Y, and Renard J. Sn3.0AgO.5Cu
respect to the allowable displacement of the PCB and the critical strain Solder Joints Lifetime Estimation for Electronic Assemblies under Random
Vibration. in: 2nd IEEE Electronics Systemintegration Technology Conference.
acting on it. The effectiveness of the proposed methodologies was in-
2008:175–180.
vestigated by comparing the calculated MoS values with the results of [22] Huaicheng L, Tong A, Tao T, and Fei Q. Vibration Reliability Test and Analysis of
fatigue tests on the PCB specimen with PBGA packages and TSSOPs Plastic Ball Grid Array. in: 17th International Conference on Electronic Packaging
Technology. 2016:1247–1250.
under a random vibration environment. The comparisons indicate that

216