Analysis on solder ball shear testing conditions with a

simple computational model

Electronic Pack aging Laborat ory, Depart ment of Mechanical Engineering ,

S.W. Ricky Lee

Hong Kong University Science & Technolog y, Kowloon, Hong Kong

Electronic Pack aging Laborat ory, Depart ment of Mechanical Engineering ,

Xingjia Huang

Hong Kong University Science & Technolog y, Kowloon, Hong Kong

Keywords Introduction packages. The substrates for ball attachment were BT

Plastic ball grid array, Solder joints, laminates with a thickness of 0.46 mm. The solder bond pads
Finite element analysis, Reliability In the past decade, BGA technologies have overwhelmed were solder-mask-deŽ ned with an opening of 0.6 mm in
the whole IC packaging industry. This is because BGA diameter. The metallization of bond pads was Cu with Ni/Au
Abstract packages have many advantages over conventional ones. electro-plating. The solder balls were attached to the bond
This paper introduces a simple Among them are larger number of I/Os, self-alignment pads using a standard SMT re ow proŽ le. After re ow, the
computational model for the capability, more robust assembly process and better average ball height and diameter were 0.64 mm and
analysis on the solder ball shear thermal/electrical performance (Lau, 1996; 2000; Lau and 0.78 mm, respectively. The ball shear tests were conducted
testing conditions. Both two- Pao, 1997). Furthermore, BGA packages are compatible using a Dage-4000S machine. In the present study, the shear
dimensional (2-D) and three- with surface mount technology (SMT), consequently, ram height was Ž xed at 10% of the solder ball height (i.e.
dimensional (3-D) Žnite element leading to high throughput and low cost assemblies for mass 64 m m), which complies with the new JEDEC standard, but
models are used to investigate the production. For surface mount components (SMCs), the the shear ram speed varied (20 m m/s, 50m m/s, 100 m m/s,
effect of shear ram speed on the solder joints are not only the passage of electrical signals, 200 m m/s and 500 m m/s). For each testing condition, 25
solder ball shear streng th of power and ground, but also act as the mechanical support to solder balls were sheared. The peak shear forces were
plastic ball grid array (PBGA) hold the module in position on the printed circuit board recorded and the calculated average was taken as the ball
packages. An effective thickness (PCB). Therefore, solder joint reliability is a major concern shear strength. The testing results of ball shear strength are
is identiŽed for the 2-D Žnite for BGA packages. presented in Table I.
element analysis. By using this Currently, the most popular method to evaluate the
effective thickness as a scale strength of solder ball attachment is the ball shear test. The
factor, it is shown that the 2D typical value of solder ball shear strength for PBGA Finite element model
model is feasible for the study of packages may be around 1000 gf (for 30-mil solder balls)
3-D problems. The computational (Coyle et al., 1999; Lee et al., 2000a; Levis and Mawer, In the present study, both 2-D and 3-D Ž nite element models
model is validated by ex perimental 2000). Although such tests are simple and convenient to were built and used in the analyses. The geometric models
data in terms of load-displacement implement, there is not much mechanics justiŽ cation to were based on the average dimensions measured from the
curves. The results from both interpret the testing results. Most people still use the method cross-section of specimens for solder ball shear tests. A
testing and modeling indicate that of A-to-B comparison to determine the acceptance of commercial Ž nite element code, ANSYS Release 5.6, was
the shear ram speed has a packages. In July 2000, JEDEC published a new standard, used to perform the computational analysis. Figure 1 shows
substantial effect on the solder JESD22-B117, for BGA solder ball shear tests (JEDEC, the 2-D model simulating the ball shear test. The Ž nite
ball shear streng th. In general, 2000). In this publication, the testing condition is speciŽ ed element meshes consisted of the solder ball, the Cu bond
faster ram speed can result in in terms of the shear ram height only. The speciŽ cation is pad, and the BT substrate, which were modeled by 8-node
higher ball shear streng th. that the gap between the shear ram tip and the ball mounting plane strain elements (PLANE183). The shear ram was
Therefore, the characterization of surface should be larger than 0.05 mm (2 mils) and smaller considered a rigid body. A feature in ANSYS using the
solder ball shear streng th is than (or equal to) 25 per cent of the ball height. However, surface-to-surface target element (TARGE169) and the
loading rate-dependent. another important testing parameter, namely, the shear ram contact element (CONTA172) was employed to simulate the
speed (loading rate), is not addressed in this standard. The contact between the shear ram and the solder ball. The left,
The Research Grant Council of lack of speciŽ cation in the shear speed may cause confusion right, and bottom sides of BT substrate were Ž xed as rigid
Hong Kong sponsore d this study for the comparison of solder ball shear strength character- boundaries. In the present 2-D model, there were 1582
through the grant of ized with different loading rates (Coyle and Solan, 2000; nodes, 490 PLANE183 elements, 8 TARGE169 elements
HKUST623101E to the Hong Kong Hasegawa et al., 2000). Therefore, further research effort is and 40 CONTA172 elements. Since this was a time-
Univers ity of Science and required for understanding the rationale of solder ball shear dependent non-linear analysis, both large deformation and
Technology (HKUST). The authors tests and for investigating the effect of shear speed. transient options were enabled in the analysis.
wish to acknowledge this support. The present study is aimed at establishing a Ž nite For the three-dimensional analysis, 20-node solid
element model for the analysis of ball shear tests in order to elements (SOLID186) were used to mesh the 3-D model.
evaluate the solder ball attachment strength of PBGA Due to the symmetry, only one half of the solder ball was
packages. In particular, the emphasis is placed on modeled as shown in Figure 2. Similar to the 2-D modeling
investigating the effect of the shear speed. The results above, the surface-to-surface target element (TARGE170)
obtained from this computational modeling analysis will be and contact element (CONTA174) were employed to
validated by experimental data. simulate the contact between the shear ram and the solder
ball during the solder ball shear test. In the 3-D model, there
were 11277 nodes, 2520 SOLID186 elements, 249
TARGE170 elements and 240 CONTA174 elements. The
Solder ball shear tests material properties and other modeling features for the 3-D
model were the same as those assumed in the 2-D analysis.
The purpose of solder ball shear tests in the present study
Soldering & Surface Mount was for model validation. Solder balls under evaluation were
Technology standard 0.76 mm (30 mils) 63Sn-37Pb spheres for PBGA
14/ 1 [2002] 45–48 Constitutive models for the materials
q MCB UP Limited Thecurrent issueandfull text archiveofthisjournal isavailableat In the present computational model, all materials, except the
[ISSN 0954-0911]
http: www.em eraldin sight.com 0954- 0911.h tm solder ball, are assumed to behave as linear elastic materials.
DOI 10.1108/09540910210416468 The material properties used in modeling are listed in

[ 45 ]
S.W. Rick y Lee and X ingjia Huang Table I
Analysis on solder ball shear Ball shear streng th under various shear speed
testing conditions with a simple
computational model Shear speed (mm/s)
Soldering & Surface Mount Fixed ram height (64 mm) 20 50 100 200 500
Technology 1131 (4%) 1171 (5%) 1244 (6%) 1350 (4%) 1418 (3%)
14/ 1 [2002] 45–48 1143 1202 1244 1291 1350
Testing
Modeling
Note: (*) –Standard deviation, unit – g f.

Figure 1 ANSYS, equation (2) is called the implicit creep model and
2-D Žnite element model the equation is further modiŽ ed as:
³ ´
de C4
= C1 [sinh(C2s)] C3 exp 2 (3)
dt T
where s is the effective stress and de/dt is the steady-state
effective creep strain rate. The values of C1, C2, C3, and C4
in equation (3) for the eutectic solder are given in Table III
(Lee et al., 2000b).

Determination of effective thickness

The direct result from the present computational analysis is
the force-displacement curve simulating the ball shear test.
It should be noted that, for the 2-D plane strain analysis, the
force output from ANSYS was in the unit of N/mm (force
per unit thickness). However, the actual diameter of solder
was 0.78 mm ( , 1 mm). In order to obtain the meaningful
Table II. In the present study, the solder is considered as a shear force for comparison, an effective thickness (scale
visco-plastic material, which follows the Garofalo-Arrhe- factor) should be deŽ ned. Based on the fact that the ball
nius creep constitutive (Lau, 1995): mounting pad has a diameter of 0.6 mm and the solder ball
³ ´ ± t² n ³ ´ has an axisymmetric geometry, it may be concluded that the
dg G Q effective thickness of the 2-D model must be smaller than
=C sinh ! exp 2 (1)
dt H G kH 0.6 mm. With such an observation, the scale factor for the
where g is the steady-state creep shear strain; dg/dt is the effective thickness could be estimated as follows:
steady-state creep shear strain rate; t is the time; C is a In the present study, the shear test condition with 64 m m
material constant; G is the temperature-dependent shear ram height and 100 m m/s shear speed was chosen as the base
modulus; H is the absolute temperature (8 K); ! deŽ nes the case. For this base case, the peak shear force obtained from
stress level at which the power law stress dependence breaks ANSYS was 2969 gf/mm. Compared with the corresponding
down; t is the shear stress, n is the stress exponent; Q is the experimental data (1244 gf, see Table I), the effective
activation energy for diffusion mechanism; and k is the thickness, t, can be determined as:
Boltzmann constant (8 617 £ 102 5 eV 8 K) Darveaux and 1244 gf
Banerji modiŽ ed equation (1) to t= = 0 42 mm (4)
2969 gf mm
± t² n ³ ´
dg Q Figure 3 shows the original shear force-displacement curve
= A sinh exp 2 (2)
dt B kT from ANSYS and the modiŽ ed curve with an effective
thickness of 0.42 mm for the base case. Subsequently, the
and have characterized the coefŽ cients, A, B, n, and Q, for a same value of effective thickness was applied to all other
few solder materials (Darveaux and Banerji, 1992). In 2-D analyses. If the same effective thickness could yield to
good agreement between testing and modeling for all cases
(which will be shown in the next section), then it is
Figure 2
3-D Žnite element model
reasonable to conclude that the selected effective thickness

Table II
Elastic material properties for modeling

Materials E (MPa) n r (g/cm3)

29,800 0.40 8.41

128,700 0.34 8.31
63Sn-37Pb

14,000 0.39 1.2

Cu pad
BT Substrate

Table III
Creep option input parameters for ANSY S

C1 (s2 1
) C2 (MPa2 1
) C3 C4 (8 K)

339.0102 0.062653 3.3 6,360

[ 46 ]
S.W. Ricky Lee and X ingjia Huang Figure 3 Model validation and discussion
A nalysis on solder ball shear ModiŽcation of force-displacement curve
testing conditions with a simple
Figures 5 – 8 present the results of 2-D analyses for the shear
computational model
speeds of 20, 50, 200 and 500 m m/s, respectively. All
corresponding experimental data are plotted as well for
Soldering & Surface Mount comparison. The peak shear forces, which are regarded as
Technology the ball shear strength, are also summarized in Table I. It is
14/ 1 [2002] 45–48 observed that the testing and modeling results are in good
agreement for all cases. Since the scale factor was
determined using the base case only, it is reasonable to
conclude that the selected effective thickness was not a
random coincidence. It should carry a certain physical
meaning, which represents the scale factor between the 2-D

Figure 5
Force displacement curves for 204 m m/ s

was not a random coincidence. It should carry a certain

physical meaning, which represents the scale factor between
the 2-D and 3-D analyses. Furthermore, the fact that the
chosen effective thickness is smaller than 0.6 mm also
enhances the conŽ dence in the selected value.

Comparison between 2-D and 3-D

analyses
The force-displacement curve obtained from the 3-D
analysis for the base case is presented in Figure 4 and
compared to the corresponding results from the 2-D analysis
and experimental data. It is observed that the 2-D modeling
Ž ts the experimental data better than the 3-D analysis.
In general, it is considered that the 3-D model should Figure 6
give more accurate results than the 2-D analysis. However, Force displacement curves for 50 m m/ s
due to the accumulation of numerical truncation errors, this
may not be always true for the large scale computation in the
non-linear time-dependent analyses. In the present study, the
degrees of freedom for the 2-D and the 3-D analyses were
3164 (1582 nodes £ 2) and 33831 (11277 nodes £ 3),
respectively. With one order of magnitude in the number of
degrees of freedom, the accumulation of numerical errors
might play a signiŽ cant role making the 3-D analysis a
worse case.
In addition, the 3-D analysis is usually very time
consuming. For the comparison shown in Figure 4, while the
3-D analysis took more than 3 days, the 2-D model only
required 3 h using the same computing facility. Therefore,
the 2-D analysis should be a better choice for the present
study, as long as the proposed effective thickness model is
validated.

Figure 4
Comparison of force-displacement curves
Figure 7
Force displacement curves for 200 m m/ s

[ 47 ]
S.W. Rick y Lee and X ingjia Huang
. The results from testing and 2-D modeling were in good
Figure 8
Analysis on solder ball shear Force displacement curves for 500 m m/ s
agreement. An effective thickness was identiŽ ed for the
testing conditions with a simple
2-D plane strain analysis. With such a scale factor, it is
computational model
feasible to study 3-D problem with a 2-D Ž nite element
Soldering & Surface Mount
model.
Technology
. The ideal solder ball shear test conditions were
14/ 1 [2002] 45–48
recommended to be the cases with shear speed slower
than 200 m m/s.
It should be noted that although the present analyses
were performed for solder balls of BGA package with 30 mil
solder balls, the methodology may be applied to other cases
such as BGA packages with smaller solder balls and the
solder bumps of  ip chips. With the assistance of
computational modeling, it seems that it is possible to
forecast the ball shear strength under various testing
conditions as long as a base case is tested in advance. In
addition, with further stress analyses, it may be possible to
achieve more meaningful comparisons among different
packages. The results obtained from this study should be
very helpful for the electronics manufacturing industry to
and 3-D analyses. Based on the good agreement in interpret their testing data and determine the acceptance
comparison, the concept of effective thickness is validated. criteria for the products with solder ball attachment.
For the ease of summarizing the general trends of ball
shear strength, the data given in Table I are plotted in
Figure 9. From this Ž gure, one can Ž nd that, the ball shear
strength increases with the increase of shear speed. Another
point to be noted is that, although in general the shear
strengths from testing and modeling are in good agreement,
there is a notable discrepancy for the cases with shear speed
References
Coyle, R.J. and Solan, P.P. (2000), “The in uence of test parameters
faster than 100 m m/s. From this point of view, it may be and package design features on ball shear test requirements”,
deduced that the ideal solder ball shear testing conditions are 26th IEEE/CPMT International Electronics Manufacturing
the cases with shear speed slower than 200 m m/s. Technology Symposium, pp. 168-77.
Coyle, R.J., Holiday, A., Mescher, P., Solan, P.P., Gahr, S.A., Cyker,
H.A., Dorey, J.K. and Ejim, T.I. (1999), “The in uence of
Summary and conclusions nickel/gold surface Ž nish on the assembly quality and long term
reliability of thermally enhanced BGA packages”, 24th
In the present study, both 2-D and 3-D Ž nite element models IEEE/CPMT International Electronics Manufacturing
were developed for the analysis of solder ball shear tests. Technology Symposium, pp. 23-35.
The effect of shear speed on solder ball shear strength was Darveaux, R. and Banerji, K. (1992), “Constitutive relations for tin-
studied not only by experimental investigation, but also by based solder joints”, Proceedings of IEEE Electronic
computational modeling. The results are summarized as Components & Technology Conference, pp. 538-51.
follows. Hasegawa, K., Takahashi, A., Noudou, T., Nakajima, S., Takahashi,
A., Nomoto, M. and Nakaso, A. (2000), “Electroless Ni-
. The shear speed has substantial effects on the solder ball P/Pd/Au plating for semiconductor package substrate”, SMTA
shear strength. The data from both testing and modeling International Proceedings of the Technical Program, Edina,
MN, pp. 225-31.
indicates that faster shear speed would results in higher
JEDEC (2000), “BGA Ball Shear”, JEDEC Solid State Technology
ball shear strength. Association JESD22-B117, July.
Lau, J.H. (1995), Ball Grid Technology, McGraw-Hill, New York,
Figure 9 NY.
Effect of shear speed on the ball shear streng th Lau, J.H. (1996), Flip Chip Technologies, McGraw-Hill, New York,
NY.
Lau, J.H. (2000), Low Cost Flip Chip Technologies, McGraw-Hill,
New York, NY.
Lau, J.H. and Pao, Y.H. (1997), Solder Joint Reliability of BGA,
CSP, Flip Chip, and Fine Pitch SMT Assemblies, McGraw-Hill,
New York, NY.
Lee, S-W.R., Newman, K. and Hu, L. (2000b), “Thermal fatigue
analysis of PBGA solder joints with the consideration of
damage evolution”, Packaging of Electronic and Photonic
Devices, EEP-Vol. 28, November 5-10, FL, pp. 207-12.
Lee, S-W.R., Yan, C.C., Karim, Z. and Huang, X. (2000a),
“Assessment on the effect of Electroless nickel plating on the
reliability of solder ball attachment to the bond pads of PBGA
substrate”, Proc. 50th ECTC, May 21-24, NV, pp. 868-73.
Levis, K.M. and Mawer, A. (2000), “Assembly and solder joint
reliability of plastic ball grid array with lead-free versus
lead-tin interconnect”, Proc. 50th ECTC, May 21-24, NV,
pp. 1198-204.

[ 48 ]