COMPONENT LEVEL RELIABILITY FOR HIGH TEMPERATURE

POWER COMPUTING WITH SAC305 AND ALTERNATIVE

HIGH RELIABILITY SOLDERS
Thomas Sanders, Sivasubramanian Thirugnanasambandam and John Evans, Ph.D.
Auburn University, Department of Industrial & Systems Engineering
Auburn, AL, USA
[email protected]

Michael Bozack, Ph.D.

Auburn University, Department of Physics
Auburn, AL, USA

Wayne Johnson, Ph.D.

Tennessee Tech University, Department of Electrical & Electronics Engineering
Cookeville, TN, USA

Jeff Suhling, Ph.D.

Auburn University, Department of Mechanical Engineering
Auburn, AL, USA

ABSTRACT
This experiment considers the reliability of a variety of Key words: BGA, PCB, Reliability, Solder, lead free, HALT
different electronic components and evaluates them on
NOMENCLATURE
0.200” power computing printed circuit boards with OSP.
Single-sided assemblies were built separately for the Top- BGA Ball Grid Array
side and Bottom-side of the boards. This data is for boards EPA Environmental Protection Agency
on the FR4-06 substrate. FC Flip Chip
FR Flame Retardant
Isothermal storage at high temperature was used to HALT High Accelerated Life Test
accelerate the aging of the assemblies. Aging Temperatures JEDEC Joint Electron Device Engineering Council
are 25oC, 50oC, and 75oC. Select data from aging times of 0- OSP Organic Solderability Preservative
Months (No Aging, baseline), 6-Months, and 12-Months will PCB Printed Circuit Board
be presented. RoHS Restriction of Hazardous Substances
SEM Scanning Electron Microscopy
The assemblies were subjected thermal cycles of -40°C to SMR Surface Mount Resistors
+125°C on a 120 minute thermal profile. The test was TC Temperature Cycling
subject to JEDEC JESD22-A104-B standard high and low
temperature test in a single-zone environmental chamber to Symbols
assess the solder joint performance. Ag Silver
Bi Bismuth
The principal test components are 5 mm, 6mm, 13mm, Cu Copper
15mm, 17mm, 31mm, 35mm and 45 mm ball grid array Ni Nickel
(BGA) packages with solder ball pitch varying from 0.4 mm Pb Lead
to 1.27 mm. Most of the BGA packages are plastic over- Sb Antimony
molded, while the 31mm and 45mm packages are Super- Sn Tin
BGAs (SBGAs). Several surface mount resistors (SMRs) are
also considered in order to understand the effect of solder Greek Symbols
paste composition on paste-only packages. β Slope
η Characteristic Life
The primary solder for package attachment in this ρ Probability plot
experiment is standard SAC305. Two solders designed for Subscripts
high-temperature reliability are also considered.
Tg Glass Transition Temperature

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 144

 INTRODUCTION EXPERIMENTAL SETUP
Electronic packages are subjected to thermally-induced Test Vehicle
stress due to power cycling and a variety of other sources. It The test vehicle (TC1-SRJ) was designed and constructed
is therefore important to test the reliability of these packages following the JEDEC specifications. The board dimensions
under such conditions in order to determine applicable are 173 mm x 254 mm with a board thickness of 5 mm (200
product lifetimes, etc. This is particularly vital in the case of mil). The tool-hole diameter is 3.8 mm diameter and the
products intended for use in harsh environments. distance from edge of package to the center of the holes is
The semiconductor and packaging industries have been 7mm. There are 6 Copper (Cu) layers with 14,607 pins, 3590
gravitating toward the use of smaller and more reliable through-hole, and 11017 SMT per board. The board
packages to meet the growing market demand for hand held plating/finish is organic solderability preservative (OSP).
electronics. Simultaneously, industry has been moving away Each board was designed to allow for placement of 249
from the use of Lead (Pb) due to the increasing awareness of components, although SMR components are daisy-chained
the health and safety concerns surrounding its use. This has together for readout through a single channel. A different
forced a move away from Eutectic Tin-Lead (63%Sn, design is used for the top and bottom side of the board. In
37%Pb) solder. total, there are 19 channel readouts and one ground for the
top-side, and an additional 39 channel readouts on the
Tin-Lead Eutectic Solder, which has been in use historically bottom-side (ground shared). The test vehicle design is
since at least the time of the Romans, was the traditional shown in Figure 1.
material used for solder interconnect since the inception of
the electronics industry [1,2]. Due to a variety of Two different board/substrate materials were tested: FR4-06
advantageous material properties – such as a good melting and Megtron6. The FR406 board material used in this
temperature and suitability for both reflow ovens and wave experiment was a high-temperature multifunctional glass
soldering processes – Tin-Lead (SnPb) solder is extremely epoxy laminate with a glass transition temperature (Tg) of
difficult to replace without affecting solder joint reliability. 170°C, whereas the Megtron6 board material used in this
On the other hand, most electronic wastes (e- wastes) are not experiment was a high temperature Polyphenylene Ether
treated properly and significant health concerns exist for blend with a glass transition temperature (Tg) of 210°C.
long-term exposure to lead even at low levels. In response to
concerns about lead contamination from e-waste, new rules
in Japan and regulations from the European Union (RoHS
and WEEE) have forced the electronics packaging industry
to use lead-free (Pb-free) solders [3, 4, 5].
Current industry standards for ball grid array (BGA) and
solder interconnect reliability testing rely mainly on pass/fail
electrical continuity functionality test criterion, with limited
knowledge of factors contributing towards the failure. A
variety of factors affect the reliability of the solder joints
used in those electronic components. Chip dimension, chip
structure, and BGA pad size are some of the factors to be
considered, in addition to the principal factor of solder
material properties.

Both the composition and microstructure of the solder joint

will affect its bulk properties. These will determine a joints
ability to provide the necessary mechanical and electrical
connection and strongly affect the reliability of the joint.
Although an initial microstructure will be present following
assembly – which will involve one or more soldering steps –
but this structure will continue to evolve over the lifetime of
the joint.

The combination of disparate coefficients of thermal

expansion (CTEs) and temperature changes can result
excessive stress, leading to weakening of the solder joints
and eventual component/package failure. During a Thermal
Cycling (TC) test, solder materials are typically subjected to
higher temperatures above half of their melting point (i.e.
greater than 0.5 in terms of their homologous temperature),
Figure 1. Test Vehicle Design: a) Top-Side and b) bottom-
facilitating thermally driven evolution and failure
side of the TC1-SRJ test vehicle.
mechanisms.

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 145

 Surface Mount Assembly Momentum. Two different aperture stencils were used for
The test boards were made by TTM Technologies (Time-To- the top and bottom sides. The top-side stencil had a 127
Market Interconnect Solutions), Chippewa Falls Division. microns (5 mil) aperture, while the bottom-side which
Dummy-die (daisy-chained) components were sourced from houses the finer-pitch components had a 76 microns (3 mil)
Practical Components. The ball grid array (BGA) design and aperture. Bottom-side boards were double-printed in order to
wiring scheme follow a simple daisy chain structure. The get adequate solder volume on the fine-pitch components.
test components were assembled at STI Electronics Inc. in Two pick-and-place machines were used for the TC1-SRJ
Madison, Alabama. test vehicle assembly: the Juki KE-2080L and Juki FX3.
Solder reflow was done using a Heller 1913 MKIII reflow
A total of 880 boards were built: 720 boards of FR406 oven. Two different reflow profiles were used: one for SnPb
substrate material, and 160 boards of Megtron6 substrate and the other for the SAC305 and Innolot solder pastes. A
material. There were also 30 additional FR406 intended for number of quality assurance steps were taken. The resistance
setup during build/assembly work. The boards are being of each daisy-chained circuit component was checked by
built (assembled) in two groups. 660 boards were built as hand following reflow in order to eliminate them from
‘Top-side’ boards, with only the top-side components inclusion in further testing (excluding the socketed
assembled. 220 boards were built as ‘Bottom-side’ boards, components, which were hand-assembled into the LGA
with only the bottom-side components assembled. sockets later). Post-assembly resistance testing showed a
100% yield. Boards were also visually inspected and x-ray
This experiment is designed to evaluate the effect of long analysis was used to determine typical solder-joint quality
term isothermal aging of several lead-free solder alloys, following reflow. Additional measurements were made of
including SAC105 and SAC305. The test matrix with the solder paste height and diameter. Finally, several
board groupings is shown in Figure 2, below. There are four components on one of the setup boards were sacrificed in a
(4) aging times (0, 6, 12, and 24 months) and three different ‘pry test’ in order to assure the mechanical strength of the
aging temperatures (25°C, 50°C, and 75°C). solder joints as reflowed. XRD voiding analysis and shear
(“pry”) testing also indicated excellent build quality.

Figure 2. Board Groupings

The principal test components are ball grid array (BGA)

packages of 5 mm, 6mm, 13mm, 15mm, 17mm, 31mm,
35mm and 45 mm with solder ball pitch varying from 0.4
mm to 1.27 mm. Three different solder paste compositions
were used, in combination with four different solder ball
compositions. Various surface mount resistors (SMRs) and a
limited number of land grid array (LGA) sockets were also
used. Figure 3 summarizes the below summarize the
component matrix for the TC1-SRJ test vehicle. The LGA
socket was used to house the memory module, a pin grid
array (PGA) which was attached by hand following
assembly. Heat sinks for the 45mm and 35mm components
were also added later by hand.

The three solder pastes used in this test were SnPb (eutectic),
SAC305, and Innolot. The SAC305 (“Type 4”) paste and the
Innolot paste were provided by Cookson, while the SnPb
eutectic paste was sourced from Kester. The screen printing
machine used was a Speed line Technologies MPM

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 146

 Component Ball Alloy Pitch Dimension A104-B standard high and low temperature test. The ramp
rate of 15℃ per minute is a stipulated condition in the
CVBGA97 SAC 305 0.4mm 5mm JEDEC JESD22. Depending on the thermal capacity, an
CVBGA97 SAC 105 0.4mm 5mm appropriate condition in the range of 10 to 15℃ can be
CVBGA97 SnPb 0.4mm 5mm selected, and for this test a ramp rate of 10℃ per minute was
selected for the single zone chamber. The single-zone
CVBGA432 SAC 305 0.4mm 13mm
thermal cycling chamber used was a Blue M ETC-16. Two
CVBGA432 SAC 105 0.4mm 13mm chambers were used, one for the 0-Month (“No Aging”) and
CVBGA432 SnPb 0.4mm 13mm 12-Month aging groups and a second chamber for the 6-
Month aging group.
CTBGA84 SAC 305 0.5mm 6mm
CTBGA84 SnPb 0.5mm 6mm The boards were placed vertically in the chamber and the
CABGA36 SAC 305 0.8mm 6mm wiring passed through the independent access ports to the
Labview-based monitoring system. The monitoring system
CABGA36 Innolot 0.8mm 6mm
utilized a Keithley 7002 switching system and Keithley 2000
CABGA36 SnPb 0.8mm 6mm and 2001 digital multi-meters (DMMs). Monitoring was
CABGA36 "SAC-Y" 0.8mm 6mm accomplished by cyclically scanning the resistance on each
CABGA208 SAC 305 0.8mm 15mm channel. The monitoring system used a ground-switching
system enabled by custom software and interface boards to
CABGA208 SAC 105 0.8mm 15mm monitor up to 3600 channels using a single switching
CABGA208 Innolot 0.8mm 15mm system. Solder joint failure was defined as increase of
CABGA208 SnPb 0.8mm 15mm electrical continuity greater than 100 ohms above baseline
(uncycled) resistance. Channels that exhibited five (5)
CABGA208 "SAC-Y" 0.8mm 15mm consecutive threshold-exceeding events were recorded as a
CABGA256 SAC 305 1.0mm 17mm failure in the monitoring system. Due to wiring limitations,
CABGA256 SAC 105 1.0mm 17mm the 6-Month aging group was hand-probed at ~50 cycle
intervals, with failures corresponding to ‘open’ resistance
CABGA256 SnPb 1.0mm 17mm values.
PBGA1156 SAC 305 1.0mm 35mm
PBGA1156 SAC 105 1.0mm 35mm Each aging group will be subjected to 3000 thermal cycles.
Data is presented for 3000 TC for the 0-Month (No Aging)
PBGA1156 SnPb 1.0mm 35mm group, ~2700 cycles for the 6-Month aging group, and ~900
SBGA304 SAC 305 1.27mm 31mm cycles for the 12-Month aging group. The failure data was
SBGA304 SnPb 1.27mm 31mm analyzed and the reliability of the solder joints was
determined in terms of the characteristic life (η) and slope
SBGA600 SAC 305 1.27mm 45mm (β) from a two parameter Weibull analysis [3,6,7].
SBGA600 SnPb 1.27mm 45mm
1206 SMR 100% Sn 3.2x1.6mm RESULTS AND DISCUSSIONS
Temperature Cycling Results
1210 SMR 100% Sn 3.2x2.6mm The temperature cycling test results below show some the
0805 SMR 100% Sn 0.8x0.5mm highlights from the reliability data from the 0-Month (No
0603 SMR 100% Sn 0.6x0.3mm Aging) group, 6-Month aging group, and 12-Month aging
groups. For analysis purposes, it is sometimes convenient to
0402 SMR 100% Sn 0.4x0.2mm look at the overall trends in the failure data and illustrate
0201 SMR 100% Sn 0.2x0.1mm particular key points using data from ‘representative’ parts in
01005 SMR 100% Sn 0.1x0.05mm cases when several components show similar overall
Figure 3. Component Matrix behavior. Specifically, the failure trends are constant in all
available data for the smaller plastic ball grid array packages
(5mm – 17mm), so we will begin by illustrating overall
Test Setup trends using data from the CABGA 208 component.
The electrical components for this experiment were daisy
chained (dummy die) for electrical continuity testing and in The CABGA 208 component is found on both the top and
situ, continuous monitoring throughout the thermal cycling bottom side of the board and within all groups, allowing for
(TC) test. The resistance for each component was multiple comparisons across various experimental
independently monitored during the temperature cycle test. parameters. Below are a few key points from the failure data
The assemblies were subjected to thermal cycles of -40°C to of this component.
+125°C on a 120 minute thermal profile in a single-zone
environmental chamber to assess the solder joint When examining the effect of various solder paste [P] and
performance. The TC was based on the JEDEC JESD22- sphere [S] combinations, the Characteristic Life values show

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 147

 the following pattern, listed from best to worst: (1) Matched
Innolot ([P] + [S])*, (2) [S]SAC305 doped with [P]Innolot,
(3) Matched SAC305 ([P]+[S]), and (4) Matched SnPb
([P]+[S])

* Note that Matched Innolot data is available only for the

CABGA 208 and CABGA 36 components.
Unsurprisingly, under our thermal cycling (TC) test,
components balled with SAC305 spheres are more reliable
than equivalent SAC105 balled components. This pattern
holds for both SAC305 and Innolot solder pastes. No
difference is seen in the pattern of failures when comparing
components found on the Top-Side of the board and the
Bottom-Side of the board, with the exception of the SAC-Y
(Bi doped) spheres.
Figure 6. Weibull Plot: CABGA208 – FR4-06 – 12-Month
Figures 4-6 show Weibull plots for the CABGA 208
component on the FR4-06 substrate. Note that every Figure 7 shows the trend in characteristic life based on the
subgroup may not appear on each graph and color coding of available solder material combinations for the CABGA 208
groups varies. Data is from 75oC aging unless otherwise component on the FR406 substrate.
marked. From the Weibull plots, there are three failure
criteria that can be used to determine the reliability, i.e. first
failure, mean life and characteristic life. The mean life and
characteristic life is preferred than the first failure since they
are statistically more accurate for life predictions [9].

Figure 7. Characteristic Life: CABGA208 – FR4-06 – No

Aging

When examining the effect of the substrate (FR4-06 vs.

Megtron6), a clear trend emerges. Smaller (5mm – 17mm)
plastic BGA components assembled on the FR4-06 substrate
are universally more reliable than identical components
assembled on the Megtron6 substrate, when controlling for
all other factors. (See Figure 12 for a comparison across
Figure 4. Weibull Plot: CABGA208 – FR4-06 – No Aging multiple packages.)

Other components, naturally, show somewhat different

trends in their failure data. Although most of the plastic
BGA paakcages follow all of the above trends, our largest
PBGA package stands out in at least one regard. The PBGA
1156 is a 35mm component found only on the top-side of the
board, with 1.0mm pitch. With a solid 34x34 I/O array, it has
by far the largest I/O count in this experiment.

The key difference between the data from the PBGA 1156
and the other plastic packages is that there is not a significant
improvement seen in the reliability of the solder joints when
doping with Innolot paste for the PBGA 1156. Characteristic
life values are similar for SAC305 and Innolot paste. (Note
that this package fails later in testing, and data is only
Figure 5. Weibull Plot: CABGA208 – FR4-06 – 6-Month avialable for the the 0-Month and 6-Month aging groups.)
Figure 8 shows the characteristic life values for the PBGA

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 148

 1156 package on the FR4-06 substrate, 75oC aging. Data is
shown for components with and without heatsinks.

Figure 9. Characteristic Life: SBGA304 – No Aging.

Hashed bars indicate that insufficient failure data exists to
form a Weibull distribution.
Figure 8. Characteristic Life: PBGA1156 – FR4-06 – No
Aging

Another interesting difference can be seen in the failure data

of the plastic packages discussed so far and the two Super-
BGA (SBGA) packages tested: the SBGA 304 and the
SBGA 600. These are cavity-down, metal-capped
components, and so are structurally quite different from the
previously discussed packages. The SBGA 304 package is a
large-pitch (1.27mm) component found only on the top-side
of the board. This package has a footprint of 31mm x 31mm.
The SBGA 600 component has the largest footprint in this
experiment (45mm x 45mm), and is found solely on the top-
side of the board. Like the SBGA 304, this is a metal-capped
package with large pitch (1.27mm).
Figure 10. Characteristic Life: SBGA600 – No Aging. (*
Figures 9 and 10 show the characteristic life values for the Some early failures are significantly affecting the
No Aging group. The failure trends shown are mirrored in characteristic life of this distribution.)
the 6-Month aging data. Data is shown for 75oC aging. There
are two key features that make this data stand out from what The overall differences in reliability trends are summarized
one might consider to be the ‘normal trends’ from the in Figures 11 and 12, which show the comparisons
smaller plastic BGA components (5mm – 17mm, as highlighting the effect of paste-doping and substrate-
represented by the CABGA 208). material, respectively.

One key difference is that – like the larger plastic BGA

component (PBGA 1156) – solder joints doped with Innolot
paste do not perform better than matched SAC305 joints. In
fact, both the SBGA 304 and SBGA 600 show higher
reliability with matched SAC305 paste/spheres that with
SAC305-spheres doped with Innolot-paste. (This behavior is
more evident for the SBGA 304 component.) A second key
difference is that the reliability of these two Super-BGA
packages is higher on the Megtron6 substrate than on the
FR4-06 substrate in this experiment, which reverses the
trend from all of the plastic packages, including both the
smaller packages and the larger PBGA 1156.

Figure 11. Weibull plot for (a) Comparison of different

paste using SAC305 sphere in CVBGA432

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 149

 For Innolot paste, SAC305 spheres have superior reliability ACKNOWLEDGEMENTS
to SAC105 spheres, as with other components. See Figure The authors would like to thank Center for Advanced
9b. Vehicle and Extreme Environment Electronics at Auburn
University for providing much of the testing equipment.

REFERENCES
[1] K. Gilleo, “The First 7,000 Years of Soldering, Part
I.” Circuit Assembly, Oct. 1994.
[2] Hunt and LePrevost. “Getting the Lead Out –
Soldering with Lead-free Solders.” Proc. Of the RCI 22nd
International Convention.
[3] S. Thirugnanasambandam, J. Evans, M. Perry, B.
Lewis, D. Baldwin, K. Stahn, and M. Roy, “Component
level reliability on different dimensions of lead free wafer
level chip scale packages subjected to extreme
temperatures,” in 13th InterSociety Conference on Thermal
and Thermomechanical Phenomena in Electronic Systems,
Figure 12. Weibull plot for (b) Comparison of different 2012, pp. 612–618.
sphere using Innolot paste in CVBGA432 [4] T. T. Nguyen, D. Yu, and S. B. Park,
“Characterizing the Mechanical Properties of Actual
SUMMARY AND CONCLUSION SAC105, SAC305, and SAC405 Solder Joints by Digital
The failure data from this test was found to follow specific Image Correlation,” Journal of Electronic Materials, vol. 40,
trends depending on the type and size of the component. The no. 6, pp. 1409-1415, Feb. 2011.
smaller plastic ball grid array (BGA) packages (5mm – [5] H. Ma, T. K. Lee, D. H. Kim, H. G. Park, S. H.
17mm) show failure data trends that are exemplified by the Kim, and K. C. Liu, “Isothermal Aging Effects on the
CABGA 208 (15mm) package. Regarding the effect of Mechanical Shock Performance of Lead-Free Solder Joints,”
various solder paste [P] and sphere [S] combinations, the Components, Packaging and Manufacturing Technology,
Characteristic Life values show the following pattern, listed IEEE Transactions on, vol. 1, no. 5, pp. 714–721, 2011.
from best to worst: [9] D. Suraski, “The current status of lead-free solder
(1) Matched Innolot ([P] + [S])*, alloys,” IEEE TRANSACTIONS ON ELECTRONICS
(2) (2) [S]SAC305 doped with [P]Innolot, PACKAGING MANUFACTURING, VOL. 24, NO. 4,
(3) (3) Matched SAC305 ([P]+[S]), and OCTOBER 2001.
(4) (4) Matched SnPb ([P]+[S]) [7] R. Chilukuri, “TECHNOLOGY SOLUTIONS FOR
(* Note that Matched Innolot data is available only for the A DYNAMIC AND DIVERSE WLCSP MARKET,” in 7th
CABGA 208 and CABGA 36 components.) Annual International Wafer Level Packaging Conference,
2010.
A very clear trend also exists regarding the substrate effect: [8] J. O. Richard Coyle, Heather McCormick, “Pb-free
smaller plastic BGA components assembled on the FR4-06 Alloy Silver Content and Thermal Fatigue Reliability of a
substrate are universally more reliable than identical Large Plastic Ball Grid Array (PBGA) Package,” SMTA
components assembled on the Megtron6 substrate, when Journal, vol. 24, no. 1, pp. 27-33, 2011.
controlling for all other factors. [9] H. Ma and J. C. Suhling, “A review of mechanical
properties of lead-free solders for electronic packaging,”
A larger plastic BGA component, the PBGA 1156, shows Journal of Materials Science, vol. 44, no. 5, pp. 1141-1158,
similar failure trends in terms of most particulars. Howerver, Jan. 2009.
one key difference exists. The PBGA does not show a [10] J. Zhang, S. Thirugnanasambandam, J. L. Evans,
significant improvement in joint reliability during Innolot M. J. Bozack, and R. Sesek, “Impact of Isothermal Aging on
paste doping. the Long-Term Reliability of Fine-Pitch Ball Grid Array
Packages With Different Sn-Ag-Cu Solder Joints,” IEEE
Two Super-BGA components, the SBGA 304 and SBGA Trans.Components, Packag. Manuf. Technol., vol. 2, no. 8,
600, also show differences in failure data trends to the pp. 1317–1328, Aug. 2012.
smaller plastic ball grid arrays. These are cavity-down,
metal-capped components, and so are structurally quite
different from the previously discussed packages. Like the
PBGA 1156, these packages do not show an improvement in
reliability via Innolot paste doping (in fact, reliability is
lower in the doped case). Moreover, both of the Super-BGA
components show a reversal of the substrate-effect seen in
the plastic packages and display higher reliabiltiy on the
Megtron6 substrate than on the FR4-06 substrate.

Proceedings of SMTA International, Sep. 27 - Oct. 1, 2015, Rosemont, IL Page 150