SMTA Pan Pacific Microelectronics Symposium, Kauai, HI, Jan. 25 – 27, 2005

FLEXURAL STRENGTH OF BGA SOLDER JOINTS WITH ENIG SUBSTRATE FINISH USING 4-POINT BEND TEST

Anurag Bansal, Sam Yoon, and Vadali Mahadev

Altera Corporation, San Jose, CA, USA abansal@altera.com

ABSTRACT PCB assembly and handling operations involving excessive monotonic bending at high strain rates can cause brittle solder joint failures in BGA packages. This paper deals with brittle fracture of flip-chip BGA packages with electroless Nickel immersion Gold (ENIG) substrate pad finish. A high speed 4-point bend test has been used to evaluate the flexural strength of different component sizes and pin-counts with eutectic Sn/Pb and lead-free Sn/Ag/Cu solder. The devices were electrically monitored for the duration of the tests, along with the load, displacement, and PCB strain. The tests were performed at three different strain rates. Results show that the high strain rate 4-point bend test is capable of replicating the brittle fracture failure modes observed in PCB assembly operations. The PCB strain to failure and the failure mode were found to be strongly dependent on the strain rate. A significant outcome of this work is the determination of the PCB strains which, if exceeded, could pose a substantial risk of brittle fracture in assembly, test, shipping, and handling operations. Key words: BGA, ENIG, brittle, fracture, bend, flex INTRODUCTION Flip-chip BGA is presently the preferred technology for high end device packaging. For high-density organic flip-chip packages, the most widely used substrate pad finish technology is electroless nickel immersion gold (ENIG). ENIG finish offers several advantages such as planarity, solderability, substrate shelf-life, corrosion resistance, a wide process window, thermal resistance over several temperature excursions, and reworkability. Over the last few years, the ENIG surface finish has demonstrated the capability to consistently meet industry quality and reliability requirements. However, with increasing volumes, there have been sporadic reports of solder joint brittle failures in PCB assembly and handling operations [1]. This paper deals with brittle fracture of BGA solder joints with ENIG finish under excessive and uncontrolled PCB flexural loads. The brittle fracture cracks are located at the interface between the Ni-Sn intermetallic and the phosphorus rich electroless Ni layer [2, 3], and the package corner solder joints are generally most susceptible. In the past, electrolytic Ni/Au substrate finish was commonly used for relatively lower density BGA packaging. However, with increasing device densities and the migration to flip-chip packaging with high density

organic built-up substrates, the substrate pad finish transitioned to an electroless nickel immersion gold process. During early stages of the ENIG development processes, the plating technology was plagued with the so-called “black pad” failure mode. The failure mechanism resulting in black pad has been studied in detail [4-6]. Black pad is essentially caused by a corrosion mechanism that can occur during the auto-catalytic immersion gold process. The gold deposition selectively attacks grain boundaries in the phosphorus enriched nickel layer, causing mud-flat cracks in the electroless nickel layer. The existence of black pad can be detected if the failure mode in conventional ball-shear testing is separation at the ENIG interface, instead of failure in the bulk solder [7]. The process variables required to eliminate black pad have subsequently been studied and optimized [7, 8]. The brittle fracture failure mechanism described in this study is unrelated to black pad. However, due to the lack of a reliable test procedure to demonstrate the brittle fracture mechanism, failures encountered in high-strain assembly operations are often incorrectly attributed to black pad based simply on the observation of fracture at the substrate pad intermetallic interface as opposed to failure in bulk solder as seen in conventional ball-shear and accelerated temperature cycling tests. Post reflow, PCB assembly operations often include a variety of steps that involve monotonic bending of the PCBs at relatively high speeds. The sporadic instances of ENIG brittle solder joint failures have occurred soon after surface mount and before the fully assembled boards are deployed in end use. In each case that the authors are aware of, investigation of the process steps involved in the PCB assembly and handling steps have resulted in the discovery of a source of uncontrolled and excessive PCB bending load applied at high strain-rates. Furthermore, in each case, simple design changes to reduce the PCB flexural strain has successfully eliminated the brittle fracture occurrences. Some examples of assembly processes which could result in PCB overloads are:

1. In-circuit testing (ICT), 2. Depaneling or breaking off of end tabs in PCB

assemblies, 3. Insertion or removal of boards in chassis, 4. Attachment or removal of fasteners, press-fit

connectors, and spring loaded heat sinks, 5. Shipping or handling with insufficient mechanical

shock protection. CF-FSB032105-1.0

2

This study evaluates ENIG solder joint integrity using the recently approved IPC/JEDEC 9702 standard [9]. The test method involves 4-point bending of PCBs under high strain rate. 4-point tests are preferred over 3-point bend tests because the 4-point loading provides a uniform bending moment on the board. Geng et al [10] have used a 4-point bend test coupon with the component oriented at 45˚ in order to induce failure at component corners. The solder joint strengths were determined to be significantly lower at high strain rates. However, the actual failure modes were not discussed. Harada et al [11] performed 4-point bend tests at different strain rates and confirmed the strain rate dependency of ENIG interfacial fracture. Their study also showed that packages tested within an hour of surface mount have lower brittle fracture strength compared to packages tested a few weeks after surface mount. The maximum compressive strain on the back side of the PCB (non-component side) was used in their study to report failure strain [11]. Recently, Geng et al [12] used high-strain rate 4-point bend tests to evaluate the effects of Sn-Pb and Sn-Ag-Cu solder, solder resist opening diameter and ball diameter. However, the actual strain values were normalized thus the reported data are not usable [12]. The IPC-9702 standard followed in this study is likely to be universally adopted in future. This will facilitate convergence of data among various sectors of the industry. PROCEDURE The flip-chip components had an integral metal heat-spreader and the substrate had a BT core with built-up layers. The ENIG finished pads had a mask opening diameter of 0.55mm. The components had either 63Sn/37Pb or lead free Sn/Ag/Cu solder balls with 1mm pitch. The peak reflow temperature was 210°C for 63Sn/37Pb and 245°C for Sn/Ag/Cu. Three different flip-chip component sizes were evaluated in this study; 1020-pin (33 x 33mm), 780-pin (29 x 29mm), and 672-pin (27 x 27mm). The packages are henceforth referred to as F1020, F780, and F672 respectively. The corner solder ball locations 1mm x 1mm away from the package corners are unpopulated. The boards used in this study were 93 mil (2.36mm) thick FR4 with 8-layers. Since thicker boards result in higher strain at equivalent deflections, the 93 mil board thickness was preferred to 63 mil (1.6mm) board thickness. The PCB pads were non-solder mask defined (NSMD) with 0.48mm (19mil) pads and 0.55mm (21.6mil) solder resist opening. The PCB pad finish was hot air solder level (HASL). Each board was designed to simultaneously test four identical components (Figure 1). The distance between the loading bars (load span L1) was 115mm and the distance between the support bars (support span L2) was 230mm. Per the recommendations of IPC9702 [9] the components were located 20mm away from the loading bars and 20mm away from the lateral edge of the board. During the tests, the components were electrically monitored to determine the displacement at failure. Based

on the loading configuration, the edges of the package parallel to the loading bars were expected to fail first, with the solder joints near the package corners suspected to be most vulnerable. Each component was monitored with 4 electrical signals. The FPGA devices were configured such that I/O pins on the external edge (nearest to the loading bar) formed one daisy chain and the inner edges farther from the loading bar formed the second chain. Along the external edge, the top and bottom corner pins were monitored separately. In all, 16 signals (4 per component) were acquired from each board. The clock frequency for the I/O daisy chains was 250Hz and the data acquisition frequency was 2kHz.

Figure 1. Test board and fixture dimensions

In addition to acquiring the load, displacement, and device monitoring signals, the strains were measured on selected boards. Single axis strain gages aligned with the board length were placed at 3 locations. The strain gage element size was 0.062” (1.57mm). Strain gage A was placed on the component side of the PCB, aligned with the solder ball closest to the corner and located midway between the loading bar and the component edge (see figure 1). This gage is meant to measure the uniform global PCB strain, thus it has to be located sufficiently far from the component and loading bar. Strain gage B was placed on the back-side of the PCB directly underneath the solder ball near the external edge corner. This gage is intended to capture the maximum strain concentration caused by solder joints near the package corner. Strain gage C was also placed on the back-side of the PCB and coincident with the center of the component. This gage is intended to capture the minimum strain on the board due to the stiffening effect of the component. Since there are some studies that have reported that the ENIG interfacial strength increases due to room temperature aging over time [11], all the boards were tested within 1 week of surface mount. For improved statistical significance, at least two boards were tested under identical test conditions. Prior to starting the test at a fixed displacement rate, a preload of 1.4 kg-f was applied so that the board could be properly aligned in the fixture. To evaluate the effect of strain rate on the failure strains and

A 20mm

20mm

B

115mm

330mm

L1 = 115mm

20mm 20mmC

L2 = 230mm

3

failure modes, the tests were carried out at three different displacements rates of approximately 2, 22, and 48 mm/sec. In order to generate failures in all four components on the board, the monotonic bending was ramped to peak displacements ranging from 20 to 25mm. Note that these displacement values are related to the movement of the loading bars with respect to the support bars. The actual flexural displacement of the board is approximately twice the measured crosshead displacement (see figure 2). Once the peak deflection was attained the boards were immediately unloaded back to the starting crosshead position.

Figure 2. 4-point bend fixture with peak PCB bending

RESULTS PCB Strain Evaluation The measured global PCB strain from gage A is shown in figure 3 for F672, F780, and F1020 components. The result shown in figure 3 and from strain measurements of additional boards (not included in figure 3) showed that the global PCB strain ramps linearly with displacement. The strain values for all three component sizes were in good agreement. If the stiffening effect of the components is ignored, a simple relationship for 4-point bending of an elastic beam has been suggested in IPC-9702 [9] for estimating the global PCB strain and strain rates:

)2)((6

1212 LLLLt

+−= δε (Equation 1)

where, ε is the global strain along the length of the board, δ is the relative displacement between the load bars and the support bars, t is the PCB thickness, L1 is the load span, and L2 is the support span. Notice that thicker boards will result in higher strains and strain rates. Figure 3 shows that the PCB strain calculated from equation 1 is in good agreement with the strain gage measurements. Strain gage readings from one board to another will often differ due to minor differences in gage location, and furthermore the gage signals could be noisy. Therefore the calculated global PCB strain is a preferred method because such inconsistencies are eliminated from the final reporting of strain data.

0

1000

2000

3000

4000

5000

6000

-1.0-0.8-0.6-0.4-0.20.0

Displacement, inch

Mic

ro-S

train

Calculated

F672 (Gage A)

F780 (Gage A)

F1020 (Gage A)

Figure 3. Measured and calculated global PCB strain versus

displacement relationship

Table 1 lists the measured strain readings from gages A, B, and C at a displacement of 12.7mm. As expected, the global strain measured at gage A is tensile, while the strains measured at gages B and C are compressive. The strain values for gages A and C have only minor differences for the three components, thus these strains are independent of the component size. In contrast, the maximum strain measured at gage B tends to increase significantly with component size. In order to evaluate the true nature of strains at the gage B location, in one F1020 board a 0-45-90° rosette strain gage was used instead of a single axis gage. The principal direction at this location was found to vary from -40° to -37°, measured from the length of the board counterclockwise towards the edge. This implies that the maximum strain concentrations measured on the PCB back-side underneath the corner solder joint can be misleading if single axis gages are used.

Table 1. Measured strain values at 0.5 inch (12.7mm) displacement. Rosette strain gage was used for the F1020 board Gage B reading.

Strain (x 10-6) Package (size) A B C F672 (27 x 27mm) 3639 -970 -975 F780 (29 x 29mm) 4016 -3515 -836 F1020 (33 x 33mm) 4002 -6247 -918

Failure Strains For the purpose of reporting the failure strain of various components, the calculated global PCB strain (equation 1) is used in this study. For the F672 components, figure 4 shows the failure strains at three different strain rate ranges. At the slow strain rates, the strain to failure was higher, and in fact on two boards there were two components that did not fail up to the peak displacement (or 6800 microstrain). At strain rates of 6000 microstrain per second and beyond, the strain to failure was lower. The results of the F780 boards are shown in figure 5. In this case the components had either eutectic Sn-Pb or lead free Sn-Ag-Cu solder balls. While there is significant scatter in the strains to failure, the Sn-Ag-Cu solder joints had similar behavior as Sn-Pb solder. This is in contradiction to Geng et al [12] where Sn-Ag-Cu

4

was found to perform worse than Sn-Pb. Figure 6 shows the strain to failure results for the F1020 boards. Once again the strain to failure data seems to be in the same range of scatter as the F672 and F780 components.

0

1000

2000

3000

4000

5000

6000

7000

0 2000 4000 6000 8000 10000 12000 14000

Strain Rate (10-6/sec)

PCB

Prin

cipa

l Stra

in (X

10-6

)

Figure 4. Failure strain versus strain rate for F672

components

0

1000

2000

3000

4000

5000

6000

7000

0 2000 4000 6000 8000 10000 12000 14000

Strain Rate (10-6/sec)

PC

B P

rinci

pal S

train

(X 1

0-6)

SnPb ENIGSnAgCu ENIG

Figure 5. Failure strain versus strain rate for F780

components with Sn-Pb and Sn-Ag-Cu solder

0

1000

2000

3000

4000

5000

6000

7000

0 2000 4000 6000 8000 10000 12000 14000

Strain Rate (10-6/sec)

PCB

Prin

cipa

l Stra

in (X

10-6

)

Figure 6. Failure strain versus strain rate for F1020

components

Failure Analysis Following the bend tests the failed components were not electrically functional, therefore electrical fault isolation

was not possible. This is usually an indicator of widespread damage as opposed to failure at just a few pins. For failure analysis, dye and pry was used to obtain a spatial distribution of solder joint failures and failure modes. This simple procedure involved application of a dye. Sufficient time was allowed for the dye to penetrate cracks and fissures, and an ultrasonic shaker was used to aid this process. The PCB with attached component was baked for a few minutes to dry the dye. The component was then mechanically pryed off the PCB to expose the failed solder joints stained by the dye. The solder joint failures were widespread along several columns near the outer edge (closest to the loading bars). At the inner edge (farthest from the loading bars) there were fewer columns of solder joint failures (see figure 7). For most of the solder balls the failure mode was PCB pad lifting, as represented by a shaded rectangle in figure 7. In the components tested at strain rates greater than 5000 microstrain per second, a few solder joints at the corners failed at the component substrate pad interface, as represented by the circles in figure 7. Figures 8 and 9 show typical images of component corners at the outer and inner edge respectively. The substrate side failures at the ENIG interface can be seen in both figures 8 and 9. Figure 9 also shows that the bulk of failures were at the PCB pad interface. In the case of components tested at slow strain rates of 500 microstrain per second, 100% of the solder joint failures were at the PCB side.

Loading barLoading bar

Figure 7. Schematic representation of solder joint failures

While the dye and pry method provides a complete spatial mapping of the failures, cross-sections and SEM examinations were required to determine the specific interface of failure. Figure 10 shows that the substrate side failures observed in high-strain rate tests were caused by complete brittle fracture between the solder ball and the ENIG pad. Higher magnification images after a delineation etch confirmed that the separation occurred between the Ni-Sn intermetallic and the phosphorus rich Ni layer. This result demonstrated that the high strain rate 4-point bend tests do indeed successfully recreate the brittle fracture failure mechanism at the ENIG interface. Figure 11 shows a cross-section of a typical PCB side solder joint failure. The

No Failure (2 units of 2 boards)

5

FR4 material in the PCB essentially cracked causing the pad to rip out of the PCB material. In these solder joints there was no evidence of damage at the ENIG interface.

Figure 8. Sample image of component corner at inner edge

following dye and pry

Figure 9. Sample image of component corner at outer edge

following dye and pry

Figure 10. Cross-section of brittle fracture at ENIG

interface

Figure 11. Cross-section showing PCB pad lifting

DISCUSSION Analysis of Failure Strains In this study, the monotonic bending was conducted to relatively large peak displacements (20 to 25mm). Such large peak displacements were necessary to generate failure in all four components on the board, as required by IPC-9702 [9]. Since the failure analysis showed that there were two distinct failure modes involved in the high strain rate tests (> 5000 microstrain) and the majority of solder joint failures occurred due to PCB pad lifting, the quantitative assessment of the reported strains at first occurrence of electrical failure needs to be understood further. At present it is not conclusively known if the ENIG interfacial fractures occurred first or the PCB pad lifting was first. Since the slow strain rate tests of ~ 500 microstrain per second had significantly higher strain to failure, and the failure was 100% due to PCB pad lifting, it is possible that the strains required to cause PCB pad lifting are higher even at the higher strain rates where brittle ENIG fractures are observed. But this would implicitly assume that the strain required for PCB pad lifting is rate independent. For future work the first occurring failure mode needs to be conclusively established with termination of the tests at lower peak strains, or perhaps implementing a feedback loop to stop the bend test at the onset of first failure. For further discussion in this study we will adopt a conservative approach by assuming that all the mixed mode failures at high strain rates were due to ENIG brittle fracture occurring prior to PCB pad lifting. While the measured failure strains had significant scatter, the data for all 3 component sizes ranging from 27mm to 33mm were in the same range of strains when the strain rates were greater than 5000 microstrain/sec. This implies that the effect of component size is not significant for brittle fracture. For improved statistical significance, the data for all three component sizes were merged to include all components with Sn-Pb solder balls tested at rates of 5000 microstrain/sec or higher. Figure 12 shows that the data fits well to a log-normal distribution. In addition to the ENIG results, figure 12 also shows results for an electroplated Ni/Au pad finish for comparison. The tests with

6

electroplated Ni/Au substrate finish were conducted on F1020 components using identical test procedures as those for ENIG. Further discussion of the electroplated Ni/Au finish is beyond the scope of this paper.

-3

-2

-1

0

1

2

Pro

bit(1

-Sur

v)

.01

.05.1

.2

.4

.6

.8

.9.95

.99

100 1000 10000Failure Strain (1E-06/sec)

Figure 12. Log-normal distribution of ENIG and Electroplated Ni/Au failure strains

Based on the available strain to failure data, it is evident that PCB strains in excess of 1000 microstrain can cause brittle fracture. The 0.1% failure points from the log-normal extrapolation are 800 microstrain and 1000 microstrain for ENIG and electrolytic Ni/Au respectively. For the board and fixture dimensions used in this study, 800 and 1000 microstrain will be achieved at displacements of 3mm and 3.7mm respectively (see figure 3). Note that in actual assembly operations, depending on the fixture design, these strain magnitudes can be achieved in localized regions at much lower displacements. Future work in this field will need to focus of quantifying the effects of additional variables such as component or PCB pad finish, pad diameter, ball diameter, pitch, time after reflow, and board thickness, etc. Available Test Methods The results presented in this study show that the 4-point bend test, conducted at speeds above 5000 microstrain per second, is in fact capable of creating brittle fracture failures as seen in high strain rate assembly and handling operations involving uncontrolled and excessive strains. Until recently there was no test method available in the industry for recreating these brittle fracture failures under controlled conditions. The traditional ball-shear tests conducted at speeds of less than 1mm/sec displacement have been unsuccessful in recreating these failures because the strain rate is too slow, as a result the predominant failure mode is solder shearing. At high strain rates the bulk solder does not have enough time to undergo plastic deformation, thus brittle interface fracture is the ensuing failure mode. At present some efforts are underway to develop high-speed component level tests to replicate ENIG brittle fractures. Solder ball pull tests at speeds above 5mm/sec is one such alternative.

Another test method that is capable of generating these brittle fracture failures is a board level mechanical shock or drop test (see figure 13). These brittle fractures in shock tests occur at corner solder joints, and the test is capable of attaining strain rates that are 2 or 3 orders of magnitude higher. This is arguably a more realistic test to simulate shipping and handling operations. However, even in drop testing, the failure is generated as a result of PCB flexure and the PCB flexural strain will need to be determined. The dynamics of the board are far more difficult to characterize at such high strain rates. Another issue with the drop test is that it is essentially a potential energy controlled test (controlled by drop height) and the PCB flexural displacement occurring in response must be characterized. In contrast, the bend test is a deflection controlled test where the global PCB strain is known in advance (equation 1). Thus multiple drops are often required to initiate brittle fracture. For reference, the brittle fracture failure shown in figure 13 was generated after 20 drops resulting in peak PCB strain of 900 microstrain with a strain rate of 600,000 microstrain/sec.

Figure 13. Brittle fracture observed in board level

mechanical shock testing

Future Direction The 4-point bend test is likely to be a useful tool for both component and substrate suppliers to determine brittle fracture resistance. The focus of component manufacturers is likely to be on different types of packages, while substrate manufacturers can play a vital role in studying various substrate pad technologies and plating process variables. While brittle fracture strains are determined on test boards, PCB assembly suppliers can perform strain and strain rate evaluations on real boards subjected to assembly and handling stress [13, 14]. With increasing PCB densities and the need for higher throughput, procedures such as in-circuit testing can easily run at strain rates approaching mechanical shock. Further, if the ICT fixture is not carefully designed, peak strains above 1000 microstrain could be seen in localized regions, posing a substantial risk for brittle fracture. Available strain characterization data have shown that the bulk of PCB assembly processes involve peak

ENIG Electroplated

7

strains below 500 microstrain [13, 15]. In a few cases where the peak strains were deemed excessive, simple fixture modifications dramatically reduced the PCB strain [15, 16]. For high end flip-chip packaging with organic substrates, ENIG is still the most robust and widely accepted substrate finish technology. This substrate finish has a stable intermetallic which consistently meets all existing industry standards in reliability tests such as ball-shear, component level and board level temperature cycling. Other substrate finish options have recently become viable for high end flip-chip packaging. These options are currently under evaluation. CONCLUSIONS 1. 4-point bend tests conducted at high-speed (> 5000

microstrain/sec) are capable of replicating brittle fracture failures seen in uncontrolled high strain rate assembly/handling operations involving large PCB deflection.

2. The strain to failure and failure modes are strain rate

dependent. At slow strain rates (~500 microstrain/sec) the strain to failure is relatively high and the failure mode is PCB pad lifting. At high strain rates (5000 to 13000 microstrain/sec) the strain to failure is relatively low. The failure mode is a combination of brittle fracture at the component substrate ENIG interface and PCB pad lifting.

3. Flip-chip packages with component sizes varying from

27mm to 33mm (square) did not show any measurable difference in brittle fracture strain.

4. With the limited data available for Sn-Ag-Cu solder

joints, the brittle fracture strains seem to be in the same range as Sn-Pb solder.

5. PCB assembly operations involving high strain rates

and peak strains in excess of 1000 microstrain pose a substantial risk for brittle fracture failures.

6. The calculated strain based on the elastic beam theory

(equation 1) can be used to determine the global PCB strain. This method will eliminate inconsistencies caused by minor strain gage misalignment, calibration errors and electrical noise. This simple approach is also conservative as it slightly underestimates the PCB strain to failure.

7. Strains measured at the back-side of the PCB and

coincident with the component corner showed that the strain field is biaxial, therefore single axis gages aligned parallel with the length of the board will underestimate the maximum strain.

REFERENCES 1. K. Lei, J. LLinas, and M. Gwaltney, “Flexure induced

failure of BGA solder joints”, Proc. Of SMTAI, Sept. 1999.

2. R. J. Coyle, T. I. Ejim, A. Holliday, P. P.Solan, and J.

K. Dorey, “Reliability Performance and Failure Mode of High I/O Thermally Enhanced Ball Grid Array Packages,” Proc. 23th IEMT, Austin, TX, 323-332, October 19-21, 1998.

3. Deepak Goyal, T. Lane, P. Kinzie, C. Panichas, K. M.

Chong, and O. Villalobos, “Failure Mechanism of Brittle Solder Joint Fracture in the Presence of Electroless Nickel Immersion Gold (ENIG) Interface”, Proc. of Engineering Components Technology Conf. (ECTC), 2002.

4. Nicholas Biunno, “A Root Cause Failure Mechanism

for Solder Joint Integrity of Electroless Nickel/Immersion Gold Surface Finishes,” Proc. IPC Printed Circuits Expo1999, Paper no. S18-5, March 14-18, 1999.

5. Zequn Mei, Pat Johnson, Matt Kaufmann, Ali

Eslambolchi, “Effects of Electroless Ni/Immersion Au Plating Parameters on PBGA Solder Joint Attachment Reliability,” Proc. 49th Electronic Component and Technology Conference, 125-134, June 1-4, 1999.

6. J. A. Roepsch, R. F. Champaign, B. M. Waller, “Black

Pad Defect: Influence of Geometry and Other Factors”, Proc. Of SMTA International, Rosemont, IL, Sept. 2003.

7. K. Yokomine, N. Shimizu, Y. Miyamoto, Y. Iwata, D.

Love and K. Newman, “Development of Electroless Ni/Au Plated Build-Up Flip Chip Package with Highly Reliable Solder Joints”, Proc. of Engineering Components Technology Conf. (ECTC), 2001.

8. K. Johal and J. Brewer, “Are you in Control of your

Electroless Nickel/Immersion Gold Process ?”, Proc. Of IPC Works, No. S03-3, Miami, Fl, Sept. 2000.

9. IPC/JEDEC 9702 Monotonic Bend Characterization of

Board Level Interconnects. 10. P. Geng, P. H. Chen, and Y. Ling, “Effect of strain rate

on solder joint failure under mechanical load”, Proc. Of 52nd Electronic Components and Technology Conference, San Diego, CA, pp. 974-978, May 2002.

11. K. Harada, S. Baba, Q. Wu, H. Matsushima, T.

Matsunaga, Y. Uegai, and M. Kimura, “Analysis of Solder Joint Fracture Under Mechanical Bending Test”, Proc. Of 53rd Electronic Components and Technology Conference, New Orleans, LA, May, 2003.

8

12. P. Geng, A. Mcallister, C. McCormick, M. Modi, and A. Nazario, “0.8mm BGA Solder Joint Reliability Under Flexural Load”, Proc. Of SMTA International, Chicago, IL, Sept. 2004.

13. K. Newman, “Characterization of PWB Flexural

Loading Using Strain Gages”, Presented at Sun Microsystems Strain Gage Test Summit, San Jose, CA, March 2004.

14. J. Goldstein, “Strain Gage Testing: Predicting and

Preventing Brittle Fracture of BGAs”, Surface Mount Technology (SMT), June 2004.

15. P. Bodmer, M. Jackson, and D. Sommer, “Strain Gage

Analysis Applications”, Presented at the Sun Microsystems Strain Gage Summit, San Jose, CA, March 2004.

16. S. White, S. Quale, G. Daniel, M. Altemus, “Board

Flex Analysis in Test Fixturing”, Presented at the Sun Microsystems Strain Gage Summit, San Jose, CA, March 2004.