Strain Gage Testing: Predicting and Preventing Brittle Fracture of BGAs

Estimated reading time: 9 minutes

Brittle fracture of components on printed circuit boards (PCB) during assembly, test or shipping operations has been a problem in the industry for some time. While such failures are rare, they are an important concern, because the problem often is not discovered until a product enters the production phase, at which point a redesign is extremely costly.

By Julia Goldstein

The trend toward more and larger BGA packages on PCBs has highlighted the problem, since the large, stiff BGAs cannot bend with the boards during processing. A typical failure mode is brittle fracture of a large BGA component with an electroless nickel/immersion gold (ENIG) finish between the Ni-P layer and the intermetallic layer on the BGA pad (Figure 1). Components with other types of surface finishes are not immune to the problem, but the ENIG finish is particularly susceptible. The larger the BGA, the higher the stress and strain on the solder joints during processing, and the more likely that deflection of the PCB will cause the component to separate from the board.

Figure 1. Common location for brittle fracture of a BGA with ENIG surface finish

It is nearly impossible to know the strain in the solder joints during different processing steps. The vast library of creep and fatigue data on solder joints is not effective in predicting their behavior when subjected to high strain rates, as can occur during manual handling, connector installation, in-circuit and board functional test, heat sink assembly, and installation of a PCB into a system. Ball pull and shear tests commonly performed on BGA packages can help determine the integrity of the ball attach process, but they are typically performed at relatively low strain rates, where the solder is ductile. When the strain rate increases beyond a certain point, solder undergoes a ductile-to-brittle transition and change in fracture mode. To predict the behavior of BGAs during high strain rate operations, testing must be performed under actual operating conditions.

Strain gage testing (SGT) can be used to determine the strain on the PCB near the BGA component. While the strain within the solder joint cannot be directly measured, results do correlate with data on brittle fractures. Strain gages are attached to a PCB at strategic locations, typically as close as possible to all corners of a BGA either on the top or underside of the board, and strain and strain rate are measured during assembly operations to pinpoint areas of the board that may be causing problems during each operation. The strain is generated either by performing an actual operation, such as in-circuit test (ICT), with strain gages attached, or by conducting a controlled four-point bend test that simulates strain during manufacturing.

For testing on PCBs, strain gages need to be small (less than 1.5 mm long), both for accuracy in measuring strain in the range of 10 to 1,000 microstrain (µe) and for ease of placement onto boards. Uniaxial gages are recommended for use during four-point bend testing and other operations where the principal (maximum) strain direction is known. Triaxial, stacked rosettes, consisting of three strain gages oriented 45° from one another, are preferred for processes in which the principal strain direction is not known. Planar rosettes are not recommended, since the calculated principal strain from a rosette gage assumes a uniform strain underneath all gages, and actual strain profiles for PCB assemblies are highly non-uniform.

Measurements of principal strain and strain rate are taken during SGT and help determine whether a particular process step is exerting excessive strain on a BGA. If the strain as a function of strain rate exceeds established guidelines, changes in design or process are made to bring strain down to acceptable levels.

Toward Standardization

Since SGT is relatively new, one problem is the lack of industry-wide standards for its implementation. Sun Microsystems presented a Strain Gage Test Summit in March with the goals of furthering collective knowledge of SGT to help predict and prevent brittle fractures on PCBs, as well as making a first step toward building an industry forum to standardize SGT practices and procedures. At the Summit, 32 companies were present, including OEMs, device and component designers, EMS providers, and strain gage equipment suppliers.

One first step toward standardization is a SGT Procedure Sun developed for use by all its suppliers. Sun's procedure requires SGT at all process steps after SMT reflow, including manual handling, rework, connector installation, board testing and mechanical assembly. Boards to be tested do not have to be electrically functional but "must mechanically represent the latest design." Any BGA with a body size measuring 27 × 27 mm or larger must be tested, regardless of surface finish, using specific strain gage attachment procedures. The document also describes test equipment parameters, methods for simulating manual handling processes and required data analysis. Sun's maximum allowable strain guidelines are given as a "line in the sand" on a plot of principal strain vs. strain rate and range from 500 to 1,100 µe.

IPC/JEDEC-9702, Monotonic Bend Characterization of Board-level Interconnects, was developed by representatives from Sun, Cisco, Intel, Celestica and Solectron and is designed to characterize component fracture strength using a four-point bend test. The test method describes how to use a universal tensile tester to monitor strain and strain rate on a test board. Uniaxial strain gages are attached at the corners of mounted components. The standard, currently under review by IPC and JEDEC, is limited to component evaluation on test boards and does not apply to SGT of fully populated PCBs during production processes.

This article describes several examples of SGT performed in the industry. When these results were presented at the Strain Gage Summit, questions on details of testing conditions and procedures were numerous, emphasizing the need for industry-wide standards that can be used for boards and components with various surface finishes and components other than BGAs.

ICT both detects component fractures and generates high levels of strain on a PCB, and therefore is a critical step in SGT. Large BGAs exacerbate the problem by requiring many test probes under an inflexible component, therefore concentrating board deflection in a relatively small area. The ability to counterbalance the force is limited because push fingers cannot be placed on top of a BGA. Old fixture designs do not work for boards with large BGAs. Using finite element analysis (FEA) in conjunction with SGT to design test fixtures is one approach. FEA results can help change probe spring forces and pushdown locations during the design phase, thereby avoiding redesign costs after a fixture is manufactured (Figure 2). It also is important for test fixtures to be designed with SGT in mind, so that they do not interfere with placement of strain gages.

Figure 2. FEA plots (a) showing microstrain (µe) on PCB exceeding acceptable values, and (b) showing strain within acceptable levels after modification. The strain scale in the two pictures is not the same.

One company* detected opens on a 1,012-pin BGA component during ICT and first assumed the failures were related to improper handling procedures. When pro-cess flow was changed to minimize handling and fractures were still occurring, the company tried SGT. SGT during ICT revealed that the test process was to blame. Reducing spring force on the test fixture probes and adding counterbalances on the opposite side of the board reduced the strain from 1,800 to 300 µe and eliminated the problem. The counterbalance blocks had to be custom-designed based on CAD data, with sections milled out over smaller components.

The test development process at the company has evolved regarding fixture specifications, leading it to shift responsibility for SGT back to their fixture vendors. The company analyzes the pre-ship data before fixture shipment is approved and also performs SGT as part of the test fixture acceptance process.

Contract manufacturers can use SGT to pinpoint processing steps causing high strains and measure the effect of corrective actions. Several have done so with guidance from Sun. One company** has had success with SGT in several instances. After observing high strain levels during heat sink attach, for example, the company made modifications to both the heat sink and the PCB. In another situation, a fixture was added to the PCB to lower stress during manual handling, allowing the board to pass SGT during a manual drop test, and support pins were added to make the PCB more rugged. SGT data were used to verify that the changes were effective in lowering the strain to acceptable levels.

A major component maker*** began a program of SGT in September 2003 and aims to have an internal standard in place later this year, as well as design guidelines for new products based on SGT results. The company has demonstrated solutions for desktop, laptop and server systems. In one case, moving a large BGA component by a few millimeters and adding a slit hole near it on the PCB resulted in a dramatic de-crease in strain during connector insertion. This example emphasizes the idea that board designers need to be made aware of SGT so they can be part of the solution.

Challenges Moving Forward

Strain gage testing can be valuable, but certain challenges appear when implementing it throughout the industry. SGT currently is a cost and resource burden for component manufacturers and PCB assemblers. Part of the difficulty is that complete equipment solutions are not readily available. Contract manufacturers often have to come up with their own custom software to analyze data, for example. The in-dustry may eventually get to the point where SGT is standard procedure during board assembly and therefore taken into account during design of both PCBs and test fixtures, but that will require a shift in philosophy. Many designers currently are not aware of strain issues.

SGT data are fairly limited to date, especially for finishes other than ENIG. While there may be a general consensus that finishes such as immersion Ag will have fewer problems, no data exist to back up that hypothesis. The effect of transition to Pb-free solder on brittle fractures also is unknown at this point.

Conclusion

It is important to know what suppliers mean if they say they apply strain gage testing to boards. The lack of standards that encompass a wide variety of component and board configurations is a large obstacle to industry-wide adoption of SGT. One of the outcomes of the Strain Gage Test Summit was a commitment by 10 participating companies to join a working group to instigate an IPC/JEDEC standard on SGT.

* Benchmark Electronics** MiTAC*** Samsung

Acknowledgements

The author would like to thank Brett Ong and Keith Newman of Sun Microsystems, Scott White of Circuit Check, and Milton Jackson of Benchmark for their assistance in writing this article, as well as all those who gave presentations at the Strain Gage Test Summit (Sun Microsystems, Newark, CA, March 8 through 9, 2004).

Julia Goldstein, Ph.D., technical editor, may be contacted at Advanced Packaging and SMT, (408) 376-3987; E-mail: julia.goldstein@sbcglobal.net.