Multilayer ceramic capacitors (MLCCs) are subjected to high strain rate flexural loading via drop tower and air gun tests, with PWB strain rates ranging from 1/s to 10/s. Three MLCC part sizes, three different manufacturers, and standard and flexible termination parts are included in the study.

Standard termination capacitors failed via the well documented flex crack failure mechanism. However in all cases this crack followed a vertical path not typical of this failure mechanism.

Flexible termination capacitors failed via a newly discovered failure mechanism involving delamination in the end cap metallization between the silver filled epoxy and the nickel-tin plating. For size 1206 parts, this delamination was seen in both end caps, and the part detached from the test board. For size 0603 parts, this delamination occurred in one end cap, while the opposite end cap fractured though the ceramic in a manner similar to a flex crack. Size 0603 parts also failed via the vertical flex cracks documented in standard termination parts.

All of the documented failures of MLCC devices at PWB strain rates of ≥ 1/s occurred at maximum PWB strain values greater than an order of magnitude lower than those seen in lower strain rate testing. This rate dependency of MLCC part failures has vast implications for products intended for high rate environments. Additionally, when the PWB strain rate was increased along with PWB maximum strain, flexible termination capacitors performed worse than their standard termination equivalents. This brings to issue the role of these next generation parts in portable consumer electronic devices as well as other designs with high rate implications.

Ball grid array (BGA) devices are subjected to four point bend tests via a servo-hydraulic testing machine at PWB strain rates ≤ 0.1/s. The resulting BGA data is found to adhere reasonably well to the Coffin-Manson low cycle fatigue relationship. Independently generated BGA data that differs with respect to many testing variables is plotted alongside the experimental data. The high correlation of the data set indicates the possibility of creating a BGA mechanical failure model that is independent of failure site, package type, and test specifications.