SUMMARY OF THE PROJECTS
Created: 5/21/95 Updated: 4/18/97

Failure Models for PWBs and CCAs

S. Ling, G. Ganguly, T. Rothman, L. Guan, A. Dasgupta, A. Christou, M. Osterman
Point of Contact: dasgupta@calce.umd.edu

Objective

Background

Several different failure mechanisms are usually active in a printed circuit assembly. Studies in previous years have determined which ones most frequently in CCAs, and the CALCE research and software roadmaps have focussed attention on these dominant mechanisms. A core task, initiated in 1992- 93 and funded for continuation in 1993-94, investigates the most frequent failure mechanisms through fundamental physics-of- failure models, and develops life prediction models for wearout failure mechanisms. The influence of these failure mechanisms on electrical signal distortion in PWB interconnects was modeled through lossy coupled transmission line theory, predicting the PWB nominal performance as a function of time.

Simultaneously, a comprehensive program has been initiated with core and extended-core support from BCAG, the U.S. Army, and the University of Maryland,to establish the Stress Margin Approach (SMA) for validating design models and to improve product ruggedness through fundamental re- engineering. The equipment available under this extended funding includes the following failure stimulation equipment:

Associated failure diagnosis equipment includes:

Approach

This is a four-unit task: 1 unit to develop new predictive failure models and to augment existing failure models for material failure mechanisms, 1 unit to model the resulting electrical failure modes, 1 unit to incorporate the results in appropriate software for technology transfer to CALCE members, and 1 unit to validate the results through highly accelerated stress testing.

Task 1: Development of Material Failure Models
Solder Fatigue Models - The goal is to provide in the CALCE software a rational alternative overcoming the well-known limitations of Engelmaier's (IPC-SM-785) model for solder joint fatigue. This is a two- step process: developing a stress-analysis method and an alternate fatigue damage model. CALCE has already developed an energy-partitioning damage model to address the second part. New activities are planned for developing an alternate stress analysis tool to address the first part. The new model involves mechanics-based analysis to handle any solder material, any solder joint geometry, any combination of materials, and arbitrary combinations of thermal loads. In the future, we intend to extend the model to include vibration loads also. When combined with the energy-partitioning damage model, this stress analysis model will provide the capability for solder joint fatigue life prediction.

Other Material Models - Fully elastic-plastic models for PTH fatigue have also been developed and have been augmented during previous years to increase their applicability to Aramid reinforced boards and to include crucial manufacturing variabilities, such as plating waviness. During this year, the accuracy of the models will be expanded to include in-plane CTE mismatches, an important factor with Aramid boards and other novel materials such as liquid crystal polymers which have a low in-plane CTE.

Task 2: Development of Electrical Failure Models
The material degradation models developed in Task 1 will be combined with circuit analysis tools to predict changes in electrical performance; reliability can then be predicted in terms of electrical failure modes. Predictive models for resistance or impedance changes due to corrosion in metal traces have been developed in previous years. These models will be used in conjunction with lossy coupled transmission-line models in SPICE to predict degradation of electrical performance over time; thus, package reliability will be predicted based on loss of electrical performance. Similar models for electrical opens due to fatigue crack propagation in solder joints and PTHs are desirable projects for future years. Empirical models for electrical shorts due to metal migration are also available from other related projects.

Task 3: CCA Level Software
The failure models developed in tasks 1 and 2 will be incorporated into a stand-alone software package, to be ultimately integrated into CALCE software. The main difference between this software and the existing CALCE software is the inclusion of the electrical performance models and the associated inputs/outputs (I/O). The design inputs and I/O formats are obtained from board layout data contained in the CALCE board manager toolkit, and from other databases, such as material properties and geometry (component footprint, lead style, etc.). New information on trace layout and electrical functionality data will be input manually at this time, and will be obtained in future years by integration with electrical CAD software. The formats for the inputs and outputs for these files have been determined in meetings with DoD. This software will be enhanced in future years by the addition of new failure models and better user interfaces. Eventually, this software will be integrated into the CALCE software.

Task 4: Experimental Validation
The extended core funds from BCAG (Task C93-06) and internal funds utilized last year to setup experimental facilities for highly accelerated stress application of test articles have been augmented this year for continued development of the stress margin approach. Error-seeded surface-mount PWB samples that were tested last year will be analyzed through simulation tools. New samples of leadless SMT PWBs and hybrid microcircuit devices will be tested this year to enhance our understanding of SMA principles and accelerated stress testing philosophy, and to further validate CALCE models. CALCE is also participating in the JPL solder fatigue round-robin program to assess the accuracy of CALCE software.

Work Accomplished

Task 1 - Progress has been made in developing a generic stress analysis tool for solder joints under thermomechanical cycling, using mechanics of materials principles. The stress analysis model for elastic stresses is complete, using an enhanced Rayleigh Ritz technique that uses dynamic sub-domain allocation. This model will be extended in future years to perform plastic and creep stress analysis required to conduct reliability analysis. This model is computationally much simpler than detailed finite element models, but provides the detailed energy history required for implementing the energy- partitioning fatigue damage model. When combined with the energy-partitioning method, this tool will predict cycles to failure for a given loading. This model can also be used to derive acceleration transforms for accelerated life testing. Details of the stress analysis model are presented in Section 1 of the enclosed report, and detailed implementation of the energy-partitioning method is illustrated in two examples provided in Sections 2a and 2b of the report.

Algorithms have been developed to include the in-plane thermal expansion mismatch as an additional stress driver. In previous CALCE models, only the out-of-plane thermal expansion mismatch was included. In future years, this algorithm will be incorporated into the CALCE software. Details of this algorithm are presented in Section 3 of the report.

Detailed simulation studies have been completed to examine the effects of manufacturing variabilities, such as plating waviness, on the reliability of PTHs/vias. The effect of the solder reflow/rework cycles have been considered in conjunction with the thermal cycles applied during accelerated testing. Summary of this study is presented in Section 4 of the report; further details are available on request.

Task 2 - Modeling of the electrical failure modes resulting from material failure mechanisms requires the modeling of system electrical parameters and electrical performance. Models have been developed for circuit traces, based on microstrip models using lossy-coupled transmission line theory. These models use package architecture information and package material properties to predict, first, the electrical parameters, then signal integrity characteristics and, finally, performance figures of merit. In previous years, these performance models were based on simple closed-form models available in the literature. This year we have increased the accuracy of these models by developing an interface to SPICE. Loss of performance due to corrosion has been modeled and a summary is presented in Section 6 of this report.

Task 3 - A comprehensive software environment has been created that uses inputs from the CALCE board manager about component positions and creates equivalent circuit elements to compute electrical parameters, signal integrity characteristics, and performance figures of merit. Models for loss of reliability due to corrosion have already been incorporated. In future years, capabilities will be enhanced by interfacing with electrical CAD environments. Section 6 of this report presents an overview.

Task 4 - The four experimental programs undertaken this year include the following:

  1. Compliant leaded surface-mount PWB, under vibrational loads.
  2. Surface mount PWBs with tailored CTEs and leadless components, under thermal cycling.
  3. Hybrid DIP components, under combined temperature and random 6-DOF vibration cycling.
  4. Hybrid couplers, under combined temperature and random 6 DOF vibration cycling.

Testing on all items other than item 1 have been completed. Item 1 has not been completed due to scheduling constraints. A detailed report has been written to describe the important generic lessons learned about accelerated testing from item 2. Section 5 of the report contains this information. The main lessons learned from items 3 and 4 are that testing at the individual component level is often futile, since many components today are extremely rugged. For example, no failure could be precipitated in the hybrid DIP components, even after more than 200 cycles of step-stress testing, with loads as high as 70 G rms vibration, and temperature cycling from -55oC to 135oC. The accelerated testing programs will be restricted in future years to the PWB/CCA level, since this is the optimal level at which large enough stresses can be generated to precipitate common failure mechanisms. Accompanying physics of failure simulations can also be performed without too much computational complexity.