Predicting Reliability of LSI Printed Circuit Carriers

The arrival of Large Scale Integration (LSI) devices forced the development of products which could reliably package these high density chip carriers. Until LSI, printed circuit technology designers were only moderately challenged to make high density products. This change occurred in the early 1970s at IBM with the printed circuit packaging for the IBM 4300 models. This was the first system application of LSI chip carriers. Since then, numerous LSI printed circuit carrier designs have been used in IBM systems. The methods to predict reliability of the LSI printed circuit carriers have not kept up with the hardware technology. This paper will describe an enhanced method of prediction which more closely represents the performance of today's LSI printed circuit cards and mother boards. Computer processed algorithms will be used to estimate the failure rates of a particular printed circuit carrier and the effects of this estimate on the complete system will be discussed. The estimation algorithms are contained in programmes resident on floppy diskettes and run on an IBM Personal Computer. The flexibility and ease of use of the PC allows the user to make any necessary additions or changes to the algorithms quickly. The algorithms are derived from standard Weibull reliability techniques but include internally derived factors which enhance the accuracy of the Weibull predictions. Data from extensive field reporting of any defects and detailed failure analysis are used to enhance the accuracy of the algorithms. Using the enhanced algorithms, one can generate new estimates easily. Quick access and fast turnaround times result in a convenient and cost effective technique to provide these estimates to system users on a timely basis. Design changes to the carrier which could affect the reliability are quickly modelled in the algorithms and new reliability estimates are provided allowing the final user to make the required system trade‐offs easily. These design changes and new reliability projections can be presented to management in a timely manner and allow them to make the cost/performance decisions which result in the optimum marketable product.