Page 45 - ChipScale_Sep-Oct_2020-digital
P. 45

Reliability testing and data analysis of lead-free


        solder joints


        By John H. Lau  [Unimicron Technology Corporation]
        T        hirty years ago, the author   the procedure starts with a design   Reliability testing and data analysis




                 asked key people in solder
                                                                                This section discusses various aspects
                 materials, solder mechanics,   of the interconnects of a particular   of reliability testing and data analysis.
                                           semiconductor integrated circuit (IC)
        and solder manufacturing to write a book   package with, e.g., the given chip size, the   Examples of thermal cycling and drop
        chapter in their areas of expertise. The   solder alloys, the molding compound, and   tests of a fan-out lead-free package are
        result was a book called Solder Joint   the corresponding printed circuit board   also given.
        Reliability: Theory and Applications   (PCB) and demonstrates that the design is   Definition of reliability. In this study,
        [1], published in 1991. Today, students,   electrically, thermally, mechanically, and   reliability of an interconnect (e.g., solder
        engineers, and researchers all over the   chemically sound. The next step in the   bump, solder joint, or microbump) of
        world still use the book as their reference.  process is for a certain number of samples   a particular semiconductor package
          In the past 30 years, however,   of the sound or reliable design to be built   in an electronic product is defined as
        soldering technology has changed   and tested under certain conditions for   the probability that the interconnect
        dramatically. In order to comply with the   a certain period of time. The test data   will perform its intended function
        restriction of the use of certain hazardous   (failures) are then analyzed and fitted   for a specified period under a given
        substances (RoHS)  regulations, the   into a life-distribution designation for   operating condition without failure [2-
        most challenging one is from tin-lead   the interconnects. Next, failure analysis   5]. Numerically speaking, reliability
        solders to lead-free solders. Therefore,   should be done on the failed samples to   is the percent of survivors; that is,
        the reliability of lead-free solder joints is   find out the root cause and understand   R(x) = 1 –  F(x), where  R(x) is the
        under scrutiny.                    the reason for their failure. In this study,   reliability (survival) function and F(x)
          Conducting reliability engineering of   the focus is on reliability testing and   is the cumulative distribution function
        lead-free solder joints consists of three   data analysis. Emphasis is placed on   (CDF). Life distribution is a theoretical
        major tasks: design for reliability (DFR),   acceleration models and factors. Some   population model used to describe
        reliability testing and data analysis, and   recommendations in these areas will also   the lifetime of an interconnect, and is
        failure analysis (see Figure 1). Usually,   be proposed.              defined as the CDF, that is, F(x) for the
                                                                              interconnect population. Therefore,
                                                                              the one and only way to determine the
                                                                              interconnect reliability is by reliability
                                                                              testing to determine the F(x).
                                                                                Objective of reliability testing. The
                                                                              objective of reliability tests is to obtain
                                                                              failures (the more, the better) and to
                                                                              best fit the failure data to determine
                                                                              the parameters of the CDF of a chosen
                                                                              probability distribution (e.g., Weibull).
                                                                              The number of items (i.e., sample size)
                                                                              to be tested should be such that the final
                                                                              data are statistically significant. The
                                                                              reliability test time is unknown, but
                                                                              usually takes a while (e.g., a few months
                                                                              for thermal cycling tests). It should be
                                                                              noted and emphasized that as soon as the
                                                                              life distribution F(x) of the interconnects
                                                                              is estimated by reliability testing, the
                                                                              reliability R(x), failure rate, cumulative
                                                                              failure rate, average failure rate, mean
                                                                              time to failure, etc., of the interconnects
                                                                              are readily determined [2-5].
                                                                                Most reliability tests are accelerated
                                                                              tests, with increased intensity of exposure
        Figure 1: Reliability engineering: design for reliability, reliability testing and data analysis, and failure analysis.

                                                                                                             43
                                                         Chip Scale Review   September  •  October  •  2020   [ChipScaleReview.com]  43
   40   41   42   43   44   45   46   47   48   49   50