Issue link: https://iconnect007.uberflip.com/i/474043
28 The PCB Magazine • March 2015 stress test (IST), and highly accelerated thermal stress (HATS, an air-to-air method). Reliabil- ity test conditions can and do vary widely by OEMs. We see resistive heating pre-condition- ing cycles at 260°C for six times, followed by resistive heating thermal cycling at 190°C, cur- rent induced thermal cycle of -23°C to +220°C, and air-to-air cycles of -60°C to +160°C [5] . Why such harsh testing conditions? What are we trying to prove? And can it be proved? At this rate, testing conditions are destined to become even more severe with the acceptance criteria shrinking to very tight tolerances. Often in reliability testing, the thresh- old value for determining fail- ure is chosen subjectively [6] . Current industry views on reliability consist of custom- er requirements and results based on a variety of reliabil- ity testing procedures. Howev- er, there is still no consensus within the industry as to what each of the reliability test pro- tocols prove and when each test protocol has to be used [5] . The challenge in reliability testing is correlating testing conditions to field use [6, 7, 11] . At the heart of the printed circuit board we find intercon- nects, plated through holes (PTH), and micro- vias. These features must be robust for today's demanding mission critical applications. The shift to lead-free soldering has had the largest impact on electronic assembly reliability, both printed circuit boards as well as components. Assembly and rework are the most critical op- erations that can consume a significant amount PTH's useful life. And the plated through-hole represents one of the most significant failure modes from a reliability perspective, with many conditions in the design of the product influ- encing the reliability of the plated through hole structure [5] . Reliability testing can also be used in the re- search lab to validate new products and process- es. Using designed experiments and reliability analysis, one can validate that a new product or process is as reliable as, or more reliable than, the current product or process [6] . How severe should reliability testing be to simulate assembly and field use conditions? Before we get into that, a review of basic reliability terms is in order. Let's define several terms in alphabetical order: Anderson-Darling Statistic: Relative goodness-of-fit measure for the selected distri- bution. One can compare the AD sta- tistic for several different distribu- tions with the same number of parameters; smaller AD values indicate that the fit is better. Note: the AD statistic is not restricted to reliability appli- cations. Bathtub Curve: A con- ceptual model for describing reliability-related phenom- ena at the component level over its life cycle. Consists of three stages: infant mortality, design life, and wearout. Beta (β): Determines the shape of the Weibull distribu- tion. Referred to as the "shape" parameter. Censoring: when exact failure times are not known (the test was stopped before a failure occurred). We call this "right" censored data. Usually "0" is used for censoring and "1" is used for actual failures. Competing Failure Modes: when there are two or more failure modes present. It is best practice to analyze multiple failure modes sepa- rately when failure modes behave differently. Confidence Intervals: a range of values, derived from sample data, which is likely to contain the value of an unknown population parameter. A 95% interval means 19 out of 20 samples (95%) from the same population will produce confidence intervals that contain the population parameter. In reliability, we gener- reliability testing can also be used in the research lab to validate new products and processes. using designed experiments and reliability analysis, one can validate that a new product or process is as reliable as, or more reliable than, the current product or process. " " RELIABILITy TESTING AND STATISTICS continues Feature