Issue link: https://iconnect007.uberflip.com/i/1509873
44 PCB007 MAGAZINE I OCTOBER 2023 the concentration of contaminants, or some combination of these factors. e difficulty is ensuring that the test reproduces and/or cor- relates to the failure mechanism in service. To use this data for making true reliability predictions—that is, the probability of failures at a given time under given conditions—test- ing must be continued until enough parts fail that a life distribution can be estimated. Unfor- tunately, this process can be time consuming, so qualification tests are oen substituted. Qualification test protocols specify a maxi- mum number of failures that may be observed in a specified period in a sample of specified size. If few or no failures occur, a qualification test provides almost no information about the failure distribution; for example, the probabil- ity of failure during the next time interval is unknown. is limitation of qualification test- ing is minimized when the life distribution for prop- erly manufactured samples is already known or can be estimated based on experience with similar designs. Many reliability or qualifi- cation testing approaches do not follow either of these schemes. Instead, they test the ability of the product to survive a sequence of tests under extremely severe condi- tions for a short time or small number of expo- sures. Again, this type of testing may be ade- quate when it is supported by long experience with both the product type and its use environ- ment; however, it is risky because it is not based on ensuring that probable failure modes will not occur in the life of the product. When new technologies or geometries are introduced, the old tests may not always be conservative. By the same token, irrelevant failure modes that would not occur in service may be introduced by the harsh test conditions. Design for Accelerated Reliability Testing Accelerated reliability test design is summa- rized into seven key steps. 1. Identify the service environment and the acceptable failure rate over a specified service life. 2. Identify actual environment of the PCA (modified service environment). e service environment should be translated into the ambient environment actually experienced by the PCA. For example, the temperature experi- enced by the PCA is influenced by both power dissipation and cooling. e mechanical envi- ronment is influenced by shock-absorbing material, resonances, and so on. 3. Identify probable failure modes (e.g., solder-joint fatigue, conductive anodic fila- ment growth). Accelerated reliability tests are based on the premise that the frequency and/ or severity of the environmental exposure can be increased to accelerate the incidence of the failure that occurs in ser- vice in a known way, that is, that the data can be used to predict the life distribu- tion for the in-service PCA environment. is assump- tion makes sense only if the same failure modes occur in the test as in real life. It cannot be overemphasized that the accelerated tests must be designed around the real failure modes. Probable failure modes may be identified from past service experi- ence, the literature, or preliminary testing or analysis. 4. For each failure mode, construct an accel- eration model. An acceleration model that allows test data to be interpreted in terms of the expected service environment is crucial to life distribution estimation. It is also extremely helpful in designing good tests, so ideally the acceleration model should be developed before the accelerated reliability tests are car- ried out. As an example, for solder-joint reli- ability modeling, Coffin-Manson relation- ship relating cycles-to-failure to strain in sol- For each failure mode, construct an acceleration model.