August_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 3 8 (continued on page 10) FAILURE ANALYSIS, STATISTICAL RISK ASSESSMENT, AND ADVANCED MODELING IN A STRUCTURED PROBLEM SOLVING APPROACH: CASE STUDY FOR A DELAMINATION DEFECT IN THE AUTOMOTIVE SEMICONDUCTOR INDUSTRY Corinne Bergès, NXP Semiconductors, France corinne.berges@nxp.com EDFAAO (2020) 3:8-15 1537-0755/$19.00 ©ASM International ® INTRODUCTION No matter the industrial context, a failure analysis has to study and present only facts, and has never to manage assumptions. Performing some assumptions about failure root causes goes beyond failure analysis, and is considered risk assessment, when root causes are studied to predict new potential failures that may be expected later in the field and to prevent reoccurrence, by implementing corrective actions. So, a risk assessment may use assumptions and statistical analysis to validate these assumptions. This paper dealswith a risk assessment, starting after a failure analysis and integrated in a real structuredproblem solving approach. In the context of semiconductor industry for automotive, an inter dielectric delamination defect is used as a case study to visualize statistics usage in risk assessment: for this failure, some field returns are observed on different products implemented in different customer applications. The first part of this paper describes the link between failure analysis and risk assessment, in a structured problem solving approach. A statistical key concept is fitting with sampling, and confidence level and intervals linked with sampling. So this concept in risk assessment is studied in the second section of the paper. In the third part, sampling is enlarged to playwith confidence interval width: an implicit idea is to use this sample enlarging to perform better predictions of expected failures. Finally, all the elements around sampling size and confidence interval width, are implemented in a real inter dielectric delamination failure example, for one product, then enlarged to several ones. FAILURE ANALYSIS AND RISK ASSESSMENT IN A STRUCTURED SOLVING APPROACH A failure analysis is always qualitative and factual: no assumption about failure root cause is made. The report contains only a failure description, what the analyst engi- neers do observe during their work, without any hypoth- esis. A failure analysis could fit with the answer ‘What is the defect?’ inside an Is-Is Not template. It could also fit with the description problem chapter in an 8D template. After failure analysis, when it comes to dealing with potential quantity of parts impacted by this failure, and with the prediction of later failures that can be faced, risk assessment is begun, and in order to ensure risk assess- ment result, it becomes important to implement a full structured solving approach, using these tools as Is-Is Not or 8D, started with failure analysis findings. The main four questions that are needed to be an- swered inside an Is-Is Not tool are reminded here: What is the defect? Where was it observed? When? and finally How did it occur? A more detailed template includes also two additional concerns about the failure reason (Why?) and its ownership (Who?). Finally, these six study topics (what, why, when, where, who, how) could address both the observeddefect and the failing product. A risk analysis could correspond to another additional one: how many new failing products may be pending and expected to be observed later? So, if the answer to the question ‘What’ is the failure analysis, ‘When’ and ‘Where’ could be linked to potential