Nik Sumikawa

Screening customer returns with multivariate test analysis

This work studies the potential of capturing customer returns with models constructed based on multivariate analysis of parametric wafer sort test measurements. In such an analysis, subsets of tests are selected to build models for making pass/fail decisions. Two approaches are considered. A preemptive approach selects correlated tests to construct multivariate test models to screen out outliers. This approach does not rely on known customer returns. In contrast, a reactive approach selects tests relevant to a given customer return and builds an outlier model specific to the return. This model is applied to capture future parts similar to the return. The study is based on test data collected over roughly 16 months of production for a high-quality SoC sold to the automotive market. The data consists of 62 customer returns belonging to 52 lots. The study shows that each approach can capture returns not captured by the other. With both approaches, the study shows that multivariate test analysis can have a significant impact on reducing customer return rates especially during the later period of the production.

An experiment of burn-in time reduction based on parametric test analysis

Burn-in is a common test approach to screen out unreliable parts. The cost of burn-in can be significant due to long burn-in periods and expensive equipment. This work studies the potential of using parametric test data to reduce the time of burn-in. The experiment focuses on developing parametric test models based on test data collected after 10 hours of burn-in to predict parts likely-to-fail after 24 and 48 hours of burn-in. Our study shows that 24-hour and 48-hour burn-in failures behave abnormally in multivariate parametric test spaces after 10 hours of burn-in. Hence, it is possible to develop multivariate test models to identify these likely-to-fail parts early in a burn-in cycle. This study is carried out on 8 lots of test data from a burn-in experiment based on a 3-axis accelerometer design. The study shows that after 10 hours of burn-in, it is possible to identify a large portion of all parts that do not require longer burn-in time, potentially providing significant cost saving.

Important test selection for screening potential customer returns

In a market where quality requirements are extremely high; the ultimate goal is to improve test quality and reduce the occurrence of test escapes. A customer return is a test escape which passes all tests but fails in the field. This paper analyzes seven lots of parametric wafer probe test data, where each lot contains one customer return. We ask a fundamental question: What subset of tests provides the best screening of customer returns? This leads us to the problem of selecting sets of important tests which contain necessary information to identify each customer return. We compare and combine three test selection methods and suggest an outlier analysis based test strategy for screening potential customer returns.

Multidimensional parametric test set optimization of wafer probe data for predicting in field failures and setting tighter test limits

This work proposes a wafer probe parametric test set optimization method for predicting dies which are likely to fail in the field based on known in-field or final test fails. Large volumes of wafer probe data across 5 lots and hundreds of parametric measurements are optimized to find test sets that help predict actually observed test escapes and final test failures. Simple rules are generated to explain how test limits can be tightened in wafer probe to prevent test escapes and final test fails with minimal overkill. The proposed method is evaluated on wafer probe data from a current automotive IC with near zero DPPM requirements resulting in improved test quality and reduced test cost.

Novel test detection to improve simulation efficiency: a commercial experiment

Novel test detection is an approach to improve simulation efficiency by selecting novel tests before their application [1]. Techniques have been proposed to apply the approach in the context of processor verification [2]. This work reports our experience in applying the approach to verifying a commercial processor. Our objectives are threefold: to implement the approach in a practical setting, to assess its effectiveness and to understand its challenges in practical application. The experiments are conducted based on a simulation environment for verifying a commercial dual-thread low-power processor core. By focusing on the complex fixed-point unit, the results show up to 96% saving in simulation time. The main limitation of the implementation is discussed based on the load-store unit with initial promising results to show how to overcome the limitation.

A pattern mining framework for inter-wafer abnormality analysis

This work presents three pattern mining methodologies for inter-wafer abnormality analysis. Given a large population of wafers, the first methodology identifies wafers with abnormal patterns based on a test or a group of tests. Given a wafer of interest, the second methodology searches for a test perspective that reveals the abnormality of the wafer. Given a particular pattern of interest, the third methodology implements a monitor to detect wafers containing similar patterns. This paper discusses key elements for implementing each of the methodologies and demonstrates their usefulness based on experiments applied to a high-quality SoC product line.

Understanding customer returns from a test perspective

Customer returns are defective parts that pass all functional and parametric tests, but fail in the field. To prevent customer returns, this paper analyzes wafer probe test data and tries to understand what it takes to screen them out during testing. Because these parts pass all tests, analyzing their signatures based on the original test perspective does not make sense. In this work, we search for a novel test perspective where the test signatures from parametric measurements can be used to separate the returned parts from the rest of population. Our study shows that in order to effectively screen customer returns during wafer test, a multivariate screening methodology is desired. This study is based on analyzing over 1000 parametric wafer probe tests and dies from seven lots, each lot containing one returned part. We demonstrate that analyzing customer returns from a multivariate test perspective leads to robust and conservative results.

Forward prediction based on wafer sort data — A case study

This paper studies the potential of using wafer probe tests to predict the outcome of future tests. The study is carried out using test data based on an SoC design for the automotive market. Given a set of known failing parts, there are two possible approaches to learn. First a single binary classification model can be learned to model all failing parts. We show that this approach can be effective if the failing parts are compatible in learning. Second, an individual outlier model can be learned for each failing part. We show that this approach is suitable for learning failing parts such as customer returns, where each may have a unique failing behavior. We also show that with Principal Component Analysis (PCA), a learning model can be visualized in two or three dimensional PC space, which facilitates an engineer to manually select or adjust the model.

Yield optimization using advanced statistical correlation methods

This work presents a novel yield optimization methodology based on establishing a strong correlation between a group of fails and an adjustable process parameter. The core of the methodology comprises three advanced statistical correlation methods. The first method performs multivariate correlation analysis to uncover linear correlation relationships between groups of fails and measurements of a process parameter. The second method partitions a dataset into multiple subsets and tries to maximize the average of the correlations each calculated based on one subset. The third method performs statistical independence test to evaluate the risk of adjusting a process parameter. The methodology was applied to an automotive product line to improve yield. Five process parameter changes were discovered which led to significant improvement of the yield and consequently significant reduction of the yield fluctuation.

Multivariate outlier modeling for capturing customer returns — How simple it can be

Univariate outlier analysis has become a popular approach for improving quality. When a customer return occurs, multivariate outlier analysis extends the univariate analysis to develop a test model for preventing similar returns from happening. In this context, this work investigates the following question: How simple multivariate outlier modeling can be? The interest for answering this question are twofold: (1) to facilitate implementation of a test model in test application and (2) to ensure robustness of the methodology. In this work, we explain that based on a Gaussian assumption, a simpler covariance-based outlier analysis approach can be sufficient over a more complex density-based approach such as one-class SVM. We show that correlation among tests can be a good metric to rank potential outlier models. Based on these observations a simple outlier analysis methodology is developed and applied to effectively analyze customer returns from two automotive product lines.