Analog/RF test ordering in the early stages of production testing

Abstract

ISBN 978-1-4673-1073-4International audienceOrdering of analog/RF tests is important for the identification of redundant tests. Most methods for test ordering are based on a representative set of defective devices. However, at the beginning of production testing, there is little or no data on defective devices. Obtaining this data through defect and fault simulation is unrealistic for most advanced analog/RF devices. In this work, we will present a method for analog/RF test ordering that uses only data from a small set of functional circuits. A statistical model of the device under test is constructed from this data. This model is next used for sampling a large number of virtual circuits which will also include defective ones. These virtual defective circuits are then used for ordering analog/RF tests using feature selection techniques. Experimental results for an IBM RF front-end have demonstrated the validity of this technique for test grading and compaction

A Classification Approach for an Accurate Analog/RF BIST Evaluation Based on the Process Parameters

“…Along the same lines, some methods are based on performance ordering [19] and start with the test of performances having a high probability to detect faulty circuits. These methods adopt die-level statistical models to approximate the original test set or to predict pass/fail labels from a reduced or alternate low-cost set of measurements.…”

Test data analytics — Exploring spatial and test-item correlations in production test data

“…In [12], [13], adaptive schemes were developed for effective test item ordering, with which re-ordered tests can detect failures earlier, benefiting stop-on-fail test programs. In [14], [15], test data were analyzed to detect outliers.…”