Single Decision Function for Cost-Optimized Device Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current testing methods for devices with multiple parametric measurements, such as mixed-signal or RF devices, are inefficient due to the inability to reflect different costs of misclassifications and ignore correlations between measurements, leading to high test costs and time consumption.
Innovation Solution
A system and method that determine a single decision function based on relevant subsets of tests, allowing for cost-effective classification by predicting test results using correlations between measurements, and adaptively selecting tests to minimize misclassification costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all tests are performed on each device under test, then measurement precision and reliability are improved, but test time and cost increase significantly
Solution Approach 1:
The patent extracts and utilizes correlations between measurements to predict test results. By identifying that certain measurements are correlated with others, the system extracts only the essential measurements needed for accurate prediction, eliminating redundant tests while maintaining reliability
Solution Approach 2:
The patent performs preliminary analysis to determine correlations between measurements before actual testing. By pre-establishing which measurements are correlated and can predict others, the system prepares a prediction model in advance that enables accurate pass/fail decisions with fewer tests
2Ease of operation
If traditional classification algorithms are used, then classification is simplified, but different costs of misclassifications cannot be reflected
Solution Approach 1:
The patent changes the parameters of the classification approach by incorporating cost weights into the decision function. Instead of using simple threshold comparisons, the system uses a weighted decision function where each measurement is multiplied by a cost factor reflecting its importance and misclassification cost, thereby preserving simplicity while adding cost sensitivity
3Ease of operation
If individual specification limits are used for pass/fail decisions, then decision making is straightforward, but correlations between measurements are ignored
Solution Approach 1:
The patent merges individual specification limits into a unified cost-driven decision function. By combining multiple measurements with their respective cost weights into a single decision metric, the system maintains straightforward decision-making while incorporating correlation information and relative importance of different measurements
4Loss of time
If sensor data is not used for pass/fail decisions, then test time overhead is minimized, but test efficiency is reduced
Solution Approach 1:
The patent introduces a cost-driven decision function as an intermediary that processes sensor data efficiently. This intermediary layer takes raw measurements and quickly computes the weighted decision metric, enabling fast pass/fail decisions that utilize sensor data without significant time overhead
Data Source
AI summary
An apparatus for determining a single decision function is configured to obtain measurements from a plurality of devices under test corresponding to stimulating signals applied to the plurality of devices under test. The stimulating signals correspond to a set of tests performed on the plurality of devices under test. The apparatus may further determine a subset of tests from the set of tests, such that the subset of tests is relevant for indicating whether the plurality of devices under test pass the set of tests. The apparatus may also determine the single decision function applicable to measurements from an additional device under test tested using the subset of tests, such that the single decision function is adapted to predict a test result for the set of tests on the basis of the subset of tests.