Circuit Test Point Selection Using GNN Coverage Influence Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing circuit testing methods face challenges in achieving high test coverage while minimizing design overhead, particularly due to uncontrollable and unobservable input/output signals, which degrade testing efficiency and increase production costs.
Innovation Solution
Utilizing a graph neural network (GNN) model to convert circuit design data into graph data, generate test coverage and influence data, and select test points for insertion based on node influence, optimizing test point insertion (TPI) to enhance test coverage without excessive design overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional circuit testing methods are used, then testing can be performed with existing infrastructure, but test coverage is insufficient due to uncontrollable and unobservable signals
Solution Approach 1:
The patent applies preliminary action by converting circuit design data into graph data structures before testing, and by using a GNN model to pre-analyze and select optimal test points. This preprocessing enables the testing system to identify and control previously uncontrollable signals, improving test coverage without requiring complex runtime modifications to the circuit design.
Solution Approach 2:
The patent introduces graph data as an intermediary representation layer between the original circuit design data and the testing process. The GNN model serves as another intermediary that processes this graph data to generate test point selections. This intermediary approach enables sophisticated analysis and control of circuit signals without directly modifying the hardware design, thereby improving test coverage while managing design overhead.
2Measurement precision
If more test points are inserted to improve test coverage, then testing accuracy increases, but design overhead and production costs increase
Solution Approach 1:
The patent applies parameter changes by using the GNN model to analyze the graph representation of the circuit and identify key parameters (test points) that have the maximum impact on test coverage. Instead of uniformly increasing the number of test points, the system changes the selection criteria based on graph-based influence analysis, thereby achieving high testing accuracy with minimal design overhead.
Solution Approach 2:
The patent applies local quality by selecting specific critical test points based on their local influence within the circuit graph, rather than uniformly distributing test points throughout the design. The GNN model identifies regions of the circuit graph that have disproportionate impact on overall test coverage, allowing concentrated testing resources to be applied where they provide maximum measurement precision with minimum overhead.
3Reliability
If comprehensive testing is performed to reduce defects, then product reliability improves, but testing time and production costs increase
Solution Approach 1:
The patent applies preliminary action by performing graph-based analysis and test point selection before actual testing execution. The GNN model pre-processes the circuit design data to identify the most critical test points, allowing the subsequent testing phase to focus only on these high-impact locations. This preliminary analysis reduces the overall testing time while maintaining high product reliability by ensuring that the most critical defects are detected.
Solution Approach 2:
The patent applies partial action by selecting a subset of critical test points rather than performing exhaustive testing at all possible locations. The GNN-based selection identifies the minimum necessary set of test points that provide comprehensive coverage for detecting critical defects, thereby achieving high product reliability without the time cost of complete exhaustive testing.
Data Source
AI summary
A method and apparatus for testing a circuit based on test coverage optimization are provided. The method includes converting design data representing circuit components and circuit nodes of a circuit to be tested into graph data, generating test coverage of the design data and influence data representing influence of the circuit nodes for the test coverage based on the graph data input in a graph neural network (GNN) model, and selecting test points for test point insertion (TPI) from the circuit nodes, based on the influence data.