LBIST Diagnostics for Sparse Error Detection in IC Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Logic Built-In Self-Test (LBIST) diagnostics for integrated circuit devices are ineffective in real-time identification of non-cyclical, non-reproducible, and speed-of-test dependent failures, often missing or misidentifying sparse error events, and require time-consuming repetitive testing.
Innovation Solution
A method that saves stump data and identifiers of scan chains during test cycles, preserving data for analysis if it doesn't match pre-defined signatures, allowing for real-time identification and analysis of failed test cycles, including virtually non-reproducible errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional LBIST diagnostics are used, then testing efficiency is improved, but detection of sparse error events deteriorates
Solution Approach 1:
The patent applies preliminary action by saving stump data and test cycle identifiers during test cycles before final analysis. This allows the system to capture sparse error events as they occur during testing, rather than requiring post-test analysis. The stump data is preserved in memory with associated metadata (test cycle identifier, scan chain identifier) so that when an error is detected, the exact timing and context are already recorded, enabling real-time identification of non-reproducible failures without compromising testing efficiency.
2Measurement precision
If repetitive testing is performed to capture sparse failures, then detection capability is improved, but testing time increases
Solution Approach 1:
The patent implements feedback by continuously monitoring test results and comparing actual stump data against expected signatures during the testing process. When a mismatch is detected, the system immediately identifies the failure and preserves the associated stump data and test cycle information. This real-time feedback mechanism eliminates the need for repetitive testing to capture sparse failures, as the system can detect and record errors during the first occurrence, significantly reducing testing time while maintaining high detection capability.
3Measurement precision
If LBIST is configured to stop on locking cycles for analysis, then diagnostic accuracy is improved, but testing speed deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-saving all stump data and test cycle identifiers during the entire LBIST execution without interrupting the test flow. This allows the complete test to run at full speed without stopping for analysis. After testing completes, the saved data can be analyzed offline to identify failures, including sparse errors. This approach maintains testing speed while preserving diagnostic accuracy, as all necessary data is captured during testing without requiring mid-test interruptions.
Data Source
AI summary
A method, system and computer program product for performing real-time LBIST diagnostics of IC devices. During LBIST, stump data and identifiers of test cycles are saved in the IC device-under-test (DUT). If compressed stump data does not match a pre-defined coded value (i.e., “signature” of the test cycle), the saved stump data and an identifier of the failed test cycle are preserved, otherwise the determination is made the DUT passed the test cycle. Identifiers and stump of the failed test cycles are used to analyze errors, including virtually non-reproducible errors.