Delay Fault Modeling, Test, and Diagnosis

Texas A&M Engineering Experiment Station, College Station TX

Investigators

Abstract

As integrated circuit speed and density increases, more circuits fail due to delay faults -- manufacturing defects that cause the circuit to operate at a speed slower than intended. Such circuits either cannot be used or must be sold at a lower price. The behavior of delay faults is complex, and traditional manufacturing test approaches are increasingly ineffective in detecting them. This research attacks this problem by developing a novel realistic delay fault model. This model is being used to develop powerful techniques for fault simulation, automatic manufacturing test generation, and diagnosis for next-generation integrated circuits. This work is being done in cooperation with U.S. semiconductor manufacturers. These test and diagnosis techniques are being integrated into a state-of-art software system and will be tested on real manufacturing problems. The new realistic delay fault model considers resistive bridges and opens, the impact of process variation on interconnect, device, and defect parameters, and the influence of interconnect parasitics. Fast layout and parasitic extraction algorithms, and model order reduction techniques are being developed to reduce model cost for a given accuracy level. The model is encapsulated in a parameterized static timing analysis engine for use in fault simulation, test generation and diagnosis. A constraint-based fault coverage analysis is used to determine fault coverage over a set of vectors. Together, these techniques can accurately predict fault coverage and achieve very high delay fault coverage for scan-based CMOS logic circuits.