Research

SLIDER: A Fast and Accurate Defect Simulation Framework

Wing Chiu Tam

Shawn Blanton

Conventionally, integrated circuit (IC) test is used to screen out weak or failing ICs. Emerging approaches use test data from faulty ICs as an essential feedback loop in understanding the failure mechanisms inherent in modern IC fabrication processes. In other words, in addition to its normal sorting function, test has assumed a new learning role. Valuable information can be extracted from the test data associated with a large population of failing ICs that include, for example, design-feature failure rates and defect-occurrence statistics. However, it is difficult to examine the accuracy of these test-learning techniques because of the unavailability of sufficient fail data where the 'answers' are known a priori.

This research describes a fast and accurate defect simulation framework that we call SLIDER (Simulation of Layout-Injected Defects for Electrical Responses). SLIDER generates virtual test data that can be used to validate a large variety of test- and yield-learning techniques. Fig. 1 shows the components within the SLIDER implementation. SLIDER injects defects at the layout-level and extracts a circuit-level netlist of the defect-affected region(s) to form an analog block. The rest of the circuit is kept in a digital block. Mixed-signal simulation is used to simulate the analog block with circuit-level accuracy, and the digital block with logic-level accuracy. This mixed-signal environment ensures accurate representation and simulation of the defect behavior without significantly increasing runtime. The cross-domain signals that connect the analog and digital blocks are transformed appropriately at the interface through the use of A2D/D2A convertors. Using test patterns of any type, the “defective” circuit is then simulated to produce the virtual tester response. This process is repeated as many times as needed to generate a failure population with any desired characteristics. The resulting failure population is useful because the ‘answer’ (i.e., defect type, defect location, and defect behavior) that corresponds to a given virtual tester response is completely known. Fig. 2 shows the result of using example failure populations to validate an in-house diagnosis tool.

Fig 1. Flow diagram implemented in SLIDER.	Fig 2. Validation of diagnosis accuracy using failures created by SLIDER.