Investigation into voltage and process variation-aware manufacturing test

Increasing integration and complexity in IC design provides challenges for manufacturing testing. This thesis studies how process and supply voltage variation influence defect behaviour to determine the impact on manufacturing test cost and quality. The focus is on logic testing of static CMOS designs with respect to two important defect types in deep submicron CMOS: resistive bridges and full opens. The first part of the thesis addresses testing for resistive bridge defects in designs with multiple supply voltage settings. To enable analysis, a fault simulator is developed using a supply voltage-aware model for bridge defect behaviour. The analysis shows that for high defect coverage it is necessary to perform test for more than one supply voltage setting, due to supply voltage-dependent behaviour. A low-cost and effective test method is presented consisting of multi-voltage test generation that achieves high defect coverage and test set size reduction without compromise to defect coverage. Experiments on synthesised benchmarks with realistic bridge locations validate the proposed method.The second part focuses on the behaviour of full open defects under supply voltage variation. The aim is to determine the appropriate value of supply voltage to use when testing. Two models are considered for the behaviour of full open defects with and without gate tunnelling leakage influence. Analysis of the supply voltage-dependent behaviour of full open defects is performed to determine if it is required to test using more than one supply voltage to detect all full open defects. Experiments on synthesised benchmarks using an extended version of the fault simulator tool mentioned above, measure the quantitative impact of supply voltage variation on defect coverage.The final part studies the impact of process variation on the behaviour of bridge defects. Detailed analysis using synthesised ISCAS benchmarks and realistic bridge model shows that process variation leads to additional faults. If process variation is not considered in test generation, the test will fail to detect some of these faults, which leads to test escapes. A novel metric to quantify the impact of process variation on test quality is employed in the development of a new test generation tool, which achieves high bridge defect coverage. The method achieves a user-specified test quality with test sets which are smaller than test sets generated without consideration of process variation