Fault modelling and accelerated simulation of integrated circuits manufacturing defects under process variation

摘要

As silicon manufacturing process scales to and beyond the 65-nm node, process variation can no longer be ignored. The impact of process variation on integrated circuit performance and power has received significant research input. Variation-aware test, on the other hand, is a relatively new research area that is currently receiving attention worldwide.Research has shown that test without considering process variation may lead to loss of test quality. Fault modelling and simulation serve as a backbone of manufacturing test. This thesis is concerned with developing efficient fault modelling techniques and simulation methodologies that take into account the effect of process variation on manufacturing defects with particular emphasis on resistive bridges and resistive opens.The first contribution of this thesis addresses the problem of long computation time required to generate logic fault of resistive bridges under process variation by developing a fast and accurate modelling technique to model logic fault behaviour of resistive bridges.The new technique is implemented by employing two efficient voltage calculation algorithms to calculate the logic threshold voltage of driven gates and critical resistance of a fault-site to enable the computation of bridge logic faults without using SPICE. Simulation results show that the technique is fast (on average 53 times faster) and accurate (worst case is 2.64% error) when compared with HSPICE. The second contribution analyses the complexity of delay fault simulation of resistive bridges to reduce the computation time of delay fault when considering process variation. An accelerated delay fault simulation methodology of resistive bridges is developed by employing a three-step strategy to speed up the calculation of transient gate output voltage which is needed to accurately compute delay faults. Simulation results show that the methodology is on average 17.4 times faster, with 5.2% error in accuracy, when compared with HSPICE. The final contribution presents an accelerated simulation methodology of resistive opens to address the problem of long simulation time of delay fault when considering process variation. The methodology is implemented by using two efficient algorithms to accelerate the computation of transient gate output voltage and timing critical resistance of an open fault-site. Simulation results show that the methodology is on average up to 52 times faster than HSPICE, with 4.2% error in accuracy.