Variation Aware Soft Error Modeling And Analysis Method For Integrated Circuits

As the semiconductor industry comes into sub-micron field, the supplyvoltage and the size of transistor continues to scaling, rendering circuits moresusceptible to electrical noise and particle strike and reliability issue becomes avital concern. This thesis focuses on soft error of integrated circuits and alsotakes the influence of parameter variation such as NBTI effect and processvariation into consideration to achieve more accurate modeling of soft error.Through the estimation of soft error rate, more responsible method to helpstrengthen the robustness of integrated circuits can be developed.According to the sources of soft error, it can be classified as electrical-noise-induced soft error and particle-strike-induced soft error. For the formerone, probability voltage transfer characteristic is employed which can acquireaccurate fault rate by a discrete probability density function approximationmethod. Compared to Monte Carlo analysis, the methodology maintains highaccuracy while speed up the whole process significantly. Based on thismethodology, the change of NBTI-aware transient error rate estimation is alsostudied.Soft error induced by particle strike is also thoroughly explored bymodeling the process of its formation, propagation and latching. The threemasking effects (logical masking, electrical masking and latching-windowmasking) of logic circuits are carefully considered. The flow of prevailing softerror rate estimation tools is also interpreted. However, the variation ofelectrical parameter may render the estimation unaccurate. Finally, SupportVector Regression is implemented to estimate the parameter of single errortransient pulse considering process variation and NBTI effect, which obtainshigh accuracy and efficiency.