An Evolutionary Negative-Correlation Framework for Robust Analog-Circuit Design Under Uncertain Faults

Abstract

Uncertain and unpredictable faults can occur in electronic systems that work in harsh environments. Analog circuits play an important part in modern electronic systems. Accordingly, it is important to investigate the reliability of analog circuits under uncertain faults. For robust analog circuits designed to be tolerant of certain faults, unexpected faults might issue great influence. For the enhancement of analog circuits' robustness generalization on unexpected and uncertain faults, this paper presents a new idea to evolve analog redundancies that are negatively correlated. An evolutionary negative-correlation framework is proposed to evolve negatively correlated redundancies. An experimental study is provided to observe the reason why the proposed framework is able to evolve negatively correlated results. In Section IV, we implement various fault simulations to statistically test and evaluate the fault-tolerant performance of the proposed negatively correlated redundancies under uncertain faults. Experimental results show that negatively correlated redundancies significantly improve the analog circuit's fault-tolerant performance under uncertain-fault environments in two aspects: 1) negatively correlated redundancies have lower performance degradation in the statistical evaluation after various fault simulations, and 2) negatively correlated redundancies have better stability in the face of unpredictable faults. A significant improvement is that negatively correlated redundancies have more outstanding robustness on various uncertain faults than analog circuits that are designed for certain faults. By this token, the proposed idea can be a promising way of robust analog circuit design for uncertain-fault environments.