M-IVC: Applying multiple input vectors to co-optimize aging and leakage

Abstract

With technology scaling, aging effect and leakage pose significant challenge on the reliability of integrated circuits. Existing techniques on co-optimizing circuit aging and leakage either implement in an intrusive way or rely on input vector control (IVC) method. However, intrusive schemes increase the design complexity and induce extra delay and area overheads. IVC method becomes ineffective as the circuit scale increasing. In this paper, a non-intrusive scheme exploiting multiple input vectors (M-IVC) is proposed to co-optimize NBTI-induced degradation and leakage when the circuit steps into standby mode. M-IVC grounds on a new co-optimization model which formulates both the NBTI-induced delay degradation and the average standby leakage as the function of duty cycle. This co-optimization model facilitates to identify a set of optimal duty cycles which result in minimum NBTI-induced degradation and leakage simultaneously. To achieve the optimal duty cycles, an ATPG-like algorithm is proposed to generate multiple input vectors and determine the applying time for each vector. Experimental results demonstrate that with only a small number of vectors, M-IVC can effectively optimize NBTI-induced delay degradation and leakage together and keep the effectiveness as the time of standby period increasing.