Testing, diagnosis and repairing methods for NBTI-induced SRAM faults.

摘要

Negative Bias Temperature Instability (NBTI) is a major Static Random Access Memory (SRAM) aging mechanism, leading to reduced read and hold static noise margins and increased soft error rate. The existing techniques include guardbanding, on-chip sensor-based detection, and recovery. In this thesis, a group of testing, diagnosis, and repairing methods for NBTI-induced memory faults are presented. The resulting observations show that NBTI leads to SRAM read errors rather than write errors. An error detection method is proposed to identify NBTI-induced memory read errors based on the existing Error Correcting Code (ECC) circuitry, differentiate them with soft errors by correction and double checking, and keep them idle for recovery. Furthermore, a predictive test method is proposed for NBTI-induced memory faults by adapting body biasing. An adaptive body biasing formula is achieved to simulate the NBTI effect. The experimental results validate the proposed methods and show that they cost little silicon area and power consumption.