Resistive random access memory variability and its mitigation schemes

Abstract

The need for design of new computing and storage paradigms has leaded to the emergence of new technologies and procedures. Among these technologies, emerging non-volatile memories such as RRAMs are getting intense attention due to their attractive characteristics such as scalability and CMOS friendly manufacturing. However, similar to any other new technology emergences, having reliability and high performance devices is a challenge, and innovative new techniques are required to make the products attractive and robust enough before entering into the semiconductor market. The research for such crucial reliability concerns and mitigation techniques are ongoing hot topic and the main motivation for this work. Therefore, in this paper, we have studied the origins of RRAM variability and reviewed some of the existing techniques to mitigate its effect at circuit level. To show the relevance of variability in RRAM memories we have further analyzed its impact in the Read/Write memory operation and have presented the memory unreliability that we measure by a parameter as probability of error can be 25% during the read operation and in presence of such resistance variations. In the next phase we have presented a conventional 1T1R memory architecture where we have proposed our reconfiguring strategies to extend the memory lifetime. These reconfiguration strategies utilize a monitoring technique, what we have implemented in order to measure the resistance ratios in RRAM memory cells. Such monitoring approach can detect the highly variability effected and differentiate the bad cells from the good cells; therefore, it can improve the overall RRAM memory reliability.