Area, Power and Speed Optimized Early Output Majority Voter for Asynchronous TMR Implementation

This paper presents a new, efficient asynchronous early output majority voter that can be used to effectively realize an asynchronous triple modular redundancy (TMR) implementation. For the input-output mode asynchronous realization, the dual-rail code was used for data encoding and four phase return-to-zero and return-to-one handshake schemes were separately used for data communication. The proposed majority voter requires 62.8% less area and dissipates 37% less power on average compared to the best of the existing asynchronous majority voters while considering both handshake schemes. Importantly, the reductions in area and power are achieved without sacrificing the speed. Example TMR implementations show that the proposed majority voter leads to simultaneous reductions in cycle time, silicon area, and power dissipation. As a result, the proposed majority voter enables improved optimization in figure-of-merits such as area–cycle time product, power–cycle time product, and area–cycle time–power product for TMR implementations utilizing it compared to TMR implementations incorporating other majority voters. The circuits were implemented using a 32/28-nm CMOS technology.