AOI-Based Novel Majority Approximators for Energy-Efficient On-Chip Data Bus Inversion

초록

This paper proposes two new AOI-based approximate encoders for data bus inversion (DBI) in on-chip data buses. Traditional accurate DBI encoders suffer from high hardware complexity, making them impractical for on-chip DBI applications. A previously proposed approximate DBI encoder, the AOI-AOI approximate majority voter (CONV-AOI), reduces hardware complexity but has low accuracy in majority condition decisions, limiting its effectiveness in minimizing switching activity. To address these limitations, we propose novel majority approximators that improve majority decision accuracy while maintaining a low-cost implementation, making them more suitable for on-chip DBI encoder designs. The conventional and proposed encoders were synthesized using Synopsys Design Compiler with a 28 nm CMOS technology under identical conditions. Synthesis results show that the proposed designs reduce power consumption by up to 63.44% compared to accurate majority voters. Additionally, our proposed approximate encoders reduce incorrect detections by up to 28.83% compared with CONV-AOI.

키워드