Improving Error Metrics in Approximate Multipliers with an Error Recovery Unit

Citations

SCOPUS

1

초록

This paper proposes a new approximate 4-2 compressor with an error recovery unit. Proposed approximate multipliers, which incorporate the proposed compressor, demonstrate significant improvements in error metrics (MED, MRED, NED) compared to a conventional approximate multiplier and existing approximate multipliers. Additionally, using this technique, a notable improvement in peak signal-to-noise ratio (PSNR) of the proposed approximate multipliers is observed. However, the proposed multipliers exhibit higher values in Power, Performance, Area (PPA) compared to the conventional approximate multiplier and existing approximate multipliers. To address this issue, a truncation technique is applied. The truncation reduces power consumption and area by cutting off a portion of the approximate multiplier and replacing it with constant values, which has a minimal impact on overall accuracy. The truncated multipliers show a slight increase in error metrics compared to the multipliers before the truncation is applied. © 2025 IEEE.

키워드

Approximate computing; Approximate multiplier; Compressor, Error recovery, Truncation
제목
Improving Error Metrics in Approximate Multipliers with an Error Recovery Unit
저자
Jeon, Jongmin; Kim, Youngmin
DOI
10.1109/ICEIC64972.2025.10879652
발행일
2025
유형
Conference paper
저널명
2025 International Conference on Electronics, Information, and Communication, ICEIC 2025