北京大学学报(自然科学版)

根据数据统计特征的低功耗FFT处理器

张师群,于敦山,盛世敏   

  1. 北京大学信息科学技术学院微电子学系,北京100871;
  • 收稿日期:2006-12-28 出版日期:2008-01-20 发布日期:2008-01-20

A Low Power Pipeline FFT Processor Based on Data Statistical Characteristics

ZHANG Shiqun . YU Dunshan. SHENG Shimin   

  1. Department of Microelectronics, School of Electronics Engineering and Computer Science, Peking University, Beijing 100871;
  • Received:2006-12-28 Online:2008-01-20 Published:2008-01-20

摘要: 提出了一种基于SDF(single-path delay feedback)结构的低功耗FFT处理器。该FFT处理器使用了根据输入数据的统计分布特征的功耗优化方案。详细分析了该方法的优缺点,并提出了相应的改进方案。使用中芯国际0.18 μm工艺设计实现了一个64点的FFT处理器,通过比较发现对于特定的数据流,大约可以节省15%的功耗。

关键词: 快速傅立叶变换, 低功耗, 数据统计特征

Abstract: A low power FFT (Fast Fourier Transform) processor based on Single-path Delay Feedback (SDF) architecture is proposed. Power optimization strategy based on statistical characteristics of input data is applied in the design. Both the advantages and the drawbacks of this strategy are analyzed in detail, and solutions to overcome the drawbacks are addressed. An experimental 64-point FFT processor model is implemented in SMIC 0.18 μm CMOS technology, and for given input signals, 15% power reduction can be achieved.

Key words: Fast Fourier Transform (FFT), low power, data statistical characteristics

中图分类号: