Scholarlink

Low performance of digital filtering devices is the main problem in solving practical problems of digital signal processing. One of the ways to improve the performance of such devices is parallel signal processing. In this paper, an approach is proposed to search for optimal sets of modules of a special type for filtering signals using truncated multiplication and accumulation (TMAC) blocks in the residual number system (RNS). Thus, the application of the proposed metric for searching for sets of modules for the range of 16-bit, 24-bit, 32-bit and 48-bit RNS filter using the proposed metric allowed us to reduce the delay to 11.33%, 25.63%, 25.63% and 15.13%, respectively. The developed filters can be used to improve the performance of digital devices, where the main priority is to minimize the delay. Further research will be aimed at implementing the proposed filters into digital signal processing devices to solve machine learning and image processing problems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving the Balance of RNS by Using the Metric of Finding Optimal Sets of Low-Cost and 2 k + 1 Modules

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Albina Abdulsalyamova,
  • Maxim Bergerman,
  • Pavel Lyakhov

摘要

Low performance of digital filtering devices is the main problem in solving practical problems of digital signal processing. One of the ways to improve the performance of such devices is parallel signal processing. In this paper, an approach is proposed to search for optimal sets of modules of a special type for filtering signals using truncated multiplication and accumulation (TMAC) blocks in the residual number system (RNS). Thus, the application of the proposed metric for searching for sets of modules for the range of 16-bit, 24-bit, 32-bit and 48-bit RNS filter using the proposed metric allowed us to reduce the delay to 11.33%, 25.63%, 25.63% and 15.13%, respectively. The developed filters can be used to improve the performance of digital devices, where the main priority is to minimize the delay. Further research will be aimed at implementing the proposed filters into digital signal processing devices to solve machine learning and image processing problems.