Electronics Engineering Department, University of Engineering and Technology, Taxila, Pakistan.
PLoS One. 2021 Jan 27;16(1):e0245943. doi: 10.1371/journal.pone.0245943. eCollection 2021.
Stochastic computing has recently gained attention due to its low hardware complexity and better fault tolerance against soft errors. However, stochastic computing based circuits suffer from different errors which affect the output accuracy of these circuits. In this paper, an accurate and area-efficient stochastic computing based digital finite impulse response filter is designed. In the proposed work, constant uniform patterns are used as stochastic numbers for the select lines of different MUXes in the filter and the error performance of filter is analysed. Based on the error performance, the combinations of these patterns are proposed for reducing the output error of stochastic computing based filters. The architectures for generating these uniform patterns are also proposed. Results show that the proposed design methodology has better error performance and comparable hardware complexity as compared to the state-of-the-art implementations.
随机计算由于其硬件复杂度低和对软错误的容错能力更好，最近受到了关注。然而，基于随机计算的电路会受到不同的误差的影响，这些误差会影响这些电路的输出精度。在本文中，设计了一种准确且面积高效的基于随机计算的数字有限脉冲响应滤波器。在提出的工作中，常数均匀模式被用作滤波器中不同多路复用器的选择线的随机数，并分析了滤波器的误差性能。基于误差性能，提出了这些模式的组合，以减少基于随机计算的滤波器的输出误差。还提出了生成这些均匀模式的架构。结果表明，与最先进的实现相比，所提出的设计方法具有更好的误差性能和可比的硬件复杂度。
