A Radix-4 Partial Product Generation-Based Approximate Multiplier for High-speed and Low-power Digital Signal Processing
A Radix-4 Partial Product Generation-Based Approximate Multiplier for High-speed and Low-power Digital Signal Processing
复制标题
DOI：
10.1109/icecs.2018.8617854
复制
发表时间：
2018-12
期刊：
影响因子：
--
通讯作者：
Xiaoting Sun;Yi Guo;Zhenhao Liu;S. Kimura
中科院分区：
文献类型：
--
作者：
Xiaoting Sun;Yi Guo;Zhenhao Liu;S. Kimura
功率和能量最小化已成为当今许多电子系统的重要设计要求。近似乘法器作为一种常用的运算器，是高速、低功耗数字信号处理的理想目标。提出了一种基于基4的部分积生成方法，将部分积的高度减半。在硬件和部分积累加精度的综合考虑下，提出了一种新的4-2压缩器。结果表明，与传统的精确乘法器相比，本文设计的8位乘法器的功耗、面积和延迟分别降低了49.4%、53.4%和48.2%。同时，将该设计的精度和硬件性能与已有的几项工作进行了比较，验证了其有效性。此外，该乘法器还可以在不影响整体图像质量的情况下应用于图像处理应用中，同时大大降低了功耗。
Power and energy minimization has become a significant design requirement for many electronic systems nowadays. As a commonly used arithmetic unit, approximate multiplier is a desirable target for high-speed and low-power digital signal processing. This paper proposes a novel design based on radix-4 partial product generation method, which reduces the height of partial product to half. A new 4-2 compressor is also proposed under the consideration of hardware and accuracy performance for partial product accumulation. As a result, the proposed design for 8-bit multiplication achieves a reduction of power, area and delay up to 49.4%, 53.4% and 48.2% compared to the conventional accurate multiplier. Meanwhile, the accuracy and hardware performance of the proposed design are also compared with several previous works and shows its effectiveness. In addition, the proposed multiplier is also applied to image processing applications without compromising the overall image quality while the power consumption is dramatically reduced.