基于分级存储并行运算的FFT处理器设计
Design of FFT Processor Based on Grade-Memory and Parallel-Computation
-
摘要: 研究了一种基于分级存储并行运算的改进快速傅里叶变换(FFT)处理器算法,通过减少对RAM存储器的读写次数降低功耗,采用并行运算方法减少数据处理时间. 基于该算法以及改进的基-4蝶形单元设计了一款4 096点FFT处理器. 该处理器采用SMIC 0.18 μm CMOS工艺设计实现,芯片核面积为9 mm<sup<2</sup<,在slow工艺角条件下,版图后仿真最高时钟频率为192.3 MHz,功耗为422 mW@100 MHz,最小处理时间为67.92 μs.Abstract: The improved FFT arithmetic based on grade-memory and parallel-computation is proposed and studied. It can reduce the power consumption by decreasing the RAM's writing-reading numbers as well as reduce the execution time by parallel-computation. A 4 096-point FFT processor has been designed using the new arithmetic and the improved radix-4 butterfly. The processor was implemented by SMIC 0.18 μm CMOS technology and the core area of the chip was 9 mm<sup<2</sup<. In the slow corner condition, the post-layout simulation shows the highest frequency is 192.3 MHz, the power consumption is 422 mW@100 MHz and the minimal execution time is 67.92 μs.
下载: