浮点指数函数的设计与实现  

Design and Implementation of Floating-Point Exponential Function

作　　者：李丽斯[1] 何虎[1] 

机构地区：[1]清华大学微电子学研究所,北京100084

出　　处：《微电子学》2012年第5期702-705,709,共5页Microelectronics

基　　金：核高基重大专项(2009ZX01034-001-001-001)

摘　　要：设计了一种用于现代雷达系统的指数函数。设计中,浮点数格式均采用IEEE-754标准32位单精度表示[1]。采用Table-driven算法实现浮点指数函数的运算。为了简化硬件的实现并提高算法实现的精度,引入了CORDIC算法。基于这两种算法结构,提出一种改进结构。采用Verilog语言描述,通过ModelSim6.5b进行仿真,结果表明,在一定的数据范围内,改进后指数函数运算结果的相对误差是传统方法的40%。在TMSC 65nm工艺下综合,频率达到483MHz,满足系统的工作频率要求。An exponential function was designed for modern radar system.Numbers of floating-point were represented in IEEE-754 single precision format,which is 32-bit in length.Table-driven method was used to implement operation of floating-point exponential function,and CORDIC algotithm was introduced to simplify the design and improve the precision in a certain range.Based on the two algorithms,an improved architecture was proposed.The design was described in Verilog HDL and simulated with ModelSim6.5b.Simulation results showed that the error range of the improved exponential function operation was reduced to 40% of the conventional method.Synthesized in 65 nm CMOS standard cell library,the circuit achieved an operating frequency of 483 MHz,satisfying requirements of the system.

关 键 词：浮点指数函数 Table-Driven算法 CORDIC算法 

分 类 号：TN492[电子电信—微电子学与固体电子学]