传统流水线CORDIC(Coordinate Rotation Digital Computer,CORDIC)算法精度不高,输出延时较大,并且需要依靠剩余角度计算进行旋转方向的判断,占用较大的资源.针对以上问题,本文采用角度二极化重编码方法消除剩余角度计算,通过折叠角度区间将角度映射于区间[0,π/4].结合查找表以及合并迭代技术,减少角度计算的迭代次数和硬件单元,降低输出时延,只需要3个周期就能完成CORDIC计算.使用结果重映射方法完成正弦和余弦的全象限实现.寄存器资源消耗为传统算法的35.37％,输出时延减少85％.基于180nm CMOS工艺,完成CORDIC算法的ASIC实现.正弦和余弦的平均绝对误差分别为2.5472 × 10-6、1.9396× 10-6,相比较于传统CORDIC算法,精度提升一个数量级.
Design and ASIC Implementation of Low-Latency CORDIC Algorithm
The traditional CORDIC(Coordinate Rotation Digital Computer)algorithm used in a pipeline has low pre-cision,long output delay,and relies on residual angle calculations to determine the rotation direction,which consumes significant resources.To address these issues,this paper proposes the Angle Bipolar Encoding method to eliminate residual angle calculations,maps the angles into[0,π/4]by folding the angle range.Combining lookup tables and merged iteration techniques,the number of iterations for angle calculations and hardware units are reduced,resulting in reduced output delay.The CORDIC calculation can be completed in just 3 cycles.The Full Quadrant Implementa-tion of sine and cosine functions is achieved using result remapping.The register resource consumption is 35.37％compared to the traditional algorithm,resulting in an 85％reduction in output delay.An ASIC implementation of the CORDIC algorithm is performed using a 180nm CMOS process.The average absolute errors for sine and cosine are 2.5472 × 10-6 and 1.9396 × 10-6 respectively.Compared to the traditional CORDIC algorithm,the precision has improved by an order of magnitude.
coordinate rotation digital computerangle bipolar encodingmerged iterationCMOSASIC
万方数据
维普
