Full-pipeline floating-point trigonometric function device based on combination of CORDIC and Taylor

Abstract

The invention discloses a full-pipeline floating-point trigonometric function device based on combination of CORDIC and Taylor algorithms. The device comprises a preprocessing module which is used for judging whether a taylor expansion algorithm is started or not according to an input angle Z and converting the input angle Z from a double-precision floating-point format into a fixed-point format; a cordic algorithm calculation path module which is used for finishing calculating sine and cosine results of the input angle Z, wherein when N is smaller than 13, obtained results are precise results and are output; a taylor algorithm calculation path module which is used for finishing calculating sine or cosine of the input angle Z, wherein only when the N is greater than or equal to 13, the taylor algorithm is selectively started; and a postprocessing module which is used for carrying out normalization processing on calculation results of the cordic algorithm and selectively outputting the sine and cosine calculation results. The device has the advantages of low delay, low error and full pipeline.

Description

technical field [0001] The invention mainly relates to the design field of a trigonometric function generator, in particular to a full-flow floating-point trigonometric function device based on the combination of CORDIC and Taylor algorithm. Background technique [0002] In signal processing systems such as audio communication, video communication, signal processing, and radar image processing, fast and high-precision numerical calculation of trigonometric functions (especially trigonometric sine / cosine functions) is required. In order to meet the requirements of image and digital signal processing and speed up data processing capability, it is necessary to explore fast trigonometric sine / cosine function hardware implementation circuits. [0003] The calculation method of trigonometric sine / cosine function is divided into: look-up table method, polynomial approximation method and digital iteration method. The traditional calculation of trigonometric sine / cosine usually adop...

AI Technical Summary

Problems solved by technology

[0016] 2. Truncation error: The truncation error is caused by the limited operand bit width. The truncation error is inevitable in the calculation process. For N iterations, the truncation error may cause log 2 (N) bits are inaccurate. In order to eliminate this error, enough guard bits must be added after the valid bits.

At the level of relative error, the above-mentioned truncation error cannot be solved by adding protection bits. When the input angle is close to 0 and π / 2, the relative error of the trigonometric function result calculated by CORDIC will become very large.

Images