Faculty of Computing, Health and Science
School of Engineering
The paper presents a novel method of evaluating the square root function in FPGA. The method uses a linear approximation subsystem with a reduced size of a look-up table. The reduction in the size of the lookup table is twofold. Firstly, a simple linear approximation subsystem uses the lookup table only for the node points. Secondly, a concept of a variable step look-up table is introduced, where the node points are not uniformly spaced, but the spacing is determined by how close to the linear function the approximated function is. The proposed method of evaluating nonlinear function and specifically the square root function is practical for word lengths of up to 24 bits. The evaluation is performed in one clock cycle.