Optimization of fast fourier transform based on twiddle factor using genetic algorithm on raspberry pi

The research work revolves around the 16-point Radix-4 Single Path Delay Feedback (R4SDF) for optimizing the pipelined Fast Fourier Transform (FFT) processor, which can be done by using both Single Objective Genetic Algorithm and Multi-Objective Genetic Algorithm , Nowadays in many areas of engineering and science are widely using FFT processors in most of their applications, thus, the modern science requires continuously new optimizations which includes the FFT processor to have a lower power consumption and a smaller size. However, both Signal to Noise Ratio (SNR) and Switching Activity (SA) values depend on the word length of the FFT processor, the bigger the word length of the FFT processor will result in a higher value for the SNR and the SA, Thus, this research aims to reduce the power consumption of the FFT processor by lowering the word length of Twiddle Factor for the FFT by using both Single Objective Genetic Algorithm (SOGA) and Multi-Objective Genetic Algorithm (MOGA) to find the optimum results for SNR and SA values while lowering the Word Length. Over the years the Genetic Algorithms (GA) proved to be one of the best methods for optimization. The proposed work will start by tasking the SOGA with modifying the SNR fitness function to secure the SNR value (which determines the accuracy factor) for the research to check if the research can obtain SNR value more than 63dB while lowering the word length of the Twiddle Factor, next is to reduce power consumption by tasking MOGA with finding the SA values below 192 (SA values determine the power consumption) while maintaining the SNR values above 63dB, then is to evaluate both of SOGA and MOGA results to compare with the results of the default parameters of the most relevant research. In this research the GA is for reducing the word length by optimizing its coefficients. The required amount of value for the SNR is to be more than 63 dB and for SA is to be lower than 192. The proposed work was done successfully in optimizing the FFT by using SOGA to lower the word length until 12 bits and obtaining a SNR value of 66.452dB which resulted in an improvement of 5.47% for SNR, also, the optimization for the FFT was done successfully by using MOGA to lower the word length until 12 bits and obtaining a SNR value of 65.65dB which resulted in an improvement of 5.47% for SNR and a SA value of 134 which resulted in reduction to SA by 30.2%.

Text FK 2020 11 IR.pdf Download (1MB)

Actions (login required)

View Item