A novel ROM-less architecture for Direct Digital Frequency Synthesizer based on Linear Neural Networks

In Conventional Direct Digital Frequency Synthesizer, the presence of ROM results in large size and complexity. Some DDFS is based on a mathematical series which also has some computational time delay. This paper deals with the Artificial Neural Network based DDFS architecture with linear transfer function, unlike the non-linear transfer function [3], which has low complexity, low computational delay and high spectral purity. The proposed DDFS has a high SNR (Signal to Noise Ratio) of about 48dB.