Area Efficient VLSI Architectures for Weak Signal Detection in Additive Generalized Cauchy Noise

Detection of a weak signal in additive Generalized Cauchy (GC) noise is important in many applications. The locally optimum detector (LOD) for a weak signal in GC noise is nonlinear in nature. When noise variance is unknown, the maximum likelihood estimator (MLE) is nonlinear and the resultant detector is complicated. Since VLSI implementation of complex nonlinearities is a challenging task, we develop an order statistics framework for detection in GC noise. We propose linear and ratio detectors, for weak signals in GC noise with known and unknown but deterministic variance, respectively. We provide extensive simulation results to show that performance loss of proposed linear and ratio detectors, is very small compared to LOD and nonlinear detector using MLE, respectively. We propose SORT - N (for running ordering of samples) and its VLSI architecture, using which we develop two-stage VLSI architectures for proposed linear and ratio detectors. Synthesis results for arbitrary waveform case indicate, for same throughput and latency, linear detector consumes lesser area to LOD, upto a sample size of N=64. For same throughput, ratio detector renders substantial area savings (approximate to 50%) over nonlinear detector using MLE, for any $N$ . Finally, we propose a reconfigurable architecture for efficient realization of all the detectors.