Partial discharge signal de-noising evaluation of SGWT

This paper presents a real-time implementation of partial discharge (PD) signal de-noising based on wavelet analysis. The results show that the Second Generation Wavelet Transform (SGWT) removes effectively white Gaussian noise (WGN), requires a smaller footprint and is less computationally intensive than the Discrete Wavelet Transform (DWT), reducing the cost for the hardware implementation. It is implemented on an OMAP-L138 processor as part of a real-time DSP-based PD monitoring system and its performance is determined. The throughput of SGWT de-noising, at different decomposition levels, using both cacheable and non-cacheable external mobile DDR memory addresses, is then evaluated.