Comparative Analysis of Envelope Representations for UHF Partial Discharge Signals

Abstract

Partial discharge monitoring using the radiometric method often requires high sampling rates due to the ultra-high frequency content of the signals. Envelope detection is widely used to reduce these requirements; however, its application in PD analysis is typically based on assumptions derived from amplitude-modulated signals, which may not hold for impulsive PD signals. This work investigates the impact of different envelope representations (positive, negative, and full-wave rectified) on the information extracted from radiometric PD signals. The analysis is based on experimental data from two defect sources combined with circuit-level simulations. Results show that positive and negative envelopes are not redundant, presenting only moderate correlation, while full-wave rectification leads to reduced peak, energy, and crest factor values. These findings indicate that, although commonly adopted, conventional envelope detection approaches may discard relevant information, potentially affecting feature extraction and degrading the performance of PD monitoring, classification, and localization systems.