SUMMARYIn this paper, a novel multi-resolution filter for on-line decomposition and analysis of partial discharge (PD) signals based on discrete wavelet transform (DWT) is developed. By applying this filter, as soon as a new sample of the original signal is drawn, the new elements of detail and approximate components at any desired level is derived, and the successive procedure of signal processing in basic DWT is eliminated. Using this filter, the PD or any non-stationary (transient) signal is decomposed to different frequency band components, simultaneously. Therefore, the decomposition components of PD signals in all frequency bandwidths are simultaneously available. The frequency characteristic of this filter is verified, then in order to make the proposed filter more suitable for on-line applications, its order is reduced without losing the accuracy of the results. Results, for simulation and field test data, have shown the appropriateness of the proposed filter for on-line PD detection and measurement.