Abstract-The proliferation of non-linear loads in both industrial and residential distribution grids leads to undesirable non-sinusoidal and fluctuating harmonic pollution on voltage and current waveforms. New analysis tools, such as wavelets, are being used to overcome the problems posed by the use of the Fourier Transform when analyzing complex waveforms. Nevertheless, the selection of the wavelet basis must be done carefully to minimize spectral leakage due to the non-exact frequency discrimination. In this context, this paper proposes an objective method for comparing different wavelet families for the measurement of harmonic contents. This methodology is applicable for determining the best filter among the 53 preselected structures according to the following requirements: frequency selectivity, computational complexity, convolution results and observed spectral leakage. With all these considerations, the Butterworth IIR filter of order 29 was found to be the best wavelet decomposition structure to achieve an effective harmonic analysis up to the 50 th order.