In this paper the influence of different wavelet filters on the harmonic analysis of power system waveforms is investigated. For this aim, wavelet packet transform-based algorithm for rms measurement is implemented. Besides commonly used Daubechies filters, Vaidyanathan with 24 coefficients and Coiflets the influence of several Johnston's filters with different characteristics is examined. Experimental tests are performed under different conditions, in case of stationary signals with different harmonics components and in case of nonstationary harmonics distortion. Experimental results show that Johnston's filters outperform the commonly used wavelet filters for harmonic analysis.