Nowadays in industry, data acquisition system signals are often compressed by piecewise linear trending. Many authors have indicated that compression with use of a wavelet transform provides better results. In this paper, the threshold selection algorithms, such as VisuShrink, SureShrink, BayesShrink, and NeighCoeffs are compared experimentally using real industrial data from the Damadics benchmark. The impact of the compression on the decoded signal is analysed. The results of the calculations are compared with a standard piecewise linear trending approach.