The article justifies the necessity of taking into consideration measurement accuracies in experimental tests of welding machines and presents the effect of systematic errors of measurement channels on the errors related to the determination of root-mean-square current, root-mean-square voltage and average values of momentary power. The study also presents errors and uncertainties in measurements of active, passive and apparent power in supply systems of one-phase and three-phase welding machines. For the study-related purposes it was necessary to assume the symmetry of three-phase supply voltage and the linearity of elements in load branches. The article provides primary information about digital measurements of the frequency and the angle of the phase shift of periodic wavelengths indicating sources of systematic errors. The work also presents dependences enabling the experimental determination of filling factors of rectangular wavelengths of TIG welding machines as well as describes measurement errors in three cases of rectangular wave shapes. The computational determination of errors and uncertainties was exemplified using the results of tests concerned with welding source power efficiency.