A numerical procedure for decomposition of time series into inherent periodic nonstochastic components in form of harmonic oscillations will be explained. It is supposed that the rhythmic process is built up by these harmonic oscillating components and by a stochastic component. The procedure is a combination of Harmonic Analyses for intervals with different numbers of measured values with a significance test elaborated by R.A. Fisher for the greatest component in the spectrum. The parameter values of the harmonic oscillating components are elevated by the procedure, and the partial oscillations will be successively eliminated from the time series. The revelant application of Harmonic Analysis in any form on the one hand and of power spectra on the other will be discussed. Hints for an efficient combination of several numerical procedures are given for a decomposition of biorhythmic processes in harmonic oscillating components, and for the statistical significance of the decomposition.