This paper presents a revised version of the new signal processing method based on the analysis of a specially constructed correlation function of amplitude and phase fluctuations (APCF). This method allows us to detect the presence of a group of equidistant frequencies in the spectrum of the original signal and to measure the difference of two adjacent frequencies Δf in such a group. The end product of the processing is a histogram of a set of Δf. The effect of noise that may be present in the original signal has been examined. It has been shown that even with a very high level of noise when its component in the spectrum completely absorbs and masks spectral peaks of equidistant frequencies of the desired signal, the APCF method copes with the problem of detecting these frequencies. This method was first applied to processing of natural signals, for which recordings of geomagnetic field disturbances of an ultralow frequency (ULF) range were used. The comparison of one of the histograms with the traditional spectrum indicates that the chaotic spectrum, which has always been considered to be a noise spectrum, actually has a strictly ordered structure. It has been found that most spectral peaks belong to one of the sets (more than 10) of equidistant frequency groups. In the entire spectrum, peaks of these groups are superimposed on each other and form a complex chaotic sequence. The analysis of peaks of all histograms allows us to conclude that the equidistant frequency groups, which correspond to peaks in each histogram, are eigenfrequencies of the 2D Alfvén wave resonator. The existence of such a resonator in the magnetosphere in the vicinity of the outer edge of the plasmopause has been predicted in theoretical studies [Guglielmi, Polyakov, 1983; Leonovich, Mazur, 1987]. The processing method APCF enables us to experimentally confirm this prediction.