A.B. Tokarev

The priori information about the presence and location of noise intervals in the analyzed frequency band usually is absent. But at least one of the noise intervals can be found by searching the frequency domain with the lowest average value of the discrete spectrum samples. This article contains the information necessary to specify the width of the noise interval, which practically guaranteed (with a probability of 99 %) should be present in the analyzed discrete spectrum at different values of the frequency axis usage. If signal frequencies have the uniform distribution in the analyzed frequency range we can guarantee the existence of more than 20 sample width noise interval between the signals discrete spectrums only at low values of the frequency axis usage. Taking this into account, it is necessary to develop data processing algorithms, based on noise intervals with the width of not more than 20 samples. High spectral resolution of radio monitoring systems increases the probability of wide noise intervals occurrence, but usually this advantage is not suitable in practice, because it increases significantly the size of the spectrum arrays (and, respectively, increase the processing complexity), or makes analyzed frequency band too narrow. It is recommended to use the radio monitoring systems with a resolution of 4-8 samples per radio channel. For spectrum arrays containing 1000 samples or more it will be guaranteed (with a probability of 99 %) that there is at least one 18 sample width noise interval for the value of frequency axis usage up to 50 %.