Measurements and statistical modelling of spectrum occupancy

Although results from a number of spectrum occupancy surveys have been published in recent years, the associated modelling has concentrated on a single quantity representative of a band (service allocation). The congestion Q, defined as the fraction of channels within an allocated band with observed levels exceeding a given threshold or as the percentage of all channel samples that exceed a given threshold is essentially equal to the unconditional probability q that a random sample is above the threshold. From this definition its complement p = 1-q can be identified. This paper presents results from two recent experimental investigations at a quiet rural site in Wiltshire. Rapid scans of small bands of frequencies were used to measure mean durations of clear channels and slower scans were used to examine the whole of the HF spectrum from 5 to 29 MHz to determine whether there are preferred sub-bands within particular service allocations. Both investigations suggested the need for a statistical model involving additional variables other than p or q