Statistical modelling of transmission holding time in PAMR systems

A statistical analysis of transmission holding time data based from two different Public Access Mobile Radio (PAMR) is presented. The data are obtained from digitised conversations by automatically identifying the spectrum of the in-band signalling that limits every transmission. It is shown how the empirical histogram is far from being exponential, which is the type most used in teletraffic issues. From the different theoretic distribution models adjusted to the transmission times a mixture of lognormals is the best fit. As transmission trunking needs far more signalling and the grade of service (GoS) is more crucial than when trunking by message, a careful choice of the distribution model is strongly recommended. The best fit distribution can be used to feed computer simulations in order to obtain accurate measurements of system performance.