New results on selection diversity

The performances of selection diversity receiver structures in a slow flat Rayleigh-fading environment are assessed. A number of new and interesting results are obtained. Binary digital signaling using noncoherent frequency-shift keying (NCFSK), differential phase-shift keying (DPSK), coherent phase-shift keying (CPSK), and coherent frequency-shift keying (CFSK) is considered. The traditional analysis (the traditional selection diversity model) of a selection diversity system is based on choosing the branch with the largest signal-to-noise (SNR) power ratio while assuming that the noise power is constant across all branches. However, many practical selection systems choose the branch based on a largest signal-plus-noise (S+N selection) sample of a filter output. This paper comprises accurate analyses of such S+N selection systems. Results show that S+N selection systems perform better than predicted by the traditional selection diversity model. This is because the former includes the statistical nature of the noise, whereas the latter does not. The performance difference between the two models increases as the number of diversity branches increases. For each of DPSK and CPSK, the dual diversity equal gain (EG) combining and S+N selection systems perform identically. For each of NCFSK and CFSK, receiver structures which are equivalent when there is no diversity perform differently in a diversity environment. Certain dual diversity S+N selection systems give the same performances as EG combining or square law combining. The results are contingent upon perfect cophasing for the EG combining. In systems where estimates of the combining carrier phases contain noise, S+N selection outperforms EG combining for dual diversity.