Reduction of the error floor of MSK by selection diversity

We examine the error floor of minimum shift keying (MSK) when the receiver uses selection diversity. Both received signal strength indication (RSSI)-driven diversity and bit error rate (BER)-driven diversity are analyzed. The considered channel is a two-delay Rayleigh-fading channel; detection is done by differential detectors with either fixed or adaptive determination of the sampling time. Analysis is based on the method of error regions, where the instantaneous impulse response is represented by phasors in the complex plane. We first compute the joint probability density function (pdf) of the phasors of the two diversity branches. Depending on the selection criterion, the error probability is then computed as an integral over certain functions of this pdf. In the limit of small delay spreads S, the integrals are evaluated analytically. For fixed sampling, the admissible delay spread for BER=10^-3 is increased by about a factor of three (as compared to the no-diversity case) for RSSI-driven diversity and a factor of four for BER-driven diversity. For adaptive sampling, RSSI-driven diversity gives little improvement in the admissible S for BER=10^-3 while BER-driven diversity increases it by some 50%. Results are confirmed by comparison with Monte Carlo simulations and measurements