Asymptotic Performance of Generalized Selection Combining in Generic Noise and Fading

In this letter, we investigate the performance of generalized selection combining in channels impaired by generalized fading and generic noise with finite moments. In particular, we derive accurate high signal-to-noise ratio approximations for the bit and symbol error rates of linear modulation schemes, which offer significant insight into the impact of the number of selected diversity branches, the type of fading, and the type of noise. Our analysis is applicable to all practically relevant types of fading including Rayleigh, Rician, Nakagami-q, and Nakagami-m fading, all types of noise with finite moments including Gaussian noise, Gaussian mixture noise, generalized Gaussian noise, and co-channel interference, and non-identically distributed fading and noise across the diversity branches. Our results reveal that while performance always improves with increasing number of selected branches for identically distributed noise, this is not necessarily true for non-identically distributed noise.