Finite-State Markov Modelling of Frequency-Selective Fading Channels with Correlated Taps

We consider the problem of modelling a randomly time-varying frequency-selective fading channel as a finite-state Markov channel (FSMC). For a fading channel with two correlated taps and given statistical parameters, the accuracy of an FSMC is assessed by comparing its information rate, when the receiver has ideal channel-state information, to that of the original continuous-valued channel. We show that in order to construct an FSMC with a given desired accuracy, fewer states are required if the correlation between taps is taken into account than if the taps are treated as being independent. These results demonstrate the suitability and accuracy of FSMCs for modelling a time-varying frequency-selective fading channel with memory.