Sadeghi, ParastooRapajic, Predrag2015-12-080090-6778http://hdl.handle.net/1885/32867We study information rates of time-varying flatfading channels (FFC) modeled as finite-state Markov channels (FSMC). FSMCs have two main applications for FFCs: modeling channel error bursts and decoding at the receiver. Our main finding in the first application is that receiver observation noise can more adversely affect higher-order FSMCs than lower-order FSMCs, resulting in lower capacities. This is despite the fact that the underlying higher-order FFC and its corresponding FSMC are more predictable. Numerical analysis shows that at low to medium SNR conditions (SNR ≲ 12 dB) and at medium to fast normalized fading rates (0.01 ≲ fDT ≲ 0.10), FSMC information rates are non-increasing functions of memory order. We conclude that BERs obtained by low-order FSMC modeling can provide optimistic results. To explain the capacity behavior, we present a methodology that enables analytical comparison of FSMC capacities with different memory orders. We establish sufficient conditions that predict higher/lower capacity of a reduced-order FSMC, compared to its original high-order FSMC counterpart. Finally, we investigate the achievable information rates in FSMC-based receivers for FFCs. We observe that high-order FSMC modeling at the receiver side results in a negligible information rate increase for normalized fading rates fDT ≲ 0.01.Keywords: Acoustic intensity; Channel capacity; Decoding; Fading (radio); Markov processes; Signal to noise ratio; Time varying systems; Finite-state Markov channels; Higher-order; Information rates; Fading channels Channel capacity; Fading channels; Finite-state Markov channels; Markov processes200810.1109/TCOMM.2008.0606222015-12-08