Error floor of unequalized wireless personal communications systems with MSK modulation and training-sequence-based adaptive sampling

We analyze the error floor of unequalized personal communications systems using minimum-shift keying (MSK) with training-sequence-based adaptive sampling in a two-delay fading channel. We include Gaussian filtering of the input, finite accuracy in determining the optimum sampling time, and filtering at the receiver. We prove that for pure MSK and low time dispersion, training-sequence-based adaptive sampling can completely avoid errors caused by intersymbol interference. The actual errors are caused by "secondary" effects (filtering and finite-resolution sampling) in conjunction with the channel time dispersion. Errors occur if the normalized phasor of the channel impulse response falls into certain error regions; the computation of these regions gives physical insights into the error mechanism and allows a highly efficient computation of the average bit-error rate (BER). The average BER varies as K/spl middot/(S/T)/sup 2/, where S is the RMS delay spread and T is the bit length. The proportionality constant K depends on the Gaussian filtering in the transmitter, the receiver filtering, and the amount of oversampling. The BER can be orders of magnitude lower than for the (quasi-) fixed sampling case, in which timing is derived solely based on the channel delay power profile. For two-branch diversity reception, we show that the BER is proportional to (S/T)/sup 4/.