Matched filter bounds without channel knowledge at the receiver

In this paper we analyze the limits on communication induced by mobility. The case of no channel state information at the receiver (CSIR) is considered. We specifically analyze the matched filter bound (MFB), which corresponds to the ML performance for the detection of a symbol assuming all other symbols are known. This setting allows to incorporate channel estimation from the known symbols. Two channel models can be considered for two types of transmission. In the case of (quasi)continuous transmission, parametric channel models are considered with a decomposition into fast and slowly fading channel parameters, with the estimation errors on the slow parameters being neglected. For the case of block-wise transmission, we assume basis expansion models with the correlations between the basis components being taken into account. On the other hand one can also distinguish between specular (pathwise) and diffuse (separable correlation) channel models. The MFB degradation due to channel estimation is characterized in terms of a misadjustment factor, which is analyzed further for SISO flat channels and MIMO OFDM channels, with specular or diffuse channel models. The channel estimation based approach is also briefly contrasted with and outperforms differential (de) modulation approaches. Finally, some links with previous work on capacity analysis are made.