Sampling and prefiltering effects on blind equalizer design

In the development of equalization algorithms for unknown channels, the effects of the sampling rate and the analog receive prefilter prior to discretization of the received signal are often overlooked. In this paper, these effects are investigated. The relationship between the fractionally spaced output samples of a noise-limiting prefilter and the symbol spaced output samples of a matched filter is studied for both the time-invariant and the timevarying channels. It is shown that the prefilter and the sampling rate can have significant effects on blind equalization algorithms. Thus, this paper provides a common framework for comparing different blind algorithms that are studied in the literature with different sampling rates. A case study of the well-known subspace method for blind channel identification is presented. The effects of the noise color due to the prefilter on equalizers is investigated, and the sensitivity of the truncation of the overall channel impulse response in terms of the mean squared error (MSE) performance criterion is investigated through numerical examples.