This paper considers LTE downlink channel estimation. The proposed technique addresses time-varying multipath fading with separable one-dimensional interpolations in frequency and in time. In the frequency dimension, we take the linear minimum mean-square error (LMMSE) approach. But LMMSE channel estimation requires knowing the channel correlation function, whose estimation poses a problem for systems with a small number of reference signals (RS) operating in time-varying channels. To solve the problem, we consider a previously proposed approach which approximates the channel power-delay profile (PDP) with one that can be described using only two parameters, namely, the mean delay and the root-mean-square (RMS) delay spread. We adapt that approach and specialize it to LTE. An approximate channel correlation function can then be obtained from the approximate PDP and used in frequency-domain channel estimation. In the time dimension, we consider several ways of interpolating time-varying channel responses, including one that uses a basis expansion model based on discrete prolate spheroidal sequences (DPSS). Various ways of putting together the frequency- and time-domain interpolations are investigated, and their performance is verified via computer simulation and analysis of the results.
- Channel estimation
- Discrete prolate spheroidal sequence (DPSS)
- Linear minimum mean-square error (LMMSE)
- Power-delay profile (PDP)