Time versus frequency domain channel tracking using Kalman filters for OFDM systems with antenna arrays

We derive decision directed Kalman tracking algorithms both in the time and the frequency domain to track the channel impulse response (CIR) and the transfer function (TF), resp., upon modeling the channel dynamics by a first-order autoregressive process for orthogonal frequency-division multiplexing systems. By comparing the covariance propagation of the TF estimate we show that, as long as no structural unobservability problem occurs, the tracking performance is identical in both domains if the same channel dynamic model is employed. The analysis is based on a spatially uncorrelated multiple-input multiple-output (MIMO) channel and can be extended to the case of spatially correlated MIMO channels.