Optimal Placement of Training for Channel Estimation and Tracking

The problem of training symbol placement for the estimation of block-fading channel is first considered. The channel is frequency selective with correlated random taps. The Cramér-Rao Bound (CRB) on the mean square error (MSE) of channel estimators is derived and minimized with respect to the pilot symbols and their placement. It is shown that, among all orthogonal pilot sequence placements, the minimum CRB is achieved by concentrating pilot and data powers in the middle of the packet. The placement of training for tracking time-varying channels is then addressed. Both flat and frequency selective fading environments are considered. We optimized training by minimizing the MSE of channel estimators chosen from optimal and sub-optimal tracking algorithms.