Estimation of time-varying digital mobile radio channels

In packet-based communication systems using dispersive channels, training symbols are usually provided to estimate the channel response. It is demonstrated that for both simple constant and adaptive least-squares estimates, the average performance of the estimator can be computed with explicit reference to channel time variations. With these analytical results, quantitative evidence can be provided for the classification of channel time variations as fast or slow. They also can be used to evaluate important system design decisions, such as total packet size, training overhead, and data rate, for particular channel conditions.<>