Theoretical bounds on the estimation and prediction of multipath time-varying channels

Starting from a parametric model of time-varying multipath channels with additive white Gaussian noise, we derive the theoretical bound on the channel parameters estimation as a function of number of samples, SNR and model mis-matching errors. We use this bound to derive an approximate closed form expression of the channel mean square prediction error, subsequently used to evaluate the maximum training period. Finally, we illustrate a simple and effective method for estimating the channel parameters (path delays, amplitudes, phases and frequencies), using chirp signals as training sequences.