Quantised decision based gradient descent algorithm for fast fading OFDM channels

For very fast fading OFDM channels robust channel tracking is needed. We propose a quantised decision based gradient descent algorithm (QDGDA) which exploits the fact that the fading in mobile environments is correlated across time. A single tap NLMS tracker is run across time for each subcarrier in an OFDM symbol to track the channel. To make the algorithm less sensitive to error propagation, the gradient estimate has to be continuously and correctly updated, (almost) irrespective of any decision error. To obtain reliable gradient information for the NLMS algorithm at time instant n, we use a weighted combination of the gradient at time n and time n-1. These time varying weights are a function of reliability of the decision made on data symbols.