Optimizing Training-Based MIMO Systems: How Much Time is Needed for Actual Transmission?

We study the design of training-based multiple-input multiple-output systems in two block-wise transmission schemes. The conventional transmission scheme has a fixed amount of energy to be used in each block, hence transmission takes place in every block. For this scheme, we study the optimality of using all available time in each block for transmission and provide bounds to significantly reduce the ranges of the possible values of the optimal training and data lengths. The second scheme, called the flashy transmission scheme, is constrained by an average amount of energy per block, and uses some but not necessarily all blocks for transmission. For this scheme, we find the optimal fraction of blocks to be used for transmission. When this optimal fraction is less than one, we show that the optimal training and data lengths are independent of the energy constraint.