Training-Based MIMO Systems: Part I - Performance Comparison

A simple and straightforward way of estimating the unknown communication channel is to transmit known training/pilot sequences. Conventionally, these non-information carrying pilots are time-multiplexed with the data and, therefore, consume valuable bandwidth. Hence in the recent past, a number of publications have instead suggested superimposing pilots to the data to perform channel estimation in, e.g., MIMO systems. However, the performance gain achieved by using superimposed pilots (SIP) compared to the more conventional time-multiplexedpilot scheme is not well studied and motivates further research. In this paper we attempt to make a fair comparison between the two schemes, their respective maximum information rate is computed and compared using lower bounds. The theory is applied to a block-wise flat-fading MIMO channel and it is found that in certain scenarios (such as many receive antennas and/or short channel coherence times), it is beneficial to use SIP. Finally, an abstract projection-based viewpoint of the training-based models using vector spaces is given