Skip to Main Content
Channel estimation is one of the key components of space-time systems design. The transmission of pilot symbols, referred to as training, is often used to aid channel acquisition. In this paper, a class of generalized training schemes that allow the superposition of training and data symbols is considered. First, the Cramer-Rao lower bound (CRLB) is derived as a function of the power allocation matrices that characterize different training schemes. Then, equivalent training schemes are obtained, and the behavior of the CRLB is analyzed under different power constraints. It is shown that for certain training schemes, superimposing data with training symbols increases CRLB, and concentrating training power reduces CRLB. On the other hand, once the channel is acquired, uniformly superimposed power allocation maximizes the mutual information and, hence, the capacity.