Superimposed Pilots Vs. Conventional Pilots for Channel Estimation

We study two competing pilot symbol based schemes, viz. superimposed pilots (SP) and conventional pilots (CP) for the estimation of a single-input multiple-output (SIMO) wireless channel. We derive expressions for the mean-squared error (MSE) and Cramer-Rao bound (CRB) of estimation of the SIMO channel. It is demonstrated that the asymptotic CRB of SP is 3dB lower than the MSE of the simplistic mean estimator. We propose a semi-blind SP scheme which asymptotically achieves this CRB. In the second part, we quantify the throughput performance of the estimation schemes by developing a framework for the worst case capacity of a channel with correlated symbols and noise. It is observed that while CP outperforms SP in terms of MSE of estimation, SP has an overall advantage over CP in terms of net throughput and is therefore bandwidth efficient. Further, we derive expressions for the optimal source to noise power ratio (SNR) as a function of the pilot to noise power ratio (PNR) in the context of SP.