Study of pilot designs for cyclic-prefix OFDM on time-varying and sparse underwater acoustic channels

Estimation of time-varying and sparse channels is a topic that has attracted considerable interest lately, especially for underwater acoustic communication. In the context of multicarrier transmission, several important questions remain unanswered: i) should pilots be placed in clusters as for time-varying, non-sparse channels, or randomly dispersed between the data as common in the Compressive Sensing literature; ii) as pilot and data subcarriers cannot be perfectly separated at the receiver due to intercarrier interference (ICI) should channel estimation be based on observations corresponding to pilot subcarriers only, or can the data subcarriers be used to practically extract additional information about the channel; and iii) how does the performance vary with the number of pilot symbols? We use data recorded at the recent CalCom'10 experiment to investigate these questions. We find that a sufficient number of adjacent observations is required to estimate the ICI, which can either be achieved by a pilot design that uses clustered pilots, or by using data subcarriers as additional observations. When using data subcarriers as observations, the effect of the unknown data should be modeled as increased noise. Finally, by varying the number of pilot subcarriers, we can trade off error performance for data rate.