Study of pilot overhead for iterative OFDM receivers on time-varying and sparse underwater acoustic channels

While the frequency-selectivity of underwater acoustic channels motivates the use of orthogonal frequency division multiplexing (OFDM), the time-selectivity leads to large pilot overhead for channel estimation and Doppler effects that cause intercarrier interference (ICI). ICI can be estimated and mitigated using an appropriate channel model, but in turn leads to even further increased demand on pilot overhead. While sparse channel estimation can alleviate this challenge, pilot resources for channel estimation are always a precious commodity. We therefore study iterative receivers that feed back tentative or soft symbol decisions from error correction to be used as additional pilots in the next iteration of channel estimation. The key point is that this can be coupled with sparse channel estimation and that the uncertainty in soft decisions is modeled accurately. We use an experimentally recorded data set to compare the performance of sparse and conventional channel estimators under varying pilot overhead. We find that when increasing the pilot overhead - at the cost of reduced data rate - the performance saturates at some point. This point varies depending on if ICI is estimated or not, and is reached much earlier when using iterative receivers.