Performance analysis of filtered multitone modulation systems for underwater communication

In Filtered Multitone Modulation (FMT), the bandwidth is split into a number of subbands in which single-carrier signals are modulated onto separate carriers with very little spectral overlap, making the waveform resilient to intercarrier interference. Demodulation is followed by a bank of parallel equalizers, one for each subband. The length of each equalizer is less than what would be needed in a single-carrier system operating over the same total bandwidth. A channel-estimation-based decision-feedback equalization method using a two-step procedure (estimation/fusion) is proposed for FMT. In this structure, channel estimates are used to cancel postcursor intersymbol interference from input signals prior to linear equalization. Parallel channel estimates of adjacent subbands are then fused, i.e. linked to a common underlying multipath model which exploits frequency correlation to improve upon the channel estimates. The sparse nature of underwater acoustic channels may be exploited by setting to zero all but the significant channel taps, thus further relieving equalization of the estimation noise. The performance of underwater FMT is assessed via simulation and using real data transmitted over 800 m in shallow water at rates of 2-6 kbit/s. The results are compared to OFDM and single-carrier QPSK modulation operating at similar bandwidth efficiencies.