Search

In this paper a multiband transceiver designed for underwater channels is presented. Multi-branch filtering at the receiver is used to leverage the diversity offered by a multi-scale multi-lag scenario. The multi-branch bank of filters is constructed by estimating scale and delay coefficients through an initial preamble composed by a maximum length sequence. An intelligent design of the pulse shaper at the transmitter and the receiver permits the reduction of the complexity of the equalization algorithm: the effective channel can be seen as a simple time-invariant finite impulse response filter possibly affected by a carrier frequency offset. Adaptive turbo equalization is chosen to deal with the high time-variability which characterizes many underwater channels and also to avoid the burden of estimating all parameters of the different paths. A phase locked loop and recursive least squares algorithm is implemented on each branch and for each subband which sweep the received sequence several times to refine decoding; in order to enhance the receiver performance the updating of the equalizer taps is obtained by making use of soft information provided as feedbacks from the turbo decoder. The proposed transceiver is tested on a realistic channel obtained by channel soundings performed in the Lyme Bay area, South England. Also comparisons with other existing turbo equalizer schemes are performed.

The joint linear precoder and decoder Minimum Mean Squared Error (MMSE) design represents a low complexity yet powerful solution for spatial multiplexing MIMO systems. Its performance can be further boosted through optimally selecting the number of spatial streams to be used according to the available Channel State Information (CSI), so-called spatial-mode selection. The performance of both the latter MMSE design and the related spatial-mode selection criteria, however, critically depends on the availability of timely CSI at both transmitter and receiver. In practice, the latter assumption can be severely challenged, due to channel time variations that lead to imperfect CSI at the transmitter. State-of-the-art designs mistakenly use this imperfect CSI to design the linear precoder and rely on the receiver to reduce the induced degradation. We have alternatively proposed a robust Bayesian joint linear precoder and decoder solution that takes into account the uncertainty on the true channel, given the channel mean feedback at the transmitter. In this paper, we further improve the performance of our aforementioned robust design using a new spatial-mode selection criterion based on channel mean feedback. We also illustrate, via Monte-Carlo analysis, the robustness of the resulting improved design to channel time variations, which outperforms the state-of-the-art approach.