Generalized equalization for underwater acoustic communications

Severe time-dispersion of the channel is one of the limiting factors for high data rate acoustic communications in shallow water. To overcome the large time spread, the research community has devoted enormous effort in the design of high efficient, low complexity equalizers. The goal of these equalizers is to convert the time-dispersive channel into a unit-tap impulse function. In another research direction, code division multiple access (CDMA) and orthogonal frequency division multiplexing (OFDM) systems are designed to combat the multi-path effect. However, using OFDM or CDMA alone to combat the multi-path effect often forces system designers to reduce the data rate in such severe time-dispersive channels. In this paper, we propose to adopt a class of generalized equalizers to generate an effective channel with a desired length. These equalizers are referred to as channel-shortening equalizers. Then, other advanced modulation schemes, such as OFDM and CDMA, can be employed to overcome the remaining time-dispersion of the channel. Compared to the unit-tap equalization, channel-shortening equalizers can provide quality performance at low computation cost. Examples of current channel-shortening equalizers are based on minimum mean-squared error (MMSE) and maximum shortening signal-to-noise ratio (MSSNR) criteria. In this paper, these algorithms are compared in the context of underwater acoustic communications. Their performance, complexity, and implementation in acoustic communications are investigated. In the digital subscriber line (DSL) channel, where channel-shortening equalizers are often applied to, MSSNR algorithms perform better than MMSE algorithms. However, we show that the MMSE algorithm is more suitable for the underwater acoustic channels with lower complexity and robustness to noise. With the MMSE method, a channel-shortening equalizer with 20 taps can convert a typical 120-tap acoustic channel into a 33-tap shortened channel. Using data obtained from t- - he Kauai experiment, June-July 2003, it shows that the mean of the effective signal-to-interference ratio (SIR) achieves 15.7 dB. For the same channel, the optimum unit-tap equalization needs more than 200 taps to achieve the same effective SIR