In this paper, an iterative multi-channel equalization and decoding technique is introduced to improve system performance in underwater acoustic communications. The turbo principle is applied to the existing canonical receiver structure including fractionally spaced decision feedback equalization with phase synchronization. The performance of multi-channel equalization and adaptive weight update algorithms are aided by son information delivered from the channel decoder. The complexity of the proposed scheme is shown to be reasonable for highly reverberant underwater channels since it has linear complexity by the use of the least mean square (LMS) algorithm. Experimental results for data transmission at 41.67k symbol/s with a carrier frequency of 100k Hz demonstrate the feasibility of the proposed algorithm.