Prof. Dr.-Ing. Werner Henkel
This Guided Research focuses on frame synchronization for the DMT-based transmission system, which is used on a proposed acoustic sensor network for sonar applications. The time offset estimate not only needs to be robust and accurate to maintain the orthogonality of the various sub-channels in the DMT system, but also computationally efficient to reduce hardware cost. The current frame synchronization method used for the proposed sensor network is the maximum likelihood (ML) estimation . This method is relatively robust and accurate, however requires a large memory at the receiver for storing the overhead of the data frames to compute the average. The overhead typically consists of hundreds of frame symbols, and the averaging of the time offset estimation for each frame symbol consumes a significant amount of computation time. This paper compares two different frame synchronization methods with the ML estimation method, namely the Schmidl and Cox method , and the Park method  in order to find an optimum solution. All three methods are evaluated in the flat channel free of noise, the AWGN channel, and the wireline bus channel simulations. The results suggest that the Schmidl and Cox method has a plateau in its proposed timing metric, which introduces great uncertainty in the estimation. The Park method, nevertheless, shows promising results in both the estimation accuracy and the computation efficiency. Two peaks of the timing metric in the Park method exhibit great robustness and accuracy in 300 simulations, and it only needs one training symbol appended at the beginning of the information frame symbol for the synchronization. Even though the peaks are sensitive to SNR, the simulation shows that in the worst-case scenario, the transmitting power per DMT sender (given the fixed noise), or equivalently the SNR, is more than suitable for the implementation of the Park method. Therefore it can be concluded that the Park method is the optimum solution out of all three in the frame synchronization for the proposed DMT wireline acoustic sensor network. .