الفهرس Only 14 pages are availabe for public view
 |  | from | 94 |  |  |
| | | from | 94 | | |
Abstract
The thesis anm ta iru ticaie and improve the perrormance of the difterent cqu.:iizalion techniLiue ruea n broadband DM1-based DSL systems which pro ide in-speed data colnmunicatioru and are able to provide hich-qualitv multimedia ontems. In these systems, the cyclic prefix CP) used is of insufficient length to simpl\ LsC one-tap FEQ per tone. Two equalization structures are considered. namely, channel shortening time domain equalizers TEQ) and per tone equalizers (PTEQ). The first equalization structure is the channel shortening TEQ, in which an FIR filter is used in order to shorten the length of’ the channel impulse response to be no longer than the length of the CP. Four TEQ design methods are discussed. Firstly, the MMSE method is thoroughly investigated which is currently the most commonly used approach in commercial ADSL modems. The drawbacks of MMSE are analyzed and the modified eigen approach method is proposed to override these drawbacks. It is shown with simulations that this method enhances the performance and improves the system bit rate. The other three design methods, namely, MSSNR, Min-ISI and MDS, are then presented. To evaluate the performance of these design methods, standard ADSL channel models and typical ADSL transmission parameters were used. All TEQ design methods discussed so far assume the knowledge of the channel impulse response as well as noise statistics at the receiver. It is found that none of these design methods can directly optimize channel capacity, and that channel capacity optimization requires costly non-linear optimization algorithms. The second equalization structure considered is the PTEQ. It is based on the transfer of the TEQ operation to the frequency domain after the FFT demodulation. It is shown through simulations that PTEQ outperforms any TEQ design for a given number of taps. The PTEQ is the only bit rate maximizing DMT equalizer that has been presented so far. It is shown that the PTEQ and the TEQ have a similar computational complexity during data transmission. The main focus here is on performance evaluation of the adaptive algorithms used for PTEQ initialization. Three adaptive algorithms are presented. The NLMS which is the simplest algorithm but has slow convergence speed,