Date of Award

Summer 1998

Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)


Electrical and Computer Engineering

First Advisor

Yeheskel Bar-Ness

Second Advisor

Hongya Ge

Third Advisor

Alexander Haimovich

Fourth Advisor

Eliza Zoi-Heleni Michalopoulou

Fifth Advisor

H. Vincent Poor


CDMA is an attractive multiple-access scheme, because of its potential capacity increase and its anti-multipath fading capability. For satisfactory performance, however, the effect of the "near-far" problem has to be resolved. This problem can be combated by using power-control, which, however, results in an overall reduction in communication ranges, and thus in a loss of capacity. Among other methods for mitigating the near-far problem is the use of decorrelating receivers, both of fixed type, which directly utilizes the cross-correlation of the users codes, and of adaptive type, which uses recursive algorithms that leads to signal decorrelation. Not to lessen the importance of other adaptive algorithms, the current research concentrates on what was termed in the literature "bootstrap algorithm" . Although the emphasis will be on applying the adaptive bootstrap decorrelator, the fixed type will be used primarily to provide comparison. Also used for comparison are both blind adaptive and training sequence based MMSE.

Most of the literature on multiuser detection has been assuming BPSK. However, a need for transferring wideband data demands using modulation schemes with high bits/cycle, such as QAM. Therefore, modification of the receiver is considered, so that QAM-modulation can be applied efficiently, using the complex signal approach of this modulation.

For the asynchronous channel, vast amounts of research have been devoted to using one-shot matched filter banks followed by conventional decorrelators which implement the inverse of some (partial) correlation matrix. In this work, an adaptive bootstrap version is presented, which is suitable for the one-shot structure shown previously to be more robust to errors in delay estimation. It has also been noted that such a correlation matrix can, depending on the channel characteristics, become ill-conditioned or even singular. Therefore, another matched filtering structure, followed by what is called a multishot conventional (fixed type) decorrelator, has been previously suggested to mitigate this singularity problem. However, the fixed type of the multishot decorrelator is expected to have similar non-robustness to errors in delay estimation as was previously shown for the one-shot. Therefore, the adaptive multishot bootstrap decorrelator is presented and evaluated. Also, by adding an adaptive canceler, an extension to the above matched filter-decorrelator combination, will be proposed and evaluated. A multipath time-variant fading environment will be used in some of these performance evaluations.

Finally, when handling multipath channels, the question is raised whether path combining should be done before or after the signals are decorrelated. For the asynchronous case, a one-shot extension of the bootstrap algorithm is presented, which is capable of decorrelating the signals from resolved paths of different users, to facilitate the decorrelate before combining case.