general information
research groups
radioscience seminars
stanford courses
oral defense abstract
industrial affiliates
Oral Defense Abstracts

Date: Thursday, June 26, 1997
Time: 1:45 pm (Refreshments at 1:30 pm)
Location: CIS 101

Special University Ph.D. Oral Examination
Blind Equalization for short burst wireless communications
Byoung-Jo Kim
Department of Electrical Engineering

Recent wide-spread use of wireless personal communications services encourages efficient use of limited spectrum to provide better service to more subscribers. Adaptive equalization or sequence estimation is often used in high speed wireless systems to mitigate inter-symbol-interference caused by multipath propagation of radio channels. However, adaptive equalization or sequence estimation requires training sequences to obtain vital information about constantly changing wireless channels. This training overhead occupies a significant portion of the available bandwidth in many TDMA systems employing relatively short burst formats.( a few tens of symbols up to one or two hundred symbols.). Short burst formats are used to limit end-to-end delay of such systems. Blind equalization is an adaptive equalization technique that does not rely on training sequences. Thus, blind equalization may recover the wasted bandwidth associated with training.

In this talk, we will discuss important issues involved in applying blind equalization to short burst wireless communication. They include slow convergence speed and local/ill convergence problems associated with blind equalization algorithms based on non-convex cost functions, e.g., the Constant Modulus Algorithm( CMA ). Considering these issues, we have developed a blind equalizer suited for short burst communications. In the process, we quantify the length of data with sufficient information for convergence, and the effect of initialization on fractionally spaced CMA equalizers. Also, we have developed a blind sequence estimator based on per-survivor-processing( PSP ) for the same application. The sequence estimator obtains its initial channel estimates blindly from the on-set of a burst with minimal complexity, and fully utilizes PSP.

< back to top