Performance analysis of turbo equalization with channel estimation
(2007)
Files
Ramon.pdf
Restricted Access - Adobe PDF
- 16.63 MB
Details
- Authors
- Supervisors
- Vandendorpe, Luc
- Abstract
- Since the discovery of turbo codes, the need for channel estimators delivering accurate estimates at very low signal-to-noise ratios (SNRs) arose. Inspired by the turbo principle, several authors then proposed to enhance the estimation quality by supplying the estimator with (soft or hard) information on data symbols output by the channel decoder (or by the equalizer). Indeed, the bad performance of conventional estimators at low SNRs was attributed to their lack of information about the data symbols. Because involved in an iterative process with the decoder, these new estimators are referred to as iterative channel estimators. Unfortunately, although performing better than standard estimators, most of the existing iterative channel estimators are developed in an ad-hoc way often without often clearly emphasizing the considered optimization criterion, their properties and complexity. Consequently, this thesis firstly derives in a systematic fashion iterative estimators of frequency selective channels. For each proposed estimator, the optimization criterion is specified and the properties as well as complexity are discussed. In addition, by making relevant assumptions on the information sent by the decoder back to the estimator, the performance of these estimators is analytically approached. This enables to highlight the influence of the key system parameters (length of data frame, number of pilots, quality of the feedback information,...). Secondly, this thesis analyzes sensitivity of equalization to channel and signal-to-noise ratio (SNR) estimation errors. The goal is to quantify how accurate the channel and SNR estimates delivered to the equalizer must be in order to obtain acceptable performance.
- Affiliations
Citations
Ramon, V. (2007). Performance analysis of turbo equalization with channel estimation. https://hdl.handle.net/2078.5/98075