Mutual information based approximation of coded BER/ FER with finite signal alphabets

Tutor: Carsten Bockelmann
Type of Thesis: Studienarbeit (Dipl.)
date of end: 06/2009
Student: Serkan Arikan
Status: finished
ANT-shelfmark: S-06/09-2

This project aims at the approximation of BER or FER after decoding by bit- oder symbolwise mutual information. Complex systems which apply channel coding can be simulated bit by bit only to a certain extend because of the limited computational resources available today. Especially in system simulations where whole systems with many users and complete protocol stacks are simulated a bit by bit approach is not feasable anymore. In Order to model the behaviour of the physical layer approximations are needed. One way to get a reliable estimate of the paket error rates after decoding is based on the constrained mutual information of the channel and heuristic mappings to the error rates of interest.

The project starts with a literature research with regard to information theoretic basics, notably the mutual information, constrained mutual information and calculation techniques. Beginning with a simple AWGN transmission, the accuray of mutual information based error rate modelling for different types of codes (convolutional codes, turbo codes, ldpc codes) shall be evaluated. A possible extension of this project could be the optimization of a power and rate allocation problem with regard to a mutual information based model.

Premises: Familiarity with Matlab, Channel Coding I+II

Last change on 08.10.2009 by C. Bockelmann
AIT ieee tzi ith Fachbereich 1
© Department of Communications Engineering - University of BremenImprint / Contact