Loading...
Please wait, while we are loading the content...
Turbo-estimation de canal pour les systèmes OFDM utilisant des préambules
| Content Provider | Semantic Scholar |
|---|---|
| Author | Seller, Olivier Lacroix, Dominique |
| Copyright Year | 2003 |
| Abstract | In this article, we present an iterative semi-blind channel estimation algorithm, for OFDM systems relying on preamble. Compared to simple estimation, based on preamble averaging, gains are 1,5dB or better. When Doppler is high, the proposed algorithm still keeps system working whereas simple estimation algorithm does not. This work is based on an optimal iterative algorithm [1][2][3], relying on the Maximum A Posteriori criterion (MAP), and uses the Expectation-Maximization algorithm (EM) [4][5] The algorithm uses a particular representation of the OFDM channel, using Karhunen-Loeve decomposition theorem [6]. Channel estimates are projected on a sub-space obtained with this representation. From this algorithm we developed six modifications. First a channel tracking algorithm was set up to cope with Doppler variations, without the help of pilots. Another modification is proposed to adapt to 16QAM case. More, receiver complexity has been reduced thanks to 2x1D projection, update formula modification, and the use of a hard output decoder. Last, we study algorithm performances under bad knowledge of channel statistics. 1. Signal OFDM et modele de canal Nous notons n l R , le symbole recu sur la sous-porteuse l du multiplex, pour le n symbole OFDM. Nous supposons qu'il n'y a ni interference entre symboles, ni interference entre sous-porteuses, et prenons comme modele: n l n l n l n l N H C R , , , , . + = (1) ou n l C , est le symbole transmis, n l H , le coefficient de canal qui affecte la sous-porteuse l pendant le temps symbole n. n l N , est un bruit additif blanc gaussien. Nous supposons que le canal est quasi-stationnaire, i.e. sa valeur ne change pas pendant un temps symbole TS. Pour chacun de ces temps symboles, les coefficients Hl,n sont la TFD du canal equivalent a temps discret hn(i.T), on a: [ ] ( ) l T i h DFT H i n n l ) ) . ( ( 63 , 0 , ∈ = (2) La TFD est de taille 64, le prefixe cyclique comporte 20 echantillons, NPM=52 sous-porteuses sont modulees. 2. Representation appropriee du canal discret a evanouissements On considere des blocs temps-frequence comportant NSYM symboles OFDM. Dans [2] un modele est derive a partir de la matrice de covariance des coefficients n l H , , en utilisant le theoreme d'expansion de Karhunen-Loeve. On note cette matrice hermitienne tt HH R , , elle comporte M=52*NSYM lignes. Le canal a 2 dimensions est represente par : |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://documents.irevues.inist.fr/bitstream/handle/2042/13512/A104.pdf?sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
