This site has been archived on

AC202 NEWTEST

List of participants
Contact

Back to "ACTS Projects by Title Letter N"

List of Domains and Chains
Mobile Domain
Broadband Access Chains

High Performance Neural Network Signal Processing Schemes for Wireless Terrestrial and Satellite Transmissions

Main Objective
The aim of this project is to propose new Neural Network based methods for adaptive equalisation of the satellite UMTS transmission channel. The use of roof top Adaptive Antennas will also be studied within the frame of the project. Objectives are to evaluate the performance issues related to the use of Neural Networks equalisation techniques and to compare with classical equalisation implementations. The technical feasability of the implementation of Neural Network based equalisers will be studied, especially at UMTS terminal level.

Technical Approach
For adaptive equalisation of non-linear satellite channels, artificial Neural Networks (NN) have shown their ability to solve various problems encountered in the related topics: non-linearity arising in satellite data links, equalisation of rapidly time varying and multipath channels, adaptive antennas.
The project aims at developing a demonstrator system for UMTS Satellite links and to carry out the feasibility study of ASIC based subsystems, up to but not including the foundry level.
To this end, the first year of the project has been devoted to the implementation of a non-linear down link satellite channel together with the first studies on Neural Network equalisers and Adaptive Antennas. During the second year of the project, performances of NN algorithms and Adaptive Antennas will be available together with the comparison of NN equalisation methods to classical equalisation techniques. When NN equalisers are validated in simulations their implementation will be done on the hardware prototype. The medium-bit rate of the UMTS satellite application (around 64KBits/s) allows the implementation of a hardware prototype, with todays available subsystems. During year three, when NN algorithms are validated on the prototype, the complete design study and simulation of an ASIC for a non-linear satellite link equaliser will be achieved. ASIC design will be carried out up to the edge of the foundry process. The last step of the project is devoted to the demonstration of the performance of adaptive NN equalisation through technology trials using actual satellite equipments.

Summary of Trial
Trials are of technological nature.
Demonstration of Neural Network equaliser performance with actual satellite non linear components will be achieved. In particular, a testbed allowing to simulate a complete satellite transmission channel including non linear distorsions will be implemented. This testbed will be used to assess the performance of the Neural Network equaliser.

Key Issues

Achievements
During first year a simulation tool for down link non-linear satellite channel has been implemented.
Second year will be devoted to the study of Neural Network based algorithms and to the demonstration of the viability of such algorithms for the implementation of simple equalisers to be used over non linear UMTS satellite channels. It is expected that the proposed methodology will lead to unexpensive implementation, hence favouring the emergence of unexpensive dual mode (terrestrial satellite) terminals.

Expected Impact
Impact on future UMTS air interface standardisation. Deployment of satellite UMTS will take advantage of unexpensive and efficient equalisation techniques. Also, satellite resource optimisation will lead to more cost efficient S-UMTS systems deployment.

Contact:

					Francis Castanie
ENSEEIHT: Ecole Nationale Superieure d'Electronique
d'Informatique et d'Hydraulique de Toulouse
2 Rue Camichel, BP 7122
31071 Toulouse cedex 7
France
Tel:    +33 5 61 58 82 90
Fax:    +33 5 61 58 82 37
E-mail:
					(email removed)
				

List of Domains and Chains
Mobile Domain
Broadband Access Chains

List of participants

F
ENSEEIHT
F
CNES
F
GIAT Industries
D
MEDAV
F
Midivaleur
UK
Mobile System Int
F
THARSYS
S
Univ College Karlskrona
UK
University of Newcastle

DCSIMG