Equalization of satellite UMTS channels using neural network devices

Abstract

Abstnzcl-The presence of non-linear devices in severnl communication channels, such as satellite channels, causes distortions of the transmitted signal. These distortions are more severe for non-constant envelope modillations such as 16-QAM. Over the last years Neural Networks (NN) have emerged txq competitive tools for linear and non-linear channel equalization. However, their main drawback is often slow convergence speed which results in poor tracking capabilities. The present pnper combines simple N N structures with conventional equalizers. The N N techniques are shown t o efficiently approxiniate the optimal decision boundaries which results in good symbol error rate (SER) performance. The paper gives simulation examples (in the context of satellite mobile channels) and compares neural network approaches t o classical equalization techniques.

Predistortion System Implementation Based On Analog Neural Networks For Linearizing High Power Amplifiers Transfer Characteristics

“…There are several solutions to operate the HPA close to its saturation point without generating non-linearities [2]- [6]. One is to implement a module directly on-board the payload, just before the HPA, in order to obtain a linear transfer characteristic at the HPA output [4]- [6].…”

Low complexity predistortion and equalization in nonlinear multicarrier satellite communications

“…In model-based mitigation techniques, the use of Volterra models [10] or pruned Volterra basis is extended [7,11]. Neural network mitigation approaches have been reported for SISO (single carrier) in satellite [12,13] and terrestrial applications [14,15]. On the other hand, MIMO neural networks are usually sensitive to the training data and due to their complexity they are not extensively used in mitigation techniques that require the compensation of dynamic effects [16].…”

Equalização não-linear de canais de comunicação.

“…Várias estruturas não-lineares têm sido estudadas, como as baseadas em séries de Volterra e as que utilizam redes neurais. As redes neurais, apesar de serem computacionalmente complexas, apresentam vantagens sobre outras estruturas não-lineares devido ao seu alto grau de paralelismo o que as tornam atraentes para implementação em circuitos integrados [45].…”

“…Em [45], foram apresentadas algumas estruturas híbridas utilizadas na equalização de canais de satélite. Essas estruturas consistem basicamente do LTE em série com uma rede RBF ou MLP utilizando ou não realimentação de decisões.…”

“…Os resultados das simulações mostrados em [45] evidenciam que a utilização de uma estrutura híbrida garante um bom desempenho na equalização de canais não-lineares. …”