Feed-forward neural networks for analog impairment mitigation in high power RF transceivers

Constantly rising capacity and increasing complexity of mobile communication systems as well as the general demand to reduce the power consumption to get the systems greener, are a big challenge especially for radio frontends. Complex modulated signals of mobile communication standards like LTE have high demands on radio frontends regarding signal requirements. the numerous variety of different standards in different frequency bands have further demands on radio transceivers and its architecture. Flexible radios, suitable for different standards, signals and frequencies with highest efficiency and dynamic are required. This paper shows possibilities to enhance the flexibility of RF transceivers and how to enable future standards and handle high requirements while relaxing configuration using neural networks for signal processing in RF transceivers.