| Amplifiers have had great demands placed on them as spectrally efficient modulation techniques become more prevalent in the industry, and bandwidth requirements increase. Not only must they have broad bandwidth, they must remain linear over a large dynamic range. Linearization has been shown to be a viable technique for improving an amplifier's performance.;Baseband predistortion is a linearization technique with excellent performance capabilities that requires very little additional RF equipment, and the system can be made adaptive resulting in reduced sensitivity to changes. Until recently the focus has been on memoryless predistortion. For narrow-band systems the memory effects of the amplifier can largely be ignored. However; as bandwidth increases the memory effects will become more significant and can't be ignored.;This thesis discusses the effect of increasing bandwidth on memoryless predistortion using neural networks. Neural networks were chosen for their ability to model black box systems by knowing only their input and output. The system can also be extended to compensate for memory using Wiener and Hammerstein models. Results from simulation and experiments are presented and compared. |
