Design Considerations for Behavioral Modeling and Pre-Distortion of Wide Bandwidth Wireless Systems

This work presents the challenges of processing wideband communication signals as well as analyses of signal processing techniques to improve the quality of non-linear RF transmitters. The behavioral modeling begins with optimizing input training signals, which are stimuli comprised of different frequency sinusoidal waves to model the properties of signal sources. Then, a Volterra-based pre-distorter hybrid model consisted of polynomial and dynamic deviation terms is chosen to validate the training. When model parameters extraction is performed under multi-tone stimuli, the resulting compensator is robust and maintains decent linearity performance under 100 MHz carrier aggregation traffic. For minimal computational cost, a model complexity reduction technique based on matching pursuit is applied to reduce the number of model parameters to 75 while maintaining similar performance. The methodology discussed in this paper is applicable to designing non-linear RF systems facing wideband requirements.