Characterization and compensation of nonlinear distortion.

摘要

Efficient methodologies for applying polynomial models and Volterra models to the characterization and compensation of nonlinear distortion in radio communication systems are investigated in this dissertation. Nonlinear characteristics associated with high power amplifiers at the transmitter result in not only undesirable amplitude and phase distortion but also undesirable intermodulation interference in neighboring frequency bands. In order to mitigate such nonlinear effects, both predistortion and equalization techniques are proposed.; We present a new parallel adaptive predistorter for the linearization of high power amplifiers based on the indirect learning architecture. We directly model the inverse of the AM/AM response and model the AM/PM response of the high power amplifier in parallel, from the amplitude and phase response of the amplifier output, respectively. The efficacy of the new predistorter is demonstrated in terms of suppression of intermodulation products and spectral regrowth, lower total degradation for a given bit error rate, higher output power, and adaptation to changes in the high power amplifier parameters. Also we apply the new predistortion linearizer to compensate nonlinear distortion in code division multiple access (CDMA) and orthogonal frequency division multiplexing (OFDM) communication systems which require highly linear amplification due to their characteristic high peak-to-average power ratio. We demonstrate improvement in CDMA systems in terms of a reduction of adjacent channel power ratio, improved waveform quality factor, and increased output power. With respect to OFDM systems, the improvement is manifested by a lower total degradation and increased output power.; We also present sparse Volterra-based predistorters and equalizers obtained by applying the orthogonal-search technique. We demonstrate that such sparse predistorters and equalizers provide comparable performance to their full counterparts, but with a considerably reduced level of implementation complexity.; Lastly we apply Volterra modeling and neural networks to investigate the linear and nonlinear properties of laboratory-generated random sea waves. A second-order Volterra filter is used to decompose the wave-elevation time series into its first- and second-order components. The time-domain decomposition shows that the generation of large-amplitude extreme-like sea waves is due to momentary phase-locking of the first and the second-order components.