Adaptive predistortion for a nonlinear data transmission system

Project Type: Master/Diploma Thesis
Student: Angeringer Ulrich
Mentor: Gernot Kubin


 The reduction of the nonlinear distortion of a simplified discrete-time model of an AFE (Analog Front End) is investigated. This model incorporates interpolation, decimation and a model of the linedriver, which shows weakly nonlinear behavior. It is identified using appropriate adaptive nonlinear system identification methods. The system identification is the pre-requisite to the linearization which is realized using an adaptive pre-processing scheme which can reduce the nonlinear distortion of a weakly nonlinear system. The adaptive identification is done using a Volterra filter, a combined Volterra/Hammerstein filter structure, a Volterra filter with parameterizable kernel structure and an LNL (linear-nonlinear-linear) parallel-cascade model. The latter two are used for linearization. The Volterra filter with parameterizable kernel structure is a compromise between filter complexity and identification accuracy. The parallel cascade of LNL filters involves linear filters only and static nonlinearities, therefore the number of coefficients is considerable reduced compared to a Volterra filter.