Dynamic nonlinear behavioral modeling and adaptive predistortion for RF transmitters.

摘要

The increasing demand on wireless transmitter systems has generated a need for efficient wideband designs capable of transferring the data usually carried via a signal modulation scheme such as OFDM, CDMA...etc with the least possible distortion during transmission. In this context one should remember that the main source of distortion in a transmitter is the power amplifier which has a tendency to saturate when the input signal power level increases beyond a certain level, thus resulting in nonlinear degradation in terms of gain and relative phase shift respectively known as AMIAM and AM/PM effects.;One other important element is the particular nature of the training signal that is used for operating the adaptive algorithm. Some signals will add up more complexity to the procedure as well as some feedback delay due to error calculation dependency on the input signal.;The objective of this thesis is the design and implementation of an adaptive predistorter based on a specific type of transmitter model and using single frequency waves as training signals with a frequency which falls within the bandwidth of the transmitter. The linearity criterion of the predistorter/amplifier combination is that a single frequency signal should reach the output with a linear gain and phase shift for a broad range of power levels. Once this goal is achieved, one is left with a merely linear dynamic system i.e. that is identifiable by means of linear system identification techniques available in standard MATLAB toolboxes.;Clearly, when using a static predistorter, the underlying key assumption is that the static nonlinearities are separable from linear dynamics, i.e. the transmitter can be modeled for example with 2-box models also known as Hammerstein or Wiener models where a static nonlinear function is followed by a linear dynamic filter or vice versa. We successfully illustrate the single sine wave based predistortion linearization for a theoretical Hammerstein type model.;One common approach to compensate for these power dependent distortions is known as predistortion whereby a nonlinear functional box with inverse nonlinear characteristics of that of the transmitter precedes the power amplifier where as a results, this combination gives rise to an identity operator. The adaptive procedure that updates the predistorter parameters at each iteration is based on a specific optimization algorithm which is in general computationally expensive. Therefore, it is important to produce a rough estimation of the predistorter parameters which can serve as a good initial point; this can help considerably in reducing the length of computations.;It is also shown that for phase distortion compensation, one needs to apply two sine waves since the relative phase information in the context of quadrature modulation cannot be revealed with only one sinusoidal signal. In one last step the constructed predistorter performance is validated by launching a WCDMA signal and the AM/AM and AM/PM curves are discussed as well as the power spectrum density of the output signal where the non linear effect known as adjacent channel power regrowth must be reduced.