Investigation On Linearization Technology Of Multi-Band Radio Frequency Transmitter

In the modern wireless communications systems,the multi-band radio frequency(RF)transmitter is used more and more widely on the strength of its high spectrum efficiency and concise implementation structure.With the development of circuit design and manufacturing technologies,the circuits of weak analog signal in transmitters are becoming more and more ideal,while power amplifiers will still cause inevitable nonlinear distortions.For the multi-band RF transmitter,as the signals located in different bands feed to the power amplifier concurrently,the nonlinear distortion will become more and more serious.In order to satisfy the requirement of communication standards,multiband RF transmitters need more effective nonlinearity compensation methods.At present,the linearity improvement technologies can be categorized into two major groups.The first group,referred to here as exterior assisted linearization technologies.The most important representative of the first group is digital predistortion(DPD).The second group,referred to here as the optimization of transmitter structure which is implemented by improving the amplification process of input signals.The main content of this thesis includes two sides.On the one hand,several key technologies of the multi-band DPD are studied.On the other hand,under the guidance of digital-analog joint-design method,the structure of dual-input Doherty power amplifier based transmitter is optimized to improve its linearity as much as possible.The main innovations of this thesis are listed as follows:1)A dual-band baseband equivalent behavioral model generation method which is by converting the RF non-linear model to baseband non-linear model is obtained.Based on the nonlinear characteristics of power amplifier with concurrent dual-band input,and the mapping relations between RF signals and baseband signals,a low pass equivalent expression of power amplifier behavioral model is obtained.A volterra series-based model for concurrent dual-band power amplifier using dynamic memory depth is obtained by pruning the above equivalent expression according to the physical characteristics of power amplifier.In order to balance the model complexity and modeling accuracy,six evolution forms of the proposed model are derived by classifying the dynamic order of the model kernels and introducing even order nonlinearity.Comparing to the conventional multi-band baseband equivalent behavioral model,the proposed model describes the interactions between different passband signals more rigorously,and is able to modeling the output characteristics more accurately.The DPD results of a concurrent dual-band power amplifier operating at 3.0 GHz and 3.5 GHz show that: when the bandwidth of input signals are 20 MHz and 40 MHz,the proposed model can improve the adjacent channel power ratio of output signal to-51.2 dBc and-47.9 dBc,respectively.2)The identification algorithms of single input DPD and non-single input DPD are compared.Comparison results show that the identification algorithm can be shared by single input DPD and non-single input DPD.The rationality and necessity of sparse parameter identification algorithm are analysed.In order to estimate the sparse parameters,two identification algorithm based on sparse-Bayesian-learning theory and sparse representation are proposed.The sparse-Bayesian-learning based identification algorithm estimate the coefficients of behavioral model from the view of probability.In this algorithm,the prior distribution of the parameters is set to a Gauss distribution to force the model parameters to be sparse.In addition,the prior distribution of the parameters can reduce the solution space of the parameters,which tends to reduce the length of sampling data that needed for parameter identification.Experiment results show that: in most instances,with the proposed algorithm,the number of model coefficients can be reduced over 50% without sacrificing performance.The sparse representation theory based identification algorithm chooses the model kernels with the idea of greedy algorithm,evaluates the importance of model kernels by the projection of the residual vector on each kernel,and avoids matrix inversion by iterative computation.The algorithm is used to estimate the parameters of the memory polynomial model with 75 parameters.Experiment results show that with 10.6% and 24% model kernels,the modeling accuracy can reach 88.4% and 98.9% of that of the complete model,respectively.3)An equivalent single input and single output DPD using baseband stitching technique is proposed to linearize the concurrent multi-band power amplifier.An important design criterion that the description bandwidth of behavioral model must equal to the bandwidth of the feedback signal is reported in this thesis.The DPD results of a concurrent dual-band power amplifier operating at 3.3 GHz and 3.8 GHz show that: for the equivalent single input architecture,when the hardware resource consumption and DPD performance of the are close to that of the conventional multi-input architecture,the required number of model coefficients can be reduced by 77.8%.4)An inter-band time-delay compensation algorithm that is based on correlation enhancement technique has been proposed to improve the performance of behavioral modeling and DPD of multi-band transmitters.The DPD results of a concurrent dualband power amplifier operating at 2.9 GHz and 3.3 GHz show that: the adjacent channel power ratio of output signal in the conditions of without inter-band time delay compensation,with the traditional off-line compensation method,and with the proposed on-line delay compensation method are-35 dBc,-48.7 dBc and-50.1 dBc,respectively.5)The meaning of joint design method that between signal preprocessing and final stage power amplifier in the optimization of transmitter architecture is discussed,and based on this,a design method of dual-input Doherty power amplifier based transmitter is proposed.There are two basic requirements for the design method: maximizing the drain efficiency and keeping the gain fixed.The generating mechanisms of nonlinear distortion of the dual-input Doherty power amplifier are discussed,and based on this,an adaptive signal separation algorithm is proposed to improve the nonlinearity.The dualinput Doherty power amplifier based transmitter which obtained through the above design and optimization methods show good linearity between input and output in the experiments.Specifically,when the bandwidth of input signal is 10 MHz,the adjacent channel power ratio of output signal can reach-45 dBc without any exterior assisted linearization technologies.