Reseach On Nonlinearity Characteristics Of Envelope Tracking Power Amplifiers With Dynamic Bias

Envelope tracking(ET) power amplifier can improve the efficiency of PA by varying the bias voltage with the amplitude of input signal. Compared to fixed power amplifier, ET PA has different nonlinearity characteristics; therefore nonlinearity of ET PA is discussed in this paper. Modeling of ET PA with Memory Polynomial(MP) and Neural Networks(NN) are researched,and Digital Predistortion(DPD) on ET PA is also discussed in this thesis.First, electronic model of FET is introduced. Source of nonlinearity and memory effect of fixed biased PA is then analyzed. Then the nonlinearity of ET PA is researched with AM-AM and AM-PM distortion through the change of input and output impedance of ET PA. Memory effect caused by signal bandwidth is discussed as well as the delay between the signal and bias voltage. This thesis also proposes a method to optimize ET design based on the signal’s power probability density.Then, envelope shaping function is discussed. Taking high power amplifier as an example, three kind of ET PAs with fixed gain, fixed gain compress and max efficiency are compared, finding that ET PA with fixed gain is the best design for its good linearity and high efficiency. This thesis also proposes a method to reduce the bandwidth of bias voltage signal and compares their characteristics.At last,modeling and DPD simulation of ET PA are researched. This thesis analyzes the modeling with Memory Polynomial and Neural Networks and compares their accuracy using a low power amplifier and a high power amplifier. A novel MP model is discussed based on the traditional MP model and achieves a 4~7dB improvement on NMSE. DPD based on Memory Polynomial increases the ACPR of ET PA by more than 10 dB with ADS and MATLAB simulation with 5MHz LTE signal. This thesis also designs the hardware platform to realize DPD design of ET PA, and DPD hardware system including up conversion, down conversion and clocking module are designed.