Research And Improvement Of Multi-Level Modulation Self-Adaptive Digital Predistortion Algorithm

With the development of wireless communication technology, QPSK,QAM,OFDM are utilized in many communication systems. These digital multi-level modulation schemes are more and more popular for the requirement of increasing data rate. But the nonlinearity of RF power amplifier induced distortion of digital multi-level modulated signal which have high peak-to-average power ratio. Amplifier can only work on very low operating point, with lower output power and efficiency. So we must linearize RF power amplifier.The research begins from analysis and comparison of amplifier linearization methods, and then simulates negative feedback,feed forward,LINC(Linear amplifier using Nonlinear Components) and so on. The performance of all linearization methods are analyzed, also their advantages and disadvantages are showed. Different techniques have variable application occasions. When people choose linearization plan they must base on specified needs, synthetically consider efficiency,modulation,bandwidth,complexity of circuit,dynamic range, etc.The superiorities of digital predistortion are increased with the development of digital techniques. Especially it can track the amplifier's slow changes after adding self-adapatation. Self-adaptive digital baseband predistortion is our emphases. First the principles of predistortion are described in detail, then this linearization technique is researched from polynomial distortion and lookup table (LUT) separately. The key point of polynomial distortion is extraction of amplifier model, which can realized by polynomial fitting,Volterra series method or neural networks. On the other hand the performance of LUT is related to entry indexing ways and self-adaptive updating algorithms. Comparisons are made between them with complexity and speed of convergence.In view of digital baseband implementing, digital predistortion based on LUT is chose. LUT predistortion in 16QAM system is carried out, scatter plot and spectrum acquired obvious improvement after LUT predistortion. Entry indexing methods is analyzed and improved. A simple indexing means called amplitude-power alternative method is raised. This method satisfied common amplifier nonlinear curve, moreover, it can be easily realized. Next self-adaptive updating algorithms for entries are simulated and studied. All common algorithms are simulated and compared with convergence speed. A new bisection-secant combination method is advanced. This algorithm reduced iteration number, speeded convergence rate. Different entry number and quantization step size caused variable effects on predistortion performance because of digital inherent finite expression. But it can be solved by the progress of digital technology. Finally normalization method of LUT is discussed. Further more an envelope normalization plan is put forward, increasing efficiency and linear gain. Simulation showed that linear gain have 0.5078dB increase when input back off is equal to 3dB, more input back off will conduce more increase in linear gain.