Study On Behavioral Model Of Power Amplifiers Based On Polynomial Structures

Because of fast development of microwave circuits simulation method and digital predistortion (DPD) technology, behavioral modeling of power amplifiers has become one of most important research fields of power amplifiers. For behavioral model approaches, polynomial, or Volterra series for memory, is the most common function which is used to fit nonlinearity of power amplifiers. In this dissertation, we represent the measurement method on nonlinear characteristic and memory effect of power amplifiers and then discuss behavioral modeling based on polynomial structure. The main content of this dissertation include: (1) The measurement on nonlinearity and memory effect of power amplifiers; (2) Behavioral modeling building from one tone and two tone test; (3) Behavioral modeling building from modulation signal; (4) Dynamical methods on estimation of value of model time delay and calculation of model parameters. At the end of the paper, two power amplifiers are measured to verify the performance of new models intrduced in our paper.The main originality in this paper can be summarized as follows:1. A new IM3 phases measurement method based on DSP technology.Since the magnitudes and phases asymmetries of IM3 components describe the memory characteristic of power amplifiers, phases of IM3, which is very hard to be measured by tradition analog instruction, can be used to evaluate memory effect of power amplifier precisely. We represent a new IM3 phase measurement method which uses digital receiver technology and estimate value of IM3 phase by signal process arithmetic from sampled data. At last, the precision of method and cause of error are analyzed.2. A multi-silce structure behavioral model building by two tones test.One tone and two tone tests are most common methods used to measure nonlinear characteristic and memory effect of power amplifiers, and manay researchers applied these measurement results to build behavioral modelings of power amplifiers. Here, we introduce a multi-slice structure model based on amplifier circuit characteristics to deal with both nonlinearity and enveliope memory of power amplifiers. One advantage of this model is the value of its parameters can be simply extracted from measurement data obtained by two tone test.3. Simplied Volterra Series model based on feedback structure model.Owning to the development of wide-band communication system, the structure of behavioral modeling of power amplifiers, which is often built by system identification theory, becomes more and more complex because more severe nonlinearity and more complex memory effect need to be represented. One of the foremost challenges to behavioral modeling approaches is model reduction technology based on the physical knowledge of power amplifiers. Here, we show a new simplied Volterra series model, which is deduced from amplifier feedback model and simplified by omitting higher order memory item in the intial model. By far, a frequency field method is been given to estimate model time delay.4. A dynamical building method for tradition behavioral modeling.Because output characteristic of power amplifiers is highly related to work environment, behavioral model of power amplifiers must be dynamic and its parameters are time-varying. Another thing should be mention is each power amplifier shows different memory characteristic and in most cases only one kind memory effect is dominating in memory effects of power amplifiers. Here, we give a new model time delay determination method depending on analysis of error-input cross-correlation sequence at first, and then introduce an RLS arithmetric to calculate model parameters dynamicly. For future application, we also describe a RLS arithmetric based on QR decomposition at same section. At last we show a dynamic model building method based on these two arithmetics and apply it to a tradition MP model to build a dynamic model.