Characterization And Modeling Of Memory Effects Of RF Power Amplifiers Based On Volterra Series

Linear modulation with non-constant envelope has been widely used in modernwireless communication, thus, RF power amplifier is asked to keep appropriate powerand efficiency, at the same time guarantee the linearity. Research found that due to thereference amplifier model reckoning without memory effects, traditional linearizationcircuits’ engineering performance are not as good as theories expected. With the rapiddevelopment of broadband communication systems, study on amplifiers’ memoryeffects, and further establish more real and practical models of RF power amplifier isvery necessary. RF devices multiport behavioral model based on vector networkanalysis is currently a trend in research field of RF. Making whether memory effects isincluded as a judgment standard, behavior models can be divided into two categories:static models and dynamic models. How to evolve gradually from static models todynamic models constitutes the research route of this topic.In terms of static nonlinear behavior models, based on classical Volterra seriestheory, this paper summarizes and puts forward the envelope domain Volterra algorithm,which is a more general static behavior model compared with X-parameter model.Envelope domain Volterra algorithm perfectly contains three nonlinear sources of RFdevices, namely, harmonic nonlinear mapping, intermodulation nonlinear mapping andAM-AM transformation. Key of this Algorithm is to clarify the mapping from inputports harmonics and their interaction terms to output ports harmonics. Experimentalresults show the ability of the proposed algorithm in predicting memory-less RFdevices’ behavioral features.In view of memory effects, the research process follows sequence from the internalphysical mechanism to the externalized characterization of features. The research objectfocus on electrical memory effects, thermal memory effects and baseband effects, inother words, by theoretical derivation and experimental verification, research on howthe three factors act on amplifiers respectively. Experiments show that trap timeconstant, thermal time constant, and the frequency and amplitude of input envelopeaffect devices’ memory characteristic directly or indirectly. In this paper, specificVolterra coefficients are used to describe the asymmetry of third order intermodulation,and thus characterize memory effects. The experimental results reveal the memorycharacteristic changes with the spacing frequency of two-tone stimulation. Combiningwith the analysis of third order cut off Volterra series, a mathematical model of memorycharacteristics changing with two-tone spacing frequency is derived, and the fittingcurve matches with characterization results.From static behavioral models to dynamic behavioral models involving memoryeffects, three innovations are proposed by this paper: envelope domain memory Volterra algorithm is put forward, compared with the memory kernel of the existing dynamicX-parameter theory, the named Volterra memory kernel shows more generalsignificance; discrete dynamic X-parameter model is put forward, the experimentalresults show that the model can truly predict amplifiers’ behavior containing memoryeffects, besides, the concept of average memory kernel and time window of memory areproposed and used, making the algorithm more concise and practical; specific totwo-tone large signal stimulation, intermodulation X-parameter model theory is putforward, and experiments show that intermodulation X-parameter extracted frommeasurement platform can be directly used in ADS circuit simulation.