Simulated And Experimental Research On The Package Based RF System

RF designer and packaging designer are facing the challenge to make electronic products with high density,high stability and high functionality.RF system-in-package (SIP)is a technique to solve this problem.An accurate electromagnetic simulation and an effective design procedure are the primary factors for the development of this technique.The topics studied in this thesis come from Digital Beamforming(DBF)channel receiver design.This thesis contributes to the development of novel methods and techniques for electromagnetics(EM)-based modeling and design of low temperature co-fired ceramic(LTCC)RF circuits exploiting space mapping(SM)and neural network (NN)modeling.We present a method of microwave circuits design combining frequency space mapped neuromodeling(FSMN)approach with Stoer-Bulirsch adaptive frequency sampling(S-B AFS)technique.In neuromodeling,frequency space-mapped technique decreases the cost of training,improve generalization and reduce the complexity of NN topology.An S-B AFS technique is used to decrease the cost of frequency sweeps in sampling.The coarse circuit models synthesized from a layout-level is exploited in frequency partial-space-mapped neuromodeling technique(FPSMN)for modeling RF circuit. FPSMN establishes a partial mapping for the physical parameters,making even more efficient use of the implicit knowledge in the coarse model.This allows us a significant reduction in NN complexity and in training time.The coarse model obtained from layout-level synthesis describes the important parameters and parasitic parameters.It can serve as a high quality coarse model in FPSMN.The S-B AFS technique is used in frequency sweep of fine model to further reduce computing time.In implementation of neural inverse space mapping(NISM)optimization,a neural network approximates the inverse mapping at each iteration and the coarse model is obtained from layout-level synthesis.Trained NN is introduced as a buffer space between the original coarse model space and the fine model space,where the ingredients of the coarse model space are completely different from those of the fine model space.By using Broyden's formula,a rapid convergence of the fine model simulation,the optimization efficiency of complicated LTCC RF positive circuit is greatly improved.A multilayer coupled stripline resonator is used to design a miniature LTCC broadband bandpass filters.By using four multilayer-layer coupled stripline resonators and the cross-shape defected ground structure(DGS),the filter exhibits wide bandwith and high attenuation in stopband.Miniaturization of RF system is more and more important,especially in DBF radar system.A C band DBF radar receiver SiP is designed and optimized in this thesis.A two-stage low noise amplifier(LNA)and IF amplifier are designed and simulated by using software.The two filters,matching network and some passive components are embedded in multilayer ceramic for miniaturizing the volume of the whole receiver.The module is suitable for fabricating the 1×2 antenna sub-array.