Embedded passive circuits design and optimization for highly integrated single package RF module

This dissertation describes highly integrated RF subsystems using embedded passive circuits such as bandpass filter, lowpass filter and diplexer as well as neural network modeling technology. A 3.6 V compact integrated RF subsystem in 13 mm x 13 mm single package has been developed using an embedded lowpass filter, diplexer and super-diplexer. Newly developed super-diplexer is based on the performance of cascaded diplexer and low/high lowpass filters. The embedded lowpass filter, diplexer and super-diplexer were built and measurement results show good agreement with the simulation data. In addition, a novel inductor resonator filters are designed. The design can achieve compact filter design even on low cost relatively low dielectric constant material. The design is based on the self-resonance frequency of the spiral inductors and electromagnetic coupling effects between resonators. The measured results have good agreement with the simulation results. The compactness of newly developed bandpass filter makes the design of bandpass filters attractive for further development and applications in single package RF system. Silicon based inductors in flip-chip structures on-package substrate are analyzed in this dissertation using full-wave electromagnetic simulation and measurement. The impact of flip-chip on inductor performance are characterized. The measurement and simulation results show that the inductance of flipped silicon-based inductor is influenced by flip-chip environment. Special design considerations are required in the flip-chip process. The equivalent circuit model of the flipped inductor was verified with measured results over a broad band of frequencies. Also, the RF flip-chip characterization technique using on-strip measurement method is presented. The transition structures of coplanar waveguide (CPW) probes to Device Under Test (DUT) on die are fabricated on low cost package substrate to measure the high-frequency response of RF flip-chip. A neural network was trained and used to model the S-parameters within single package RF system condition. One of advantages of applying neural networks to embedded passive circuit optimization was that only a small number of measured data (target data) was necessary to quickly and accurately model the S-parameter response of the circuits in complex package environment. The neural network is based on fitting calculated parameters to measure/simulated data. The full understanding of the parasitic effects from complex package environment was not required to optimize embedded diplexer in single package RF system. Using the calculated S-parameters obtained from the neural network, the single package RF system was successfully designed and then tested at the design frequency. A single package RF system allows PA power control for GSM systems. It combines all required features to support GSM power control loop. The modeled GSM power control loop will be applied to the modeling of GSM system Tx/RX chain. The highly integrated RF system using embedded passives in single package has been evaluated in terms of system level. Evaluation results are the output power of 32.68 dBm and noise figure of 7.6 dB. These results of the system evaluation indicate that the RF system in single package meets the GSM specification. As a result, reduced board area also allows phone manufactures to add additional functions such as RF radio, digital TV, Bluetooth, WiFi, image sensors, and multimedia application.