CHARACTERIZATION AND DESIGN OF BROADBAND MICROWAVE AMPLIFIERS

Abstract

The device model of a transistor, which can include the small-signal model, or the large-signal model, is extremely important for active microwave circuit work. The model provides a vital link between the measured S-parameters and the electrical processes occurring within the device. In this dissertation, the results of the investigation of some new algorithms for the characterization of both the microwave transistor and the microstrip discontinuities are reported. The derived model is subsequently applied in some new design methods for broadband low and high power microwave amplifiers. Conventionally, the parameter extraction of the small-signal has been mostly approached using the Gradient Descent algorithms. In this dissertation, tl1e Simulated Annealing method is employed in the small-signal equivalent circuit parameter extraction. Various variants of Simulated Annealing are thoroughly investigated. Two novel methods, namely: (i) the New Adaptive Very Fast Simulated Annealing methods and (ii) the hybrid Gauss-Newton Simulated Annealing algorithm, are in turn presented. Simulated results for the various variants of annealing methods are compared. In this dissertation, two approaches, namely, one using directly the small-signal S-parameters has been used to design a distributed amplifier and the other, utilizing the small-signal equivalent circuit model has been used to design a linear broadband, high power amplifier. As an application of the first approach, a new method of analysis based on the S-parameter formulation for the distributed amplifier design is proposed. The proposed approach, which is independent of the type of configuration or equivalent circuit used, allows circuit designers the ease to design one's own configuration without ever re-deriving the gain response or pelfonn parameter extraction as compared to the conventional analyses. This work is verified by an experimental design. A prototype of a 1-7 GHz, 10 dB gain distributed amplifier is fabricated and tested with the proposed method. Since the noise model and the small-signal equivalent circuit model are closely correlated, the correlated, the use of the device's S-parameters for noise analysis is thus investigated. In this dissertation, some new noise analysis methods that allow concurrent evaluation of the noise properties and the S-parameters of the device without any domain transformation are studied Two methods, namely, (a) the T-wave method and (b) the S-wave method, are proposed, and of these, the S-wave method, which is recursive, is found to be suitable for CAD implementation. Numerical comparisons on two simple examples for each of the two new methods are performed in this dissertation. For the second approach which makes use of the small-signal equivalent circuit model, a linear broadband, high power amplifier design is investigated. A prototype of a 1-6.4 GHz , 25 dB gain, 1 Watt power amplifier is fabricated and theoretical explanations for the nonlinear behaviour of the measured currents are also given. In addition, a new large-signal model is derived for completeness. Two new current expressions for both the Igs and Ids are presented for the first time. In the design of the distributed amplifier, it is found that the microstrip T unction plays a very critical role in the detennination of the amplifier's response. As a generalized case of the microstrip T-junction, the arbitrary angle microstrip Y-junction is selected for analysis. A full-wave electromagnetic analysis using the spectral domain method is adopted. Some novel basis functions are implemented and a new pole extraction method is presented in the dissertation. Two dedicated numerical schemes that allow fust convergence are also suggested. Intensive numerical comparisons with published data and experimental results are also performed. Two microstrip Y-junctions of varying angles are also fabricated and tested with the new method of analysis. The resulting numerical results for the two cases agree quite well with the simulated results.