A new approach for the extraction of an FET equivalent circuit from measured S parameters

Abstract

Figure 4 Characteristic impedance and dispersion characteristicsfor the dominant and first two higher-order modes of shielded dielectric-loaded edge-coupled CBCPW structure with a = 4 mm, h , = h , = 1 mm, E ,~ = 2.22, and other parameters as in Figure 3 choosing appropriate loading and structural parameters, one can optimize the behavior with respect to isolation, singlemode bandwidth, and uniformity of coupling. ABSTRACT A new method for the entraction of the 13-element GaAs FET model from hot S parameters is presented. The method reduces the uniqueness problem that mists between the parasitic gate resistance R, and the channel resistance K,. Ihirteen error functions are used, with the order of o~~rimiza~ion determined through a principle-components semihi& analysis. 0 ABSTRACT In this work we present a new optical& controlled microwave matching technique. In this technique the gap between two microwave coupled

Performance and limitations of decomposition-based parameter extraction procedures for FET small-signal models

“…It is important to note that Kondoh did not optimize all the model parameters across the complete set of frequency points. This subdivision in frequency is not used in the new method of Van Niekerk and Meyer [5] since it was found to not provide any accuracy improvements. Kondoh subdivided his optimization problem to obtain eight subproblems.…”

“…A principle components sensitivity analysis [5], [8] is used to order the model elements in descending order of their influence on the global-error function. The suboptimization problems are then optimized in this order.…”

“…The principle components' sensitivity analysis can, therefore, be viewed as the study of the eigenvalues and eigenvectors of the Hessian matrix. An example illustrating how this is done is presented in [5].…”

“…Both achieve good results. Van Niekerk and Meyer [5] recently demonstrated a very robust and powerful parameter-extractor based on the method proposed by Kondoh [6]. Kondoh showed that good results can be achieved by breaking the optimization problem into eight subfunctions and optimizing the different model elements only with respect to specific subfunctions.…”

“…In particular, the effects of starting values and the final distribution of results have been largely neglected. In this paper, the results of such tests are presented for the algorithm proposed in [5], and for the original algorithm proposed by Kondoh [6]. Robustness tests are performed using a large number of randomly chosen element starting values for each extraction.…”