Type of Document	Thesis
Author	Langoni, Diego
Author's Email Address	langodi@eng.fsu.edu
URN	etd-11212005-171421
Title	Gallium Arsenide MESFET Small-Signal Modeling Using Backpropagation & RBF Neural Networks
Degree	Master of Science
Department	Electrical and Computer Engineering, Department of
Advisory Committee	Advisor Name Title Mark H. Weatherspoon Committee Chair Anke Meyer-Baese Committee Member Simon Y. Foo Committee Member
Keywords	MESFET Modeling RBF Backpropagation Artificial Neural Networks Intrinsic Ecps
Date of Defense	2005-10-19
Availability	unrestricted
Abstract The small-signal intrinsic ECPs (equivalent circuit parameters) of a 4x50 µm gate width, 0.25 µm gate length GaAs (gallium arsenide) MESFET (metal semiconductor field-effect transistor) were modeled versus bias (voltage and current) and temperature using backpropagation and RBF (radial basis function) ANNs (artificial neural networks). The resulting ANNs consisted of 3-input, 8-output models of the MESFET ECPs and were compared to each other in terms of memory usage, convergence speed, and accuracy. Also, each network’s performance was evaluated under “normal” training conditions (75% training data with a uniform distribution) and “stressed” training conditions (50% and 25% training data with a uniform distribution, 75%, 50%, and 25% training data with a skewed distribution). The results showed that for the RBF network, much better overall convergence speed as well as better accuracy under both “normal” and “moderately stressed” training conditions were obtained. However, the backpropagation network yielded better accuracy for the “extremely stressed” training conditions and better overall memory usage.
Files	Filename       Size       Approximate Download Time (Hours:Minutes:Seconds)     28.8 Modem   56K Modem   ISDN (64 Kb)   ISDN (128 Kb)   Higher-speed Access    Thesis_manuscript.pdf 921.19 Kb 00:04:15 00:02:11 00:01:55 00:00:57 00:00:04