Yapay sinir ağlari ile mikrodalga transistörünün indirgenmiş veri ile modellenmesi

Attaining today?s high technologic level of electronics systems appeared after the invention of semiconductor devices. In electronics, not only the transistors are of importance but also are the signal to noise parameters. This thesis tackles with works which are realized by occupying block box modelling in order to obtain mentioned parameters effectively. The work of the thesis is performed by using VMMK-1218, VMMK-1225, ATF-551M4, ATF-35143, ATF-55143 microwave transistors and the results which are obtained from the tests in which the transistors VMMK-1225, ATF-551M4 are used shown in this study. This study can be summarized as follows: (i.i) Noise characteristic of a microwave transistor is modelled by one polar trainning model and the practical results. (ii.i) Small signal characteristics modelling of a microwave transistor by means of two polar trainning model and practical results is performed and then interpolation and extrapolation applications are realized. (ii.ii) Small signal characteristics modelling of a microwave transistor by means of one polar trainning model. (ii.iii) Small signal characteristics modelling of a microwave transistor is performed by comparison of other modelling examples and practical results. Black box modelling of the VMMK-1225 microwave transistor is performed with the values of Vds= 1,5; 2; 3; 4V and Ids= 5; 10; 15; 20mA also frequency ranges 2-18 GHz for N-parameters, 2-45 for S-parameters. As for ATF-551M4 microwave transistor, values Vds= 2; 2,7; 3V and Ids= 10; 15; 20mA also the frequency gaps 2-18 GHz for N and S-parameters are used for its black box modelling. Resulting models are compared with each other to check the consistency and similar error rates are observed for each. These comparisons are depicted on cartesian coordinate system, polar system and smith chart. In the 4.3.3 section of the thesis work, in the topic ?Comparison of training and practical values with other models?, VMMK-1225 microwave transistor as a test device and the values of Vds= 1,5; 4V training data and Vds= 3V practical test value are used and seperated incidences are realized for MLP, RB, RBE and GRNN and finally the results are demonstrated on the same graphic. Either traning-practical test results performances (drawing 4.15 and 4.16) or graphics (figures 4.53-60) show the success of the occupied neural network. Especially, just as the achievement of the modelling technique are observed in the figures 4.6 and 4.17, so too instantaneous flactuations are detected by GRNN. Moreover, figure 4.54 can be referred to observe the success of comparative models. Also optimum source reflection angles can also be seen. Keywords: Noise Figure, Scattering Parameters, Reciprocal, Optimum Noise Reflaction Coefficient, Artificial Neural Networks, Regression.