Automatic Extractionand Optimization Of The BJT Device Model Parameter

With the development of BCD (Bipolar-CMOS-DMOS) technology,there is a higher demand for the device characterization of BJT (BipolarJunction Transistor) and for the extraction of model parameters. On onehand, circuit designers desire that the model is able to reflect thecapability of IC (Integrated Circuit) fabrication techniques, which wouldhelp to increase the product yield. On the other hand, designers wish touse the model as early as possible in order to simulate the circuits andrelease the products. Unfortunately, current extraction method of BJT isnot as advanced as that of MOS (Metal-Oxide-Semiconductor), and it canhardly meet the requirement for both accuracy and efficiency.Under such circumstances, this thesis illustrates an optimized methodto extract BJT model parameters, where electrical parameter fitting isutilized in addition to the traditional method of characteristic fittingcurves. In this way, the model is more likely to simulate the devicecharacterization in reality. Besides, parameters on circuit level areextracted as well, which make the simulation results of individual devicescloser to the simulation of that in overall circuits. This would furtherenhance the accuracy of model and assist IC designers in various ways.Furthermore, an automated scheme is also given in the thesis, so as tocompensate for the prevailing lack of customization of commercialsoftware. This scheme not only integrates the parameter extraction ofboth device level and circuit level but also improves the traditional Levenberg-Marquard and Nelder-Mead optimization algorithms with acustomized Perl program, thus make the BJT model possible for bettercustomization, easier automation and wider application.