Measurement forecast of anomalous threshold voltages in BCD LV submicron n-MOSFETs with two artificial intelligence methods

In this study, two intelligent methodologies were used to estimate the anomalous threshold-voltage (Vth) measured behaviors in sub-micrometer Bipolar-CMOS-DMOS (BCD) low-voltage (LV) MOSFETs by using the grey system (GS) GM (1,1) model and a fuzzy-neural network (FNN). This paper describes the implementation procedures of these two models for making Vth predictions. Moreover, discrepant comparisons between the GS and FNN output data are also demonstrated. Eventually, only the outputs of FNN can have the complex action of reverse short-channel property. Then, it will be developed to analyze the Vth inclination of submicron n-channel MOSFETs due to the device geometric effect. A comparison between the measured characteristics of Vth and the characteristics of Vth predicted by the FNN shows good agreement for a wide range of channel lengths, widths and bias conditions. And, the maximum error percentage was less than 0.08%. As such, the developed procedure may be well suited for the data estimation of the complicated BSIM-model parameters in foundry fabrications.