Electrical Characteristics of Neuron Oscillation Circuits Composed of MOSFETs and Complementary Unijunction Transistors

Abstract

A pulse frequency modulation (PFM) type neuron circuit composed of a metal-oxide-semiconductor field effect transistor (MOSFET) and a complementary unijunction transistor (CUJT) was fabricated on a silicon-on-insulator (SOI) structure as an approach to an adaptive learning neuron circuit using a metal-ferroelectric-semiconductor field effect transistor (MFSFET). The output pulse interval was controlled by changing the pulse duty ratio of input pulse signals as well as the magnitude of the DC input voltage. These results demonstrate that the electrical properties of MOSFET neuron circuits are good enough for future neural networks, that can be expected to be realized by replacing the MOSFET with an MFSFET. A new circuit using a complementary MOS (CMOS) Schmitt-trigger configuration is also proposed for improving output pulse amplitude.