Design of RF MEMS Based Oscillatory Neural Network for Ultra High Speed Associative Memories

• DOI: 10.1007/s11063-013-9312-y
• Publication Date: 2013-06-28T11:15:16Z
• Authors: Masoud Baghelani; Afshin Ebrahimi; Habib Badri Ghavi...
• Keywords: Initialization; Hebbian theory; Memristor; Resistive touchscreen

Microelectromechanical systems are utilized alongside with transistor amplifiers and resistive connections for implementing of oscillatory associative memories. Phase locking is studied in such a network and all requirements of the circuit level implementation are satisfied. A very high gain trans-impedance amplifier operating in 1 GHz in addition to a novel automatic amplitude control circuit is employed to remove amplitude dynamics of the system. Requiring resonator characteristics are extracted and calculated as well. A new method for initialization of the network is proposed. Each neuron consumes 1.08 mW from a 1.8 V power supply. The convergence time of a typical network trained by Hebbian rule is less than 1.5 ns which results in an ultra high speed analog signal processing system.