STDP learning in saturated differential amplifier circuits: design and analysis

Efficient synaptic and neuron circuit design is essential for the progress of cognitive computing architectures. Spiking Neural Network architectures are being considered due to their biological importance and application potential. The traditional STDP circuit design has been challenged with issues of power consumption and operating voltage. The proposed research aims to introduce a novel circuit design for the STDP learning rule, DIFFSAT, utilizing the non-linear characteristics of differential amplifiers. The proposed method involves the design of an analog circuit for STDP with a differential amplifier operating with decaying input signals provided by a time-delay module. The proposed DIFFSAT circuit simulation results show successful operation of the circuit with both potentiation and depression effects, with reduced power consumption when compared with traditional STDP circuit design. The DIFFSAT circuit design shows successful operation within a voltage range of 1–3 V, and reduced power consumption, and the DIFFSAT-STDP circuit design provides an efficient and effective solution for biological and biologically inspired synaptic learning and its application potential with regard to cognitive computing architectures.