Modeling and Comparing the Impact of Resistive and Capacitive Crossbar Associated Parasitics of Neuromorphic Circuits

Neuro-inspired computing systems have shown a promising future in artificial intelligence. To meet the performance requirement of neuro-inspired computing systems, various synaptic devices are being used to store the weights. Resistive synapses have been in focus, but recently, capacitive synapses have emerged as a potential candidate for the synaptic device. Increasing complexity in a neuromorphic system causes the interconnect parasitics to significantly affect the circuit's performance. Therefore, this paper provides a comprehensive study on how the resistive and capacitive synapse-based crossbar parasitics affect the performance of the system. To extract the parasitic components, a Q3D model is designed considering both the resistive and capacitive synapses. A complete crossbar-associated parasitics architecture is designed that can be used to analyze the crossbar of any dimension. The analysis is done to compare the parasitic effect of both the resistive and capacitive synapses on different spiking patterns of the Izhikevich neuron for different crossbar sizes. Based on the results, the capacitive synapse-based crossbar array outperforms the resistive synapse in terms of reduced signal distortion and delay. The circuit implementation is done using 180 nm standard CMOS technology.