A Fully Analog Circuit Topology for a Conductance-Based Two-Compartmental Neuron Model in 65 nm CMOS Technology.

Neuromorphic circuits offer a means to emulate or interface with the nervous system, yet detailed model implementations remain underexplored. This paper proposes an analog CMOS realization of the Pinsky-Rinzel (PR) neuron model, a conductance-based two-compartmental model of hippocampal CA3 pyramidal neurons capable of producing spike and complex bursting. To achieve first-order dynamics, we introduce a seven-segment approximation circuit with a customizable architecture based on the variable's behavior. For second-order dynamics, we modify the architectures of the log-domain low pass filter and Tau-cell. To enhance power and area efficiency, we implement conductances and reversal potentials using current mirrors within circuit blocks. Additionally, we propose four compact subthreshold multipliers specifically tailored for neuromorphic systems, while eliminating any unnecessary DC current across circuit blocks. These modifications led to a 150-fold reduction in power consumption compared to state-of-the-art conductance-based implementations. Simulation results demonstrate the independent operation of soma and dendrite compartments for spike generation, while coupling generates complex bursting. Realized in TSMC 65 nm CMOS technology with a 0.2 V subthreshold power supply, the circuit operates in the 50–300 kHz frequency range, showcasing an average power consumption of 125.1 nW. Finally, Monte Carlo simulations for both spiking and complex bursting modes are evaluated in 200 points. These results not only validate the functionality of the proposed circuit but also position it as a versatile platform for implementing various conductance-based models, including astrocytes. [ABSTRACT FROM AUTHOR]