A programmable analog subthreshold biomimetic model for bi-directional communication with the brain.

Area of Science:

• Neuromorphic Engineering
• Analog Circuit Design
• Computational Neuroscience

Background:

• Volterra Kernel models are crucial for describing nonlinear systems, including neural dynamics.
• Hardware implementations are needed for efficient, large-scale neural simulations.
• Low-power analog circuits offer advantages for neuromorphic computing.

Purpose of the Study:

• To present a hardware implementation of a second-order Laguerre Expansion of Volterra Kernel (LEV) model.
• To demonstrate the model's versatility across different abstraction levels (synapse, neuron, network).
• To develop a modular analog circuit using low-power subthreshold CMOS transistors.

Main Methods:

• Implementation of a second-order LEV model using four basis functions.
• Design of modular analog building blocks with low-power subthreshold CMOS transistors.
• Evaluation of the circuit's performance by comparing its output to the ideal LEV model.

Main Results:

• Achieved a normalized mean square error of 8.15% between the circuit and ideal LEV model.
• Demonstrated a total power consumption of less than 33nW for the analog circuitry.
• The modular design facilitates scalability for large-scale systems.

Conclusions:

• The developed analog LEV circuit is a viable hardware implementation for nonlinear neural modeling.
• The low-power, modular design is suitable for building large-scale neuromorphic systems.
• This approach enables efficient emulation of multi-input multi-output (MIMO) spike transformations in neuronal populations.