Charge-Redistribution Based Quadratic Operators for Neural Feature Extraction.

Area of Science:

• Mixed-signal integrated circuit design
• Neural signal processing
• Biomedical engineering

Background:

• Feature extraction is crucial for analyzing neural signals.
• Quadratic operators are effective for characterizing neural activity.
• Existing methods often require significant power and area.

Purpose of the Study:

• To propose a SAR converter-based mixed-signal multiplier architecture.
• To implement and evaluate two quadratic operators (MAE and NEO) for neural signal analysis.
• To demonstrate energy computation and action potential detection capabilities.

Main Methods:

• Designed a mixed-signal multiplier architecture utilizing a SAR converter.
• Implemented programmable chips for Moving Average Energy (MAE) and Nonlinear Energy Operator (NEO).
• Fabricated prototypes in a HV-180 nm CMOS process.

Main Results:

• Experimental results validate the suitability of MAE and NEO for energy computation and action potential detection.
• Prototypes achieved low power consumption: 116 nW for MAE and 178 nW for NEO at 30 kS/s.
• The design digitizes both input neural signals and operator outputs without requiring digital multipliers.

Conclusions:

• The proposed SAR converter-based mixed-signal multiplier offers efficient neural signal feature extraction.
• The implemented MAE and NEO operators demonstrate superior area×power performance compared to existing solutions.
• This architecture is promising for low-power, high-performance neural signal processing applications.