A frequency shaping neural recorder with 3 pF input capacitance and 11 plus 4.5 bits dynamic range.

Area of Science:

• Biomedical Engineering
• Neuroscience Technology
• Integrated Circuit Design

Background:

• Conventional neural recording systems face limitations in electrode offset rejection, input impedance, and dynamic range compression.
• Simultaneous recording of local field potentials (LFPs) and extracellular spikes is crucial for comprehensive neural data analysis.
• Long-term neural recording experiments require robust architectures with low power consumption and high data fidelity.

Purpose of the Study:

• To present a novel frequency-shaping (FS) neural recording architecture.
• To implement and characterize the FS architecture in a 0.13 μm CMOS process.
• To evaluate its performance against conventional counterparts for enhanced neural data acquisition.

Main Methods:

• Designed and fabricated a neural recording architecture utilizing a frequency-shaping (FS) technique in a 0.13 μm CMOS process.
• Integrated FS amplifiers, buffers, and an 11-bit successive approximation register analog-to-digital converter (SAR ADC).
• Measured power consumption, input impedance, noise performance, and data dynamic range (DR) at a 40 kHz sampling clock.

Main Results:

• The FS architecture inherently rejects electrode offset and increases input impedance 5-10 fold.
• Achieved neural data dynamic range compression by 4.5-bit, enabling simultaneous LFP and spike recording.
• Demonstrated low power consumption (50 μW/ch) with high-fidelity input-referred noise (13 μVrms for LFPs, 7 μVrms for spikes).

Conclusions:

• The proposed FS neural recording architecture offers significant advantages for long-term neural monitoring.
• The design provides high-fidelity, full-spectrum neural data with enhanced dynamic range and efficiency.
• The implemented CMOS chip demonstrates suitability for advanced brain-computer interfaces and neuroscience research.