Statistically Reconstructed Multiplexing for Very Dense, High-Channel-Count Acquisition Systems.

Area of Science:

• Electrical Engineering
• Biomedical Engineering
• Signal Processing

Background:

• Traditional multiplexing architectures for multichannel systems are limited by per-channel antialiasing filters, hindering scalability for high-density, low-noise applications like neural recording.
• Existing methods face tradeoffs between recording density and noise performance, failing to achieve ideal neuron-to-sensor mapping.

Purpose of the Study:

• To present a novel multiplexing architecture that overcomes the limitations of per-channel antialiasing filters.
• To enable high-density, high-channel-count neural recording systems with improved signal-to-noise ratios.

Main Methods:

• Developed a multiplexing architecture that omits per-channel antialiasing filters.
• Employed a compressed sensing strategy for data recovery, including statistical reconstruction and undersampled thermal noise removal.
• Replaced bulky analog components with scalable CMOS-compatible digital signal processing blocks.

Main Results:

• The new architecture achieves significantly improved signal-to-noise ratios compared to conventional multiplexing with antialiasing filters at the same per-channel area.
• Implemented in a 65,536-channel neural recording system, demonstrating signal recovery performance comparable to high-performance single-channel systems.
• Achieved a four-orders-of-magnitude increase in channel density without compromising signal quality.

Conclusions:

• The proposed statistically reconstructed multiplexing architecture offers a scalable solution for high-density multichannel acquisition systems.
• This filter-less approach significantly enhances signal-to-noise ratio and channel density, paving the way for advanced neural interfaces.