An Area- and Energy-Efficient 16-Channel, AC-Coupled Neural Recording Analog Frontend for High-Density Multichannel Neural Recordings

Hyeon-June Kim,Kyungsik Eom,Younghoon Park,Sung-Yun Park

doi:10.3390/electronics10161972

Abstract

We present an AC-coupled modular 16-channel analog frontend with 1.774 fJ/c-s∙mm2 energy- and area-product for a multichannel recording of broadband neural signals including local field potentials (LFPs) and extracellular action potentials (EAPs). To achieve such a small area- and energy-product, we employed an operational transconductance amplifier (OTA) with local positive feedback, instead of a widely-used folded cascode OTA (FC-OTA) or current mirror OTA for conventional neural recordings, while optimizing the design parameters affecting performance, power, and area trade-offs. In addition, a second pole was strategically introduced in the LNA to reduce the noise bandwidth without an in-channel low-pass filter. Compared to conventional works, the presented method shows better performance in terms of noise, power, and area usages. The performance of the fabricated 16-channel analog frontend is fully characterized in a benchtop and an in vitro setup. The 16-channel frontend embraces LFPs and EAPs with 4.27 μVrms input referred noise (0.5–10 kHz) and 53.17 dB dynamic range, consuming 3.44 μW and 0.012 mm2 per channel. The channel figure of merit (FoM) of the prototype is 147.87 fJ/c-s and the energy-area FoM (E-A FoM) is 1.774 fJ/c-s∙mm2.

Highlights

Published: 16 August 2021An in-depth understanding of the brain’s activities will require large-scale recordings from multiple neuronal structures
The requirements for extracellular neural recording are high-density and high-quality signal acquisitions without unnecessary interventions for long time periods, which poses huge challenges to the engineering works for it. Those challenges when focusing on the integrated circuit design for neural recording frontends can be enumerated as follows: considering the dense population of neurons in the brain (e.g., >1000 neurons within a radius of 140 μm in the rat cortex [1]), multichannel neural recording frontend circuits that fit into a small area of the brain are highly required; due to the tiny amplitude of extracellular neural signals (~100 μV) and their high dynamic range (DR) (~60 dB), decent quality recording must be provided [4]; low-power operation of the neural recording frontends is essential because heat dissipation from the highdensity neural recording frontend circuit can negatively affect living issues
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations

Summary

Introduction

An in-depth understanding of the brain’s activities will require large-scale recordings from multiple neuronal structures For such large population recordings, extracellular neural recording has been recognized as one of the most powerful techniques due to its high spatial and temporal resolutions, related research tools for extracellular neural recording have been steadily advanced [1,2,3]. The requirements for extracellular neural recording are high-density and high-quality signal acquisitions without unnecessary interventions for long time periods, which poses huge challenges to the engineering works for it. Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Methods

Results

Conclusion