A TDMA Neural Recording SoC With IIR-RLS Adaptive Filters for 83.4 dB Artifact Suppression Across 256 Channels

Abstract

This article introduces a digitally-assisted, multi- electrode neural recording system equipped with a multi-channel stimulation artifact canceller (SAC) module. The system employs a 16-electrode time division multiple access (TDMA) scheme, allowing multiplexed neural signals to be recorded through a single shared analog front end (AFE). Simultaneously, it cancels stimulation artifacts from 16 stimulation electrodes using an infinite impulse response (IIR) recursive least squares (RLS) adaptive filter (AF). The proposed system-on-chip (SoC) is validated in vitro, demonstrating an instantaneous, real-time sup- pression of 100 mV stimulation artifacts by 83.4 dB, with rapid AF calibration convergence times under 5.1 s for 256 channel transfer function (TF) combinations. Fabricated using a 65 nm process, each recording electrode AFE occupies 0.0018 mm 2, achieves an input-referred noise of 4.9/4.8 µVrms for local field potential (LFP) and action potential (AP) bands respectively, and consumes 5.8 µW in recording mode. The SAC module consumes 2.9 µW/TF and minimally impacts recording signal- to-noise and distortion ratio (SNDR) by 0.8 dB. This results in a high-performing SAC chip, optimized for energy and area efficiency.