Performance Evaluation of Chopping and Auto-Zeroing based Time-Division Multiplexed High-Density Analog Front-End for Neural Recording

A wide range of neural recording analog front-end (AFE) architectures in the context of closed-loop neuro-modulation SoCs have been presented in the literature. A large majority of the reported time-division multiplexed (TDM) neural AFE operate with the chopping technique to break the trade-off associated with area, offset, noise and power. While only one architecture proposes a non-conventional auto-zero (AZ) based AFE, the proposed work aims to provide a comparative performance evaluation of chopping and auto-zeroing architecture. The comparison is provided in the context of merging the analog multiplexer with the auto-zero and chopping switches for an efficient TDM-based high-density AFE. Post-layout results of the two 32 -channel designs implemented in a 65 nm CMOS process at 1.2 V supply voltage show that the chopper approach has the best noise performance while suffering from reduced input impedance. Post-layout performance metrics, along with design choices for a fair comparison of the two architectures, are detailed in this work.