A new architecture for multi-channel neural recording microsystems based on delta-sigma modulation

A new multi-channel neural recording architecture is proposed. In this architecture, preconditioned neural signals are digitized based on delta-sigma (¿¿) analog-to-digital (A/D) conversion technique. A/D conversion in the proposed approach is divided into two steps. In the first step, all the channels are converted into serial bit streams using per-channel ¿¿-modulators. All the neural bit streams are then packaged and sent to an external setup, where the second step of ¿¿-A/D conversion, i.e. decimation filtering, is performed. Moving the decimation filters to the external side will significantly save power and silicon area for the implantable device, both of which are of crucial importance in implantable microsystem design. The system presented in this paper, is an 8-channel neural recording system, designed using first-order ¿¿-Modulators in a 0.5-¿m 2P3M n-well CMOS process. The system, with 8 bits of resolution for quantization, demonstrates Signal-to-Noise and Distortion Ratio (SNDR) of 49 dB, and dissipates as low as 16.5 ¿W power per channel at 3 V.