Low-power low-data-rate analog front-end for neural recording system.

摘要

An analog stage for multi-channel neural recording requires a number of wires comparable to the number of channels, leading to restriction of movement and poor scalability. To overcome this problem, an implantable neural recording system is desirable which is operated by a power and data transfer through wireless inductive coupling.;However, a data-rate problem arises when recording from a large number of multi-channel sensors in transcutaneous wireless systems, since the data rate is typically limited to few Mbps. With the limited bandwidth, a neural recording system having sampling rate of even few tens of KHz with 10-bit quantization resolution could handle only few channels of neural information simultaneously over a wireless link. A low-power low-data-rate analog front-end, for implantable neural recording system that can process 32-channel neural spikes as well as local field potential (LFP), is proposed.;32-channel low-power low-noise amplifiers with bandpass-characteristic filter stages were fabricated on AMI CMOS 0.6um technology and tested with a custom designed general purpose test station interfacing with PC. Additionally, a low-power low-noise preamplifier stage was fabricated on MITLL FD-SOI (Silicon on Insulator) 0.18um technology and tested. An efficient power harvesting system in CMOS for bio-implantable devices was designed and simulated. Based on experimental testing of the fabricated IC chips, improved designs of the analog front-end circuits were simulated, and an alternative approach for low-data-rate neural recording system was proposed.