An integrated, low noise neural recording system

Simultaneous recording of neural signals from multiple electrodes is of great interest to electro-physiologists and neuroscientists. An integrated, low noise neural recording system enables the recording of neural signals in a moving subject. This is an important subsystem which enables useful applications such as neural prosthetics and brain-computer interfaces.;For such subsystems and applications it is desirable to have as many channels as possible in one chip. Additionally the die area and power consumption of each channel has to be as small as possible when considering in-vivo applications. Low voltage operation is desirable for portable applications using batteries. Here we describe a neural recording chip and the design considerations that are important for the realization of these specifications.;First, we summarize the requirements for the preamplifier and survey the designs in the literature. We then describe our 1st and 2nd generation, low power, low noise, capacitive feedback preamplifiers.;Second, we discuss the system design issues for multi-channel neural recording systems. We describe tradeoffs among analog power consumption, digital power consumption and silicon area for the optimal system design.;Third, we describe our 1.5-V, 8-bit single channel recording unit prototype, including a novel preamplifier and an ADC. The widely used large gain, capacitive feedback preamplifier involves use of large on-chip capacitors, which translates to a large die area. We propose an AC-coupled, current mode based, variable gain preamplifier which occupies small silicon area and has high input impedance. Artifact suppression functionality is also included for in-vivo application. After we evaluation of various ADC architectures, including algorithmic ADCs, incremental Delta-Sigma ADCs and SAR ADCs we chose to implement a successive approximation (SAR) ADC with a split capacitor. Our one channel prototype system only occupies 0.07 mm2 and consumes 50 muW.;Finally in order to investigate neuronal behavioral, electrical stimulation is used in a number of research and clinic fields as well as electrical recording. After stimulation, the stimulation artifact might last more than 100 ms. We further propose a preamplifier with artifact suppression circuitry to solve this problem. The design is verified in simulation and results in ten times less discharge time.