Study And Design Of Front-End Integrated Circuits For Implantable Neural Recording System

Neural recording system has become the main methodology in neuroscience research. In the system, one of the most critical parts is the front-end circuits, which directly affect the performance of the system. A lot of new challenges in designing front-end circuits with lower power consumption, higher noise performance and smaller chip area are introduced by the new implanting application and the characteristics of neural signal. The main works of this thesis are designing neural front-end circuits for specific implanting applications.Firstly, according to the requirements of specific neural recording experiments, a new low-power and low-noise neural signal recording front-end module with variable gain and bandwidth is designed using 0.18μm standard CMOS technology. This module is consist of a preamplifier, a variable gain amplifier(VGA) and a Gm-C filter. In the design of preamplifier, the input-referred noise is further decreased without increasing the power consumption through current-reuse technology. The Gm-C filter can generate a low-pass corner frequency about 600 Hz without using large capacitor. By varing the tranconductance of the operational transconductance amplifier(OTA) in the Gm-C filter, the low-pass corner frequency can be adjusted. The testing results of the prototype chip verified the design. The chip(including the bias circuits) draws a total current of 22.3μA from a 1.8V supply voltage. The input-referred noise voltages of the front amplifier are 4.43μVrms and 6.74μVrms with the bandwith set as 200 Hz to 5kHz and 1Hz to 5kHz respectively. The gain of the front-end module can be varied in the range of 46dB~64dB. The adjustable high-pass corner frequency and low-pass corner frequency are 0.1Hz~600Hz and 600Hz~8kHz respectively.Secondly, in order to improve the spatial resolutions of multi-channel neural recording systems based on Utah electrode neural sensors or Michigan electrode neural sensors, a compact, low-noise and high-input-impedance neural recording front-end circuitshas been designed based on 0.18μm CMOS technology. This front-end module includes a preamplifier and a band-pass filter. In the design of the preamplifier, current feedback is achieved in a modified fully differential folded cascode amplifier to improve the input impedance of the preamplifier. Meanwhile, by using this structure, the uses of large on-chip capacitors is eliminated in this design, and thus the chip area of the preamplifier is futher reduced. The simulation results are presented in this paper. The mid-band gain of the amplifier is 46 dB. The high-frequency roll-off is 10 kHz and the low-frequency roll-off is adjustable in the range of <1Hz to 600 Hz. The entire module consumes a total current of 4.67μA from a 1.8V supply voltage(not including the bias circuits). The input-referred noise voltages are 4.8μVrms and 3.89μVrms integrated from 1Hz to 10.6kHz and 200 Hz to 10.6kHz respectively. The whole front-end circuits occupied a chip area of 0.0397mm2.