A system for extracellular neural interfacing that had the capability for stimulation and recording at multiple electrodes was presented. As the core of this system was a custom integrated circuit (IC) that contained low-noise amplifiers, stimulation buffers, and artifact-elimination circuitry. The artifact-elimination circuitry was necessary to prevent the activity of the stimulation buffers from interfering with the normal functioning of the low-noise amplifiers. The integrated circuits were fabricated in in a 0.35 micron CMOS process. We measured input-referred noise levels for the amplifiers as low as 3.50 microvolts (rms) in the in the bandwidth 30 Hz-3 kHz, corresponding to the frequency range of neural action potentials. The power consumption was 120 microwatts, corresponding to a noise-efficiency factor of 14.5. It was possible to resume recording signals within 2 ms of a stimulation, using the same electrode for both stimulation and recording. A filtering algorithm to remove the post-discharge artifact was also presented. The filtering was implemented using a field-programmable gate array (FPGA). The filtering algorithm itself consisted of blanking for the duration of the stimulation and artifact-elimination, followed by a wavelet de-noising. The wavelet de-noising split the signal into frequency ranges, discarded those ranges that did not correspond to neural signals, applied a threshold to the retained signals, and recombined the different frequency ranges into a single signal. The combination of the filtering with the artifact-elimination IC resulted in the capability for artifact-free recordings.

No description available.