An Ultra-Low-Power Front-end Neural Interface with Automatic Gain for Uncalibrated Monitoring

This paper presents a dynamic front-end towards achieving unsupervised single-neuron activity monitoring. By implementing at the front-end, an automatic gain control that is optimized for neural signal dynamics, subsequent processing can be achieved without the need for calibration. The system uses three amplification stages (low-noise first stage, variable-gain second stage and high-gain third stage), a tuneable high-pass filter, and a feedback loop to tune the variable gain. The circuit has been implemented in a commercially-available 0.18 mu m CMOS technology with total power consumption between 1.79 and 1.95 mu W. The front-end achieves a variable gain from 52 to 86.4 dB with 3 kHz bandwidth and a high-pass filter that is tuneable from 100-300 Hz. The input referred noise is 9.66 mu V with a total harmonic distortion of under 1%.