A low-power neuromorphic bandpass filter for biosignal processing

Various types of biosignals originating from the human body are being extensively used for diagnostics as well as therapeutic interventions. Low-power biological signal processing necessitates energy-efficient filter blocks for time-frequency analysis. In an attempt to reduce the power consumption of an implantable biosignal processor, this paper presents a neuromorphic low-power bandpass filter with excellent figure-of-merit. The charging and discharging profiles of different ionic channels of a Si neuron are utilized to achieve the bandpass filter characteristics. The entire filter structure constitutes 5 transistors working in the weak-inversion saturation regions. Designed in a standard 0.13-μm CMOS process, the proposed bandpass filter consumes only 5 nW with a 0.5 V supply for a center frequency of 200 Hz. The center frequency can be tuned from 150 Hz to 1.5 KHz. The Monte Carlo simulation reveals 58 μV
_rms
input-referred noise and 1% THD for 7 mV
_p-p
of input signal. The proposed architecture also demonstrates excellent figure-of-merit.