Compressed Sensing Analog Front-End for Bio-Sensor Applications

Abstract

In a conventional bio-sensor, key signal features are acquired using Nyquist-rate anal og-to-digital conversion without exploiting the typical bio-signal characteristic of sparsity in some domain (e.g., time, frequency, etc.). Compressed sensing (CS) is a signal processing paradigm that exploits this sparsity for com- mensurate power savings by enabling alias-free sub-Nyquist ac- quisition. In a severely energy constrained sensor, CS also elim- inates the need for digital signa l processing (DSP). A fully-inte- grated low-power CS analog front-end (CS-AFE) is described for an electrocardiogram (ECG) senso r. Switched-capacitor circuits are used to achieve high accuracy and low power. Implemented in 0.13 µm CMOS in 2×3 mm², the prototype comprises a 384-bit Fibonacci-Galois hybrid linear feedback shift register and 64 dig- itally-selectable CS channels with a 6-bit C-2C MDAC/integrator and a 10-bit C-2C SAR ADC in each. Clocked at 2 kHz, the total power dissipation is 28 nW and 1.8 µW for one and 64 active chan- nels, respectively. CS-AFE enables compressive sampling of bio- signals that are sparse in an arbitrary domain.