Design of a low-power compressive sampling circuit for Gaussian sensing matrices

The new theory of compressed sensing concerns the acquisition and recovery of sparse signals from their sub-Nyquist sampled data. This paper presents an analog implementation of compressed sensing with Gaussian sensing matrices. The proposed circuit is based on charge redistribution similar to the successive approximation analog-to-digital converters and produces a random sensing matrix with Gaussian entries of 8-bit resolution. This circuit was implemented in 0.18μm CMOS process. Simulation results show that the circuit achieves a percentage root-mean-square difference of 2% or less when the input signal is composed of up to 47 different sine waves all having frequencies less than 25 kHz. The total (simulated) power consumption of the circuit is approximately 12.5μW at 1-volt supply voltage and sampling rate of 50 kS/s.