E-Nose: Multichannel Analog Signal Conditioning Circuit With Pattern Recognition for Explosive Sensing

Abstract

This paper presents E-Nose, a novel cost-effective, field-deployable portable system that constitutes a 4-channel signal conditioning circuit and multi-coated piezo-resistive micro-cantilever sensors for explosive sensing. E-Nose also features an embedded PCA and K-means based pattern recognition (PR) algorithm for the classification of explosives from non-explosives. The 4-channel configuration is a stack of two 2-channel circuits that are capable of measuring the change in the sensor resistance or capacitance in four optional modes of ΔR-Δ R,ΔR-ΔC,Δ C-ΔR, and ΔC-ΔC by using time multiplexing. The circuit uses a bidirectional AC current excitation method to drive the sensor bridge for significant reduction of DC offset errors, 1/f noise, line noise, and DC drifts. The proposed signal conditioning circuit uses the phase-sensitive synchronous rectification(PSSR) method for AC-to-DC conversion by using balanced demodulation. The circuit can measure a wide range of resistors that range from 100 Ω to 4 MΩ, with a sensitivity of 0.4mV/ppm and the worst relative error of 2.6%. The capacitive measurement range is from 100pF to 100 μF with the worst relative error of 3.3%. The entire data processing and the PR algorithms run on Raspberry Pi (R-Pi), which is integrated into the E-Nose system. The system performance is tested with MEMS cantilevers for the detection of explosive compounds, such as TNT and its derivatives, RDX and PETN in a controlled environment at a concentration that was as low as 16ppb TNT, 56ppb RDX and 134ppb of PETN. Measurements show that the E-Nose can detect explosives with 77% as true positive results without considering the environmental and mixed vapor effects.