An analog integrated signal processing circuit for on-chip diffusion-based gas analysis

In diffusion-based gas analysis, the transient of gas diffusion process is recorded by a generic gas sensor to serve as a fingerprint for qualitative and quantitative analysis of gaseous samples. Following the acquisition of these specific signals, any standalone gas analyzer requires a pattern recognition system for pattern classification. The classic digital pattern recognition methods require computing hardware of adequate computational throughput. In this paper, we have followed a straightforward mathematical procedure to relate the signals to their associated target gases. We have shown that the procedure can be implemented by a set of analog functions. Based on the results, we have designed an analog integrated circuit, in 0.18 µm standard CMOS process, for processing the diffusion-based transient signals. The main circuit components are a low-pass filter, the differentiator, the feature extractor and an artificial neural network. The output of the circuit is a 2-bit binary code that specifies the target gas. The circuit successfully classified four alcoholic vapors by processing the experimentally obtained response patterns. The proposed signal processing circuit, the semiconductor gas sensor and the diffusion channel can all be implemented on a single substrate to fabricate an integrated micro gas analyzer. (paper)