The reliable and low-cost quantitative detection of ethylene for food/fruit applications remains an unsolved problem.
Existing commercial systems are able to quantify ethylene (at sub ppm levels) but either they are off-line: require
periodic sample collection and use of reagents or high-cost. We will report on the development of an RFID reader with
onboard micro-machined metal oxide gas sensors aimed at monitoring climacteric fruit during transport and vending.
The developed platform integrates a commercial off the shelf inductive coupling RF transceiver in the 13.56MHz band,
fully compliant with the ISO15693 standard, micro-hotplate gas sensors, driving and readout electronics. If the sensors
are operated at a fixed temperature, the reader could work as an alarm level monitor able to assess the conservation stage
of apples. On the other hand, when the sensors are operated under an optimised temperature-modulation mode, accurate
calibration models for the species that are relevant to assess the conservation stage of apples (i.e., ethylene, acetaldehyde
and ethanol) can be built. Finally, different feature extraction techniques such as the FFT and the Energy Vector will be
used in combination with pattern recognition tools like PLS and PLS-DA to show that our system is able to identify and
quantify the species that are relevant for the application considered.
Metal oxide gas sensors suffer from lack of selectivity and response drift. The use of sensor dynamics has been introduced to ameliorate sensor performance. The usual approach consists of modulating the operating temperature of gas sensors. Temperature modulation alters the kinetics of the adsorption and reaction processes taking place at sensors' surface. This results in response patterns that are characteristic of gas/sensor pairs. Despite the fact that a great deal of work has been done, the selection of the modulating frequencies remains an obscure and non-systematic method. A new approach to systematically select frequencies is discussed. The method is based on the use of pseudo-random binary sequences (MLS) to modulate the working temperature of gas sensors in a wide frequency range. The impulse response of a pair sensor-gas can be estimated from the circular cross-correlation of the MLS and the sensor response sequences. From the study of the impulse response in the frequency domain, an identification of the modulating frequencies that
convey important information to both identify and quantify gases is obtained.