1324 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 58, NO. 5, MAY2009 Model and Experimental Characterization of the Dynamic Behavior of Low-Power Carbon Monoxide MOX Sensors Operated With Pulsed Temperature Proﬁles Sebastian Bicelli, Student Member, IEEE, Alessandro Depari, Member, IEEE, Guido Faglia, Alessandra Flammini, Member, IEEE, Ada Fort, Member, IEEE, Marco Mugnaini, Member, IEEE, Andrea Ponzoni, Valerio Vignoli, Member, IEEE, and Santina Rocchi, Member, IEEE Abstract—Wireless sensor networks for home automation or environment monitoring require low-cost low-power sensors. Car- bon monoxide (CO) metal–oxide (MOX) sensors could be suitable in terms of device cost, but they show some severe limits, such as the need to be heated, which means large power consumption and the need for complex and frequent calibration procedures, which increases the overall cost. This paper investigates the pos- sibility to partially overcome these limits by a low-cost detection system based on a suitable commercial sensor (TGS 2442, Figaro, Inc.) and an ad hoc measurement technique exploiting speciﬁ- cally tailored temperature proﬁles. To this aim, the authors study the dynamic behavior of low-power CO MOX sensors operated with pulsed temperature proﬁles by means of two approaches: 1) sensor modeling and 2) experimental evaluation. To analyze how the sensor dynamic response changes as a function of the CO concentration, the authors individuate a temperature proﬁle, which ensures satisfactory sensitivity to the target gas and very low power consumption. Moreover, some parameters describing the sensor response shape are selected, which prove to be signiﬁ- cant in terms of both robustness to environmental conditions and calibration simplicity. Index Terms—Carbon monoxide (CO) detection, gas sensors, low-power gas sensors, metal–oxide (MOX) sensors, sensor model. I. I NTRODUCTION C ARBON monoxide (CO) detection is required in several applications, from home automation to environmental monitoring; most of these applications could take advantage of emerging low-power low-data-rate wireless technologies (e.g., IEEE802.15.4) [1]. In this context, there is an evident need for low-power, low-cost sensors. Manuscript received June 30, 2008; revised November 18, 2008. First published February 6, 2009; current version published April 7, 2009. The Associate Editor coordinating the review process for this paper was Dr. Emil Petriu. S. Bicelli was with the Department of Electronics for Automation, University of Brescia, 25123 Brescia, Italy. He is now with Camozzi Group Research Center, 25126 Brescia, Italy. A. Depari and A. Flammini are with the Department of Electronics for Automation, University of Brescia, 25123 Brescia, Italy. G. Faglia and A. Ponzoni are with the CNR-INFM Sensor Laboratory, Department of Chemistry and Physics, University of Brescia, 25133 Brescia, Italy. A. Fort, M. Mugnaini, V. Vignoli, and S. Rocchi are with the Department of Information Engineering, University of Siena, 53100 Siena, Italy (e-mail: ada@dii.unisi.it). Digital Object Identiﬁer 10.1109/TIM.2009.2012940 Some low-cost metal–oxide (MOX) gas sensors for low- power operation are already available, but they still have some well-known problems, such as the requirement of a complex calibration and a critical dependence on the environmental conditions [2], [3]. For instance, the CO sensor TGS2442 by Figaro, Inc., can be operated by a very simple circuit and man- aged by a low-cost microcontroller, but it requires at least four calibration points. In fact, the resistance value R must be mea- sured under four different conditions: 1) R with 100 ppm of CO at 20 ◦ C; 2) R with 300 ppm of CO at 20 ◦ C; 3) R with 100 ppm of CO at a temperature lower than 20 ◦ C; and 4) R with 100 ppm of CO at a temperature higher than 20 ◦ C. In addition, to limit the inﬂuence of the humidity and the interfering gas effects, the sensor is provided with a charcoal ﬁlter. Such ﬁlter could be affected by poisoning with a consequent sensor drift with time. To solve this problem, frequent recalibrations are required. This sensor has a very fast thermal response, and it is proposed by Figaro, Inc., as a low-power sensor. Nevertheless, if heated as recommended (i.e., a 14-ms-wide heater voltage pulse of 5 V with a 1 Hz repetition rate), its current consumption limits battery life to less than one month [4]. However, it must be pointed out that the heating power cannot be reduced without a previous analysis on the impact of any temperature proﬁle change on the performance of the sensor. Recently, special temperature proﬁles have been proposed to reduce power dissipation by a factor of 30 [5]. First experimental tests show a satisfactory behavior in terms of sensitivity, particularly for low concentrations (< 100 ppm). On this basis, the authors’ intention is to further investigate the dependence of the TGS2442 sensor behavior on the temperature proﬁle characteristics with the aim to enhance the sensor performance. Nevertheless, this study, if performed by experimental trials, is extremely time consuming, involving long measurement processes. To overcome this problem, and to guide the selection of an efﬁcient temperature proﬁle, we propose the application of a simpliﬁed gray-box model that was presented by the authors in some previous works [6]–[8], which could be exploited for simulating the sensor behavior when operated with different temperature proﬁles. The developed model is robust because it depends on a few parameters, and it gives satisfactorily accurate results when tested in the 0018-9456/$25.00 © 2009 IEEE