CALORIMETRIC SENSORS 305 Gall (1993) showed that with a MEMS device, three methods of signal enhancement were possible. Figure 7.14 shows results he obtained for different gases as a function of inverse temperature. These results were used to define the operating temperature for an array of sensors with identical Pt catalyst set to different temperatures. The tested gases were methanol, ethanol, pentane, trichloroethylene, carbon monoxide, and methane. The results were analyzed with a neural network. All test gases were separated, although the results were only barely resolved for methanol and ethanol (as expected, given the similarity of results for those two gases in Figure 7.14). The second method of enhancement was to utilize different catalysts in the array. By using an additional Ir-coated sensor in the array, it was possible to achieve a larger separation between methanol and ethanol. The third method was to vary temperature as a func- tion of time on a single device. Using this technique, Gall was able to separately identify four different alcohol vapors. Niebling et al. (1996) obtained similar results With a MEMS device, Engel et al. (2004) used stepped temperature operation of a calorimeter to distinguish between gasoline and diesel fuels. In this work, linear discriminant analysis combined was used to select the temperatures that produced the greatest information content for discrimination; these temperatures were then used in an artificial neural network to provide discrimination in one cycle (2 s) of data. In a different approach, sinusoidal modulation of the temperature can be obtained using a feed- back network to control the delivered power to a micro-heater (Aigner et al. 1996; Niebling et al. 1996). The sinusoidal modulation of temperature is imposed in addition to a gradual sweep in temperature by a steadily increasing unmodulated component to the power. In the absence of a combustible gas, the